Exp Brain Res (2001) 139:160167

DOI 10.1007/s002210100743

R E S E A R C H A RT I C L E

Steven P. Tipper Nicola Phillips Chris Dancer

Donna Lloyd Louise A. Howard Francis McGlone

Vision influences tactile perception at body sites

that cannot be viewed directly
Received: 29 November 1999 / Accepted: 6 March 2001 / Published online: 30 May 2001
Springer-Verlag 2001

Abstract Previous research has demonstrated that vision of a body site, without proprioceptive orienting of
eye and head to that site, could affect tactile perception.
The body site viewed was the hand, which can be seen
directly under normal viewing conditions. The current
research asked three further questions: First, can vision
similarly affect tactile perception at a body site that cannot normally be viewed directly such as the face or
neck? Second, does prior experience of seeing a body
site, such as occurs when viewing the face in mirrors,
produce larger effects of viewing than body sites rarely
seen such as the back of the neck? And third, how quickly can visual information affect tactile target detection?
We observe that: detection of tactile targets at these body
sites was influenced by whether or not they were viewed,
this effect was greater when viewing the more familiar
site of the face than that of the neck, and significant effects were observed when the stimulus onset asynchrony
between visual display and tactile target was as little as
200 ms.
Keywords Proprioceptive orienting Target detection
Vision-touch interactions Prior experience Human

Introduction
It is essential for goal-directed behaviour that perceptual
systems are able to rapidly deliver accurate information
about the external environment. This information can arS.P. Tipper () L.A. Howard
School of Psychology, University of Wales, Bangor, Gwynedd,
Wales, LL57 2DG, UK
e-mail: s.tipper@bangor.ac.uk
Fax: +44-1248-382599
N. Phillips C. Dancer D. Lloyd F. McGlone
Cognitive Neuroscience Group, Unilever Research, Port Sunlight,
Wirral, CH63 3JW, UK
F. McGlone
Centre for Cognitive Neuroscience, University of Wales, Bangor,
Gwynedd, Wales, LL57 2DG, UK

rive via multiple sensory modalities, and these have to

be integrated to produce a coherent internal representation of the world (Driver and Spence 1998). For example, a touch to a body surface will trigger a rapid orienting of eyes and head towards the stimulated site (Groh
and Sparks 1996), in an effort to gather further information, and will determine subsequent behaviour. Whether
the source of the stimulation is the touch of a blade of
grass or the movement of a potentially dangerous spider
can only be ascertained by integrating vision and touch.
A number of studies have shown that orienting the
eyes and head to a body site facilitates touch (Larmande
and Cambier 1981; Pierson et al. 1991). This facilitation
is observed even when subjects cannot see the body site
being stimulated, such as when orienting while blindfolded or in complete darkness (Driver and Grossenbacher 1996; Honore et al. 1988). However, from previous
research it is not known whether vision itself facilitates
the sense of touch, because it is always confounded with
proprioceptive orienting of eyes and head to the body
site.
Recently, Tipper et al. (1998) have examined the effects of vision of a body site while proprioceptive orienting of the eyes and/or head to that body site was prevented. This pure vision condition was achieved by means of
a video camera. While the hands were occluded from direct view by being placed within concealing covers, a
camera presented a real-time image of one of the hands
on a monitor placed at the body midline. It was discovered that detection of tactile targets is faster when the
stimulated hand is visible than when the other hand is
visible. Subsequent work by Ladavas et al. (2000) has
confirmed the critical role of vision for tactile perception
on the hand independent of proprioceptive orienting.
The present research takes the investigation of the effects of vision on tactile perception further in two ways.
First, in the study by Tipper et al. (1998), the stimulated
body site was the hand. During normal visuomotor processes, this site is directly viewed by either orienting eye
and head or by moving the hand into the line of sight.
Therefore, there has been a long evolutionary and person-

161

al history of integrating vision, proprioception and touch

at such visible body sites. The study shows that vision
can be influential even in the absence of proprioceptive
orienting, but it could be argued that this is the case only
because the hand is a site that is habitually seen via proprioceptive orienting. Stronger evidence for a direct role
of vision in tactile perception would be provided if such
effects were also obtained at body sites that are not available for proprioceptive orienting. Many body sites such
as the face and back cannot be oriented towards, in that
we cannot directly see them. Therefore, if pure visual effects were to be obtained at these sites it would be strong
evidence for the effects of vision on touch independent of
the learning acquired during proprioceptive orienting.
The second issue concerns the effects of prior experience in linking visual and tactile inputs. As discussed,
some body sites (e.g. hand) have a long history of integrated visual, proprioceptive and tactile inputs, because
they can be viewed directly from the earliest stages of development. However, body sites that can never be viewed
directly also have different levels of prior visual experience. Our hypothesis is that prior links between vision and
touch will produce larger experimental effects. To this
end, we compared tactile target detection at the face and
back of the neck. In Western cultures, the face is indirectly
viewed via mirrors on many occasions. Throughout any
day, most individuals see their face at least once in a mirror, and numerous occurrences can take place every day.
Furthermore, many of these occasions involve both visual
and tactile stimulation, as when shaving, putting on makeup, dressing blemishes, etc. In sharp contrast, vision of the
back of the neck is very rare. This is especially the case
among the women taking part in our experiments who
tended to have long hair covering this body site.
In the present study, subjects were required to detect the
presence of a tactile target at one body site, for example the
face, while ignoring distractors at the other possible body
site, i.e. the neck. Whilst undertaking this task, the subjects
were provided with three forms of visual information:
compatible, for example detecting a tactile target at the
face while viewing the face; neutral, for example detecting
a tactile target at the face while viewing the hand, which
was never stimulated; incompatible, for example detecting
tactile targets on the face while viewing the neck.
The difficulty of the targets to be detected was also
manipulated. In one condition, pure tactile stimuli were
presented, as in the experiments of Tipper et al. (1998).
This was achieved by presenting white noise at a level
that masked the auditory properties of the tactile stimulators. In a second condition, both tactile and auditory information was presented, in which the volume of the
white noise was turned down. It was assumed that the
task would be generally easier when there were two
compatible sources of information (tactile + auditory)
that subjects had to detect and localize, than when there
was only tactile input. This enabled us to determine
whether the effects of vision were magnified in a more
demanding target-detection task (pure tactile) than a less
demanding one (tactile + auditory).

Experiment 1
Materials and methods
Subjects
Subjects were 24 right-handed women (12 in the pure tactile condition and 12 in the tactile + auditory condition) selected from the
Unilever subject panel, who were paid a small fee. They were in
the age range 2235 years (mean 25.6 years) and had normal or
corrected-to-normal visual acuity. All subjects gave informed consent.
Stimuli
Stimuli were delivered via two electromagnetic contact transducers (bone conductors normally used in hearing aids) secured to the
skin with double-sided adhesive tape in two positions: one on the
centre of the right cheek just below the cheekbone, the other on
the back of the neck to the right of the spine (see Fig. 1 for details). A third stimulator was fixed to the thenar eminence (fleshy
base of thumb) of the right hand, but subjects did not receive any
stimulation to this site, as it was placed there to maintain consistency when viewing this neutral condition. The target was an isointense 200-Hz, 150-ms stimulus, set at 5 times the detection
threshold for each site (14 dB attenuation), reflecting the relative
sensitivities of these areas. Headphones providing white noise
masked any auditory cues generated by the stimulators.
Design
Whilst undertaking the tactile target-detection task, subjects experienced three visual-tactile compatibility conditions: (1) compatible, for example detecting a tactile target at the face while viewing
the face, or detecting a tactile target at the neck while viewing the
neck; (2) neutral, for example detecting a tactile target at the face
or neck while viewing the hand which was never stimulated; (3)
incompatible, for example detecting tactile targets on the face
while viewing the neck, or vice versa.
Each subject was tested in all three conditions (compatible,
neutral and incompatible) at both test sites on each of two consecutive days. Six test blocks (three test conditions per stimulated
site) were undertaken each day, which took approximately 30 min.
Within each block, 20 stimuli were randomly presented: 18 to the
target site and 2 to the distractor site (totaling 60 trials per test site
per day); however, trials were repeated in the case of errors. The
order of test blocks was counterbalanced between subjects and
across days.
Procedure
After depressing the foot-pedal, tactile targets were presented randomly between 850 and 1850 ms later. This variability prevented
prediction of tactile-target onset. The subject's primary task was to
release the foot pedal when a vibration was detected at the target
site and to refrain from responding when stimulated at the distractor site (10% of all trials). Error feedback was provided in the
form of a beep if subjects responded to a distractor, responded before onset of the target or failed to respond to a target (more than
1,000 ms) and the trial was repeated. Vision of the three body sites
(face, neck and hand) was achieved by means of a monochrome
video camera, trained only on the site to be tested, which presented a real-time image on a monitor situated at eye level 50 cm in
front of the subject at their midline. Subjects rested their heads on
an adjustable chin rest. The surrounding body areas, and especially the eyes, were never seen on the monitor, as a pilot study
showed that this was too distracting for the subjects and increased
error rates. In addition, when the camera was moved between
sites, subjects were told to close their eyes.

162

Fig. 1AC A participant is shown viewing the monitor placed at

the midline for each of the three viewing conditions: A neutral;
B neck; C face. In the face condition, care was taken to exclude
the eyes from the viewed image. Stimulators are shown in position
on the face, neck, and hand. There were three visual-tactile relationships: (1) neutral; for example, viewing the hand (A) while detecting tactile targets to the face or neck; (2) compatible; for example, viewing the neck (B) while detecting tactile targets to the
neck; (3) incompatible; for example, viewing the neck (B) while
detecting tactile targets to the face

discarding values more than 150% or less than 66% of the median,
and taking the mean of the remaining sample.
A practice session was included at the start of each day's test
session (consisting of five stimuli to the face and five to the neck)
in which subjects also rated the intensity of the stimuli at both
sites so that these could be equated. In tactile-only conditions, the
auditory mask was raised above the level of the noise from the
stimulators for each subject during the practice trials.

Results
Subjects were advised at the beginning of each block as to
where the target site was and to respond only to stimuli at that site
and ignore stimuli to the distractor site. They were also instructed
to keep looking at the video monitor throughout each test block
during each of the three test conditions (compatible, looking at the
target site; neutral, looking at the hand; and incompatible, looking
at the distractor site). Trimmed means were obtained for each subject in each cell of the analyses by identifying their median score,
Table 1 Mean and standard deviation of foot-pedal response
times in milliseconds to tactile
or tactile + auditory stimuli on
the face and neck under compatible, neutral or incompatible
viewing conditions over 2 days
of testing. The proportion of errors are given as misses (miss)
and false alarms (FA)

Stimulation site

Means and standard deviations of response times (RTs)

and proportional error rates are reported for all experimental conditions in Table 1. RT analyses were conducted using a 2223 ANOVA, with Stimulus type (tactile
or tactile+auditory) as a between-group factor, and Day
(1 and 2), Stimulated site (face and neck) and Compati-

Face

Neck

Compatible
(face)

Neutral
(hand)

Incompatible Compatible
(neck)
(neck)

Neutral
(hand)

Incomatible
(face)

Tactile and auditory

Day 1
Mean
SD
Miss
FA
Day 2
Mean
SD
Miss
FA

432
90
0.02
0.20
433
69
0.02
0.20

444
77
0.03
0.14
445
81
0.02
0.11

453
89
0.01
0.20
437
77
0.03
0.11

441
70
0.02
0.17
420
64
0.02
0.14

441
71
0.02
0.17
408
54
0.01
0.23

460
64
0.02
0.33
429
71
0.01
0.17

Tactile only
Day 1
Mean
SD
Miss
FA
Day 2
Mean
SD
Miss
FA

481
129
0.12
0.33
461
115
0.05
0.38

482
126
0.08
0.25
469
98
0.14
0.23

489
132
0.15
0.29
466
91
0.10
0.31

402
76
0.06
0.27
433
99
0.07
0.14

421
77
0.06
0.29
413
85
0.06
0.25

460
83
0.03
0.33
446
104
0.09
0.25

163

Fig. 2 Response time (in milliseconds) to detect stimuli on the

face and neck under compatible, neutral or incompatible viewing
conditions for experiment 1

bility (compatible, neutral and incompatible) as repeated

factors. Compatible stimuli were those in which the view
was the same as the stimulated site, neutral stimuli were
those in which the view was of the hand, and the incompatible stimuli were those in which the view was of the
other potential stimulus site.
There were three types of error: The proportion of anticipatory responses or delayed responses falling outside
the temporal window was 0.05 and of failing to respond
to a target (misses), 0.02, while the proportion of responses to distractor stimuli (false alarms) was 0.17. The
proportion of misses were arcsine transformed and analysed using ANOVA with the same four factors as those
used to analyse RTs. Only significant effects are reported. The low number of distractor items in the experiment
rendered the false-alarm data too discrete to analyse.
In the analysis of errors, there was a significant main
effect of stimulus type (F1,22=8.097; P<0.01), indicating
that errors were more common in the tactile-only condition (mean 0.04) than the tactile + auditory condition
(mean 0.01). No other main effects or interactions
reached significance.
In analysis of RTs, there was a significant main effect of
stimulated site (F1,22=7.314; P<0.02), in which detection of
targets on the face was slower than on the neck. Of most
importance to our two main research questions, there was a
highly significant effect of visual compatibility (F2,44=
6.334; P<0.004), and this visual compatibility interacted
with body site (F2,44=4.350; P<0.02). As predicted, vision
of the more familiar face site had greater effects on detection of tactile and tactile+auditory targets. As can be seen in
Fig. 2, when detecting targets on the face, vision of the face
produced a trend towards a facilitatory effect (F=1.718;
n.s.), as one might expect from compatible visual and tactile
inputs. In contrast, when attempting to detect targets on the
neck, vision of the incompatible face significantly impaired
performance (F=18.764; P<0.0001). On the other hand,
viewing the neck did not facilitate target detection on the
neck or impair target detection on the face (for both, F<1).
No other main effects or interactions reached significance.
Discussion
The results of this study have confirmed and extended
our previous findings in interesting ways. First, they

have shown that vision of a body site, independent of

proprioceptive orienting, can influence tactile detection.
This conclusion is reinforced by the fact that these crossmodal interactions are produced at body sites that can
never be directly viewed. That is, they are sites that have
no history of proprioceptive orienting of eyes and head
towards them.
The data have also confirmed our second hypothesis:
the effects of viewing the face are significantly more robust than those produced by viewing the back of the
neck. In modern cultures, viewing the face via mirrors is
a relatively common experience, whereas vision of the
back of the neck is very rare indeed. This is reflected in
the significant interaction between visual compatibility
and stimulation site.
In particular, when detecting tactile targets presented
on the neck, vision of the face substantially interferes
with performance. This interference is not simply due to
the sight of a face causing general interference, because
a trend towards facilitation is observed when detecting
tactile targets on the face. That vision of the face interferes with detection of tactile targets on the neck seems
to be a particularly robust effect, as it is observed for
both pure tactile targets and tactile+auditory targets in
this study. Thus, adding the extra stimulus dimension of
sound does not alter the effects of vision of a body site.
And furthermore, in other studies we have repeatedly observed this cross-modal interference effect from vision
of the face to non-face somatosensory processing. In
those studies target discrimination, rather than detection,
was required (unpublished data).
However, before accepting these conclusions, we felt
it worthwhile to undertake a further study for a variety of
reasons. First, in this experiment and that reported by
Tipper et al. (1998), subjects viewed a particular body
site continuously for approximately 3 min. It is therefore
somewhat surprising that effects of vision were observed, when it is likely that subjects habituated to the
non-informative visual display. A second problem with
this procedure is that we had no way of controlling eye
movements over such a long viewing period.

Experiment 2
Therefore, in experiment 2, three cameras were used to
project displays of the three body sites (face, neck,
hand). At the beginning of each trial, one of these visual
displays was randomly presented and then the tactile target (to the face or neck) was presented 200 ms or 700 ms
later. This procedure had a number of advantages: First,
we predicted that the sudden onset of an unpredictable
visual stimulus, rather than continuous viewing for
3 min, would engage attention and produce more robust
cross-modal interactions. Second, we could test whether
visual inputs affected tactile processing at short stimulus-onset asynchronies (SOA). Third, the short, 200-ms
SOA between vision and touch helps reduce proprioceptive orienting of eyes and head to particular sides of

164

space. Fourth, the shorter viewing time helps reduce, to

some extent, effects of movement of body areas during
such real-time viewing.
Materials and methods
Experiment 2 employed the same basic paradigm as experiment 1,
with the following experiment-specific information and modifications.

test block, with attention being switched to the other site stimulated for the next block. Attention to body sites was alternated and
counterbalanced across subjects: ABAB (BABA). Ninety trials
were presented per block, of which 72 were target trials, and 18
distractor trials (presenting tactile stimulation to the non-attended
site). Of these trials, half were the short (200 ms) SOA between
visual onset and tactile target, while half were the long SOA
(700 ms). Presentation of these trials was randomised within each
block.
Procedure

Subjects
Subjects were 20 na, right-handed female volunteers with a mean
age of 33 years, all employees of Unilever Research, Port Sunlight. Subjects were paid 5 (GBP5) for their time.
Stimuli
Tactile stimuli were delivered to two body sites the left back of
the neck, and the left cheek below the cheekbone via two moving-magnet tactile stimulators attached to the skin with double-sided adhesive tape. These tactile stimulators were smaller and quieter
devices than those used during experiment 1, and capable of delivering a more readily controllable tactile stimulation. A third tactile
stimulator, from which a tactile stimulation was not delivered, was
attached to the top of the forearm (close to the wrist), to attain consistency in the neutral Compatibility condition. Targets were
100-Hz, 50-ms stimulations delivered via the tactors.
Unlike experiment 1, in which the relative intensity levels for
the tactors were pre-set at 5 times the detection threshold for each
site, in experiment 2 the relative intensity of the stimuli delivered
to the face and neck sites was controlled through a pre-trial, using
an improved procedure for equalising stimulus intensity at these
body sites with their differing sensory acuities. This was attained
through a novel, subjective level equalisation (SLEEQ) paradigm.
This employed a modified parameter estimation by sequential testing (PEST) paradigm, using a suprathreshold stimulus delivered to
the neck as a reference level (see Taylor et al. 1983). During this
test, subjects were asked to rate the intensity of a stimulus delivered to the face as stronger or weaker than a reference stimulus to
the neck, using a simple yes-or-no response button. Multiple
pairs of individual face and/or neck stimulations were delivered.
SLEEQ uses a tracking paradigm to home in on the stimulus level
for the face that is subjectively equivalent to the reference stimulus (delivered to the neck), attaining a rapid equalisation of levels
for the stimulation at the face and neck.
The sound produced by the tactors was masked by white noise,
played into the soundproof booth through speakers. Such sound
was significantly quieter in experiment 2 than in experiment 1,
owing to the use of more sophisticated tactors. Unlike experiment
1, which used a single camera, experiment 2 used three CCD colour cameras, trained on the three body sites, capturing real-time
images in colour. The images from these three cameras were presented not in blocks, as in experiment 1, but with switching within
each block between the three body sites. These images were presented, as in experiment 1, at subjects' midlines but, unlike experiment 1, these images were presented in colour and as mirror images.
Design
Subjects were tested individually in single sessions (lasting
4555 min), each tested under all test conditions. Each participant
was presented with six blocks of trials, including two practice
blocks (with the participant attending to the face or the neck during separate practice blocks) and four test blocks (two each with
the participant attending to the face and neck). Subjects sustained
attention to one of the stimulated body sites throughout a single

The procedure was similar to that of experiment 1, except for the

following. Rather than vision of one body site being continuous
throughout a testing block, the monitor remained blank until a trial
began. When subjects depressed the foot-pedal to begin a trial, an
image of one of the three body sites was presented between 850
and 1850 ms later. It was stressed to subjects that these images
were random and did not predict tactile targets. The tactile target
was presented either 200 ms or 700 ms after visual onset, and release of the foot-pedal was required to detect the target at the attended body site. No response was required for a stimulus at the
ignored body site. The visible image was terminated after
1500 ms, at which point the next trial could begin.

Results and discussion

Error rates and mean and standard deviations of RTs for all
conditions in the first and second half of the experiment
are shown in Table 2. Errors were too low to analyse.
RTs were analysed in a 2232 within-subjects
ANOVA. The first main effect of SOA was highly significant (F1, 19=38.83; P<0.0001), showing that RTs to
detect tactile targets were faster in the longer SOA condition. The main effect of tactile body site was non-significant (F1, 19=2.89; n.s.), so the possible confound observed in experiment 1, where neck targets were detected
faster, was no longer observed. Of most importance, the
compatibility between visual display and tactile target
was highly significant (F2, 38=8.53; P<0.001): visual-tactile compatible, 392 ms; neutral, 401 ms; incompatible,
404 ms. Finally there was no significant effect of firsthalf versus second-half of the experiment (F1, 19=2.05;
see Fig. 3).
None of the interactions were significant; however,
two of particular theoretical importance deserve further
mention. The SOA Compatibility interaction was far
from significant (F2, 38=0.46; n.s.). Therefore the effects
of vision on touch are similar at the short (200 ms) and
long (700 ms) SOAs. To confirm this, two separate
ANOVAs were undertaken on short and long SOA data.
Visual-tactile compatibility effects were highly significant at both short SOA (F2, 38=5.44; P<0.01) and long
SOA (F2, 38=4.99; P<0.02). Therefore, in answer to one
of our main questions, visual-tactile integration can take
place within 200 ms.
The second interaction between tactile body site and
visual compatibility was important because of the significant interaction observed in experiment 1. Analysis of
experiment 1 revealed that vision of the face had more
impact on detection of tactile targets than vision of the
neck, as we predicted. In the current experiment, this in-

165

the neutral hand), whereas tactile target detection at the

incompatible neck site was inhibited. Analysis of these
difference scores in experiment 2 shows a highly significant effect of viewing the face upon detection of tactile
targets on the face and neck: short SOA: F1, 19=4.7;
P<0.05; long SOA: F1, 19=22.8, P<0.0002. In sharp contrast, consider Fig. 4B. This reveals that viewing the
neck has virtually no effect on detection of tactile targets. Statistical analysis confirms this: short SOA: F1,
19=0.13; n.s.; long SOA: F1, 19=0.08, n.s.

General discussion

Fig. 3A, B Response time in experiment 2 to detect stimuli on the

face and neck under compatible, neutral or incompatible viewing conditions at short (A) and long (B) stimulus-onset asynchronies (SOA)

teraction was marginally significant (F2, 38=2.39; P=0.1).

Because of the theroetical importance of the issue of prior experience on vision-tactile interactions, further scrutiny was necessary. In Fig. 4, the difference between
viewing a stimulated body site and the neutral hand condition is plotted. For comparison, the data from experiment 1 are also shown in this format. In Fig. 4A, the effects of viewing the face are presented. In each condition, it can be seen that detection of tactile targets on the
compatible face site were facilitated (relative to viewing

Table 2 Mean and standard deviation of foot pedal response

times in milliseconds to tactile
stimuli on the face and neck
under compatible, neutral or incompatible viewing conditions
over the first and second halves
of the experiment. The proportion of errors are given as misses (miss) and false alarms (FA)

Tipper et al. (1998) have demonstrated that vision of a

body site, without proprioceptive orienting of eyes and
head to the body site, could facilitate tactile perception.
Subsequent work has confirmed these original observations (Maravita et al. 2000). The results of the current experiments have also confirmed and extended our previous
findings in interesting ways. First, they have shown that
vision of a body site, independent of proprioceptive orienting, can influence tactile detection. This conclusion is
reinforced by the fact that these cross-modal interactions
are produced at body sites that can never be directly
viewed. That is, they are sites that have no history of proprioceptive orienting of eyes and head towards them.
The data have also confirmed our second hypothesis:
The effects of viewing the face are significantly more robust than those produced by viewing the back of the neck
(see Fig. 4). In modern cultures, viewing the face via mirrors is a relatively common experience; and this experience
often involves interactions between vision and somatosensation, as when shaving or putting on make-up. In sharp
contrast, vision of the back of the neck is very rare indeed.

Stimulation site

Face

Camera

Compatible
(face)

Neutral
(hand)

Mean
SD
Miss
FA
Mean
SD
Miss
FA

406
69
0.004
0.083
398
78
0.008
0.017

420
416
72
71
0.000
0.000
0.083
0.100
419
407
76
65
0.017
0.013
0.050
0.117

423
63
0.004
0.133
419
62
0.004
0.067

420
77
0.008
0.117
429
68
0.000
0.050

Mean
SD
Miss
FA
Mean
SD
Miss
FA

366
75
0.004
0.017
364
87
0.004
0.067

382
385
71
63
0.000
0.004
0.050
0.033
377
375
83
74
0.004
0.008
0.100
0.083

387
53
0.013
0.067
369
65
0.004
0.067

387
400
63
63
0.004
0.000
0.067
0.017
373
380
59
61
0.008
0.000
0.067
0.017

SOA 1
Block 1

Block 2

SOA 2
Block 1

Block 2

Neck
Incompatible Compatible Neutral
(neck)
(neck)
(hand)

Incompatible
(face)

437
70
0.009
0.117
428
69
0.004
0.067

166

Fig. 4A, B Difference scores (viewing neutral hand minus viewing face or neck) revealing overall facilitation (positive) and inhibition (negative) scores. A represents the facilitation and inhibition effects when viewing the face for experiment 1, and the short
and long SOA conditions for experiment 2. B shows the data when
viewing the neck

Finally the data have provided an initial answer to our

third question, concerning the speed with which vision can
affect touch. In experiment 2, significant effects of vision
on tactile target detection were obtained when the interval
between visual onset and tactile target was only 200 ms.
Clearly this is only an initial exploration of the temporal
properties of visual-tactile interactions; exactly how quickly vision can affect touch remains to be discovered.
These results have implications for our understanding of
the neurophysiological mechanisms mediating sensory integration. A number of neural sites such as the superior colliculus, parietal lobes and putamen are known to be involved in sensory integration. For example, in the superior
colliculus there are sensory maps for visual, tactile and auditory inputs and these maps are spatially aligned (Stein and
Meredith 1993). Furthermore they project directly to motor
maps controlling eye and head orientation (Harris 1980).
Neural representations of different sensory modalities
must be capable of being flexibly realigned. For example, bi-modal cells in the putamen respond to tactile
stimulation of the hand and visual inputs in receptive
fields around the hand (Graziano and Gross 1996). Importantly, the visual receptive fields appear to be locked
to the somatosensory fields: when the hand is moved to a
new locus, the visual receptive field moves with it.
Hence complex interactions between somatosensation,
proprioception and vision are reflected in the response of

such cells. More recently, the flexibility of these bi-modal neural units has been demonstrated. When the animal
uses a tool to reach for a stimulus, the visual receptive
field expands around the full length of the tool (Iriki et
al. 1996). Interestingly, Farne and Ladavas (2000) have
recently demonstrated the same phenomenon in humans.
As proposed by Tipper et al. (1998), the experiments reported here suggest that this flexibility may be even more
remarkable when proprioception and vision are dissociated.
Integration can normally be achieved because the environmental loci of somatosensory and visual inputs remain constant. Thus the spider walking across the hand is spatially
localised in the visual system (e.g. retinotopic and/or headcentred frames) and in the somatosensory system via proprioceptive inputs concerning hand location. That is, the visual and somatosensory properties of the spider come from
the same place in the world. But facilitation and interference between modalities can also take place when the spatial link is broken, when visual and somatosensory inputs
from the same body site come from different locations.
Rorden et al. (1999) have similarly argued that effects
of vision on touch cannot be attributed solely to the simple
spatial proximity between visual and tactile stimuli. Rather,
tactile perception can be modulated by high-level semantic
information. Of course, the ability of humans to fluently
integrate spatially dissociated vision and action (proprioception and somatosensation) should not come as such a
surprise. Such fluent interactions between somatosensation
or proprioception and vision can easily be observed in our
daily use of mirrors and in human-machine interactions.
Take, for example, the use of a computer mouse: the hand
moving across a horizontal surface and vision of the cursor
moving across the vertically oriented surface of the computer screen are in completely different locations.

167

Our current results have implications for other research

areas. For example, in both monkey and human there are
bi-modal somatosensory and visual cells encoding information around the face (see Gross and Graziano 1995 for
review of single-cell recording in the monkey, and Ladavas et al. 1998 for evidence in humans). Importantly, for
such bi-modal cells to respond to visual input, the visual
stimulus has to be close to the somatosensory receptive
field. Visual stimuli presented a few feet away have no effect. However, based on our current findings, we predict
that if the visual stimuli were presented on an image of the
face via a mirror or video system that was positioned
some distance from the subject, cross-modal effects would
still be obtained. Of course, studies with human subjects
such as patients with neglect (Ladavas et al. 1998; di
Pellegrino and Frassinetti 2000) would be necessary, as it
is still unclear whether monkeys possess recognition of
the self via mirrors (Tomasello and Call 1997).
The integration of vision and somatosensation seems
to be automatic in this present study. There is no strategic advantage in actively attempting to integrate vision
and touch, because the visual information does not predict the location of tactile targets. Furthermore, the relatively fast effects of vision on touch (200 ms SOA) also
support the notion of automatic orienting. It will be of
value to further manipulate the SOA between visual and
tactile onsets to identify the temporal limits of these
cross-modal interactions.
Further work will also be required to more formally
identify the mechanisms mediating these intersensory interactions. Certainly the influence of prior learning
should be investigated further: this seems to be critical, as
viewing of the face causes the main effect. Another critical issue that remains unresolved concerns the neural systems mediating these cross-modal interactions. Substantial information concerning brain structures involved in
cross-modal interactions exists from tasks where somatosensory and visual information arises from the same location in space. This information arises from single-unit recording studies with monkeys (Gross and Graziano 1995)
and imaging studies with humans (Macaluso et al. 2000).
Whether the multi-modal areas of intraparietal sulcus and
superior occipital gyrus identified by Macaluso et al.
(2000) are also activated in tasks where vision and touch
are spatially separated, or whether completely different
neural systems are involved in such multi-modal interactions, seem to be important questions.
Conclusion
This study confirmed that vision of a body site, without
proprioceptive orienting of eye and/or head, influences somatosensory processes in a relatively fast and automatic
way. Such cross-modal effects are observed at body sites
such as the face and back of the neck, which cannot ever
be directly viewed. Such a finding rules out any subtle eye
and/or head movements that may have explained previous
effects of vision on the detection of tactile targets on the

hand. The results also suggest that prior experience of

viewing a body site may facilitate such cross-modal interactions. Thus viewing the face, which is familiar to subjects via self-viewing in mirrors, produces a facilitation effect when detecting tactile targets on the face, and a robust
cross-modal interference effect on detection of tactile targets on the neck (see Fig. 4A). In sharp contrast, vision of
the rarely seen neck seems to have little impact (facilitatory or inhibitory) on somatosensory processes (see Fig. 4A).

References
Driver J, Grossenbacher PG (1996) Multimodal spatial constraints
on tactile selective attention. In: Inui T, McClelland JL (eds) Information integration in perception and communication. (Attention and performance XVI) MIT, Cambridge MA, pp 209235
Driver J, Spence C (1998) Attention and the crossmodal construction of space. Trends Cogn Sci 2:254262
Farne A, Ladavas E (2000) Dynamic size-change of hand peripersonal space following tool use. Neuroreport 11:16451649
Graziano MS, Gross GG (1996) Multiple pathways for processing
visual space. In: Inui T, McClelland JL (eds) Information integration in perception and communication. (Attention and performance XVI:) MIT Press, Cambridge MA, pp 181207
Groh JM, Sparks DL (1996) Saccades to somatosensory targets. 1.
Behavioural characteristics. J Neurophysiol 75:412427
Gross GG, Graziano MS (1995) Multiple representations of space
in the brain. Neuroscientist 1:4350
Harris LR (1980) Superior colliculus and movement of the head
and eyes in cats. J Physiol (Lond) 300:376391
Honore J, Bourdeaud'hui M, Sparrow L (1988) Reduction of cutaneous reaction time by directing eyes towards the source of
stimulation. Neuropsychologia 27:367371
Iriki A, Tanaka M, Iwamura Y (1996) Coding of modified body
schema during tool use by macaque postcentral neurones.
Neuroreport 7:23252330
Ladavas E, Zeloni G, Farne A (1998) Visual peripersonal space on
the face in humans. Brain 121:23172326
Ladavas E, Farne A, Zeloni G, Pellegrino G di (2000) Seeing or
not seeing where your hands are. Exp Brain Res 31:458467
Larmande P, Cambier J (1981) Effect of the state of activation of the
cerebral hemisphere on sensory extinction: a study in 10 patients with right-hemisphere lesions. J Rev Neurol 137:285290
Macaluso E, Frith C, Driver J (2000) Selective spatial attention in
vision and touch: unimodal and multimodal mechanisms revealed by PET. J Neurophysiol 83:30623075
Maravita A, Spence C, Clarke K, Husain M, Driver J (2000) Vision and touch through the looking glass in a case of crossmodal extinction. Neuroreport 11:35213526
Pellegrino G di, Frassinetti F (2000) Direct evidence from parietal
extinction of enhancement of visual attention near a visible
hand. Curr Biol 10:14751477
Pierson JM, Bradshaw JL, Meyer TF, Howard, MJ, Bradshaw, JA
(1991) Direction of gaze during vibrotactile choice reactiontime tasks. Neuropsychologia 29:925928
Rorden C, Heutink J, Greenfield E, Robertson IH (1999) When
the rubber hand feels what the real hand cannot. Neuroreport 10: 135138
Stein BE, Meredith MA (1993) The merging of the senses. MIT
Press, Cambridge, MA
Taylor MM, Forbes SM, Creelman CD (1983) PEST reduces bias in
forced choice psychophysics. J Acoust Soc Am 74:13671374
Tipper SP, Lloyd D, Shorland B, Dancer C, Howard LA, McGlone
F (1998) Vision influences tactile perception without proprioceptive orienting. Neuroreport 9:17411744
Tomasello M, Call J (1997) Primate cognition. Oxford University
Press, London