Separate neural pathways for the visual analysis of object
shape in perception and prehension
Melvyn A. Goodale*, John Paul Meenan*, Heinrich H. Bulthofft,
David A. Nicolle*, Kelly 1.Murphy* and Carolynn . Racicot*
'Department of Psychology, University of Western Ontario, London, Ontario N6A 5C2, Canada. t
Max-Planck Institut fr Biologische
Kybernetik, Spemanstrasse 38, 72076 Tubingen, Germany. Department of Ophthalmology, University of Western Ontario,
London, Ontario N6A 4G5, Canada.
Background: Earlier work with neurological patients
has shown that the visual perception of object size and
orientation depends on visual pathways in the cerebral
cortex that are separate from those mediating the use of
these same object properties in the control of goaldirected
grasping. We present evidence suggesting that
the same dissociation between perception and action is
evident in the visual processing of object shape. In other
words, discrimination between objects on the basis of
their shape appears to be mediated by visual mechanisms
that are functionally and neurally distinct from those controlling
the pre-shaping of the hand during grasping
movements directed at those same objects.
Results: We studied two patients with lesions in different
parts of the cerebral visual pathways. One patient (RV),
who had sustained bilateral lesions of the occipitoparietal
cortex, was unable to use visual information to place her
fingers correctly on the circumference of irregularly
shaped objects when asked to pick them up, even though
she had no difficulty in visually discriminating one such
object from another. Conversely, a second patient (DF),
who had bilateral damage in the ventrolateral occipital
region, had no difficulty in placing her fingers on appropriate
opposition points during grasping, even though she
was unable to discriminate visually amongst such objects.
Conclusions: This double dissociation lends strong
support to the idea that the visual mechanisms mediating
the perception of objects are functionally and neurally
distinct from those mediating the control of skilled actions
directed at those objects. It also supports the recent
proposal of Goodale and Milner that visual perception
depends on a ventral stream of projections from the
primary visual cortex to the inferotemporal cortex,
whereas the visual control of skilled actions depends on
a dorsal stream from the primary visual cortex to the
posterior parietal cortex.
Current Biology 1994, 4:604-610
Background
Humans are capable of reaching and grasping objects
with great dexterity, and vision plays an important role
in the control of this fundamental skill. Thus, when we
reach out to pick up an unfamiliar object, the opening of
our fingers and the orientation of our hand reflect the
size of the object and its orientation in egocentric space
well before we make contact with it [1,2]. Recent
neuropsychological studies suggest that the visual mechanisms
responsible for controlling this skill are largely
independent of those that support the visual perception
of object size and orientation.
Thus, for example, Goodale et al. [3] have reported
remarkably intact visuomotor performance in DF, a 34year-old
woman who developed visual-form agnosia
following anoxia from carbon monoxide poisoning. So
profound is DF's visual-form agnosia that she is unable to
identify or recognize familiar faces, line drawings of
common objects or even simple geometric shapes. Her
elementary visual abilities, however, are still relatively
intact and she shows only moderate deficits in contrast
sensitivity and only a slight shrinkage of her visual fields
[4]. Magnetic resonance imaging (MRI), carried out 13
months after the accident, revealed evidence of damage
in ventrolateral regions of DF's occipital lobe, although
her primary visual cortex appeared to be spared (for
details of the lesion, see [4]). Nevertheless, despite the
fact that DF was unable to discriminate, in perceptual
tasks, between objects that differed in size and orientation,
the pre-shaping and orientation of her hand during
the execution ofgrasping movements accurately reflected
those same object attributes.
Complementary dissociations have been reported in
patients with bilateral lesions of the posterior parietal
cortex [5,6]. Such patients are often unable to calibrate
their grasp or orient their hand appropriately when
reaching out towards target objects, even though they have
no difficulty in reporting the size and orientation of those
same objects, or in discriminating between them. Thus,
damage to one part of the cortical visual pathways, the
ventrolateral region of the occipital lobe, results in perceptual
deficits with preserved visuomotor abilities, whereas
damage to another part, the occipitoparietal region, results
in visuomotor deficits with preserved perceptual abilities.
These neuropsychological observations, together with
evidence from electrophysiological and behavioural
Correspondence to: Melvyn A. Goodale.
0 Current Biology 1994, Vol 4 No 7604
Shape for perception and prehension Goodale et al.
studies in the monkey, have led Goodale and Milner
[7,8] to propose a new interpretation of the division of
labour between the two streams of visual pathways that
leave the primary visual cortex and project to distinct
regions of the primate cerebral cortex. These two
streams, which were identified in the macaque monkey
over ten years ago by Ungerleider and Mishkin [9],
consist of a ventral stream and a dorsal stream. The
ventral stream leaves the primary visual cortex and
projects via a series of cortico-cortical projections to the
inferotemporal cortex; the dorsal stream projects from
the primary visual cortex to the posterior parietal cortex
(see inset in Fig. 1). Although one must always be
cautious when drawing homologies between monkey
and human neuroanatomy [10], it seems likely that the
visual projections from the primary visual cortex to the
temporal and parietal lobes in the human brain may
involve a separation into ventral and dorsal streams
similar to that seen in the macaque brain. Ungerleider
and Mishkin [9] originally proposed that the ventral
stream plays a special role in the identification of objects,
whereas the dorsal stream is responsible for localizing
objects in visual space.
Goodale and Milner's [7,8] re-interpretation of this story
places less emphasis on the differences in the visual information
that is received by the two streams (object
features versus spatial location) than it does on the differences
in the transformations that the streams perform
upon that information. According to Goodale and
Milner's hypothesis, both streams process information
about object features and about their spatial relations, but
each stream uses this visual information in different ways.
In the ventral stream, the transformations focus on the
enduring characteristics of objects and their relations,
and permit the formation of long-term perceptual
representations that are used to identify and recognize
objects. In the dorsal stream, the transformations use the
instantaneous and egocentric coordinates of objects, and
mediate the visual control of skilled actions - such as
manual prehension - that are directed at those objects.
This division of labour in the cortical visual pathways
can comfortably account for the dissociations between
the perception of object size and orientation, and the use
of these object features in the control of prehension that
were both described earlier.
Object size and orientation are not the only object
features that control the parameters of grasping
movements. A remarkable sensitivity to the shape of an
object is also evident from merely casual observation of
grasping. Is the visual analysis of object shape, in order to
control a grasping movement, also dependent on neural
mechanisms that are relatively independent of those
underlying the perceptual identification of objects, in the
same way as the related analysis of object size and orientation?
To answer this question, we compared the ability
of the patient DF to discriminate objects of different
shape with her ability to position her fingers correctly on
the boundaries of the same objects when she was asked
to pick them up.
We also contrasted DF's performance with that of
another patient, RV, a 55-year-old woman who had
developed optic ataxia after strokes that left her with
large bilateral lesions of the occipitoparietal cortex. The
extent of these lesions can be seen in Figure 2. The optic
ataxia, which was evident clinically as a difficulty in
directing grasping or pointing movements towards
objects presented in different parts of the visual field,
Fig. 1.Schematic diagram of the main routes whereby retinal inputs reach the dorsal and ventral streams. The larger arrows along the
LGNd route to the cortex reflect the fact that these inputs are more numerous than those that arrive via the superior colliculus and
pulvinar. The diagram of the macaque brain (right hemisphere) on the right of the figure shows the approximate routes of the dorsal
and ventral streams of cortico-cortical projections from the primary visual cortex to the posterior parietal and inferotemporal cortex,
respectively. LGNd, dorsal lateral geniculate nucleus.
RESEARCH PAPER 605
606 Current Biology 1994,'Vol 4 No 7
Fig. 2.Magnetic resonance images of the bilateral occipitoparietal lesion inpatient RV. (a)Left and (b)right parasaggital (T -weighted)
sections. Note the extensive damage to cortex and underlying white matter in posterior parietal and occipital lobes. Damage ismore
extensive on the left than on the right. (c) An axial (T2-weighted) section in which the bilateral occipital infarcts can be seen.
(Following the usual convention, the left side of the brain is on the right side of the photograph.) Note calcarine involvement on the
left but not the right.
had resolved to some extent before we saw RV, although
it was still clear that she had visuomotor problems.
Nevertheless, despite her visuomotor deficits, RV's resolution
acuity was in the normal range, she was not
apraxic (that is, she could follow simple commands such
as "show me how you would eat soup with a spoon"),
and her hand strength and finger-tapping abilities were
normal. She also achieved a perfect score on a 20-item
test of object recognition in which she was asked to
identify line drawings of common objects; DF could
identify only two items correctly on this same test.
Results
Twelve different shapes (two of each) were used to
compare DF's and RV's ability to discriminate between
shapes and to use shape information to control grasping
(see Fig. 3). The shapes were constructed out of wood
0.6 cm thick, were painted enamel white and were
presented on a flat black background. The shapes were
based on the templates used by Blake [11] to develop
algorithms for the control of grasping in two-fingered
robots working in novel environments. These shapes
have smoothly bounded contours and lack clear
symmetry; the determination of stable grasp points
therefore requires an analysis of the entire contour
envelope of the shape.
When DF and RV were presented with pairs of these
shapes in which the two shapes in any pair were either
the same or different, they showed very different discrimination
abilities. DF was unable to determine
whether the two shapes were the same or different (see
Fig. 4). Moreover, this failure was evident regardless of
whether the two shapes on 'same' trials had the same
Shape for perception and prehension Goodale et al.
the grasp points were often located on regions of the
object boundary that would be expected to yield the
most stable grip [11,12] - regions of maximum
convexity or concavity.
RV's performance was very different (see Fig. 5). In
contrast to both DF and the control subject, she often
chose very unstable grasp points, and she stabilized her
grasp only after her finger and thumb had made contact
with the object. Thus, despite her apparent ability to
perceive the shape of an object, RV was unable to use
visual information about object shape to control the
placement of her finger and thumb as she attempted to
pick up that object. Once she had made contact with
the object, however, her manipulation of it appeared
essentially normal.
In order to quantify the differences in the performance
of the three subjects, we measured the shortest distance
between the grasp line (connecting opposing grasp
points) on each trial and the object's centre of mass.
Although this measure was similar for DF and the
control subject, they both differed significantly from RV
(p<0.001; see Fig. 6).
Discussion
Fig. 3. The twelve different shapes used in both the same/
different visual discrimination tasks and inthe grasping task.
relative orientation or different orientations. In contrast,
RV achieved a score of 90% correct for this same/
different discrimination task when the relative orientation
of the two shapes on 'same' trials was identical,
falling to 80% correct when the relative orientation of
the twin shapes was varied from trial to trial. In other
words, whereas DF apparently failed to perceive
whether two objects had the same or different outline
shapes, RV had relatively little difficulty in making such
a discrimination.
Quite the opposite pattern was observed when DF and
RV were asked to pick up these objects using their index
finger and thumb in a 'precision grip'. Even though DF
failed to discriminate between these different objects, she
had no difficulty in placing her finger and thumb on
stable grasp points on the circumference of the object.
The grasp points she selected were remarkably similar to
those chosen by a neurologically intact control subject
(see Fig. 5). In addition, DF showed the same systematic
shift in the selection of grasp points as the control subject
when the egocentric orientation of the object was
changed. Moreover, the following observations,
regarding both DF and the control subject, were made.
Firstly, the line joining the two grasp points tended to
pass through the centre of mass of the object. Secondly,
these 'grasp lines' often corresponded to the axes of
minimum or maximum diameter of the object. Finally,
Despite DF's visual form agnosia, which made it impossible
for her to distinguish one object from another on
the basis of their shape, she had no difficulty using shape
information to guide the placement of her finger and
thumb as she reached out to pick up those objects. In
fact, her grasping movements were quite normal. This
Fig. 4.Performance of DF and RV on the same/different discrimination
tests. The open bars show performance on the task in
which the relative orientation of the two shapes on the 'same'
trial was identical; the blue bars show performance on the task
in which the relative orientation of the twin shapes varied
between 'same' trials. The control subject scored perfectly on
both tests, although she took longer to respond when the two
shapes on same trials were presented at different orientations.
The dotted line indicates chance performance.
RESEARCH PAPER 607
608 Current Biology 1994, Vol 4 No 7
of shape and contour information by her visuomotor
control systems. Where is the damage in DF's brain? If,
as was discussed earlier, the perception of objects and
events is mediated by the ventral stream of visual projections
to the human equivalent of the monkey
inferotemporal cortex, then DF should show evidence
for damage relatively early in this pathway. Certainly, the
pattern of damage revealed by the MRI/is consistent
with-this interpretation; the major focus of cortical
damage is in the ventrolateral region of the occipital
cortex, an area that is thought to be part of the human
homologue of the ventral stream. At the same time, her
primary visual cortex, which provides input for both the
dorsal and ventral streams, appears to be largely intact. In
addition, the dorsal stream, unlike the ventral stream,
also receives input from the superior colliculus via the
pulvinar, a nucleus in the thalamus (see Fig. 1). Thus,
input from both these routes to the dorsal stream could
continue to mediate the production of well-formed
visuomotor responses in DE
Fig. 5. The 'grasp lines' (joining points where the thumb and
index finger first made contact with the shape) that were selected
by the optic ataxic patient (RV), the visual agnosic patient (DF),
and the control subject when picking up three of the twelve
shapes. The four different orientations inwhich each shape was
presented have been rotated so that they are aligned.
result suggests that some part of DF's remaining visual
system was able to analyse the outline shape of the object
and deliver the appropriate coordinates to the visuomotor
networks that control the movements of the hand
and fingers. This dissociation between profoundly
disturbed perception and intact visuomotor control
parallels the earlier observations about DF [3,4], which
concentrated on object size and orientation rather than
object shape.
In summary then, the brain damage that DF suffered as
a consequence of anoxia appears to have interrupted
the normal flow of shape and contour information into
her perceptual systems without affecting the processing
One must be cautious, however, about drawing strong
conclusions about anatomy and processing pathways
from patients like DE Her deficits arose, not from a
discrete lesion, but from anoxia. As a consequence, the
brain damage in DF, while localized to some extent, is
much more diffuse than it would be in a patient with a
stroke or tumour. Thus, although the striking dissociation
between DF's perceptual and visuomotor abilities
can be mapped onto the functional distinction between
the ventral and dorsal streams proposed by Goodale and
Milner [7,8], that mapping can be only tentative.
Observations in patients such as RV, whose pattern of
deficits is complementary to DF's and whose brain
damage can be localized to the dorsal stream with confidence,
have, however, strengthened the support for
Goodale and Milner's proposal.
RV, who had no difficulty discriminating between the
different shapes used in this experiment, could not use
information about object shape to guide the placement
of her fingers. Yet, as soon as her hand made contact
with the object, she corrected the position of her fingers
Fig. 6. The frequency distributions of
the distances between the grasp lines
and the centre of mass of the shape for
DF, RV and the control subject for all
twelve shapes. The inset shows how
those distances were calculated for two
different grasp lines.
Shape for perception and prehension Goodale et al. RESEARCH PAPER 609
before picking up the object. This suggests that, despite
RV's problems in visuomotor control, she was able to use
tactile and haptic information to control the placement
of her fingers. As the MRI shows, the damage in RV's
brain is largely confined to the occipitoparietal region
(see Fig. 2), an area which is thought to be homologous
to the dorsal stream in monkeys. RV's inability to use
shape information to guide her precision grip is probably
due to damage to these dorsal stream pathways; her
ability to discriminate between different shapes (and to
recognize line drawings) can be attributed to the fact that
the ventral stream was spared.
The dissociation between visual perception and the
visual control of skilled actions that was observed in
these neurological patients (and the mapping of this distinction
onto the ventral and dorsal streams) is supported
by work in the macaque monkey. Thus, monkeys that
show profound deficits in object recognition following
inferotemporal lesions are nevertheless as capable as
normal animals at picking up small food objects [13], at
catching flying insects [14] and at orienting their fingers
in a precision grip to grasp morsels of food embedded in
small oriented slots (S. Buchbinder, B. Dixon, Y-W.
Hyang, J.G. May, M. Glickstein, Soc Neurosci Abstr 1980,
6:675). In short, these animals behave in much the same
way as DF: they are unable to discriminate between
objects on the basis of visual features that they can clearly
use to direct their grasping movements. Moreover, there
is a long history of electrophysiological studies showing
that cells in the inferotemporal cortex are tuned to
specific objects or object features, and that the activity of
these cells is not affected by the motor behaviour of the
animal (for reviews, see [7,8,15]).
Conversely, ever since the pioneering work of Hyvirinen
and Poranen [16] and Mountcastle and colleagues [17],
we have known that visually sensitive cells in the
posterior parietal cortex, which is the major recipient
zone for dorsal stream projections in the monkey, are
modulated by the concurrent motor behaviour of the
animal. Thus, the activity of some visually driven cells in
this region has been shown to be linked to saccadic eye
movements; the activity of others to whether or not the
animal is fixating a stimulus; and the activity of other cells
to whether or not the animal is engaged in visual pursuit
or is making goal-directed reaching movements. Of most
interest, however, in the present context, is that some
cells in the posterior parietal area that fire when monkeys
reach out to pick up objects are selective not for the
spatially directed movement of the arm, but for the
movements of the wrist, hand, and fingers that are made
prior to and during the act of grasping the target [16,17].
Furthermore, it has been shown recently that many of
these cells are visually selective and are tuned for objects
of a particular shape and/or orientation [18,19]. These
manipulation neurons in the posterior parietal area thus
appear to be tied to the properties of the goal object as
well as to the distal movements that are required for
grasping that object.
Conclusions
The pattern of visual deficits and spared visual abilities in
RV and DF lends further support to the functional distinction
that Goodale and Milner [7,8] have made
between the two streams of visual processing in the
primate cerebral cortex. According to their hypothesis,
the ventral stream mediates the visual perception of
objects whereas the dorsal stream plays the major role in
the visual control of motor acts, such as manual prehension,
directed at those objects. Both streams, it would
appear, have access to visual information about the shape
of objects but each stream uses this information for
different purposes.
Materials and methods
Discriminationtests
For the same/different discrimination tasks, two complete sets
of the twelve shapes were used. In the first discrimination task,
the relative orientation of the two shapes on a 'same' trial was
identical; in the second task, the relative orientation of the two
shapes varied from trial to trial, with the principal axis of one
shape being rotated 900, 1800 or 2700 with respect to the
other. For both tasks, the subject was seated at a table covered
with a black cloth, and the pairs of shapes were placed in front
of her, two at a time and 11 cm apart, 30cm from the edge of
the table. The order of testing was quasi-randomized: the
number of same and different trials was counterbalanced but all
other factors (shape, orientation, left-right position) were
completely randomized. For each task, there were three
practice trials and 56 test trials.
Graspingtasks
For the grasping task, subjects wore rubber thimbles, one on
the thumb and one on the index finger, that were inked with
different colours. Each shape was placed, one at a time, at one
of the four orientations used for the same/different discrimination
(0°
, 900, 1800, and 2700) 30cm from the edge of the table
in front of the subject, or 20cm to the left or right of this
midline position. Each shape was presented four times at these
different orientations and positions according to a quasirandom
schedule for a total of 96 trials. Between trials, subjects
were asked to close their eyes and then, with the instruction
'open your eyes', to reach out naturally and pick up the shape
in front of them with a 'precision grip'. The centres of the two
marks on the object boundary where the thumb and index
finger first contacted the shape were then recorded on
prepared templates. The ink marks were then removed from
the shapes with alcohol.
Acknowledgements: The authors would like to thank 1)F and RV
for the cooperation and good humour they showed throughout
testing. This research was supported by a grant from the Medical
IResearch Council of Canada to M.A. Goodale.
References
1. Jeannerod M. The Neural and Behavioural Organization of GoalDirected
Movements. Oxford: Clarendon Press; 1988.
2. Jakobson LS, Goodale MA: Factors affecting higher.order
movement planning: a kinematic analysis of human prehension.
Exp Brain Res 1991, 86:199-208.
Shape for perception and prehension Goodale et al. RESEARCH PAPER 609
610 Current Biology 1994, Vol 4 No 7
3. Goodale MA, Milner AD, Jakobson LS, Carey DP: A neurological
dissociation between perceiving objects and grasping them. Nature
1991, 349:154-156.
4. Milner AD, Perrett DI, Johnston RS, Benson PJ,Jordan TR, Heeley
DW, Bettucci D, Mortara F,Mutani R,Terazzi E, Davidson DLW:
Perception and action in visual form agnosia. Brain 1991,
114:405-428.
5. Perenin M-T, Vighetto A: Optic ataxia: a specific disruption in
visuomotor mechanisms. . Different aspects of the deficit in
reaching for objects. Brain 1988, 111:643-674.
6. Jakobson LS, Archibald YM, Carey DP, Goodale MA: A kinematic
analysis of reaching and grasping movements in a patient recovering
from optic ataxia. Neuropsychologia 1991, 29:803-809.
7. Goodale MA, Milner AD: Separate visual pathways for perception
and action. Trends Neurosci 1992, 15:20-25.
8. Milner AD, Goodale, MA: Visual pathways to perception and
action. In The Visually Responsive Neuron: From Basic Neurophysiology
to Behavior. Progress in Brain Research. Vol 95. Edited
by Hicks TP, Molotchnikoff S, Ono T. Amsterdam: Elsevier;
1993:317-338.
9. Ungerleider LG, Mishkin M: Two cortical visual systems. In
Analysis of Visual Behavior. Edited by Ingle D, Goodale MA,
Mansfield RJW. Cambridge, Massachusetts: MIT Press:
1982:549-585.
10. Crick F,Jones E: Backwardness of human neuroanatomy. Nature
1993, 361:109-110.
11. Blake A: Computational modelling of hand-eye coordination. Phil
Trans RSoc Lond 1992, 337:351-360.
12. Iberall T, Bingham G, Arbib MA: Opposition space as a structuring
concept for the analysis of skilled hand movements. In Generation
and Modulation of Action Patterns. Experimental Brain Research
Series. Vol 15. Edited by Heuer H, Fromm C. Berlin: SpringerVerlag;
1986:158-173.
13. Kluver H, Bucy PC: Preliminary analysis of functions of
the temporal lobes of monkeys. Arch Neurol Psychiatr 1939,
42:979-1000.
14. Pribram KH: Memory and the organization of attention. In Brain
Function. Vol IV. UCLA Forum in Medical Sciences 6. Edited by
Lindsley DB, Lumsdaine AA. Berkeley: University of California
Press; 1967:79-112.
15. Goodale MA: Visual pathways supporting perception and action in
the primate cerebral cortex. Curr Opin Neurobiol 1993,
3:578-585.
16. HyvArinen J, Poranen A: Function of the parietal associative area 7
as revealed from cellular discharges in alert monkeys. Brain 1974,
97:673-692.
17. Mountcastle VB, Motter BC, Steinmetz MA, Duffy CJ: Looking and
seeing: the visual functions of the parietal lobe. In Dynamic
Aspects of Neocortical Function. Edited by Edelman G, Gall WE,
Cowan WM. New York: Wiley; 1984:160-193.
18. Taira M, Mine S, Georgopoulos AP, Murata A, Sakata H: Parietal
cortex neurons of the monkey related to the visual guidance of
hand movement. Exp Brain Res 1990, 83:29-36.
19. Sakata H, Taira M, Mine S, Murata A: Hand-movement-related
neurons of the posterior parietal cortex ofthe monkey: their role in
visual guidance of hand movements. In Control of Arm Movement
in Space: Neurophysiological and Computational Approaches.
Edited by Caminiti R, Johnson PB, Burnod Y. Berlin: SpringerVerlag;
1992:185-198.
Received: 21 March 1994; revised: 27 May 1994.
Accepted: I June 1994.