Neuropsychologia 47 (2009) 1621–1626
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/neuropsychologia
Vision in the palm of your hand
Liana E. Browna, Brendan F. Morrisseyb,c,d, Melvyn A. Goodaleb,c,d,∗
aDepartment of Psychology, Trent University, Peterborough, Ontario, Canada K9J 7B8
bDepartment of Psychology, University of Western Ontario, London, Ontario, Canada N6A 5C2
cDepartment of Physiology and Pharmacology, University of Western Ontario, London, Ontario, Canada N6A 5C2
dCIHR Group on Action and Perception, University of Western Ontario, London, Ontario, Canada N6A 5C2
a r t i c l ei n f o
Received 23 May 2008
Received in revised form 11 August 2008
Accepted 21 November 2008
Available online 28 November 2008
a b s t r a c t
Here we show that pointing movements made to visual targets projected onto the palm of the hand are
more precise and accurate than those made to targets projected onto back of the hand. This advantage
may be related to the fact that the number of cortical bimodal neurons coding both visual and tactile
stimuli increases with tactile receptor density, which is known to be higher in glabrous than in hairy
© 2008 Elsevier Ltd. All rights reserved.
The hand and face are among the most densely innervated
regions of the body with respect to touch. There are a greater num-
ber of tactile receptors embedded in the skin of the hand and face
than in other regions of the body (Edin, Essick, Trulsson, & Olsson,
1995) and there is a greater proportion of somatosensory cortex
devoted to these areas relative to other body parts (Penfield &
tactile perceptual resolution as evaluated by two-point discrimina-
tion tests (Weinstein, 1968).
Neurophysiological studies in the macaque monkey have
revealed neurons in different brain regions that respond both to
visual and tactile stimulation with overlapping receptive fields
(RFs). These bimodal neurons have been found in the dorsal half of
the ventral premotor cortex [PMv; a.k.a. the polysensory zone (PZ);
the putamen (Graziano, Yap, & Gross, 1994). The skin and pericuta-
neous space near the hands and face are represented more densely
by these bimodal neurons than are the skin and surrounding space
of other parts of the body. In short, there appears to be a correspon-
dence between tactile receptor density and bimodal-cell density.
Body parts with higher tactile receptor density have greater visual-
∗Corresponding author at: CIHR Group on Action and Perception, University of
Western Ontario, London, Ontario, Canada N6A 5C2. Tel.: +1 519 661 2070;
fax: +1 519 661 3961.
E-mail address: email@example.com (M.A. Goodale).
tactile bimodal representation than body parts with lower tactile
density and perceptual resolution.
An interesting feature of visual-tactile neurons is that they can
be recruited by a visual stimulus presented alone (i.e. without any
the hand or face. Thus, the question arises as to whether or not the
‘quality’ of the representation of a purely visual stimulus presented
on the skin is correlated with the tactile resolution of that skin. In
with high tactile receptor density could mean that visual stimuli
that appear on or near these regions would be represented more
robustly than visual stimuli presented on or near skin regions with
low tactile receptor density.
To test this possibility, we asked healthy undergraduates to per-
form visually guided pointing movements with their right hand
to targets presented on their palm or on the back of their left
hand. We chose this task to measure target quality because we
know that measures of pointing accuracy and precision are sen-
sitive to the location of targets defined using tactile and noxious
stimuli (Koltzenburg, Handwerker, & Torebjork, 1993; Moore &
Schady, 1995). Moreover, it seems likely that bimodal neurons
evolved for the control of visually guided movements in concert
with tactile information (e.g. Graziano & Cooke, 2006). Palmar
(Johansson & Vallbo, 1979). In contrast, the less sensitive hairy skin
on the back of the hand has an estimated tactile receptor density of
less than 5units/cm2(Macefield, 1998). If the relationship between
0028-3932/$ – see front matter © 2008 Elsevier Ltd. All rights reserved.
L.E. Brown et al. / Neuropsychologia 47 (2009) 1621–1626
Fig. 1. Experimental Setup. The projector and mirror were used to project target circles onto the glabrous or hairy skin of the participant’s left hand or a fake hand. The Plato
A shows the array of target locations that were presented singly in random order on each trial. Inset B shows a participant making a leftward movement to a target location
on her own hairy skin without visual feedback (the PLATO goggles are occluding vision). This picture differs from the experimental situation in that during the experiment,
the only lighting was from the projector.
tactile innervation and bimodal-cell representation of skin region
extends to the different surfaces of the hand, the difference in tac-
tile receptor density between the glabrous and hairy skin could
be linked to a difference in bimodal-cell representation, and in
turn to differential processing of visual stimuli presented on the
glabrous and hairy skin. Thus, we predicted that visually guided
pointing movements directed at visual stimuli presented on the
palm would be more accurate and precise than similar movements
made to visual stimuli presented on the back of the hand, even in
the absence of tactile feedback. To establish a baseline and to con-
trol for the possibility that participants could use visual landmarks
to locate targets on the glabrous- or hairy-skin surfaces, we had
participants point to targets on a convincingly realistic fake hand
as well as to targets on their own hand (see Fig. 2). We predicted
Fig. 2. Participants pointed to visual targets projected onto the hairy or glabrous skin of their own left hand (inset, Panels A and C, respectively) or onto the hairy and glabrous
surfaces of a realistic replica of a left hand (inset, Panels B and D, respectively).
L.E. Brown et al. / Neuropsychologia 47 (2009) 1621–1626
provided informed consent before participating in the study.
Visual targets were projected by an LCD projector (InFocus) onto an angled
mirror that was suspended over the target hand (see Fig. 1). The mirror reflected
these visual targets onto either the participant’s own real left hand or a realistic left-
hand replica (Anytime Costumes) that was secured comfortably to the table below
with a VelcroTMstrap. The fake hand was constructed of dense, heavy foam. On
the glabrous-skin side, it was detailed with skin-like surface features that included
the high-spatial-frequency finger- and palm-print texture and all of the major low-
frequency crease lines. Although the hairy-skin surface did not have hair, it did have
knuckle indentations and raised areas depicting the finger-extensor tendons and
surface veins (see Fig. 2). The room was dark except for the light coming from the
play goggles (PLATO, Translucent Technologies) that were worn by the participants.
The PLATO goggles were to control stimulus-exposure time. An infrared-emitting
diode (IRED), attached to the tip of the participant’s right index finger, was used
to track the trajectory of the pointing movements. The IRED was tracked with an
OPTOTRAKTM3D motion analysis system (Northern Digital Inc., Waterloo, Ontario,
Canada). The OPTOTRAKTMrecorded the three-dimensional position of the IRED at
a frequency of 200Hz. The positions were saved and later analyzed off-line.
The targets appeared as bright, white circles, 1, 2, and 3cm in diameter. A visual
texture was also projected onto the target surface in an additional effort to prevent
participants from using glabrous- and hairy-skin surface features as cues for coding
target location on the skin surface. Each target was projected to one of six different
randomly chosen locations on the glabrous- or hairy-skin surface of the palm of the
hand. Six target locations were used to minimize the participants’ tactile memory
for target location over the course of the experiment. A center target position was
encircled by five surrounding positions, which were each 1cm away from the center
position (see Fig. 1). The center of the palm and back of the hand were measured
and aligned with the central target at the beginning of each block. Although the
targets were projected onto the hand, the location of the targets in space was fixed
and did not vary with the projection surface. The distance between the start button
and the center target position was 35cm. The center target position and the start
button were both 20cm away from the participant in depth (Fig. 1). It should be
emphasized once more that the stimuli were entirely visual and there was no tactile
information about target location.
1.4. Experimental design and procedure
The experiment followed a 2 hand-type (real, fake) by 2 skin-side (glabrous,
hairy) by 3 target-size (1, 2, 3cm diameter)×6 target-location within-subjects
left hand. The four hand-skin conditions, real-glabrous, real-hairy, fake-glabrous,
and fake-hairy, were presented in eight blocks of 36 trials. Block order was coun-
terbalanced across participants. Within each block, target size and location were
presented pseudo randomly such that each combination of target size and location
was repeated twice. During the real-hand conditions, the fake hand was placed out
of view and during the fake-hand conditions, participants’ were instructed to place
their own left hand below the table across their lap.
At the beginning of each trial, participants’ vision was occluded. The target was
projected onto the hand, and after a variable foreperiod (500–1250ms), the PLATO
goggles were opened and the participant heard an audible tone. Participants were
instructed to initiate a quick and accurate pointing response to the center of the
chronized with the closure of liquid-crystal display goggles worn by the participant,
occluding participants’ vision and removing any opportunity to use vision to adjust
the hand’s trajectory en route. Although the instructions stressed movement speed
for longer than 1000ms. Participants were also instructed to make smooth point-
ing movements and to avoid making secondary, corrective movements. In any case,
as described below, if corrective movements did occur, they were not included in
the definition of the movement. Although participants did touch what they thought
was the target location at the end of each pointing movement, they were never
given the opportunity to compare their pointing performance to the actual target
location. That is, they received no post-performance feedback about the accuracy of
the movement they had just executed.
Although the physical location of the targets was fixed in space, it was possible
that the projected location of the target may have differed between participants due
to factors like the thickness of the hand. To address this possibility, we measured
the location of each target on each participant’s hand at the end of the experiment.
Participants were given as much time as they needed to point to the center of the
1-cm target as accurately as possible. These pointing trials were performed with full
vision of the target. These measurements were used as the standard for calculating
pointing error within each individual participant.
1.5. Data analysis and dependent measures
The OPTOTRAKTMsystem recorded the position of the IRED in the x, y, and z
dimensions over time. Analysis programs written with Matlab (The Mathworks,
Inc.) were used to define the beginning and end of each pointing movement. The
onset and the end of the movement were defined as the time when the resultant
velocity of the index finger first exceeded or and then first fell below (respec-
error in both the horizontal and depth dimensions, and endpoint error variability
(standard deviation) of the right index finger were calculated. Each of these mea-
sures were submitted to repeated-measures analysis of variance. Interactions were
decomposed by conducting simple main effects analyses. Main effects for factors
involving more than two levels (target and location) were further investigated with
Tukey’s HSD post-hoc tests.
Fig. 3 shows the final position of the pointing movements per-
formed by all 16 participants in each of our four target-surface
els A and C) and the hairy and glabrous surfaces of the fake hand
(panels B and D). We compared the resolution of the target repre-
Fig. 3. Here we show the end position with respect to the target location (i.e. the
pointing error) for every pointing movement for all participants (?) in each of the
four target-presentation conditions—the hairy or glabrous (palm) skin of their own
left hand (Panels A and C, respectively) or the hairy and palm “skin” surfaces of the
fake left hand (inset, Panels B and D, respectively). The black cross (+) represents the
in each condition. The grey ellipses represent error variability, the 95% confidence
interval of the error in the horizontal and depth dimensions.
L.E. Brown et al. / Neuropsychologia 47 (2009) 1621–1626
Fig. 4. (Panel A) Error variability as a function of hand type and skin side. Error
the hairy skin only when those targets were projected onto the real hand. (Panel B)
Error variability as a function of target size and hand type. Error variability was
significantly less on the real hand than the fake hand, and a significant increase in
variability with target size was marginally attenuated when targets were presented
on the real hand.
sentation in these four conditions by analyzing pointing variability,
grey circles and the center of the black circle for each of condi-
tions in Fig. 3). This analysis revealed an interaction between hand
type and skin side, F(1, 15)=9.56, p=.007. As illustrated in Fig. 4A,
pointing movements to the real hand were more precise to tar-
than on the hairy skin (14.95±.20mm), F(1, 15)=5.65, p=.031,
but when pointing movements were made to the fake hand,
the precision of pointing to the glabrous (16.45±.21) and hairy
skin (16.20±.20mm) did not differ significantly, F(1, 15)=.40,
Predictably, pointing variability increased with target size, F(2,
15)=4.29, p=.023, and this effect interacted marginally with hand
were attenuated for targets presented on the real hand in com-
parison to the fake hand (see Fig. 4B). There were no interactions
(16.94±.23mm) were more variable than movements to all other
locations (mean at 5 other locations: 14.71±.24mm). This effect
interacted with both hand and skin, F(5, 15)=2.54, p=.035, such
that the variable error at this location was significantly reduced
(13.18±.47mm) in the real-hand, glabrous-skin condition only.
These patterns were also reflected in the directional error vari-
ability. Directional error variability is represented by the ellipses
shown in Fig. 3, whose vertical and horizontal radii represent
the 95% confidence interval of the signed error along the depth
(away from the body) and horizontal (along the axis of move-
ment) dimensions, respectively. There was an interaction of hand
type and skin side for pointing variability in depth, F(1, 15)=6.66,
p=.021. Again, for movements to the real hand, pointing preci-
sion in depth was greater for movements made to targets on the
15)=8.57, p=.01, but this skin effect was not present for the fake
hand, F(1, 15)=.00, p=.993 (“glabrous side”: 8.32±.24mm; “hairy
such that horizontal movement precision was greater for move-
ments made to the glabrous side (7.26±.25mm) than to the hairy
side (8.71±.24mm) of either the real or fake hand.
Analysis of the mean signed error along the horizontal axis
of movement (from right to left) revealed that although partici-
pants were more likely to undershoot the target when pointing
to their own left hand (5.44±.18mm) than to the fake left hand
(.80±.18mm), F(1, 15)=25.40, p<.001, this error was signifi-
cantly smaller on the glabrous skin (3.88±.25mm) than on the
hairy skin (7.04±.25mm), F(1, 15)=11.94, p=.004. In the depth
dimension, the mean signed error was smaller for targets on the
real hand (7.11±.17mm) than on the fake hand (10.45±.17mm),
F(1, 15)=6.06, p<.026. These differences in accuracy cannot be
attributed to a speed-accuracy trade-off, as movement times were
marginally shorter for the real-hand targets than for fake-hand
targets, F(1, 15)=3.04, p=.056, and were marginally shorter for
glabrous-skin targets than for hairy-skin targets, F(1, 15)=4.02,
Participants may have used tactile information from either the
pointing finger (in all conditions) or from the target surface (in the
real-hand conditions) to learn the tactile location of targets as tri-
als were repeated, and it is possible that this learning was better
in the real-hand, glabrous-skin condition than in other conditions.
We used two methods to address the possibility that participants
learning was better for the glabrous than the hairy skin. First, when
we analyzed only the first trial in each condition, we still found
that movements directed at targets on the glabrous skin of the real
hand (12.01±.49mm) were significantly more precise than those
directed to targets on the hairy skin (14.94±.49mm), F(1, 15)=7.11,
p=.018, or on either side of the fake hand, F(1, 15)=.87, p=.367
(“glabrous side”: 15.19±.50; “hairy side”: 16.37±.49). Thus, pre-
cision was better in the real-hand, glabrous-skin condition even
we looked for evidence of tactile learning by assessing changes
in error variability as trials were repeated within each condition
(see Fig. 5). We found that there was no significant overall change
in error variability as a function of trial repetition, F(5, 79)=2.25,
p=.056, and there was no systematic decrease in error variability
ses, there were no significant interactions involving trial repetition.
learning as a strategy for reducing error variability in the real-hand
conditions. Moreover, there is no evidence that participants used
tactile learning as a strategy for reducing error variability in the
real-hand-glabrous-skin condition over the real-hand-hairy-skin
L.E. Brown et al. / Neuropsychologia 47 (2009) 1621–1626
Fig. 5. Error variability as a function of trial repetition in each hand-skin condition.
There was no systematic decrease in error variability between Trial 1 and Trial 2,
or between Trial 1 and Trial 6. Therefore, we found no evidence suggesting that
participants used tactile learning as a strategy for reducing error variability in the
real-hand conditions over the fake-hand conditions, or in the real-hand-glabrous-
skin condition over the real-hand-hairy skin condition.
the hand are represented more robustly because of links between
tactile receptor and cortical density and visual-tactile bimodal-cell
tactile receptor density and cortical representation are also more
densely represented by bimodal neurons. Visual stimuli shown in
the palm of the hand recruit bimodal neurons and these neurons
contribute to a more robust visual representation of the target,
allowing pointing movements to be performed more precisely to
targets that appear in the palm of the hand. Kennett, Taylor-Clarke,
and Haggard (2001) used a magnifying glass to exaggerate vision of
tion there. The complementary result we show here demonstrates
that the more naturally occurring variations in skin tactile recep-
tor and cortical density can have a tangible impact on how visual
information on and perhaps near the skin is represented.
One alternative explanation for our results is that participants
used skin-surface landmarks to locate the target with more preci-
sion and accuracy on the glabrous skin than the hairy skin of the
allow us to be confident that our effects cannot be explained by the
possible visual surface-feature differences between the glabrous
and hairy skin.
We used a convincing fake hand to control for the visual mark-
ings on the glabrous and hairy skin. This fake hand is so detailed
that on the glabrous skin it has all of the major low-frequency
crease lines on the palm and it had the high-frequency textural
hairy surface does not have hair, it does have knuckle indentations
and raised areas representing finger-extensor tendons and surface
the glabrous surfaces, yet there were no differences in precision or
there were significant advantages for the glabrous skin of the real
hand, contrary to the predictions of a landmark-based localization
Finally, to reduce the possible impact of landmarks on perfor-
mance, we masked all of these surface details by projecting a high
contrast grid pattern onto each target surface. We believe the grid
was equally effective at obscuring skin landmarks on the real and
fake hands. Moreover, because the grid and the target were pro-
jected onto the skin in a darkened room, the contrast between the
target and the grid was identical on the palm and hairy surface of
both the real and the fake hands. Taken together then, the differ-
ences in the visual features of the target surfaces do not appear to
capture the pattern of data presented, i.e. the superior precision
and accuracy of pointing to targets presented on the glabrous skin
of the real hand.
The superior performance with targets presented on the real
as compared to the fake hand may reflect not only a difference in
visual processing but also the fact that proprioceptive information
been particularly useful for determining hand position in depth
(van Beers, Sittig, & Van Der Gon, 1999). Proprioceptive and tactile
contributions to hand localization cannot, however, account for the
fact that participants were both more accurate and more precise
in pointing to targets on the glabrous rather than the hairy skin
of their real hand. Although muscle length- and force-receptors,
and proximal-joint receptors and skin deformation, provide pro-
prioceptive information about the relative position of distal limb
segments, it is unlikely that they provide differential information
about the location of particular areas of skin on those distal limb
segments. Moreover, because presenting targets on the hairy skin
of the real hand required that the glabrous skin be compressed
as the hand was stabilized, and vice versa, tactile contributions to
targets were presented on the hairy skin rather than the glabrous
tile imagery strategy. If this was true, intersubject variability would
have been large due to individual differences in imagery ability and
strategy. One would still need to map the visual stimulus onto the
tactile map, without the benefit of any tactile stimulation. In short,
this seems a far less parsimonious explanation, given the primacy
of vision for target-directed movements and the complete absence
of any tactile cues in this task.
The superior performance with visual stimuli presented on the
real hand, particularly on the glabrous surface of the hand, may
reflect the role of bimodal cells in the fine motor control related to
object manipulation. In many tasks, the left hand is used to stabi-
lize the object of interest while the right hand is used to act upon
and manipulate that object and its parts (Sainburg, 2005). Bimodal
cells may therefore have evolved in the context of manual inter-
actions. In addition, the important role of the hands in feeding,
particularly in primates, may also have been a significant factor
in the emergence of visuotactile cells not only in association with
the hands but also the mouth and face. In future studies, it would
be of interest to explore differences in the engagement of these
putative bimodal mechanisms between the left and right hands,
Edin, B. B., Essick, G. K., Trulsson, M., & Olsson, K. A. (1995). Receptor encoding of
moving tactile stimuli in humans. I. Temporal pattern of discharge of individual
low-threshold mechanoreceptors. Journal of Neuroscience, 15, 830–847.
Fogassi, L., Gallese, V., Fadiga, L., Luppino, G., Matelli, M., & Rizzolatti, G. (1996).
Coding of peripersonal space in inferior premotor cortex (area F4). Journal of
Neurophysiology, 76, 141–157.
Graziano, M. S. A., & Cooke, D. F. (2006). Parieto-frontal interactions, personal space,
and defensive behavior. Neuropsychologia, 44, 845–859.
Graziano, M. S. A., & Gandhi, S. (2000). Location of the polysensory zone in the
precentral gyrus of anesthetized monkeys. Experimental Brain Research, 135,
1626 Download full-text
L.E. Brown et al. / Neuropsychologia 47 (2009) 1621–1626
Graziano, M., Yap, G., & Gross, C. (1994). Coding of visual space by premotor neurons.
Science, 266, 1054–1057.
Johansson, R., & Vallbo, A. (1979). Tactile sensibility in the human hand: relative
and absolute densities of four types of mechanoreceptive units in glabrous skin.
Journal of Physiology, 286, 283–300.
Kennett, S., Taylor-Clarke, M., & Haggard, P. (2001). Noninformative vision improves
the spatial resolution of touch in humans. Current Biology, 11, 1188–1191.
Koltzenburg, M., Handwerker, H. O., & Torebjork, H. E. (1993). The ability of humans
to localise noxious stimuli. Neuroscience Letters, 150, 219–222.
Macefield, V. (1998). The signaling of touch, finger movements and manipulation
forces by mechanoreceptors in human skin. In J. Morley (Ed.), Neural aspects of
tactile stimulation (pp. 89–130). The Netherlands: Elsevier Science B.V.
Moore, C. E., & Schady, W. (1995). Cutaneous localisation of laser induced pain in
humans. Neuroscience Letters, 193, 208–210.
Penfield, W., & Boldrey, E. (1937). Somatic motor and sensory representation in
the cerebral cortex of man as studied by electrical stimulation. Brain, 60,
Sainburg, R. L. (2005). Handedness: Differential specialization for control of trajec-
tory and position. Exercise and Sport Science Reviews, 33, 206–213.
van Beers, R., Sittig, A., & Van Der Gon, D. (1999). Integration of proprioceptive
and visual position information: An experimentally supported model. Journal
of Neurophysiology, 81, 1355–1364.
Weinstein, S. (1968). Intensive and extensive aspects of tactile sensitivity as a func-
tion of body part, sex, and laterality. In D. Kenshalo (Ed.), The skin senses (pp.
195–222). Springfield, Illinois: Thomas.