Desmurget and Grafton – Forward modeling Review

15 Miller, G.A. (1956) The magical number seven, plus or minus two: some 27 Baddeley, A.D. (1993) Working memory and conscious awareness. In
limits on our capacity for processing information. Psychol. Rev. 63, 81–97 Theories of Memory (Collins, A.F. et al., eds), pp. 11–28, Lawrence Erlbaum
16 Vallar, G. and Baddeley, A.D. (1984) Fractionation of working memory: 28 Tulving, E. (1989) Memory: performance, knowledge and experience.
neuropsychological evidence for a phonological short–term store. European J. of Cog. Psychol. 1, 3–26
J. Verbal Learn. Verbal Behav. 23, 151–161 29 Hummel, J. (1999) The binding problem. In The MIT Encyclopedia of
17 Wilson, B.A. and Baddeley, A.D. (1988) Semantic, episodic and auto- the Cognitive Sciences (Wilson, R.A. and Keil, F.C., eds), pp. 85–86, MIT
biographical memory in a post–meningitic amnesic patient. Brain Press
Cognit. 8, 31–46 30 Singer, W. (1999) Binding by neural synchrony. In The MIT
18 Baddeley, A.D. Levels of working memory. In Perspectives on Human Encyclopedia of the Cognitive Sciences (Wilson, R.A. and Keil, F.C.,
Memory and Cognitive Aging: Essays in Honor of Fergus Craik (Naveh eds), pp. 81–84, MIT Press
Benjamin, M. et al., eds), Psychology Press (in press) 31 Roberts, A.C. et al. (1996) Executive and cognitive functions of the
19 Ericsson, K.A. and Kintsch, W. (1995) Long-term working memory. prefrontal cortex. Philos. Trans. R. Soc. London Ser. B 351, 1387–1527
Psychol. Rev. 102, 211–245 32 Prabhakaran, V. et al. (2000) Integration of diverse information in
20 Bartlett, F.C. (1932) Remembering, Cambridge University Press working memory within the frontal lobe. Nat. Neurosci. 3, 85–90
21 Schank, R.C. (1982) Dynamic memory, Cambridge University Press 33 Cowan, N. (1999) An embedded-processes model of working memory.
22 Neath, I. and Nairne, J.S. (1995) Word-length effects in immediate memory: In Models of Working Memory (Miyake, A. and Shah, P., eds),
overwriting trace-decay theory. Psychonomic Bull. Rev. 2, 429–441 pp. 62–101, Cambridge University Press
23 Gathercole, S.E. and Hitch, G.J. (1993) Developmental changes in short- 34 Jones, D.M. (1993) Objects, streams and threads of auditory attention.
term memory: a revised working memory perspective. In Theories of In Attention: Selection, Awareness and Control (Baddeley, A.D. and
Memory (Collins, A. et al., eds), pp. 189–209, Lawrence Erlbaum Weiskrantz, L., eds), pp. 87–104, Clarendon Press
24 Hickok, G. and Poeppel, D. (2000) Towards a functional neuroanatomy 35 Allport, D.A. (1984) Auditory–verbal short-term memory and
of speech perception. Trends Cognit. Sci. 4, 131–138 conduction aphasia. In Attention and Performance Vol. X: Control of
25 Baddeley, A.D. and Logie, R.H. (1992) Auditory imagery and working memory. Language Processes (Bouma, H. and Bouwhuis, D.G., eds), pp. 313–326,
In Auditory Imagery (Reisberg, D., ed.), pp. 179–197, Lawrence Erlbaum Lawrence Erlbaum
26 Baddeley, A.D. and Andrade, J. (2000) Working memory and the 36 Conway, M.A. et al. (1996) Recollections of true and false
vividness of imagery. J. Exp. Psychol. Gen. 129, 126–145 autobiographical memories. J. Exp. Psychol. Gen. 125, 69–95

Forward modeling allows

feedback control for fast
reaching movements
Michel Desmurget and Scott Grafton

Delays in sensorimotor loops have led to the proposal that reaching movements are
primarily under pre-programmed control and that sensory feedback loops exert an
influence only at the very end of a trajectory. The present review challenges this view.
Although behavioral data suggest that a motor plan is assembled prior to the onset of
movement, more recent studies have indicated that this initial plan does not unfold
unaltered, but is updated continuously by internal feedback loops. These loops rely on
M. Desmurget is at
a forward model that integrates the sensory inflow and motor outflow to evaluate the
INSERM U534,
consequence of the motor commands sent to a limb, such as the arm. In such a model, ‘Space and Action’,
the probable position and velocity of an effector can be estimated with negligible 16 av. du Doyen
Lépine, 69500, Bron,
delays and even predicted in advance, thus making feedback strategies possible for fast France.
reaching movements. The parietal lobe and cerebellum appear to play a crucial role in S. Grafton is at the
Center for Cognitive
this process. The ability of the motor system to estimate the future state of the limb
Neuroscience, 6162
might be an evolutionary substrate for mental operations that require an estimate of Moore Hall,
sequelae in the immediate future. Dartmouth College,
Hanover, NH
03755, USA.

S ince the early contribution of Woodworth1, the degree

to which visually-directed movements are planned in ad-
century of controversy, the relative importance of three dif-
ferent models, namely the feedforward, feedback and hy-
tel: +1 603 646 0038
fax: +1 603 646 1181
e-mail :
vance or controlled online during their actual execution has brid, continues to be argued. Feedforward models propose Scott.T.Grafton@
been an issue of considerable debate2–6. After almost a that a motor command is defined in advance of the onset of dartmouth.edu

1364-6613/00/$ – see front matter © 2000 Elsevier Science Ltd. All rights reserved. PII: S1364-6613(00)01537-0
423
Trends in Cognitive Sciences – Vol. 4, No. 11, November 2000
Review Desmurget and Grafton – Forward modeling

(feedforward component). This initial plan does not unfold

Box 1. Different types of unattended, because it is imprecise12–15. Rather, it remains
under the constant ‘supervision’ of powerful internal feed-
feedback
back loops that adjust and refine it in real time (feedback
The term ‘feedback’ has multiple meanings in the literature. In component). In this paper, we examine the validity of these
this review, it refers to processes that mediate hand path cor-
three models for reaching movements of the hand in light of
rections by comparing the target position and an estimation of
recent psychophysical, lesion-based and functional imaging
the hand location. This is the primary feedback mechanism
that allows modulation of the initial motor command when it
studies. We show that both the feedforward and feedback
is inaccurate. From a theoretical point of view, we can segre- hypotheses, when considered in isolation, are overly reduc-
gate feedback loops into three categories: (1) sensory feedback, tionistic. Consequently, we propose an integrative hybrid
in which the location of the hand is estimated on the basis of model of motor control in which preplanning and feedback
sensory information; (2) non-sensory feedback, in which the control are both used by the nervous system. We first review
location of the hand is estimated on the basis of efferent infor- the evidence that is generally believed to have established a
mation; and (3) internal feedback, in which the location of the dominant role for feedforward movement control, namely
hand is estimated on the basis of both efferent and afferent that sensory feedback loops are too slow to allow efficient
signals. The term feedback alone refers to any of these loops. trajectory control. Second, we show that feedback mech-
anisms can rely on much more than sensory inflow than has
been thought traditionally. Feedback control strategies be-
movement. Within this context, the role of feedback loops come viable if the instantaneous location of the hand can be
is, at most, marginal and circumscribed to the very end of inferred by the nervous system through a forward model that
the trajectory3,4,7–9 when hand velocity is low. Feedback integrates efferent and afferent signals to infer, with no
models (Box 1), when regarded as the conceptual opposite delay, the current state of the motor system (Box 2). Third,
of feedforward models, propose that the pattern of muscle we argue against the plausibility that purely feedforward
activation that is required to point to the target is not strategies generate rapid reaching movements. To this end,
defined prior to the onset of movement, but rather during we show that online control by visual and non-visual infor-
the course of arm displacement. Thus, there is no a priori mation occurs early in a hand movement. We also provide
motor plan and the muscle command is generated in real evidence that the motor command is not generated exclu-
time through an error signal that continuously compares sively in real time, which is contrary to the suggestion of
the relative locations of the hand and target10,11. purely feedback models. Finally, we discuss briefly the func-
Hybrid models represent a trade-off between the feed- tional anatomy of internal feedback loops with emphasis on
forward and feedback hypotheses. In a hybrid model, a crude the potential contribution of the posterior parietal cortex
motor plan is assembled prior to the onset of movement (PPC) and the cerebellum.

Box 2. From robots to humans

Many of the concepts that are used to describe the planning and command and allows the CNS to estimate the current and
execution of human arm movements have been borrowed from future state of the effector immediately and without peripheral
robotic manipulators. Among these concepts, three are espe- information. Forward models are particularly interesting in the
cially relevant for the present review, namely feedforward and context of feedback control systems. For instance, a forward
feedback control and internal models. model can produce an estimate of the movement end-point lo-
In a feedforward control system, the set of muscle activations cation as output, which can be compared to the target lo-cation.
that drives a reach towards a target is defined prior to arm dis- In the case of a discrepancy, a corrective command can be gen-
placement. After the onset of movement, the motor command erated. An inverse model takes into account the inertial and vis-
unfolds unaltered until the movement is completed. Computer cous properties of the arm to estimate the motor command that
programs that specify a series of commands to be performed in will produce the desired displacement. Inverse models are there-
advance are a good analogy of feedforward control mechanisms. fore critical for feedforward control systems (Refs e,f).
In a feedback control system, the current state of the system
is compared to reference values. In the case of a discrepancy References
between these two parameters, an error signal is generated and a Hinton, G. (1984) Parallel computations for controlling an arm.
used to modulate the behavior of the controlled system. The J. Mot. Behav. 16, 171–194
most common example of a feedback system is the thermostat, b Flanagan, J.R. et al. (1993) Control of trajectory modifications in
which compares the current temperature to a reference value target-directed reaching. J. Mot. Behav. 25, 140–152

and then modulates the response of the radiator. For reaching c Wolpert, D.M. et al. (1998) Internal models in the cerebellum.
Trends in Cognitive Sciences 2, 338–347
movements, the controlled system could be the arm, the refer-
d Kawato, M. (1999) Internal models for motor control and
ence state the target position and the current state the location
trajectory planning. Curr. Opin. Neurobiol. 9, 718–727
of the hand. As long as the hand has not reached the target, a
e Katayama, M. and Kawato, M. (1993) Virtual trajectory and
motor command is generated in real time (Refs a,b). Thus, stiffness ellipse during multijoint arm movement predicted by
muscle activations do not have to be specified in advance. neural inverse models. Biol. Cybern. 69, 353–362
Internal models can be segregated into two categories, namely f Gomi, H. and Kawato, M. (1996) Equilibrium-point control
forward models and inverse models (Refs c,d). A forward model hypothesis examined by measured arm stiffness during multijoint
predicts the behavior of the motor system in response to a movement. Science 272, 117–120

424
Trends in Cognitive Sciences – Vol. 4, No. 11, November 2000
 Desmurget and Grafton – Forward modeling Review

Rapid ballistic Slow terminal

arm transport adjustment
Inverse Motor
Desired
model plan Motor command
displacement

Corrective
command Sensor

Hand Target Hand

location location ES location

Initial conditions Sensory feedback module

trends in Cognitive Sciences

Fig. 1. The classical dual model of movement control. The required arm displacement is estimated based on the respective locations of
the hand and target. This displacement is then converted into a motor plan through an inverse model. The main part of the movement un-
folds under the rigid control of this plan (ballistic arm transport). Sensory feedback loops become active at the very end of t he movement,
when velocity is low. The current location of the hand is then compared to the target position. In case of a discrepancy, an error signal (ES)
is issued and a series of corrective sub-movements is generated. The movement stops when the hand reaches the target (circle in diamond).

Sensory information and feedback control reaching movements is typically 300–700 ms. On the basis of
During the last three decades, feedforward models of move- these results, it was concluded that sensory feedback loops
ment control have indisputably been the most influential. could not be used to control hand trajectory21,22.
This dominance was based on the theoretical alternative that To reconcile the abovementioned results with the fact
feedback loops should rely exclusively on sensory information, that goal-directed movements are more accurate when pro-
an assumption that was explicitly formulated in Keele’s pio- prioceptive or visual information is present, a dual model of
neering monograph, as follows. ‘The concept of motor pro- motor control was proposed3,4,7–9. According to this model
gram might be viewed as a set of muscle commands that are (Fig. 1), reaching movements are segmented into two com-
structured before a movement sequence begins, and that al- ponents. The first is driven entirely by a motor plan and en-
lows the entire sequence to be carried out uninfluenced by sures rapid transport of the hand to the vicinity of the tar-
peripheral feedback’16. Following this definition, the contri- get. The second depends on sensory feedback loops and
bution of sensory information to movement control was allows corrections at the very end of the trajectory, when the
widely investigated. Three types of results led to the conclu- movement velocity becomes low. Classically, these correc-
sion that sensory feedback played a marginal role in move- tions are viewed as a series of one or more sub-movements
ment accuracy. First, somatic deafferentation did not prevent that are generated at discrete time intervals on the basis of a
subjects from executing relatively accurate movements in the retinal error signal5–8,23. The dual model found strong sup-
dark. Second, some movements of short duration could be port in the demonstration that viewing the hand during the
completed before the minimum delay required to process first half of the trajectory did not improve movement accu-
sensory information. Third, online corrections that were racy more than when the hand was never visible24,25. In ad-
based on sensory feedback could produce unstable trajec- dition, tendon vibration experiments showed that altering
tories for pointing that was performed at a high or medium proprioceptive feedback affected movement accuracy only
velocity. Among these arguments, the first is probably the when the vibration was applied at the end of the trajec-
least convincing, because deafferentation experiments have tory26. We shall examine alternative interpretations of these
produced inconsistent results. Specifically, while some stud- observations after developing the idea of forward models.
ies have indicated that humans or animals deprived of pro-
prioception are able to perform relatively accurate move- Feedback and the need for forward modeling
ments5,17,18, other experiments have shown that deafferented A displacement of the body with respect to the environment
subjects exhibit dramatic motor impairments5,19,20. and vice versa generates the same retinal stimulation. To ac-
The strongest evidence against the use of sensory feedback count for the ability of the nervous system to discriminate
to control movement trajectory is based on the physiological between these two situations, von Holst and Mittelstaedt27
delay that is inherent in sensorimotor loops. If the processing proposed that a ‘copy’ of the motor command was stored
of sensory information is long with respect to the duration of somewhere in the brain and used to interpret the perceptual
movement, the position of the hand will change dramatically input. This conclusion was extended and generalized, lead-
by the time the feedback signal starts to influence the ongoing ing to the concept of the forward model14,22,28–31. The idea
motor command, thus rendering the implemented correction behind this concept is that the nervous system can progres-
inappropriate21,22. Behavioural experiments have shown that sively ‘learn’ to estimate the behavior of the motor plan in
the minimum delay needed for a visual or proprioceptive response to a given command. By integrating information
signal to influence an ongoing movement is 80–100 ms that is related to initial movement conditions, motor outflow
(Refs 5,6), while that for the duration of visually-directed and sensory inflow, the probable position and velocity of the

425
Trends in Cognitive Sciences – Vol. 4, No. 11, November 2000
Review Desmurget and Grafton – Forward modeling

Continuous control
Inverse Motor
Desired
model plan Motor command
displacement

Corrective Motor outflow

command sensory inflow

Hand Target Forward

location location ES Final state estimate model
(end-point location)

Intitial conditions Feedback module

trends in Cognitive Sciences

Fig. 2. Simplified hybrid circuits that makes use of a forward model of arm dynamics for controlling hand movements. A
motor plan is initially defined based on the respective locations of the hand and target. During the movement, a forward model of the
dynamics of the arm is generated. This model receives the sensory inflow and a copy of the motor outflow as inputs and generates an es-
timate of the movement end-point location as output. This estimate is compared to the target location. In case of discrepancy, an error
signal (ES) is generated, triggering a modulation of the ongoing motor command.

effector can be determined (and even predicted). When a for- deafferentation of extraocular proprioception in monkeys34.
ward model is used to feed an internal feedback loop, control These results have, thus, revealed the existence of a non-sen-
performance is improved significantly inasmuch as large sory feedback loop that can account for the remarkable ac-
delays that are associated with sensory feedback loops can be curacy of the saccadic system under normal conditions35,36.
avoided. The advantage of forward modeling for movement For arm movements, the existence of non-sensory feed-
control is illustrated schematically in Fig. 2. When required back loops that make use of motor outflow was initially sug-
to reach a target, a subject first elaborates a motor plan, based gested by behavioral studies that showed that hand trajec-
on the initial movement conditions (i.e. the respective loca- tory could be amended with a shorter latency than the
tions of the hand and target). During the realization of the minimal latency required to process peripheral information.
movement, a forward model of the dynamics of the arm is For instance, Higgins and Angel37 observed that the reac-
generated. In its simplest version, this model receives as input tion time to an unexpected modification of the target
a copy of the motor outflow. Based on this information, the trajectory in a manual tracking task was shorter than the
end-point of the movement can be predicted and contin- proprioceptive reaction time. A similar result was reported
uously compared to the target location. Discrepancies cause by Jaeger et al.38, who found that altering the proprioceptive
an error signal to be generated, which triggers a modulation signal through vibration did not modify the reaction time to
of the motor command. More complex approaches have sug- a visual perturbation. Using a pointing task, Cook and
gested that forward models do not only use motor outflow, Diggles39 observed corrections to the hand path within
but also use sensory inflow29,30. This view is supported by 45 ms when the initial direction of movement was incor-
behavioral observations (see below) and has been modeled by rect. This value was close to that reported by van Sonderen
Miall et al.28 with reference to an engineering control scheme et al.40 (30 ms) in a double-step task in which the initial tar-
that is known as the Smith predictor. In this scheme, forward get location was changed during or after the initiation of
prediction is delayed by a period that is comparable to the movement.
sensory delay, thus making it possible to compare Convincing arguments were reported in an elegant
directly the predicted and sensory-based estimates. Any error study by Wolpert et al.29, which suggested that a reliable es-
that results from this comparison can then be used to update timation of the location of the hand could be obtained by
the current forward model of the dynamics of the arm. combining efferent and afferent signals in a forward model.
One line of evidence that shows that non-sensory feed- These authors required subjects to move their hand along a
back loops can be used to guide biological actuators is found line while holding a manipulator. The hand was allowed to
with eye movements. There is now considerable evidence be viewed for 2 s prior to the onset of movement. The ma-
that the oculomotor system uses an efferent signal to con- nipulator was connected to a torque motor that induced re-
trol saccadic eye movements. Perturbation experiments sistive or assistive force to the movement. At the end of the
indicate that if gaze is shifted during the preparation or ex- trial, the subjects estimated the location of their hand using
ecution of a saccade toward a flashed target, then a com- a visual spot controlled by the other hand. The temporal
pensatory saccade which accurately brings gaze onto the propagation of measured errors exhibited by the subjects
remembered target location is generated (head-fixed sac- could be fully accounted for by assuming that the motor
cade32 and head-free gaze shifts33). Strikingly, such compen- control system integrates both motor outflow and sensory
sation does not require visual or proprioceptive feedback, inflow to estimate the location of the hand. By contrast,
because it can occur in complete darkness and after surgical models based exclusively on either sensory inflow or motor

426
Trends in Cognitive Sciences – Vol. 4, No. 11, November 2000
 Desmurget and Grafton – Forward modeling Review

outflow were unable to predict the observed pattern of Arguments that support this claim can be found in studies
error. Hoff and Arbib14 reached a compatible conclusion. showing that vision of the hand at rest, prior to movement,
They showed, for reaching movements, that control models improves movement accuracy through an optimization of
that combine efferent signals and afferent information to es- online feedback loops45.
timate the current location of the hand and adjust the The studies reviewed in the foregoing section have es-
planned pattern of muscle activation successfully captured tablished that forward models can combine motor outflow
the kinematic characteristics of visually directed reaching. and sensory inflow to estimate the current and future states
In particular, this model, which used a look-ahead predictor of the motor apparatus with negligible delays. These find-
to compensate for delays, was able to account for the trajec- ings, therefore, obliterate the key argument against the use
tory revision that is observed in behavioral experiments in of feedback mechanisms for fast reaching movements.
which the target location is modified at the beginning of the
hand movement or during the saccadic response. Non-sequential, non-ballistic control of reaching
Further arguments in favor of the conclusion that effer- Two concepts, ‘sequential control’ and ‘ballistic reaching’,
ent and afferent signals are combined to generate a reliable have strongly influenced our thinking of how motor plans
forward estimation have been provided by recent studies on are generated over time. For an external examiner, the rela-
interjoint coordination. Gribble and Ostry41 showed that tive coordination of the eyes, head and hand during goal-
electromyographic (EMG) activity in the shoulder and directed reaching appears to be sequential. When a subject
elbow joints varies in a predictive manner during reaching points to a visual target in peripheral space, the eyes move
movements to compensate for interaction torque that arises first, followed by the head and finally the hand. The gaze ar-
from multijoint dynamics. This anticipatory response indi- rives at the target before or at about the same time as the
cates that the nervous system can use a forward model to onset of the hand movement, because the duration of eye
predict and offset the kinematic consequences of interseg- movement is brief 13,46. Several researchers hypothesized that
mental dynamics. Interestingly, recent results indicate that this sequential organization has a functional foundation23,46
sensory information is critical to set parameters for and up- and consequently suggested that the nervous system had to
date such a forward model. For example, Sainburg et al.20 achieve target foveation before building a reliable motor plan
required two patients that presented with large-fiber sensory for the arm, because the extra-foveal visual signal did not
neuropathy to make a gesture similar to slicing a loaf of allow for an accurate estimation of the target location46,47.
bread. Without being able to see their limb moving, these This hypothesis was challenged by studies that showed that
patients could not compensate for interaction torque, the serial organization of the eye, head and arm at the behav-
which led to severe impairments in interjoint coordination. ioral level results primarily from inertial factors. As shown by
The best evidence for the presence of a forward model Biguer et al.5,48, the EMG discharge for the eye, head and arm
during reaching was provided recently by Bard et al.18 A during fast reaching movements is nearly synchronous, indi-
deafferented patient was instructed to look at and point to cating that the motor command is sent to these different ef-
visual targets that were displayed in the peripheral field of fectors in parallel (the arm moves last simply because it has
vision, but was not allowed to look at the moving limb. In the greatest inertia). If one considers that the onset of an ag-
some trials, the target location was changed slightly during onistic muscle contraction occurs 50–100 ms before the ac-
the course of the ocular saccade. Saccadic suppression pre- tual motion in a reaching movement49, this observation con-
vented the patient from consciously detecting this manipu- curs with psychophysical studies that have demonstrated that
lation, who was convinced that she pointed to a stationary the arm movement generally follows the saccadic response
target. The patient was able to correct her movement online with a lag of 60 to 100 ms (Refs 13,46).
to reach the new target location despite the absence of pe- The abovementioned observations indicate that the ini-
ripheral information. It is important to note that her cor- tial motor command can be issued on the basis of an imper-
rections were not as accurate as those of control subjects, fect estimation of the target location. At the end of the ocu-
which suggests that motor outflow had to be combined lar saccade, after commands for arm movements have been
with at least some sensory inflow to generate an optimal es- initiated, the initial estimation of the target position is up-
timate of location of the hand. dated on the basis of foveal information. This updating is
Several studies have attempted to relate the pattern of clearly demonstrated by the finding that arm movements
end-point errors with key variables that are used to plan are less accurate when the eyes are not free to move to the
reaching movements2,42,43. For instance, Vindras et al.44 target46,47. We replicated and extended this finding recently
showed that final errors in visually-directed movements that (unpublished), using a protocol and apparatus similar to the
were performed without being able to view the moving limb one described in Prablanc et al.46 Focusing on hand kine-
reflected systematic biases in the estimation of the initial matics, we observed the following. First, the initial acceler-
state of the motor apparatus. This kind of result might ap- ation vector was not affected by the ability to move the eyes,
pear to contradict the idea that movement trajectory is con- which concurs with the idea that the initial motor plan is
trolled by internal feedback loops. However, this is not the assembled on the basis of peripheral visual information.
case, because the estimation of the current location of the Second, the maximal velocity vector had a smaller magni-
hand by a forward model will be affected in a systematic tude in the eye-free condition than in the eye-fixed condi-
manner if either the estimation of the initial state of the tion, which is in agreement with the observation that
motor apparatus or the inverse model that transforms the distances are usually overestimated in the peripheral visual
desired displacement into a motor command, is biased. field46,47. These observations suggest that an inaccurate

427
Trends in Cognitive Sciences – Vol. 4, No. 11, November 2000
Review Desmurget and Grafton – Forward modeling

second target12,13,15 (Fig. 3). Trajectory amendments were

(a) Open loop pointing (b)
500 smooth, as indicated by the absence of discontinuities in
500
wrist velocity curves, which exhibited the same bell-shaped

Planar hand path

20 30
Planar hand path

10 profile for both the perturbed and control movements.

300 20 300 Interestingly, corrections were detectable up to 110 ms after

(mm)
(mm)

30
the onset of hand movement, which was roughly synchro-
nized with the end of the ocular saccade, suggesting that
100 100
hand trajectory was amended very early in the movement.
0 0 The pattern of correction and the reaction time to the per-
0 100 300 500 0 100 300 500 turbation were similar, irrespective of whether the moving
Distance (mm) Distance (mm)
(c) (d) limb could be viewed13. This suggests that trajectory modifi-
250
velocity (deg. sec–1)
500 cations that were observed in the double-step trials mainly
velocity (deg. sec–1)

20 20 10
Hand tangential
Hand tangential

200 depended on non-visual feedback loops, which is compatible

with observations mentioned earlier that showed that vision
150 300 of the limb during the first half of the trajectory has no effect
100 on movement final accuracy24,25. However, the generality of
50 100 this conclusion was challenged by Bard and colleagues, who
designed a series of experiments based on the idea that pe-
0 0
0 400 800 1200 0 400 800 1200 ripheral vision might be particularly sensitive to the direction
Time (ms) Time (ms) of the movement6,23. In their first experiment, the authors
showed that the directional component of very fast aiming
trends in Cognitive Sciences
movements (,130 ms) could be corrected under the control
Fig. 3. Smooth and early path corrections during reaching performed without
of peripheral vision even when vision was only available dur-
vision of the limb. (a) Mean hand path of a subject reaching towards a stationary
target (208; broken line) and displaced targets (208 to 308 and 208 to 108; unbroken lines).
ing the first half of the trajectory50. In light of these and other
(b–d) Corresponding velocity profiles. (Figure reproduced, with permission, from Ref. 74.) data6, the preservation of end-point accuracy that is observed
when visual information is removed at the beginning of a
motor plan assembled on the basis of an imperfect estimate movement might be explained by the fact that the motor
of the target location is corrected early in the course of the command can still be optimized by potent terminal feedback
movement, specifically during the acceleration phase. This loops.
conclusion clearly challenges the idea that visually directed
movements are ballistic during their early stage. Feedforward specification of the motor command
To investigate directly whether an intended action is a Given the powerful ability of forward models to adjust
result of a preset pattern of non-modifiable commands, movement online, one might wonder whether any of the
Prablanc and colleagues designed a set of behavioral experi- movement needs to be planned in advance10,11. For exam-
ments in which the initial inaccurate estimation of the target ple, it has been shown that a control scheme that involves a
location during movement planning was artificially in- progressive definition of the arm motor command, without
creased12,13. To achieve this, the authors used a double-step any preplanning adequately predicts trajectories that cap-
pointing paradigm in which the target location was modified ture the kinematic characteristics of visually-directed reach-
slightly (target ‘jump’ paradigm) during the course of the ing11. This computational result, however, is not echoed by
ocular saccade. This procedure has the following three major other experimental findings. Two main lines of evidence
advantages: (1) the target jump is not perceived consciously suggest that a representation of the upcoming motor com-
by the subject, because of saccadic suppression; (2) the target mand exists prior to the onset of reaching movements.
jump does not alter the organization of the oculomotor sys- The first line of evidence comes from the study of fine
tem, because saccadic responses to stationary targets involve predictive compensatory adjustments in single muscles.
an initial saccade that undershoots the target position and a Gribble and Ostry41 recently reported that EMG activity in
single corrective saccade35 to achieve accurate target acquisi- the shoulder and elbow joints varied in a predictive manner
tion; and (3) the target jump does not alter the organization to compensate for interaction torque that arises from multi-
of the manual response, because pointing movements to sta- joint dynamics. Similar results were reported for the adjust-
tionary targets involve an update of the target location at the ment of grip force during arm movements performed with
end of the saccadic displacement, which is taken into account a hand-held load51 and for postural compensations that sta-
to amend the ongoing arm movement (see above). These bilize rapid arm movements52. Such adjustments can only
points can be summarized by saying that unconscious dou- be explained if the kinematic consequences of the upcom-
ble- and single-step situations are identical from a functional ing motor command can be predicted precisely, i.e. if this
point of view. The intrasaccadic modification of the target motor command is, to some extent, ‘known’ in advance.
location simply increases an error that is already present in The second line of evidence that reveals the preplan-
the system, which supports the hypothesis that the same cor- ning process comes from a recent transcranial magnetic
rective processes are engaged in the ‘jump’ and ‘stationary’ stimulation (TMS)-reaching study that was carried out by
trials. Using the target jump paradigm, Prablanc and col- our group15 (Fig. 4). Subjects pointed to visual targets with
leagues observed that the hand path, which was initially di- their right hand, but vision of the arm was not allowed dur-
rected to the first target, diverged progressively toward the ing this movement. In some trials, the target location was

428
Trends in Cognitive Sciences – Vol. 4, No. 11, November 2000
 Desmurget and Grafton – Forward modeling Review

displaced during the saccadic response, whereas in other tri-

No-stimulation Left stimulation
als, it remained stationary. As observed in earlier studies12,13,
right hand right hand
the target jump elicited a smooth and progressive adjust-
ment of the hand path. Strikingly, when a single TMS was
applied over the left intraparietal sulcus (IPS) at the onset of
hand movement, these smooth path corrections were dis-
rupted and the subject pointed to the first target location.
However, the hand trajectory to stationary targets did not
become erratic, suggesting that relatively accurate move-
ments can be performed in the absence of online feedback
loops. If continuous control loops, in which the relative lo-
cations of the hand and target are compared, were used to
generate the motor command in real time, disrupting these trends in Cognitive Sciences
loops should have resulted in either errant or dramatically Fig. 4. Critical role of the posterior parietal cortex for online movement correc-
inaccurate trajectories to stationary targets. However, a con- tions. Mean hand paths of one subject, either with (left panel) or without (middle panel)
transcranial magnetic stimulation (TMS). The continuous curves represent the mean paths
trol session that involved the same stimulation site but the directed at stationary targets (208, 308 and 408). The dashed curves represent the mean paths
other hand (left hand, ipsilateral to the stimulated site) directed at jumping targets (308 to 22.58 and 308 to 37.58). Black circles indicate stationary tar-
failed to reveal any disruption of hand path corrections. get locations and white circles represent jumping target locations. Location of TMS site
(right panel), determined by three-dimensional MRI (black circle). When TMS is applied,
This indicates that the feedback disruption observed for the
path corrections that normally occur in response to the target jump are disrupted. In addi-
right hand was not related to oculomotor deficits or to the tion, movements directed at stationary targets become less accurate (although not erratic).
inability of the subject to update the target location. (Figure reproduced, with permission, from Ref. 15.)
Unfortunately, the accuracy and the degree of detail of
the initial motor plan cannot be established from the above- The results indicated that the patient was able to reach the
mentioned data. The fact that the motor deficits observed in target properly in the stationary condition, but she pre-
some deafferented patients cannot be explained solely by sented a dramatic inability to correct her ongoing move-
their incapacity to define the initial state of the motor sys- ments in the perturbed condition. In the latter case, the pa-
tem2,20,53 suggests that the initial motor plan is only crudely tient pointed generally to the initial target location before
defined prior to the onset of movement and subsequently initiating a second movement to the final target position.
updated through internal feedback loops as the movement Control subjects exhibited early modifications of the hand
progresses. The loss of accuracy that is observed when online path, as expected from earlier studies12,13,15.
feedback loops are disrupted supports this view15 (Fig. 4). Recent studies in the monkey and human have shown
that the parietal cortex is highly differentiated with many
Functional anatomy of internal error corrections functional subdivisions58–60. Unfortunately, the function of
Movements such as reaching are controlled by widely dis- many of these subdivisions is not entirely understood and it is
tributed cortical and subcortical sensorimotor areas. It is not clear which specific areas might be associated with inter-
generally thought that parietal and pre-motor systems are nal feedback loops. In addition, progress in this area is ham-
essential for the selection, preparation and execution of a pered by the marked differences between the parietal lobe
movement. Although the functional anatomy of internal cytoarchitectonic organization of human and non-human
feedback loops is not fully known, two areas within the dis- primates59. Despite these obstacles, two potential roles for the
tributed sensorimotor system are hypothesized to be critical PPC have been evoked in the literature, particularly in areas
for updating hand trajectory, namely the PPC in the region that involve the IPS. It was first suggested that the PPC might
of the IPS and the anterior parasagittal cortex of the cer- be involved in computing motor errors by comparing the ac-
ebellum54,55. In this section, we examine the potential con- tual target location with the location of the hand15. This view
tribution of these two structures to internal feedback. is supported mainly by the demonstration that the PPC is
Indirect evidence to suggest that internal feedback crucial for establishing stable relationships between hetero-
loops rely on the PPC comes from the observation that the geneous information58, i.e. for merging the arm- and target-
reach responsiveness of neurons in area 7a of non-human related signals into a common frame of reference, which con-
primates changes as the unseen hand approaches a visual stitutes a crucial step in defining a motor error. An alternative
target56. A more convincing argument is provided by the (but not exclusive) hypothesis proposes that the PPC gener-
TMS study described in the previous section. When the ates a forward model of the location of the hand. Indirect
normal functioning of the PPC in the region of the IPS is support for this view comes from the observation that several
perturbed after the onset of hand movement, feedback parietal structures are concerned with various types of predic-
loops that allow correction of the ongoing movement are tive mechanisms61–63. Stronger support emerged with the
disrupted15 (Fig. 4). This focal deficit was recently repli- finding that sensory signals from many modalities (e.g. visual,
cated in clinical studies that involved a patient that pre- proprioceptive, auditory and vestibular), as well as efferent
sented with bilateral ischemic lesions of the PPC (Ref. 57). copy signals from motor structures, are integrated in the PPC
This patient was asked to look at and point to visual targets (Ref. 64). This concurs with the idea that sensori-motor inte-
presented on a computer screen in front of her. In some tri- gration is a crucial feature of forward models.
als, the target remained stationary, whereas in other trials it Despite the abovementioned findings that implicate the
‘jumped’ to a new location at the onset of arm movement. PPC in feedforward models, several lines of evidence support

429
Trends in Cognitive Sciences – Vol. 4, No. 11, November 2000
Review Desmurget and Grafton – Forward modeling

the idea that forwards models might rely on the cerebellum nature of their reciprocal interactions and their potential
more than the PPC. Unlike the PPC, the cerebellum has connections with other areas, remain to be investigated.
long been associated with feedback control28,65. The original
idea, proposed by Holmes66, was that this structure is partic- Concluding remarks
ularly important in the visual guidance of movement. This The present review indicates that no single computational
hypothesis was recently incorporated into a more general algorithm can adequately describe the control processes that
scheme in which it is assumed that the cerebellum plays a are used to perform goal-directed movements. Rather,
crucial role in elaborating forward and inverse internal reaching towards a target requires an integrative control
models. Based on this idea (which has been reviewed else- scheme in which feedforward specification of the motor
where30,31), one could assign two functions to the cerebellum command, forward modeling of the dynamics of the arm
for movement guidance. First, this structure might con- and online updating of the initial pattern of muscle acti-
tribute to the conversion of the error signal generated by the vation are synthesized in reliable feedback loops, which are
PPC into a motor command (inverse trans-formation). In thought to involve the cerebellum and PPC.
support of with this view, it has been shown that inverse
models are enclosed within the cerebellum30,67,68 and that pa-
Acknowledgements
tients with cerebellar lesions are unable to compensate for We thank Helene Gréa, Dr Claude Prablanc and Dr Denis Pélisson for
multijoint interaction torques during movement planning69. their helpful remarks and their critical reading of this manuscript. We are
A second potential function of the cer-ebellum is involve- also grateful to Laura Payne for editing this manuscript and to Serge
ment in estimating the probable position and velocity of the Terrones for technical assistance. Supported by Public Health Service
Grants NS3504 and NS37470.
effector (forward model). Of the supportive evidence,
reviewed elsewhere28,30,70, the following three ideas are espe-
cially convincing. First, in patients with cerebellar lesions, References
path corrections that are based on visual sensory information 1 Woodworth, R.S. (1899) The accuracy of voluntary movement. Psychol.
are characterized by excessive deviations and abnormal oscil- Rev. Monogr. 3 (Issue 13)
2 Desmurget, M. et al. (1998) From eye to hand: planning goal-directed
lations, as would be expected if feedback loops relied exclu-
movements. Neurosci. Biobehav. Rev. 22, 761–788
sively on delayed sensory information71. Second, an erro- 3 Keele, S.W. (1981) Behavioral analysis of movement. In Handbook of
neous (or absent) forward model leads to tracking deficits Physiology, Section 1: The Nervous System (Motor Control) (Vol. 2)
that are similar to those observed in cerebellar ataxia28. (Brooks, W.B., ed.), pp. 1391–1414, Williams and Wilkins
Finally, cerebellar pathology leads to major deficits in motor 4 Arbib, M.A. (1981) Perceptual structures and distributed motor
control. In Handbook of Physiology, Section 1: The Nervous System
tasks that rely on the generation of a forward model. For
(Motor Control) (Vol. 2) (Brooks, W.B., ed.), pp. 1449–1480, Williams
example, when normal subjects lift an object, the load on the and Wilkins
hand increases with movement acceleration and variations in 5 Jeannerod, M. (1988) The Neural and Behavioral Organization of Goal-
grip force anticipate this increase of load force51,72. This can Directed Movements, Clarendon Press
only be explained if the kinematic consequences of an 6 Paillard, J. (1996) Fast and slow feedback loops for the visual correction
of spatial errors in a pointing task: a re-appraisal. Can. J. Physiol.
upcoming motor command can be predicted precisely
Pharmacol. 74, 401–417
through a forward model. Interestingly, the close coupling 7 Meyer, D.E. et al. (1988) Optimality in human motor performance:
between grip and load forces is absent from patients with ideal control of rapid aimed movements. Psychol. Rev. 95, 340–370
cerebellar lesions73. 8 Milner, T.E. (1992) A model for the generation of movements
The abovementioned data suggest that internal feed- requiring endpoint precision. Neurosci. 49, 487–496
9 Plamondon, R. and Alimi, A.M. (1997) Speed/accuracy trade-offs in
back loops rely on both the cerebellum and the PPC. The
target-directed movements. Behav. Brain Sci. 20, 1–21
exact functional role of these two structures, as well as the 10 Hinton, G. (1984) Parallel computations for controlling an arm. J. Mot.
Behav. 16, 171–194

Outstanding questions 11 Flanagan, J.R. et al. (1993) Control of trajectory modifications in

target-directed reaching. J. Mot. Behav. 25, 140–152

• Classical models of motor control divide reaching movements into two 12 Pélisson, D. et al. (1986) Visual control of reaching movements without

components, namely a pre-programmed (or ballistic) component, which vision of the limb. II. Evidence of fast unconscious processes correcting

ensures rapid transport of the hand to the vicinity of the target, and a the trajectory of the hand to the final position of a double step

controlled component, which allows fine spatial adjustments at the end stimulus. Exp. Brain Res. 62, 303–311

of the trajectory. Is this view still tenable? 13 Prablanc, C. and Martin, O. (1992) Automatic control during hand

• Does a delay in sensorimotor loops prevent the use of feedback to reaching at undetected two-dimensional target displacements.

control the trajectory of the hand during reaching movements? J. Neurophysiol. 67, 455–469

• Forward models allow prediction of the behavior of a controlled system 14 Hoff, B. and Arbib, M.A. (1993) Models of trajectory formation and

(e.g. arm) in response to a given command in advance. What is the temporal interaction of reach and grasp. J. Mot. Behav. 25, 175–192

evidence that the nervous system uses this type of forward model to 15 Desmurget, M. et al. (1999) Role of the posterior parietal cortex in

control reaching movements? updating reaching movements to a visual target. Nat. Neurosci. 2, 563–567

• What is the evidence that a motor plan is effectively assembled prior to 16 Keele, S.W. (1968) Movement control in skilled motor performance.

the onset of movement? Is it plausible that the motor command is not Psychol. Bull. 70, 387–403

planned in advance, but rather generated in real time via internal 17 Taub, E. et al. (1975) Deafferentation in monkeys: pointing at a target

feedback loops? without visual feedback. Exp. Neurol. 46, 178–186

• What are the anatomical structures that are associated with internal 18 Bard, C. et al. (1999) Deafferentation and pointing with visual double-

feedback loops that control ongoing movements? step perturbations. Exp. Brain Res. 125, 410–416

• Are forward models used in other cognitive domains? 19 Rothwell, J.C. et al. (1982) Manual motor performance in a
deafferented man. Brain 105, 515–542

430
Trends in Cognitive Sciences – Vol. 4, No. 11, November 2000
 Desmurget and Grafton – Forward modeling Review
20 Sainburg, R.L. et al. (1993) Loss of proprioception produces deficits in 48 Biguer, B. et al. (1982) The coordination of eye, head and arm
interjoint coordination. J. Neurophysiol. 70, 2136–2147 movements during reaching at a single visual target. Exp. Brain Res.
21 Hollerbach, J.M. (1982) Computers, brains and the control of 46, 301–304
movement. Trends Neurosci. 5, 189–192 49 Turner, R.S. et al. (1995) Directional variation of spatial and temporal
22 Gerdes, V.G. and Happee, R. (1994) The use of internal representation characteristics of limb movements made by monkeys in a two-
in fast gold-directed movements: a modeling approach. Biol. Cybern. dimensional work space. J. Neurophysiol. 74, 684–697
70, 513–524 50 Bard, C. et al. (1985) Role of peripheral vision in the directional control
23 Paillard, J. (1982) The contribution of peripheral and central vision to of rapid aiming movements. Can. J. Psychol. 39, 151–161
visually guided reaching. In Analysis of Visual Behavior (Ingle,D.J. 51 Flanagan, J.R. and Wing, A.M. (1997) The role of internal models in
et al., eds), pp. 367–382, MIT Press motion planning and control: evidence from grip force adjustments
24 Beaubaton, D. and Hay, L. (1986) Contribution of visual information to during movements of hand-held loads. J. Neurosci. 17, 1519–1528
feedforward and feedback processes in rapid pointing movements. 52 De Wolf, S. et al. (1998) Anticipatory postural adjustments during
Hum. Mov. Sci. 5, 19–34 self-paced and reaction-time movements. Exp. Brain Res.
25 Carlton, L.G. (1981) Processing visual feedback information for 121, 7–19
movement control. J. Exp. Psychol. 7, 1019–1030 53 Ghez, C. et al. (1995) Impairments of reaching movements in patients
26 Redon, C. et al. (1991) Proprioceptive control of goal directed without proprioception. II. Effects of visual information on accuracy.
movements in man studied by means of vibratory muscle tendon J. Neurophysiol. 73, 361–372
stimulation. J. Mot. Behav. 23, 101–108 54 Desmurget, M. et al. (1999) Functional anatomy of error correction
27 Von Holst, E. and Mittelstaedt, H. (1950) Das reafferenzprinzip. during reaching movements in the dark. Soc. Neurosci. Abstr. 25, 787
Wechelwirkung Zwischen Zentralnerven system und peripherie. 55 Inoue, K. et al. (1998) PET study of pointing with visual feedback of
Naturwis. 37, 464–476 moving hands. J. Neurophysiol. 79, 117–125
28 Miall, R.C. et al. (1993) Is the cerebellum a Smith predictor? J. Mot. 56 MacKay, W.A. (1992) Properties of reach-related neuronal activity in
Behav. 25, 203–216 cortical area 7A. J. Neurophysiol. 67, 1335–1345
29 Wolpert, D.M. et al. (1995) An internal model for sensorimotor 57 Pisella, L. et al. (2000) An ‘automatic pilot’ for the hand in human
integration. Science 269, 1880–1882 posterior parietal cortex: toward reinterpreting optic ataxia. Nat.
30 Wolpert, D.M. et al. (1998) Internal models in the cerebellum. Trends Neurosci. 3, 729–736
Cognit. Sci. 2, 338–347 58 Colby, C.L. (1998) Action-oriented spatial reference frames in cortex.
31 Kawato, M. (1999) Internal models for motor control and trajectory Neuron 20, 15–24
planning. Curr. Opin. Neurobiol. 9, 718–727 59 Milner, A.D. and Dijkerman, H.C. (1998) Visual processing in the
32 Keller, E.L. et al. (1996) Endpoint accuracy in saccades interrupted by primate parietal lobe. In Comparative Neuropsychology (Milner, A.D.,
stimulation in the omnipause region in monkey. Vis. Neurosci. ed.), pp. 70–94, Oxford University Press
13, 1059–1067 60 Sakata, H. et al. (1997) The TINS Lecture. The parietal association
33 Pelisson, D. et al. (1995) Online compensation of gaze shifts perturbed cortex in depth perception and visual control of hand action. Trends
by micro-stimulation of the superior colliculus in the cat with Neurosci. 20, 350–357
unrestrained head. Exp. Brain Res. 106, 196–204 61 Eskandar, E.N. and Assad, J.A. (1999) Dissociation of visual, motor and
34 Guthrie, B.L. et al. (1983) Corollary discharge provides accurate eye predictive signals in parietal cortex during visual guidance. Nat.
position information to the oculomotor system. Science Neurosci. 2, 88–93
221, 1193–1195 62 Duhamel, J.R. et al. (1992) The updating of the representation of visual
35 Robinson, D. (1975) Oculomotor control signals. In Basic Mechanisms space in parietal cortex by intended eye movements. Science
of Ocular Motility and their Clinical Implications (Lennerstrand, G. and 255, 90–92
Bach-y-Rita, P., eds), pp. 337–344, Pergamon Press 63 Colby, C.L. et al. (1996) Visual, presaccadic, and cognitive activation of
36 Guitton, D. (1992) Control of eye–head coordination during orienting single neurons in monkey lateral intraparietal area. J. Neurophysiol.
gaze shifts. Trends Neurosci. 15, 174–179 76, 2841–2852
37 Higgins, J.R. and Angel, R.W. (1970) Correction of tracking errors 64 Andersen, R.A. et al. (1997) Multimodal representation of space in the
without sensory feedback. J. Exp. Psychol. 84, 412–416 posterior parietal cortex and its use in planning movements. Annu.
38 Jaeger, R.J. et al. (1979) Directional errors of movement and their Rev. Neurosci. 20, 303–330
correction in a discrete tracking task. J. Mot. Behav. 11, 123–133 65 Stein, J.F. (1986) Role of the cerebellum in the visual guidance of
39 Cooke, J.D. and Diggles, V.A. (1984) Rapid error correction during movement. Nature 323, 217–221
human arm movements: evidence for central monitoring. J. Mot. 66 Holmes, G. (1917) The symptoms of acute cerebellar injuries due to
Behav. 16, 348–363 gunshot injuries. Brain 40, 461–535
40 Van Sonderen, J.F. et al. (1989) Motor programs for goal-directed 67 Kawato, M. et al. (1987) A hierarchical neural-network model for
movements are continuously adjusted according to changes in target control and learning of voluntary movement. Biol. Cybern. 57, 169–185
location. Exp. Brain Res. 78, 139–146 68 Schweighofer, N. et al. (1998) Role of the cerebellum in reaching
41 Gribble, P.L. and Ostry, D.J. (1999) Compensation for interaction movements in humans. I. Distributed inverse dynamics control. Eur. J.
torques during single- and multijoint limb movement. J. Neurophysiol. Neurosci. 10, 86–94
82, 2310–2326 69 Bastian, A.J. et al. (1996) Cerebellar ataxia: abnormal control
42 Flanders, M. et al. (1992) Early stages in sensorimotor transformations. of interaction torques across multiple joints. J. Neurophysiol.
Behav. Brain Sci. 15, 309–362 76, 492–509
43 Gordon, J. et al. (1994) Accuracy of planar reaching movements. 1. 70 Miall, R.C. and Wolpert, D.M. (1996) Forwards models for physiological
Independence of direction and extent variability. Exp. Brain Res. motor control. Neural Netw. 9, 1265–1279
99, 97–111 71 Day, B.L. et al. (1998) Influence of vision on upper limb reaching
44 Vindras, P. et al. (1998) Pointing errors reflect biases in the perception movements in patients with cerebellar ataxia. Brain 121, 357–372
of the initial hand position. J. Neurophysiol. 79, 3290–3294 72 Johansson, R.S. and Westling, G. (1984) Roles of glabrous skin
45 Desmurget, M. et al. (1995) Representation of hand position prior to receptors and sensorimotor memory in automatic control of precision
movement and motor variability. Can. J. Physiol. Pharmacol. grip when rougher or more slippery. Exp. Brain Res. 56, 550–564
73, 262–272 73 Muller, F. and Dichgans, J. (1994) Dyscoordination of pinch and lift
46 Prablanc, C. et al. (1979) Optimal response of eye and hand motor forces during grasp in patients with cerebellar lesions. Exp. Brain Res.
system in pointing at visual target. I. Spatio-temporal characteristics of 101, 485–492
eye and hand movements and their relationships when varying the 74 Martin, O. and Prablanc, C. (1991) Two-dimensional control of
amount of visual information. Biol. Cybern. 35, 113–124 trajectories towards unconsciously detected double step targets. In
47 Bock, O. (1993) Localization of objects in the peripheral visual field. Tutorials in Motor Neuroscience (Requin, J. and Stelmach, G.E., eds),
Behav. Brain Res. 56, 77–84 pp. 581–598, Kluwer

431
Trends in Cognitive Sciences – Vol. 4, No. 11, November 2000