Training Position
Resume:
Exp Brain Res (****) ***:*** ***
RESEARCH ARTICLE
Guidance of visual direction by topographical vibrotactile
cues on the torso
Francois Asseman Adolfo M. Bronstein
Michael A. Gresty
Received: 29 March 2007 / Accepted: 21 November 2007 / Published online: 11 December 2007
Springer-Verlag 2007
Abstract Vibration on localised areas of skin can be used loci but variability and saccade latencies remained high.
to signal spatial orientation, multi-directional motion and The uncertainty in the relationship between vibration locus
also to guide arm and hand movements. This study inves- and gaze direction and the prolonged latencies of responses
tigated the possibility that vibration at loci on the skin indicate circuitous neuronal processing. There appears to
might also be used to cue gaze direction. Eight subjects be no pre-existing stimulus-response compatibility map-
made eye or (head + eye) gaze saccades in the dark cued ping between loci on the skin and gaze direction.
by vibration stimulation at discrete loci spaced on a hori- Vibrotactile cues on the skin of the trunk only serve a gross
zontal contour across the chest. Saccade and gaze indication of visual direction in space.
amplitudes, latencies, and directions were analysed. In the
Keywords Gaze Ocular Spatial orientation
rst experiment, performed without training, subjects could
Somatosensory Vibrotactile-cueing Prostheses
only use vibration cues to direct their gaze in cardinal
directions and gross quadrature. There was a high vari-
ability in the relationship between locus on the trunk and
gaze direction in space, both within and between subjects. Introduction
Saccade latencies ranged from 377 to 433 ms and were
related to the loci of vibration; the further from the body Orientation in space through the use of tactile cues has
midline the quicker the response. Since the association of been the subject of much applied research over the last
skin loci with gaze direction did not appear intuitive a sub- 15 years (Kaczmarek et al. 1991). Perhaps the most
group of four subjects were retested after intensive training developed application has been the use of vibration cues on
with feedback until they attained criterion on midline : 0 the torso to signal to pilots the orientation of their aircraft
and 15 cm (to right/left of midline) : 45 gaze shifts right (Rupert 2000a, b) and in aiding helicopter pilots to inter-
and left. Training gave a moderate improvement in direc- pret the motion of their craft in fore-aft, lateral and vertical
tional speci city of gaze to a particular locus on the skin. directions. Sensationally, the world water speed record has
Gaze direction was linearly rescaled with respect to skin recently been taken by a congenitally blind pilot whose
left right steering of the power craft was cued by a vi-
brotactile feedback of the boat s position in the channel
F. Asseman A. M. Bronstein M. A. Gresty (Dobbins et al. 2001). At a more mundane level, Tan et al.
Division of Neuroscience and Mental, Department of Clinical (1997) successfully guided car drivers by using vibration of
Neuroscience, Imperial College London, Charing Cross Campus,
the back-seat and with a similar in vehicle device, Van
Fulham Palace Road, London W6 8RF, UK
Erp and Verschoor (2004) showed that directional vibration
e-mail: *.******@********.**.**
cues could decrease driver s workload. Amongst wearable
F. Asseman devices an active belt has been used to provide directional
e-mail: ****************@*****.**
information in large environment by combining a GPS
receptor to vibrotactors mounted in a belt (Tsukada and
A. M. Bronstein
Yasumura 2004; Lindeman et al. 2005).
e-mail: *.*********@********.**.**
123
284 Exp Brain Res (2008) 186:283 292
learn the required association in a single intensive session in
Since it has been shown that vibrations on the torso can
which they were trained to associate speci cally indicated
indicate which way to go, our study was a natural
gaze directions with points on the torso. Failure to improve
extension to the question of whether or not vibrations on
speci city and accuracy after such training would be a evi-
the torso can indicate which way to look . An analogy is
dence against the notion that directing gaze by vibration
how points on a hedgehog s skin might be coded as
could be developed into an elegant and reliable way to cue
directions in space according to the orientation of the
gaze direction for machine or prosthetic control.
spines that originate at those points. This would require the
ability to interpret a single skin locus as being the origin of
a vector pointing out into external space and to align
attention and eye movements with such vectors (Corbetta Methods and protocol
et al. 1998). A stimulus-response mapping of cues pre-
Two groups of eight male subjects gave their informed
sented at loci on the skin onto directions of gaze could nd
consent to take part in the experiments. The Riverside
multiple applications ranging from assisting patients with
Research Ethics Committee approved the protocol and tests
visual defects to negotiate obstacles to machine control.
were performed according to the 1964 Declaration of Hel-
The investigation of tactile-gaze mapping would also
sinki. The rst group was tested with the head xed
extend our basic understanding of the relationships
(25 2 years; 90 7 cm chest circumference) and the
between sensory and motor frames of reference.
second group with the head free (26 3 years; 88 7 cm
The interpretation of a point on the skin as the origin of
chest circumference). Static two-point discrimination testing
a directional vector radiating out from that point is neither
of the skin of the different loci of vibration on the torso was
necessarily intuitive nor directionally speci c. A touch
undertaken using a discriminator or Weber s compass to
stimulus to the skin could come from many directions
ensure that all subjects had normal sensation. The subject s
including tangential. We made the assumption that the
task was to determine whether he was touched by one or two
most likely interpretation is that the direction of action of a
prongs: the smaller the distance between the prongs that the
touch is approximately normal to the plane of the skin
subject could detect the more sensitive the sense of touch. On
because of the quotidian experience that when touched, say
average, the subjects were able to differentiate two prongs
on the shoulder, to gain attention one turns to look out-
distanced by 4 1 cm in the frontal part of the trunk.
wards from the point of touch. As a corollary, a common
practical joke is to reach across and touch someone from
the opposite side whereupon they turn to the touched side
Equipment
to nd no-one there! Hence the most likely interpretation of
a touch as cuing direction suggests good stimulus response
compatibility. Compatibility implies that speci c re ex Movements of the eyes were recorded with infrared
re ection (IRIS Skalar1) which recorded the left eye
associations between skin loci and gaze direction might
pre-exist or at least, be learned readily. Evidence for vertically and right eye horizontally with freely moving
compatible stimulus-response mapping between skin loci head. The system was used because of its dynamic mea-
suring range of 30 with a low noise level of 1 min of arc
and directions in space is given by Van Erp (2005) who
tested a device with numerous vibrotactors placed in the (Reulen et al. 1988). The linearity of the system in this
horizontal plane around the trunk. Subjects indicated the range of eyes movements has been ascertained. The eye
direction of the vibration with a manual cursor mounted on movements were calibrated to ve LEDs in the horizontal
and vertical planes (Central; 30 left and right; and 20 up
an arc. Arm directions made to the vibrotactile cue were
quite accurate ( 4 ) for vibration near the midline but with and down) with the head xed on a chin rest. The LEDs
higher variability for more lateral tactors. We also chose were mounted on a cross bar positioned against a blank
the skin of the upper torso to test for vibration loci because wall at 1 m distance.
of the many precedents (described above) and because it
forms an extended surface closest to the level of the eyes.
As gaze shifts frequently involve both head and eye Assumptions of viewing distance
movements (Gresty 1974), both eye movements in isola-
tion and gaze without head restriction were studied. How the eyes might be guided by an imaginary vector
In contrast to van Erp s results our rst experiment found extending from the body into external space will depend on
only gross relationships between skin loci and gaze direction how the vector is interpreted as de ning the location and
suggesting that there was no intuitive, automatic association distance of an external target. For a target at in nity the
between skin geometry and spatial direction. Accordingly, a eyes would be ideally aligned parallel to the vector whereas
second experiment was undertaken to see if subjects could for nearer targets the eye direction would not be parallel
123
Exp Brain Res (2008) 186:283 292 285
but slightly displaced to the left or right of the vector and selected to be greater than the acuity across the torso which
right and left eyes would move by different amounts. We is a uniform 2 3 cm (Van Erp 2005) to avoid ambiguity.
attempted some control for this possible variability by All subjects were all able to discriminate between different
placing the eye calibration bar at the approximate distance tactors. In the experiments the subjects were seated in the
of dark vergence for young adults (1 m, dark focus 1.08 dark and with sound isolating earphones; they were
dioptres; Jaschinski et al. 1998) and stressing that eye instructed Imagine the vibration comes from an external
movements be made as if tactor stimuli referred to the point in space, similar to someone touching you to get your
directions of targets placed at this distance. Compliance attention so that you look at where the touch originated.
could not be monitored since eye movement recordings Similarly when you feel vibration in a certain place look at
were monocular and there was no measure of accommo- where you think the target is coming from . Subjects
dation, however, the marked inconsistency and variability were also told to feel free to move their head if they need
of the results rendered considerations of notional viewing to. The target is their interpretation of the locus of
distance to be of little relevance. vibration as a vector radiating outwards from that point on
Angular displacement of the head in yaw and pitch were the skin.
recorded with an electromagnetic device (Fastrack Polhe- Three experiments were conducted: the rst two studied
mus ) with an update rate of 120 Hz. The tactors were the intuitive use of the skin loci as indicating direction
DC pancake vibrating motors, as used in mobile phones whereas the third trained subjects on interpreting certain
(speed 10,000 12,000 rpm, vibration frequency &200 Hz) loci as the origin of a vector and assessed how this might
appropriate for stimulation of the rapid adapting Pacinian generalise to other tactors.
corpuscles, for which the most sensitive frequency range is
1. Eye direction guided by torso-tactile stimulation. The
100 500 Hz. Vibration stimuli were 100 ms duration.
head was xed by a chin rest at a comfortable level.
2. Gaze direction guided by torso-tactile stimulation
(Head free to move).
Overall design (Fig. 1a)
3. Gaze direction guided by torso-tactile stimulation after
a training session (head free to move).
Seven tactors were placed horizontally on the torso of the
During this third experiment, four subjects from the second
subject at the sternum level. Tactors were equally distanced
group have been asked to come back again. They received
(5 cm) for all the subjects independently of their somato-
a training consisting of training the 10 cm (positive for
type. Consistent with the eye movement directional
the right and negative for the left) tactors to a gaze
convention, negative distances correspond to the tactors
direction of 45 by presenting the visual stimuli at the
placed on the left of the navel. The 5 cm separation was
Fig. 1 Experimental setup (a). VIB on for VIB on for
A B 100ms 100ms
Timing sequence of the
different stimuli (b). Sample of
Fastrack emitter
a horizontal gaze direction LED LED
response to a vibration stimulus on on
Tim e
Skalar Infrared system
at 15 cm to the right from the (s) 0 1 2 3.5 5.1 6.1
Earp h o n e s
navel (c). H Horizontal, L and D
saccade latency and duration (in
ms) from the saccade horizontal C
velocity signal, F nal gaze L
D
direction in degrees, a.u. Tactors
arbitrary units
7 .5 d e g
Saccade H
151 deg/s
Velocity
S ac c a d e H
Stimulus a.u.
7 . 5 de g
Head H
F
GAZ E H 7 . 5 d eg
Time
(ms) 0 1 50 0
1 0 00 2 50 0
500 20 00
123
286 Exp Brain Res (2008) 186:283 292
same time. This was repeated until each subject could Assuming:
execute 45 gaze movements cued by the tactor alone in
DirecExp = 360 (distance of the tactor from midline)/
darkness.
Trunk circumference.
In each experiment the subject received a sequence of
70 vibration signals at approximately 5 s intervals. A
For example, the DirecExp equals 360 9 5/90 = 20 for
central led visual stimulus was illuminated during
the tactor placed at 5 cm on subject with a trunk
1,000 ms between each stimulus so that the subject could
circumference of 90 cm.
orientate back to centre and ensured that the subject was
Direction error, DirecError, was computed as the
looking straight ahead before all vibrations. Then when
difference between the gaze directions expected from
turned off and after 1,300 ms, the rst vibrator is switched
radiations of a protractor and the gaze directions measured.
on for 100 ms. Figure 1b shows the timing details of one
If the DirecError is small then the experimentally deter-
sample sequence.
mined gaze directions are aligned with protractor radiations.
Measurements (Fig. 1c)
Statistics
Horizontal gaze direction was computed by adding hori-
An ANOVA for repeated measures was used to test the
zontal eye movements and yaw angular rotation of the
effect of tactor position (-15, -10, -5, 0, 5, 10, 15 cm) on
head. Vertical gaze direction was similarly computed with
eye and gaze latencies and durations. A two-factor
vertical eye movement and pitch signals.
ANOVA for repeated measures was used to study the effect
Saccade latency was measured as the time taken from
of training (before and after), tactor position (-15, -10,
the vibration onset to the beginning of the saccade as
-5, 0, 5, 10, 15 cm) and their interaction (train-
detected by differentiation of the eye or gaze movement
ing 9 position) on gaze directions, latencies and durations.
signal. Saccade or gaze transfer duration was measured,
Since ANOVA tests difference between means gaze
also from the differentiated signals, as the time taken to
direction, values were normalised to +ve to avoid means
complete the reorientation commencing from the beginning
around zero. The William s T2 statistic that tests for dif-
of the saccade until the gaze/eye reached the target and
ferences between two dependent correlations sharing a
remained stable.
common variable (Recommended by Steiger 1980) was
In a further analysis of gaze direction, the data were
also used to compare the linear regression between tactor
normalised with respect to trunk circumference to test the
position and gaze direction before and after training.
hypothesis that there may be a common origin within the
Because we sought a decisive effect that would improve
torso for eye saccades cued by vibration. We computed the
performance a level of signi cance was chosen at P = 0.05
expected gaze direction or DirecExp as if the circum-
to test for a moderate effect size in just eight subjects
ferential arc of the trunk to which the tactors were xed
(Table 1).
were the curved edge of a protractor :
Table 1 Means and standard
Tactors position
deviations in degrees from
primary gaze for eye and gaze Left Navel midline Right
movements made to vibration
-15 cm -10 cm -5 cm 0 cm 5 cm 10 cm 15 cm
cues from tactors positioned
across the chest from -15 cm
Horizontal eye movement
(extreme left) to 15 cm
Mean -29.76 -22.21 -12.13 0.96 11.56 19.83 26.16
(extreme right)
SD 9.5 11.3 6.1 1.8 6.5 8.1 8.3
Vertical eye movement
Mean 6.84 5.43 3.25 -0.23 1.05 2.33 3.01
SD 5.1 3.1 1.1 1.3 2.9 2.9 3.2
Horizontal gaze
Mean -32.54 -27.76 -16.68 -0.003 17.37 25.81 24.62
SD 13.7 12.3 8.8 1.7 9.0 15.4 11.9
Vertical gaze
Mean -5.24 -6.8 -6.98 -3.51 -3.86 -2.91 -4.67
SD 6.7 6.7 5.9 5.7 4.7 4.0 4.6
123
Exp Brain Res (2008) 186:283 292 287
Relationship between saccade direction and tactor location
Results
Subjects accurately directed their eyes straight ahead fol-
Untrained eye and gaze movements to vibrotactile cues
lowing activation of the central tactor (Mean SD
0.96 0.9). For the two tactors placed at 5 cm to the
Subjects reported no problems with understanding the
left or to the right, eyes saccades were directed to
instructions neither did they nd the task dif cult
-12.13 1.5 and 11.56 0.8, respectively. Vertical
although performance was poor. The ability to interpret
eyes movements ranged from 1.05 2.9 to 6.84 5.1
the locus of a vibration stimulus as a direction in space
across tactors loci except for the central tactor for which
varied greatly between subjects as can be seen in Fig. 2.
the eyes were directed downwards by -0.23 1.3 .
The lines drawn for the mean results (all subjects inclu-
ded) gave no indication of a central point within the torso
from which a vector could be projected through the locus
Relationship between gaze direction and tactor location
of stimulation on the skin to form a direction in external
space. The examples shown of two subjects responses
Similarly, subjects accurately directed their gaze straight
illustrate the individual differences and the poor rela-
ahead following activation of the central tactor
tionship between a locus on the skin and a spatial
(0.00 1.2 ). For the two tactors placed at 5 cm to the
direction. Subject S1 was able to differentiate between
left or to the right, gaze was oriented to -16.68 3.6
different vibrators and generated gaze saccades propor-
and 17.37 5.7, respectively. Vertical eyes movement
tional to the laterality of the vibrotactile stimuli.
ranged downwards from -2.91 4 to -6.8 6.7
However, subject S2 looked in the same direction what-
across tactor loci.
ever the stimulus locus.
Fig. 2 View from overhead of
gaze responses to vibration at
loci across the chest showing
the means of all subjects (top)
and gaze responses of two
subjects (S1 and S2). Lines of
sights are drawn by linking the
virtual visual targets
presented on the bold line to its
corresponding vibrotactile
simulation on the subject s torso
123
288 Exp Brain Res (2008) 186:283 292
60 A
50
600
EYE GAZE
40
30 500
DirecErr (deg)
20
Latency (ms)
400
10
300
0
-15 -15 -5 0 5 10 15
200
-10
-20
100
-30
0
-40 -15 -10 -5 0 5 10 15
Tactor position from midline (cm) Tactor position (cm)
Fig. 3 Mean Gaze Errors (DirecErr) and standard deviations for each
B
tactor. The negative sign on the X-axis correspond to tactors on the 600
left side of the subject, positive to the right and 0 for the tactor on the
midline 500
Duration (ms)
The average expected gaze directions, DirecExp, were 400
respectively [from the extreme left tactor (-15 cm) to the 300
extreme right tactor (+15 cm)]: -61.7, -41.1, -20.6,
0, 20.6, 41.1, 61.7 . 200
Gaze DirecErrors were minimal for the central (Fig. 3b; 100
0.00 ) and for the closest tactors at 5 cm (-3.87 for the left
0
and 3.2 for the right). The DirecErrors abruptly increased -15 -10 -5 0 5 10 15
for the tactors placed at 10 cm from the midline (-13.25 Tactor position (cm)
for the left and 15.34 for the right) and more for those at
Fig. 4 Mean eye saccade ( lled diamonds) and gaze (empty squares)
15 cm (-28.75 for the left and 37.1 for the right).
( SD) Latencies (a) and Durations (b) for each vibrotactors position
We conclude from these results that anthropometric ranging from -15 cm (furthest left) to +15 cm (furthest right)
normalisation does not improve directional relationships
not on durations (P [ 0.05). The latencies ranged from 357
between gaze direction and tactor location. It appears that
to 433 ms. These were dependent on eccentricity: shorter
subjects do not, or cannot envisage the arc of skin over
latencies are seen for the higher eccentricities; 357 ms for
which the tactors are located as being the edge of a pro-
the vibrator placed -15 cm lateral and 361 ms for +15 cm
tractor with its origin at some point within the trunk.
lateral.
Amount of head movement deployed
Latency of gaze with respect to tactor location (Fig. 4a)
The portion of head movement in gaze displacement varied
Tactor position and thus gaze eccentricity had a signi cant
across subjects and two subjects did not move their head at
effect on gaze latencies (F = 3.01, P \ 0.05) but not on
all. For the remaining six subjects, the average proportion of
durations (P [ 0.05). Gaze latencies were in the same
head direction in space to the nal gaze direction was,
range as saccades, 380 454 ms. There was also a decrease
respectively [from the extreme left tactor (-15 cm) to the
in the latency with the target eccentricity: the slowest gaze
extreme right tactor (+15 cm) and excluding the central
shifts were for the vibrators closest to the midline; 405 ms
tactor]: 45 19, 48 23, 48 41, 41 24, 44 19 and
for the -5 cm and 454 ms for the +5 cm.
55 27%. Overall, for the six subjects who used their head,
the head movements comprised to almost half the total gaze
transference irrespective of the individual tactor stimulus.
Overall duration of the saccadic reorientation with respect
to tactor location (Fig. 4b)
Latency of saccades with respect to tactor location
(Fig. 4a) With the head xed, the subjects needed more time to get
to the most extreme gaze directions: from 179 ms for
Tactor position and thus saccade eccentricity, had a sig- -5 cm to 299 ms for -15 cm, and from 128 ms for +5 cm
ni cant effect on saccade latencies (F = 2.8, P \ 0.05) but to 195 ms for +10 cm.
123
Exp Brain Res (2008) 186:283 292 289
Overall duration of gaze reorientation with respect to A Gaze Directions amplitude (degrees)
tactor location (Fig. 4b)
Before After
70
When the head was free gaze durations remained in a
similar range to those described above and similar to eyes 50
only: a shorter duration for the closest target (179 and
30
142 ms for -5 and +5 cm, respectively) and longer time to
10
get to the target for the extreme stimuli (268 and 199 ms
for -15 and +15 cm, respectively). -15 -10 -5 -10 0 5 10 15
-30
-50
Trained gaze movements to vibrotactile cues
-70
Relationship between gaze direction and tactor location
(Fig. 5a) B Gaze Latencies (ms)
700
The subjects improved their performance through the 600
training by directing their gaze to -42.9 9.3 for the 500
-10 cm tactor and to 47.8 4.1 for the 10 cm tactor.
400
The analysis revealed a main effect of training
300
(F = 711.3, P \ 0.05), a main effect of tactor position
(F = 65.3, P \ 0.05) and an interaction effect between 200
training and tactor position (F = 4.7, P \ 0.05). 100
After the training all gaze directions (except the central
0
one) were readjusted to the right for the right-hand side -15 -10 -5 0 5 10 15
tactors and to the left for the left-hand ones. This was more
pronounced for the furthest tactors at -15 and 15 cm C Gaze Durations (ms)
shifting respectively from -40.7 8.6 before to 700
-56.2 9.4 after training and from 29.8 11 before 600
to 58 4.8 after training. The coef cient of correlations 500
between tactor position and gaze direction before training
400
(0.97) and after training (0.99) were not statistically dif-
300
ferent (t = -0.87, df = 3). Training both rescaled and more
200
importantly linearised the relationship between gaze
direction and tactor eccentricity. However, when responses 100
to the central tactor were excluded standard deviations 0
remain in the same range from 4.6 15.6 before to 4.1 -15 -10 -5 0 5 10 15
Tactor Position (cm)
12.4 after training.
Fig. 5 Mean ( SD) gaze directions (a), Latencies (b), and Durations
(c) for each vibrotactors position ranging from -15 cm (furthest left)
Latency of gaze with respect to tactor location (Fig. 5b) to +15 cm (furthest right). Filled diamonds means before training,
Empty squares means after training
There was no main effect of training (F = 2.1, P = 0.24),
P = 0.24) and no interaction effect between training and
no main effect of tactor position (F = 2.2, P = 0.11) and no
tactor position (F = 1.1, P = 0.39).
interaction effect between training and tactor position
(F = 1.02, P = 0.44).
Discussion
Overall duration of gaze reorientation with respect
We have shown that the directions of eye or gaze move-
to tactor location (Fig. 5c)
ments cued by stimulating loci on the skin of the torso are
inaccurate and with high variability. Variability of the eye
There was a main effect of tactor position (F = 7.5,
P \0.05) but no main effect of training (F = 0.635, saccade direction increased with tactor eccentricity from
123
290 Exp Brain Res (2008) 186:283 292
eye-centred coordinates (Groh and Sparks 1996a). Vibra-
the midline and showed wide inter and intra individual
tion on the ngers apparently has a greater somatotopic
differences (Fig. 2) suggesting that there is no automatic
resolution than from the trunk (Cholewiak et al. 2004).
mapping between skin loci and spatial direction. Vibration
The increasing inaccuracy of gaze direction with target
cued saccades were also slow with a long latencies
([350 ms, Fig. 4a) and durations ([125 ms, Fig. 4b) in eccentricity shown in our results could be explained partly
by a rise in the perceptual threshold to vibration stimuli
comparison with visually guided saccades and also with an
with eccentricity of loci on the trunk. Cholewiak et al.
inverse latency amplitude relation (Fig. 4a). Following a
(2004) showed a variation of sensitivity to a vibration of
training session, the relation between vibration loci and
250 Hz over the different sites tested around the abdomen,
gaze direction became more linearly related variability was
e.g. a lower sensitivity on sites further from the navel.
not reduced. The combination of poor tactile-gaze map-
Nevertheless, it seems that the perceptual threshold
ping, inconsistency of scaling and long latencies indicate
decreases with upper trunk sites, at the level of sternum for
the lack of any pre-existing aptitude in associating the
example (Yewlett et al. 2003). However, this latter study
position of the tactor with a direction in space to which
did not test sites at the level of the sternum all around the
gaze might be aligned.
trunk so the receptors around this area remain unexamined.
Saccadic eye movements can be cued by visual, auditory
Cholewiak et al. (2004) also found an accuracy of locali-
(Zambarbieri 2002) and somatosensory stimuli (Amlot and
zation around the abdomen of 92% by using 8 tactors and
Walker 2006; Blanke and Grusser 2001; Grusser 1982;
also found anchor points (navel and spine) which were
Groh and Sparks 1996a; Neggers and Bekkering 1999) and
consistently identi ed indicating the potential spatial res-
may be re exive (guided without attention) or voluntary
olution that might be attained with stimulation of the skin
(Amlot and Walker 2006). It has also been found that the
of the torso. Overall the evidence on perceptual thresholds
characteristics of saccades guided by different sensory
modalities can differ in velocity, accuracy and even tra- indicates that if subjects could be trained to map, with
jectory (Groh and Sparks 1996a) so that the main sequence consistency, gaze direction onto skin loci then accuracy
established for visual saccades does not universally apply. could be high and particularly for more medial loci.
These characteristics have implications for the mapping of
spatially patterned sensory stimuli to spatial and speci -
cally, gaze, directions. Latency
The properties of saccades to skin loci were different from
Accuracy those of visual evoked saccades. Visually triggered sac-
cades have an average latency of circa 200 ms to a 10
saccade and 240 ms for a 40 saccade with a linear relation
J Van Erp (2005) studied how a localized vibration could
be interpreted as a direction by hand positioning of a cursor between saccade latency and amplitude (Fuller 1996). The
in a pointing direction task. He showed that the duration of visually evoked saccades do not last longer than
observers do not use the body midline as the origin for the 100 ms (Leigh and Zee 2006). In contrast, the latencies of
direction vector but used two internal references, one for saccades to vibration stimuli were almost twice as long as
each body mid-axis. Our study gives no evidence for two for visual stimuli and thus similar to saccades evoked by
points of origin (Fig. 2). The variability of saccade and vibrations on the ngers and auditory stimuli (Sullivan
gaze directions, as high as 15, precludes being able to et al. 2004; Amlot et al. 2003; Groh and Sparks 1996a;
identify any single or dual points of origin within the torso. Zambarbieri 2002). There are other similarities between
This result was impervious to training, showing that vibration triggered saccades and those cued by auditory
although directions could be rescaled, latencies remained stimuli; they are slow, in the same range of our results, for
long, which would suggest that there are no convenient pre- small target eccentricities saccade latency is much greater
existing maps relating skin loci to directions in external than the latency for visual targets at the same eccentricity
space. Apart from localisation around the midline, the (Zambarbieri 2002) and there is also a decrease in latency
association of a direction in space with localised contact on with saccade eccentricity (Groh and Sparks 1996a). These
the skin of the torso does not seem to be intuitive and latter characteristics are particularly relevant to our study
requires higher order processing to construct such associ- but the authors used a paradigm with only three (centre, left
ations de novo. and right) stimuli and did not investigate a more general
The inaccuracy of somatosensory saccades has been topographical distribution.
attributed to processes upstream from the colliculus, Two main hypotheses have been raised to explain the
including imperfect somatotopic localisation and problems latencies of saccades to auditory or somatosensory stimuli.
with the translation of target position signals from body- to The rst is linked to the process of localisation of the
123
Exp Brain Res (2008) 186:283 292 291
the reason for the large inter-subject variability of head
target. Auditory targets in the midline produce small dif-
movement tendencies remains unknown (Stahl 1999) and
ferences of intensity and timing at the ears which increases
our study gives no new insights on the topic.
uncertainty and prolongs latency. For somatosensory tar-
gets at the cortical level, it has been shown that the
cutaneous regions adjacent to the trunk midline are repre-
sented bilaterally in the primary somatosensory cortex
Conclusions
(Fabri et al. 2005) which means that there could also be
potential uncertainty in the localisation of stimuli close to
Our results are disappointing and unforeseen since the skin
the midline. This could cause longer latencies, as we found.
of the torso has good tactile spatial resolution and the
The second hypothesis relating to latency concerns
normal to a point on its surface projects fairly unambigu-
delays around the peripheral central processing-motor
ously out into space. In comparison, many applications of
output loop. Peripheral sensory conduction for vibration is
vibration cues have been more successful. For example,
fast at circa 30 m/s however, at the superior colliculus
using similar vibration cues to detect lateral and vertical
where target position information is elaborated from the
translations and tilt of an aircraft can be learned within an
somatosensory reference frame to the saccade eye-centred
hour. It would seem that apart from gross cardinal direction
coordinate, the latency of somatosensory responses is
and quadrature there is no inbuilt relationship between the
39 ms longer than the latency of visual responses in the
extension of the surface of the skin in space and the
same neurons (Groh and Sparks 1996c; Sparks 1989). The
mechanisms controlling gaze direction that can be exploi-
delay in our latencies could then explained, in part, by the
ted for tactile cueing of gaze direction.
delay in neuronal responses for sensorimotor stimuli,
In terms of possible applications this does not appear at
however, there remains a lot of time to account for (circa
present to be a promising direction for arti cial aids since
200 ms) which must be attributable to higher order
intuitive coding and good stimulus response compatibility
processing.
are desirable attributes for any control system, both of
Where does the higher-order delay in somatosensory-
which are lacking here. Perhaps most surprisingly, training
gaze transformation occur? First, the model of Posner et al.
to associate selected tactors in the array to certain direc-
(1980) proposes two ways in which attention can be ori-
tions did not generalise across the array. The relationship
ented to a potential source of perceptual input, i.e.
between loci of vibration and eye/gaze direction in space
exogenously and endogenously. The former is an automatic
appears to require higher order processing to construct
system to orient the attention whereas, for the latter,
associations de novo. This could be a protracted process
attention is under the strategic control of the subject.
with signi cant individual differences in performance.
Somatosensory cues exhibit both properties when subjects
are asked to look at a vibrated hand (Amlot and Walker Acknowledgments This study was founded by a Medical Research
2006). We nd no evidence for an automatic stimulus- Council component grant (G0300405).
response mechanism for directing gaze by tactile cues.
Using Posner s terminology, associating saccade directions
References
with vibration loci is an endogenous process.
There is neurophysiological evidence for where some
Amlot R, Walker R (2006) Are somatosensory saccades voluntary or
rescaling gaze direction through training might take place.
re exive? Exp Brain Res 168:557 565
A study on primates has shown that auditory-saccades Amlot R, Walker R, Driver J, Spence C (2003) Multimodal visual-
training generated a de novo responsiveness to auditory somatosensory integration in saccade generation. Neuropsycho-
stimuli in lateral parietal area neurons (Grunewald et al. logia 41:1 15
Andersen RA, Essick GK, Siegel RM (1985) Encoding of spatial
1999). This area is known to be involved in coordinate
location by posterior parietal neurons. Science 230:456 458
transformation (Andersen et al. 1985) playing a direct role Blanke O, Grusser OJ (2001) Saccades guided by somatosensory
in the processing of saccadic eye movements and also in stimuli. Vision Res 41:2407 2412
the guidance of visual attention (Gottlieb 2007). This area Cholewiak RW, Brill JC, Schwab A (2004) Vibrotactile localization
on the abdomen: effects of place and space. Percept Psychophys
has not yet been shown to respond to somatosensory
66(6):970 987
stimulation but is clearly a candidate for brain activation Corbetta M, Akbudak E, Conturo TE, Snyder AZ, Ollinger JM, Drury
studies on training saccadic gaze direction to somatosen- HA, Linenweber MR, Petersen SE, Raichle ME, Van Essen DC,
sory cues. Shulman GL (1998) A common network of functional
Contact this candidate