Content uploaded by Bettina Forster
Author content
All content in this area was uploaded by Bettina Forster
Content may be subject to copyright.
Copyright 2005 Psychonomic Society, Inc. 458
Perception & Psychophysics
2005, 67 (3), 458-468
Our conscious perception of sensory stimulation can be
strongly biased by the arrival of subsequent events. For ex-
ample, when a series of tactile stimuli are presented in
rapid succession to different locations, perceived locations
are shifted toward subsequent stimuli. This robust illusion,
termed cutaneous saltation, can be produced when pre-
senting five taps close to the wrist, followed by five taps
at the center of the forearm, and another five taps close to
the elbow (Geldard & Sherrick, 1972). Instead of these
taps being perceived at their actual locations, they are felt
to be uniformly distributed, as if a rabbit were hopping
along the arm. This cutaneous rabbit illusion is also pres-
ent when two successive taps are delivered at one location,
followed by a third at another location. The second tap is
mislocalized, and the size of its illusory displacement to-
ward the third tap depends on the interstimulus interval
(ISI) separating these stimuli (Cholewiak & Collins, 2000;
Geldard, 1975, 1982; Kilgard & Merzenich, 1995).
The cutaneous rabbit illusion has provoked heated
philosophical debates about the timing of subjective expe-
rience, since it seems to suggest that conscious percep-
tion is influenced by stimuli that have not yet occurred
(cf. Dennett, 1991). At the same time, experimental psy-
chologists and neuroscientists have tried to identify the
mechanisms responsible for sensory saltation. When re-
searchers mapped the area on the body surface where
saltation can be elicited, the cutaneous rabbit was found
to be constrained by the anatomical organization of the
somatosensory system (Geldard, 1982; Geldard & Sher-
rick, 1983). Saltation on the limbs was more pronounced
in the longitudinal direction (e.g., from elbow to wrist)
than laterally and was eliminated altogether when succes-
sive stimuli were delivered to nonadjacent dermatomes,
or to the left and right of the body midline.
These findings have led to the proposal that the saltation
illusion is produced by strictly local (dermatome-based)
interactions between metastable tactile representations at
early stages of perceptual processing in the primary so-
matosensory cortex (Cholewiak, 1999; Geldard, 1975,
1982). However, Kilgard and Merzenich (1995) have
challenged the view that sensory saltation is a perceptual
illusion. These authors asked participants to attend to
proximal or distal regions of the forearm and found that
perceived loci of tactile events were systematically shifted
toward attended locations. On the basis of these obser-
vations, Kilgard and Merzenich argued that saltation re-
flects a postperceptual attentional bias, resulting from
participants attending to predictable locations of antici-
pated tactile events.
In previous investigations of sensory saltation, partici-
pants were typically instructed to point to the perceived lo-
cation of a tactile event or to adjust the interval between
successive taps in order to produce a saltation illusion
(psychophysical method of production; Geldard, 1975).
The use of subjective location judgments has made it dif-
ficult to objectively quantify the strength of this illusion
and to investigate how its strength is affected by manip-
ulations of temporal, spatial, or attentional factors. This
lack of an objective quantitative measure of saltation may
This research was supported by a grant from the Biotechnology and
Biological Sciences Research Council. M.E. holds a Royal Society–
Wolfson Research Merit Award. Correspondence should be addressed to
M. Eimer, Department of Psychology, Birkbeck College, University of
London, Malet Street, London WC1E 7HX, England (e-mail: m.eimer@
bbk.ac.uk).
Cutaneous saltation within and across arms:
A new measure of the saltation illusion
in somatosensation
MARTIN EIMER and BETTINA FORSTER
Birkbeck College, University of London, London, England
and
JONAS VIBELL
University of Oxford, Oxford, England
A new objective procedure was used to measure the strength of cutaneous saltation, in order to clar-
ify current debates about the nature of this illusion. Three taps were presented successively to three
possible forearm locations. Participants attended to the middle location and reported whether a tap
was perceived there. When all stimuli were delivered to the same arm and intertap intervals were short,
participants were unable to distinguish real and illusory stimuli at the middle location. When both arms
were stimulated, location judgments on one arm were shifted toward a tap subsequently delivered to
the other arm. These results challenge the view that saltation is a purely attentional phenomenon, but
they are inconsistent with the idea that this illusion is produced in the primary somatosensory cortex.
CUTANEOUS SALTATION 459
help explain why debates with respect to the mechanisms
underlying cutaneous saltation remain unresolved. The
purpose of the present experiments was to develop and
test a new objective measure of the saltation illusion and
to use this measure to shed new light on current contro-
versies about the nature of cutaneous saltation.
We developed a novel procedure to measure the pres-
ence and strength of the saltation illusion while manipu-
lating temporal parameters of tactile stimulation, stimula-
tion locations, and body posture. On each trial, a sequence
of three taps (T1, T2, T3) was presented successively to
three possible forearm locations (L1, L2, L3). Partici-
pants were instructed to attend to the middle location
(L2) and to report at the end of each trial whether or not
they had felt a tap at that location. The dependent variable
was the number of taps reported to be present at L2 for
different types of trials. In Experiment 1, three trial types
were compared. On tap trials, all three locations (includ-
ing L2) were successively stimulated, and present re-
sponses were correct. On rabbit trials, T1 and T2 at one
outer location (L1 or L3) were followed by T3 at the op-
posite outer location (L3 or L1); this should result in an
illusory displacement of T2 toward L2. Thus, participants
should be likely to incorrectly report the presence of a tap
at L2. In control trials, all three taps were successively
delivered to one of the two outer locations (L1 or L3). Ex-
periment 2 also contained a different type of control trial
(motion control ), where T1 was delivered at one outer lo-
cation and T2 as well as T3 at the opposite outer location.
According to Geldard (1975, 1982), cutaneous saltation
should be absent in both types of control trials, and the
number of taps incorrectly reported at L2 should be low.
On the basis of the assumption that the strength of the
saltation illusion is reflected by the frequency of taps mis-
located at L2 on rabbit trials, the presence and strength of
saltation were quantified as follows. First, the number of
taps correctly perceived at L2 on tap trials was compared
with the number of taps incorrectly localized at L2 on rab-
bit trials. Identical perceptual reports on these two types of
trials (i.e., the absence of significant differences in the
percentage of taps reported to be present at L2 on tap and
rabbit trials) would suggest that participants are unable to
discriminate between real and illusory tap locations (see
Cholewiak & Collins, 2000, for an analogous argument).
The question of whether real and illusory percepts are
phenomenologically indistinguishable was further ex-
plored in Experiment 3 with the use of a two-alternative
forced choice procedure. Also, and equally important, the
percentage of taps incorrectly reported at L2 was com-
pared for rabbit and control trials. The presence of a cuta-
neous saltation illusion was inferred whenever a signifi-
cantly higher percentage of taps was incorrectly reported
to be present at L2 on rabbit trials than on control trials.
This general procedure was employed to study two
central questions related to the mechanisms responsible
for cutaneous saltation. First, the fact that participants
were instructed to continuously focus their attention on
L2, in order to detect and report any tactile events deliv-
ered to this location, allowed us to reinvestigate the claim
that the saltation illusion is exclusively determined by
expectations and attention (Kilgard & Merzenich, 1995).
If saltation was linked to the current focus of spatial atten-
tion, illusory displacements of T2 should occur equally
often on rabbit trials and on control trials. With attention
constantly focused on L2 in these two types of trials, the
perceived location of T2 will be subject to the same an-
ticipatory spatial bias. Therefore, the finding that taps
are more likely to be mistakenly located at L2 on rabbit
trials than on control trials would suggest that the saltation
illusion remains present even when the attentional fac-
tors identified by Kilgard and Merzenich are neutralized.
Second, we reevaluated the hypothesis that the body
midline represents an uncrossable boundary for cutaneous
saltation by investigating whether this illusion remains
present when tactile events are successively delivered to
opposite arms. Previous studies reporting that saltation
is confined to one side of the body (e.g., Geldard, 1982)
have investigated body sites such as left and right fore-
head or left and right anterior thorax. These areas do not
include joints and therefore cannot be independently
repositioned with respect to each other. It is possible that
saltation will be elicited across the body midline when
successive tactile stimuli are delivered to left and right
body parts that move separately in external space and
can be independently realigned relative to each other and
to the body midline. To investigate this possibility, we
applied the general procedures described above under
conditions where successive taps were delivered to op-
posite arms, with both arms positioned in parallel and
perpendicular to the main body axis (Figure 1B).
EXPERIMENT 1
Experiment 1 contained two conditions, delivered in
separate successive blocks. In the single-arm condition,
all taps were delivered to the same arm (Figure 1A). Pre-
vious studies (Cholewiak, 1976; Geldard, 1975) have
demonstrated that illusory sensations of T2 at L1 as being
located midway between L1 and L3 can be most reliably
observed for T2–T3 ISIs in the range of 70–150 msec. In
order to evaluate the validity of our new measure of
saltation, T2–T3 ISIs were set to either 100 msec (where
the saltation illusion should be strongest), 200 msec
(where the strength of this illusion should be reduced),
and 400 msec (where the illusion should be largely ab-
sent). Unlike in most previous investigations of saltation,
where the T1–T2 ISI was typically longer than that sep-
arating T2 and T3, T1–T2 ISIs were identical to T2–T3
ISIs in Experiment 1. This was done to preclude the pos-
sibility that perceptual reports of T2 locations would be
influenced by the tau effect (Helson & King, 1931).
When successive tactile events are separated by unequal
intervals, their perceived spatial separation can be biased,
with stimuli separated by short intervals judged to be
closer together than those separated by longer intervals.
1
The both-arms condition was included so that we could
reinvestigate the claim that the cutaneous rabbit is unable
to cross the body midline. Here, L1 and L2 were located
460 EIMER, FORSTER, AND VIBELL
on one arm and L3 on the other (Figure 1B). If saltation
was restricted to one side of the body, it should be com-
pletely eliminated when T2 and T3 were delivered to dif-
ferent arms. In the both-arms condition, the existence of
a saltation illusion was assessed separately for trials start-
ing at L1 and L3. In rabbit trials starting at L1, T2 was
also delivered at L1, whereas T3 was presented at L3 on
the other arm. Here, an illusory perception of T2 at L2
would demonstrate that perceived tap locations on one
arm can be biased by a subsequent tactile event on the
Tap
(123, 321)
Rabbit
(113, 331)
Control
(111, 333)
Tap (123)
Rabbit (113)
Control (111)
Tap (321)
Rabbit (331)
Control (333)
L1 L2 L3
L1L2L3
(A)
(B)
100
80
60
40
20
0
ISI (msec) 100 200 400
% Present Response
100
80
60
40
20
0
% Present Response
Across-Arms Pull
Across-Arms Jump
Figure 1. Stimulation locations and perceptual reports in the single-arm con-
dition (A) and in the both-arms condition (B) of Experiment 1. Bar graphs
show the mean percentage of trials where a tap was reported to be present at
L2 (errors bars indicate SEM ), displayed separately for tap, rabbit, and con-
trol trials. (A) Percentage of pr esent responses in the single-arm condition for
ISIs of 100 msec, 200 msec, and 400 msec. (B) Percentage of present responses
in the both-arms condition, shown separately for tap sequences starting at L1
(with T2 delivered at L1 on rabbit trials: across-arms pull, left) and for those
starting at L3 (with T2 delivered at L3 on rabbit trials: across-arms jump,
right).
CUTANEOUS SALTATION 461
other arm (“across-arms pull” in Figure 1B). In rabbit tri-
als starting at L3, T2 was delivered at L3, followed by T3
at L1 on the other arm. Here, L3 (where T2 was pre-
sented) and L2 were located on opposite arms. Thus, per-
ceived illusory shifts of T2 from L3 to L2 would indicate
a mislocalization of T2 on the other arm (“across-arms
jump” in Figure 1B), thereby suggesting that the cuta-
neous rabbit can hop across the body midline. Because
the strength of saltation was expected to be reduced in the
both-arms condition, T2–T3 ISIs were set to values close
to the interval reported to be optimal for eliciting this il-
lusion (50 and 100 msec). As in the single-arm condition,
T1–T2 ISIs were always identical to T2–T3 ISIs.
Method
Participants
. Sixteen paid participants (9 female, 7 male; 19–32
years of age) were tested.
Stimuli, Apparatus, and Procedure
. The participants received
taps delivered via solenoids attached to three locations (L1, L2, L3)
on the upper forearm. Solenoids drove a metal rod that contacted
the arm for 6 msec. On each trial, three successive taps (T1, T2, T3)
were delivered. The participants were instructed to direct their at-
tention to the middle location (L2) and to report at the end of each
trial, by pressing a left or right footpad, whether or not they had per-
ceived a tap at this location. The assignment of present and absent
responses to left and right footpads was balanced across partici-
pants. On tap trials, all three locations were successively stimulated
(L1→L2→L3; L3→L2→L1). On rabbit trials, T1 and T2 at one
outer location were followed by T3 at the other outer location
(L1→L1→L3; L3→L3→L1). On control trials, all three stimuli
were delivered successively at L1 or L3.
In the single-arm condition, all taps were delivered to the same
arm. Stimulator locations corresponded to 33%, 50%, and 67% of
the elbow–wrist distance (Figure 1A). Eight participants received
stimulation of the left arm, and the other 8 participants received right-
arm stimulation. The ISI between successive taps was 100, 200, or
400 msec. Five successive blocks, consisting of 90 trials, were deliv-
ered. Each block contained 5 randomly intermingled trials for each
combination of trial type (tap vs. rabbit vs. control), ISI (100 vs. 200
vs. 400 msec), and direction (sequences starting at L1 vs. L3).
In the both-arms condition, L1 and L2 were located on one arm
(at 33% and 50% of the elbow–wrist distance) and L3 on the other
arm (at 33% of the elbow–wrist distance; Figure 1B). For 8 partic-
ipants, L1 and L2 were on the left arm and L3 on the right arm, and
this assignment was reversed for the other 8 participants. Intertap
interval was 50 or 100 msec. Four successive blocks, consisting of
84 trials, were delivered. Each block contained 7 randomly inter-
mingled trials for each combination of trial type, ISI, and direction.
Data Analysis
. Trials without responses were excluded from
analysis. Repeated measures analyses of variance (ANOVAs) were
conducted on the percentage of present responses measured in the
single-arm and both-arms conditions, separately for tap and rabbit
trials (excluding control trials) and for rabbit and control trials (ex-
cluding tap trials). In the single-arm condition, analyses included
the factors trial type (tap vs. rabbit; rabbit vs. control), direction,
and ISI. For the both-arms condition, separate analyses were con-
ducted for trial sequences starting at L1 (across-arms pull) and for
trial sequences starting at L3 (across-arms jump), including the fac-
tors trial type and ISI. Additional paired t tests were used to com-
pare perceptual reports for different trial types. For all analyses,
α
level was .05, and only statistically significant effects are reported.
Results
Single arm
. Figure 1A shows percentages of present
responses on tap, rabbit, and control trials in the single-arm
condition. When tap and rabbit trials were analyzed to-
gether, an effect of trial type [tap vs. rabbit trials: F(1,15) ⫽
57.3, p ⬍ .001] showed that present responses were more
frequent on tap trials than on rabbit trials. A trial type ⫻ ISI
interaction [F(2,30) ⫽ 45.0, p ⬍ .001] indicated that the
number of taps incorrectly reported at L2 on rabbit trials
increased with decreasing ISIs. Paired t tests revealed that
taps at L2 were perceived less frequently on rabbit than on
tap trials with 200- and 400-msec ISIs [ts(14) ⬎ 8.0, ps ⬍
.001]. In contrast, for 100-msec ISIs, the percentage of
present responses on tap trials (88%) and on rabbit trials
(83%) did not differ significantly, suggesting that real and
illusory percepts were indiscriminable. When rabbit and
control trials were analyzed together, an effect of trial type
[rabbit vs. control trials: F(1,15) ⫽ 139.8, p ⬍ .001] dem-
onstrated that present responses were much more frequent
on rabbit than on control trials. Although a trial type ⫻ ISI
interaction [F(2,30) ⫽ 23.4, p ⬍ .001] reflected the de-
crease of this saltation illusion with increasing ISI, paired
t tests revealed that participants reported illusory taps at L2
more frequently on rabbit than on control trials for all three
ISIs [ts(15) ⬎ 4.9, ps ⬍ .001].
Both arms
. Figure 1B shows percentages of present
responses of tap, rabbit, and control trials in the both-
arms condition, displayed separately for tap sequences
starting at L1 (with illusory displacements of T2 from
L1 to L2 on the same arm representing an across-arms
pull; left side) and for sequences starting at L3 (with il-
lusory displacements of T2 from L3 to L2 on the oppo-
site arm representing an across-arms jump; right side).
Because no main effects of ISI or interactions including
this factor were present, the data shown in Figure 1B are
averaged across both ISIs (50 and 100 msec). When tap
and rabbit trials were analyzed together, main effects of
trial type (tap vs. rabbit trials) were obtained for both
movement directions [F(1,15) ⫽ 19.3 and 15.2, ps ⬍
.001, respectively], reflecting the fact that present re-
sponses were more frequent on tap than on rabbit trials.
When rabbit and control trials were analyzed together,
main effects of trial type (rabbit vs. control trials) were
again obtained for both movement directions [F(1,15) ⫽
114.2 and 63.9, ps ⬍ .001, respectively]. This finding
that present responses were much more likely on rabbit
than on control trials indicates the presence of a salta-
tion illusion in the both-arms condition and suggests that
this illusion was elicited regardless of whether T2 was
presented at L1, and therefore could be mislocated on the
same arm, or was presented at L3 and thus could be mis-
located on the opposite arm (Figure 1B). Paired t tests
confirmed that present responses were more frequent on
rabbit than on control trials for sequences starting at L1
as well as at L3 [ts(15) ⬎ 8.0, ps ⬍ .001].
Discussion
In Experiment 1, we used a newly developed proce-
dure to objectively measure the strength of the cutaneous
saltation illusion, producing several noteworthy find-
ings. The results from the single-arm condition showed
that the strength of the saltation illusion decreased with
462 EIMER, FORSTER, AND VIBELL
increasing T2–T3 ISIs. For 100-msec intervals, illusory
taps at L2 on rabbit trials were reported as frequently as
were real taps on tap trials, thereby suggesting that real
and illusory tap locations were perceptually very similar.
With longer T2–T3 ISIs (200 and 400 msec), partici-
pants were less likely to report the presence of a tap at L2
on rabbit trials than on tap trials. In line with earlier ex-
periments using subjective location judgments (Geldard,
1975, 1982), this result shows that the strength of the
saltation illusion is sensitive to the T2–T3 ISI. Further-
more, illusory taps at L2 were generally reported much
more frequently on rabbit trials than on control trials in
the single-arm condition. Since these trials were de-
signed to be equivalent with respect to spatial biases pro-
duced by expectation and attention, this finding strongly
suggests that tactile saltation cannot be explained solely in
terms of an attentional bias (Kilgard & Merzenich, 1995).
The results obtained in the both-arms condition pro-
vide evidence that illusory displacements of stimuli de-
livered to one side of the body can be elicited by tactile
stimuli subsequently delivered to the opposite side. Illu-
sory taps at L2 were reported on about 50% of all rabbit
trials (and on less than 10% of control trials). The fact
that participants were more likely to report the presence
of a tap at L2 on rabbit trials than on control trials when
stimulus sequences started at L1 suggests that perceived
tap locations on one arm are systematically shifted to-
ward a subsequent tap on the other arm (“across-arms
pull” in Figure 1B). Similar findings were also obtained
for sequences starting at L3, suggesting that taps were
frequently mislocalized on the wrong arm when this arm
was subsequently stimulated (“across-arms jump” in
Figure 1B). This result seriously challenges the hypoth-
esis that the cutaneous rabbit is unable to cross the body
midline. Overall, the perceptual reports obtained in the
both-arms condition indicate the presence of a strong bi-
asing effect exerted by tactile stimuli presented to one
arm on location judgments in response to stimuli previ-
ously delivered to the other arm.
Before these interpretations can be accepted, poten-
tial shortcomings of the experimental procedure used in
Experiment 1 need to be considered. In the single-arm
condition, taps were more likely to be reported at L2 on
rabbit trials than on control trials even when the T2–T3
ISI was 400 msec, which is clearly outside the temporal
range for which cutaneous saltation was observed in ear-
lier studies (Cholewiak, 1976; Geldard, 1975). This un-
expected finding suggests that perceptual reports on rab-
bit trials may have been affected by factors other than
cutaneous saltation. One possibly problematic aspect of
the procedure employed in Experiment 1 was that stim-
ulus sequences were stationary on control trials (where
all three taps were presented at the same location), whereas
on rabbit and on tap trials, successive taps were deliv-
ered to different locations. It is possible that participants
may have been biased to report a tap at L2 whenever a se-
quence was not stationary, especially when ISIs were
short, and that this response bias caused different per-
ceptual reports on rabbit and control trials. The fact that
ISIs were randomized within blocks may even have en-
hanced the differential salience of stationary versus mov-
ing stimulus sequences, thereby potentially exacerbating
possible response bias effects. In order to interpret the
findings of Experiment 1 as evidence for cutaneous salta-
tion, the potential impact of response bias needs to be con-
trolled. Experiment 2 was conducted to address this issue.
Another possible problem of Experiment 1 is related
to the fact that the T1–T2 and T2–T3 ISI were identical.
This procedure differs somewhat from the procedure
used in earlier investigations of cutaneous saltation (e.g.,
Geldard, 1982), where the T1–T2 ISI was typically much
longer (800 msec) than the T2–T3 interval. As described
earlier, identical T1–T2 and T2–T3 ISIs were chosen in
order to neutralize any biasing influence of the tau effect
(Helson & King, 1931) on perceptual reports. However,
presenting successive tactile stimuli with ISIs of 100 msec
or less can give rise to apparent motion perceptions (tactual
phi; Sherrick & Rogers, 1966), which could have con-
tributed to the effects observed in Experiment 1. To reduce
any possible impact of induced apparent motion and to
test our new measure of saltation with temporal stimula-
tion procedures identical to those used in earlier studies,
Experiment 2 featured T1–T2 intervals of 800 msec in
all trials.
EXPERIMENT 2
The aim of Experiment 2 was to confirm the results ob-
served in Experiment 1, while at the same time controlling
for the possible impact of response bias effects induced
by the presence or absence of stimulus movement. In ad-
dition, Experiment 2 employed T1–T2 intervals identical
to those used in classical investigations of cutaneous salta-
tion (e.g., Geldard, 1975, 1982). The procedures were
identical to those of Experiment 1, with the following
exceptions. First, the T1–T2 ISI was now 800 msec in all
trials. The T2–T3 ISI was set to 100 msec (where the
saltation illusion should be strong) or 300 msec (where
this illusion should be weak or absent), for both the single-
arm and both-arms conditions. Short and long T2–T3
ISIs were now presented in separate blocks, rather than
randomized within blocks, as in Experiment 1. Finally,
and most important, an additional control condition was
introduced. On motion control trials, T1 at one of the
outer locations was followed by T2 and T3 at the other
outer location (L1→L3→L3; L3→L1→L1). These trials
were similar to rabbit and tap trials in that tactile sequences
were not stationary, but involved the successive stimula-
tion of different locations.
2
If the perceptual reports ob-
served for rabbit trials in Experiment 1 were affected by
a differential response bias induced by stationary versus
moving stimulation sequences, this bias should also be
present on motion control trials. If response bias was
solely responsible for the difference between perceptual
reports obtained on rabbit and control trials in Experi-
ment 1, taps at L2 should now be reported equally often
on rabbit trials and on motion control trials. In contrast,
genuine cutaneous saltation effects will be reflected by
CUTANEOUS SALTATION 463
more frequent present responses on rabbit trials than on
motion control trials.
Method
Participants
. Ten participants (6 female, 4 male; 21– 42 years of
age) were tested. None had participated in Experiment 1.
Stimuli, Apparatus, Procedure, and Data Analysis
. The lo-
cations of tactile stimulators, arm postures, and participants’ task
were the same as in Experiment 1. The T1–T2 ISI was now always
800 msec, and the T2–T3 ISI was set to 100 or 300 msec, in differ-
ent blocks. In addition to the three trial types (tap, rabbit, control)
included in Experiment 1, motion control trials were included,
where T1 at one outer location was followed by T2 and T3, which
were both presented at the opposite outer location (L1→L3→L3;
L3→L1→L1). The control condition already included in Experi-
ment 1 (L1→L1→L1; L3→L3→L3) is now referred to as static
control.
The experiment consisted of 12 blocks, and each block contained
40 trials. Six successive blocks were delivered for the single-arm
and both-arms conditions, respectively. Each block contained 5 ran-
domly intermingled trials for each combination of trial type (tap vs.
rabbit vs. static control vs. motion control) and direction. In three
successive blocks, the T2–T3 ISI was 100 msec, and in the other
three successive blocks, the ISI was 300 msec. In all other respects,
the procedures were identical to those of Experiment 1. Data analy-
ses were analogous to Experiment 1, except that additional repeated
measures ANOVAs were conducted to compare perceptual reports
for rabbit and for motion control trials.
Results
Single arm
. Figure 2 shows the percentage of present
responses on tap, rabbit, static control, and motion con-
trol trials in the single-arm condition for T2–T3 ISIs of
100 msec (left) and 300 msec (right). When tap and rab-
bit trials were analyzed together, an effect of trial type
[tap vs. rabbit trials: F(1,9) ⫽ 9.3, p ⬍ .02] showed that
present responses were more frequent on tap than on rab-
bit trials. A trial type ⫻ ISI interaction [F(1,9) ⫽ 30.5,
p ⬍ .001] indicated that incorrectly reported taps at L2
on rabbit trials were more frequent for the shorter ISI.
Paired t tests revealed that taps at L2 were perceived less
frequently on rabbit than on tap trials at ISIs of 300 msec
[t(9) ⫽ 4.6, p ⬍ .01]. In contrast, for 100-msec ISIs, the
percentage of present responses on tap trials (81%) and
rabbit trials (82%) did not differ significantly, thus again
suggesting that real and illusory percepts may be phe-
nomenologically indistinguishable when T2 and T3 are
separated by 100 msec. When rabbit and static control
trials were analyzed together, an effect of trial type [rab-
bit vs. static control trials: F(1,9) ⫽ 95.5, p ⬍ .001]
demonstrated that present responses were more frequent
on rabbit than on static control trials. Although a trial
type ⫻ ISI interaction [F(1,9) ⫽ 17.2, p ⬍ .01] indicated
that this difference was larger with ISIs of 100 msec,
paired t tests revealed that participants reported illusory
taps at L2 more frequently on rabbit trials than on static
control trials for both ISIs [ts(9) ⬎ 4.5, ps ⬍ .001]. Im-
portantly, when rabbit and motion control trials were an-
alyzed together, an effect of trial type [rabbit vs. motion
control trials: F(1,9) ⫽ 8.9, p ⬍ .02] demonstrated that
present responses were more frequent on rabbit trials
Tap
(123, 321)
Rabbit
(113, 331)
Control-S
(111, 333)
Control-M
(133, 311)
L1 L2 L3
100
80
60
40
20
0
S2–S3 ISI (msec)
100
300
% Present Responses
Single-Arm Condition
Figure 2. Stimulation locations and perceptual reports in the single-arm con-
dition of Experiment 2. Bar graphs show the mean percentage of trials where a
tap was reported to be present at L2 (errors bars indicate SEM), displayed sep-
arately for tap, rabbit, static control (control-S), and motion control (control-M)
trials and for T2–T3 ISIs of 100 msec (left) and 300 msec (right).
464 EIMER, FORSTER, AND VIBELL
than on motion control trials. Again, a trial type ⫻ ISI
interaction [F(1,9) ⫽ 8.2, p ⬍ .02] reflected the decrease
of this difference with increasing ISI. Paired t tests re-
vealed that the difference in the number of illusory taps
reported at L2 on rabbit and motion control trials was not
significant for 300-msec ISIs. In contrast, for ISIs of
100 msec, participants reported illusory taps at L2 sig-
nificantly more often on rabbit trials than on motion con-
trol trials [t(9) ⫽ 3.9, p ⬍ .01].
Both arms
. Figure 3 shows percentages of present re-
sponses of tap, rabbit, static control, and motion control
trials in the both-arms condition for blocks where the
T2–T3 ISIs were 100 msec (top panel) and 300 msec
(bottom panel). Results are shown separately for trials
where the tap sequence started at L1 (with any across-
arms pull resulting in illusory displacements of T2 from
L1 to L2 on the same arm; Figure 3, left side) and for se-
quences starting at L3 (with any across-arms jump re-
sulting in perceived displacements of T2 across arms
from L3 to L2; Figure 3, right side). When tap and rab-
bit trials were analyzed together, significant effects of
trial type (tap vs. rabbit trials) were present for both
movement directions [F(1,9) ⫽ 31.4 and 21.9, ps ⬍ .001,
respectively], demonstrating that present responses were
more frequent on tap than on rabbit trials. A trial type ⫻
ISI interaction was significant for trials starting at L1
[F(1,9) ⫽ 7.6, p ⬍ .03], indicating that taps were more
likely to be incorrectly reported at L2 on rabbit trials
when ISI was 100 msec. When rabbit and static control
trials were analyzed together, significant effects of trial
type (rabbit vs. static control trials) were again present
for both movement directions [F(1,9) ⫽ 13.9 and 21.7,
ps ⬍ .005, respectively]. As in Experiment 1, present re-
sponses were more frequent on rabbit trials than on sta-
tic control trials for both movement directions, thus ap-
parently again suggesting the presence of an across-arms
pull effect as well as an across-arms jump.
However, when rabbit trials and the newly introduced
motion control trials were analyzed together, a different
picture emerged. Importantly, no significant main effect
of trial type (rabbit vs. motion control trials) or trial
type ⫻ ISI interaction was found for trials starting at L3,
thus demonstrating that on these trials, taps were no
more likely to be incorrectly reported at L2 on rabbit tri-
als than on motion control trials (see Figure 3, right:
across-arms jump). In contrast, a significant effect of
trial type (rabbit vs. motion control trials) was present
for trials starting at L1 [F(1,9) ⫽ 9.8, p ⬍ .02], since
present responses were more frequent on rabbit trials
than on motion control trials (see Figure 3, left: across-
arms pull). Although the trial type ⫻ ISI interaction was
not significant, this difference in perceptual judgments
between rabbit and motion control trials tended to be
much more pronounced with ISIs of 100 msec. Follow-up
paired t tests conducted for trials starting at L1 confirmed
that for the 100-msec ISI, the percentage of present re-
sponses on rabbit trials (68%) and on motion control trials
(38%) differed significantly [t(9) ⫽ 2.4, p ⬍ .04], thereby
demonstrating the presence of a reliable across-arms pull
effect. In contrast, this difference failed to reach signifi-
cance for the 300-msec ISI (36% and 20% present re-
sponses on rabbit and motion control trials, respectively).
Discussion
The results observed in the single-arm condition of
Experiment 2 were very similar to those previously found
in Experiment 1. As in this first experiment, perceptual
reports were identical for rabbit trials and tap trials when
the T2–T3 ISI was 100 msec, thereby suggesting again
that these two types of trials are perceptually very simi-
lar, or even indistinguishable. The latter claim was fur-
ther investigated in Experiment 3. An important differ-
ence in Experiment 1 was that the T1–T2 ISI was now
800 msec (as in earlier investigations of saltation; e.g.,
Geldard, 1975, 1982). Although it has previously been
claimed that T1–T2 ISIs need to be at least 300 msec to
produce a reliable saltation illusion (Geldard, 1982), the
present results do not support this view. The fact that per-
ceptual reports for rabbit and tap trials were statistically
equivalent in Experiment 1 for T1–T2 ISIs of 100 msec,
as well as in Experiment 2 for ISIs of 800 msec, suggests
that variations in the T1–T2 interval do not affect the
strength of the saltation illusion.
As in Experiment 1, present responses were more fre-
quent on rabbit trials than on static control trials in the
single-arm condition, both when the T2–T3 ISI was
100 msec, but also for ISIs of 300 msec. Crucially, the
newly introduced motion control condition made it pos-
sible to dissociate effects of response bias triggered by
the difference between static and moving stimulation se-
quences on rabbit trials from genuine cutaneous salta-
tion effects. As shown in Figure 2, participants were
more likely to report the presence of a tap at L2 on mo-
tion control trials (where successive taps were delivered
to different locations; 39.5% present responses) than on
static control trials (where all taps were presented at the
same location; 15.3% present responses). This differ-
ence strongly suggests that response bias is likely to be
responsible for at least some of the present responses on
rabbit trials. However, and importantly, when the T2–T3
ISI was 100 msec, taps were significantly more often re-
ported at L2 on rabbit trials (82%) than on motion con-
trol trials (43%). This result demonstrates the presence
of a genuine saltation illusion, above and beyond any bi-
asing effects of perceived motion. In contrast, when the
T2–T3 ISI was longer (300 msec), no significant differ-
ences in perceptual reports between rabbit and motion
control trials were obtained. Thus, for this longer inter-
val, differences in perceptual reports between rabbit and
static control trials are primarily caused by response
bias, and not by cutaneous saltation.
The results obtained in the both-arms condition of Ex-
periment 2 for tap, rabbit, and static control trials when
the T2–T3 ISI was 100 msec were again very similar to
those obtained in Experiment 1 (compare Figure 1B and
Figure 3A). Taps were less likely to be reported at L2 on
rabbit trials than on tap trials, but these reports were con-
siderably more frequent on rabbit trials than on static
CUTANEOUS SALTATION 465
Tap (123)
Rabbit (113)
Control-S (111)
Control-M (133)
Tap (321)
Rabbit (331)
Control-S (333)
Control-M (311)
(B)
100
80
60
40
20
0
% Present Responses
Across-Arms Pull
Across-Arms Jump
T2–T3 ISI: 300 msec
T2–T3 ISI: 100 msec
Tap (123)
Rabbit (113)
Control-S (111)
Control-M (133)
Tap (321)
Rabbit (331)
Control-S (333)
Control-M (311)
Across-Arms Pull Across-Arms Jump
(A)
100
80
60
40
20
0
% Present Responses
L3 L1L2
Figure 3. Stimulation locations and perceptual reports in the both-arms con-
dition of Experiment 2. Bar graphs show the mean percentage of trials where a
tap was reported to be present at L2 (errors bars indicate SEM ), displayed sep-
arately for tap, rabbit, static control (control-S), and motion control (control-M)
trials. Data are shown separately for tap sequences starting at L1 (with T2 de-
livered at L1 on rabbit trials: “across-arms pull”; left) and for those starting at
L3 (with T2 delivered at L3 on rabbit trials: “across-arms jump”; right). (A)
Percentage of present responses for T2–T3 ISI of 100 msec. (B) Percentage of
present responses for T2–T3 ISI of 300 msec.
466 EIMER, FORSTER, AND VIBELL
control trials. We tentatively interpreted analogous find-
ings in Experiment 1 as evidence for the ability of the
cutaneous rabbit to hop across arms. However, the re-
sults obtained in Experiment 2 for motion control trials
qualify this interpretation in an important way. For trials
starting at L3 (trials where cutaneous saltation would
imply the mislocalization of a tactile event on the oppo-
site arm; Figure 3: across-arms jump), perceptual reports
on rabbit trials and motion control trials did not differ
significantly. This finding indicates that the differences
observed in both experiments between rabbit and static
control trials starting at L3 can be fully accounted for by
response bias triggered by perceived stimulus movement
and thus should not be regarded as evidence that the cu-
taneous rabbit can jump across arms.
However, a different pattern of results was obtained in
the both-arms condition for trials starting at L1 (trials
where cutaneous saltation would indicate that tactile
events delivered to one arm affect the localization of an
earlier tactile event on the opposite arm; Figure 3: across-
arms pull). When the T2–T3 ISI was 100 msec, partici-
pants were significantly more likely to report the pres-
ence of a tap at L2 on rabbit trials than on motion control
trials. Since perceptual reports on these trials should
have been equally affected by response bias induced by
stimulus motion, this difference appears to reflect the
presence of a genuine across-arms localization bias. This
finding supports our tentative conclusion regarding the
findings in Experiment 1 that sensory saltation effects
on one side of the body midline can be triggered by a
subsequent tactile stimulus presented to the opposite side.
EXPERIMENT 3
In the single-arm condition of Experiments 1 and 2,
participants were equally likely to report the presence of
a tap at L2 on rabbit trials and on tap trials when the
T2–T3 ISI was 100 msec. This fact was tentatively inter-
preted as indicating that illusory tap locations on rabbit
trials were perceptually similar or perhaps even identical
to real taps at L2 on tap trials. However, the presence of
statistically equivalent perceptual reports on tap and on
rabbit trials should not be interpreted as conclusive evi-
dence for the strong claim that observers’ percepts on
these two types of trials were phenomenologically iden-
tical. It is entirely conceivable that perceived tap loca-
tions at L2 were generally more salient on tap trials when
L2 was in fact stimulated than on rabbit trials.
3
In Experiment 3, we employed a two-alternative forced
choice procedure in order to investigate directly whether
or not perceived tap locations on rabbit trials and on tap
trials are perceptually indistinguishable. On every trial,
two tactile stimulation sequences (a tap sequence in-
cluding stimulation of L2 and a rabbit sequence without
L2 stimulation, as defined in Experiments 1 and 2) were
presented successively on the same arm and in random
order across trials. Participants had to decide at the end of
each trial whether the first or the second stimulus sequence
contained a tap at L2. If their percepts on rabbit and on
tap sequences were phenomenologically identical, forced
choice discrimination performance should be at chance
level. In contrast, if perceived tap locations at L2 were
more salient during tap sequences than during rabbit se-
quences, observers should be able to use this phenome-
nological difference to reliably identify the stimulation
sequence that contained a real stimulation of L2.
In the first part of Experiment 3, each trial contained
one tap and one rabbit sequence, and observers were
asked at the end of each block to rate their confidence
with respect to their ability to identify the sequence con-
taining a tap at L2. In the second part, trials containing
tap and rabbit sequences were randomly intermixed with
two-alternative forced choice trials containing tap and
motion control sequences (as defined in Experiment 2).
Since no saltation illusion should be elicited during mo-
tion control sequences, these trials were included to ob-
tain a baseline measure of participants’ two-alternative
forced choice accuracy. In addition, their presence might
provide observers with a clearly discriminable percep-
tual difference, which could be used to anchor their pre-
sumably more subtle perceptual judgments on trials in-
cluding tap and rabbit sequences.
Method
Participants
. Ten participants (4 female, 6 male; 24–35 years of
age) were tested. None had participated in Experiments 1 or 2.
Stimuli, Apparatus, and Procedure
. The locations of the tac-
tile stimulators were the same as in Experiments 1 and 2. All taps
were delivered to the same arm (the left arm for 5 participants, the
right arm for the other 5 participants). As in Experiment 2, the T1–T2
ISI was 800 msec, and the T2–T3 ISI was 100 msec. On each trial,
two tactile stimulus sequences were presented successively, separated
by an empty interval of 1,000 msec. Both sequences always started
at the same location (L1 or L3), and one of them was a tap sequence
(L1→L2→L3; L3→L2→L1), whereas the other was either a rabbit
sequence (L1→L1→L3; L3→L3→L1) or a motion control sequence,
as defined in Experiment 2 (L1→L3→L3; L3→L1→L1). The two
possible serial positions of a tap sequence within a trial (first vs. sec-
ond) were equiprobable and randomized across trials. The partici-
pants were instructed to report at the end of each trial which of the
two stimulus sequences included a tap at the middle location (L2) by
pressing a left or right footpad. Five participants used the left footpad
to signal the presence of a tap at L2 in the first sequence and the right
footpad for a tap at L2 in the second sequence; these response as-
signments were reversed for the other 5 participants.
The experiment consisted of two parts. In the first part, tap se-
quences were always paired with rabbit sequences. Three blocks
were run, each consisting of 12 trials. At the end of each block, the
participants rated their overall confidence with respect to the accu-
racy of their two-alternative forced choice judgments on a 7-point
scale (1 ⫽ I was guessing all the time, 4 ⫽ I was guessing on half
of all trials, and 7 ⫽ I am confident that I was always correct). The
second part of the experiment consisted of three blocks (48 trials
per block). Tap sequences were paired with rabbit sequences on 32
trials per block and with motion control sequences on 16 trials per
block. The order in which these two types of trials were presented
was completely randomized. No confidence ratings were required.
Results and Discussion
In the first part of Experiment 3, where tap and rabbit
sequences were always paired, participants correctly
identified which of the two successively presented se-
CUTANEOUS SALTATION 467
quences contained a tap at L2 on 59% of all trials. This
detection performance did not differ significantly from
chance [50% correct; t(9) ⫽ 1.0, p ⫽ .32]. Observers’
mean confidence rating was 4.0, indicating that on aver-
age, they judged themselves to have been able to cor-
rectly identify the tap sequence on 50% of all trials.
However, confidence was not predictive of actual dis-
crimination performance, because participants’ confi-
dence ratings after each experimental block and their ac-
tual detection accuracy in these blocks were completely
uncorrelated (r ⫽ .04).
In the second part of the experiment, where tap se-
quences were paired with rabbit sequences on two thirds
of all trials and with motion control sequences on the re-
maining trials, forced choice discriminative judgments
were significantly more accurate when trials contained a
tap–motion control pairing than when trials contained a
tap–rabbit pairing [t(9) ⫽ 4.6, p ⬍ .01]. Participants cor-
rectly identified the tap sequence on 85% of all trials
when it was paired with a motion control sequence, which
was clearly above chance [t(9) ⫽ 13.4, p ⬍ .001]. In con-
trast, tap detection accuracy was again only 59% on tri-
als where tap and rabbit sequences were paired, which
was not different from chance performance [t(9) ⫽ 1.1,
p ⫽ .20].
Overall, the two-alternative forced choice results from
Experiment 3 provide support for the hypothesis that ob-
servers’ percepts on rabbit trials and on tap trials are
phenomenologically very similar. Although there was a
tendency in favor of observers choosing the tap sequence,
overall discrimination performance across observers did
not differ significantly from chance, regardless of whether
tap–rabbit sequence pairings were presented in isolation
or randomly intermingled with tap–motion control pairs.
Overall, there was substantial variability in observers’
discriminative judgments in response to tap–rabbit se-
quence pairings, suggesting considerable interindividual
differences in the strength of the saltation illustration. In
fact, several observers (4 in the first part and 2 in the sec-
ond part of Experiment 3) were actually more likely to
pick the rabbit sequence than the tap sequence. However,
all 10 observers performed better (and consistently well
above chance) when distinguishing between tap and mo-
tion control sequences as opposed to tap–rabbit sequences.
If perceived tap locations at L2 had been clearly more
salient during tap sequences than during rabbit sequences,
observers could have based their discriminative judg-
ments on this perceptual difference and therefore should
have been able to choose tap sequences in the majority
of trials. The fact that overall discrimination perfor-
mance was at chance level provides additional evidence
for the strength of the saltation illusion by objectively
demonstrating observers’ difficulty of discriminating
between real and illusory tap locations.
GENERAL DISCUSSION
The aim of the present experiments was to introduce
and evaluate a new objective procedure to measure the
strength of the cutaneous rabbit illusion and to use this
measure to investigate and clarify controversies with re-
spect to the nature of this illusion. On each trial, a se-
quence of three taps (T1, T2, T3) was presented to three
possible locations (L1, L2, L3) on the left or right fore-
arm, and participants had to report whether or not a tap
was delivered to the middle location (L2). Perceptual re-
ports on trials where L2 was in fact stimulated (tap tri-
als) were compared with perceptual reports on trials
where two successive taps at one of the outer locations
were followed by one tap at the other outer location (this
was expected to produce an illusory perception of T2 at
L2; rabbit trials). Identical perceptual reports on these
two types of trials would suggest that real and illusory
taps at L2 are perceptually equivalent, and this was in-
deed found in Experiments 1 and 2 when all stimuli were
delivered to the same arm, and the T2–T3 ISI was
100 msec. Under these conditions, participants were just
as likely to report a tap at L2 in rabbit trials as in tap tri-
als. We used a two-alternative forced choice procedure in
Experiment 3 to demonstrate that observers were gener-
ally unable to distinguish between real and illusory tap
locations, thereby suggesting that the percepts elicited
on tap and rabbit trials were phenomenologically very
similar, and for most observers possibly even equivalent.
Taken together, these findings emphatically underline
the strength of the cutaneous saltation illusion.
The finding that taps were more likely to be reported
at L2 on rabbit trials than on control trials (static control
in Experiments 1 and 2, motion control in Experiment 2)
rules out the hypothesis that saltation can be explained
exclusively in terms of biases produced by anticipation
and selective attention (Kilgard & Merzenich, 1995).
The participants were explicitly instructed to attend to
L2 in order to detect the presence or absence of a tap at
that location. With attention continuously focused at L2,
perceived T2 locations should have been affected by equiv-
alent spatial biases in rabbit trials and in control trials, re-
sulting in similar perceptual reports. The fact that percep-
tual judgments were in fact markedly different strongly
suggests that cutaneous saltation is not just an artifact of
spatial attention.
4
In line with earlier findings that sensory saltation crit-
ically depends on the interval separating T2 and T3 (Gel-
dard, 1975, 1982), the percentage of taps incorrectly re-
ported at L2 on rabbit trials decreased as this interval
was increased. The results obtained in the single-arm
condition of Experiment 2 suggest that cutaneous salta-
tion is virtually absent in 300-msec ISIs, and present re-
sponses on rabbit trials reflect a response bias caused by
perceived stimulus movement. Response bias effects are
also likely to be responsible for the substantial number of
present responses for rabbit trials in the single-arm con-
dition of Experiment 1 when ISIs were 400 msec (and
therefore well outside the range where the saltation illu-
sion is normally elicited).
To reevaluate the claim that the cutaneous rabbit is
confined to one side of the body and never crosses the
body midline (Geldard, 1982; Geldard & Sherrick, 1983),
468 EIMER, FORSTER, AND VIBELL
successive stimuli were delivered to different arms in the
both-arms condition. Here, the presence of the saltation
illusion was evaluated independently for trials starting at
L3 (where present responses on rabbit trials would indi-
cate the illusory localization of taps on the wrong arm;
across-arms jump) and for trials starting at L1 (where
present responses on rabbit trials would indicate a mis-
localization of taps on one arm caused by a subsequent
tap on the opposite arm; across-arms pull). The results
obtained in Experiment 1 suggested the presence of an
across-arms pull as well as an across-arms jump, but Ex-
periment 2 indicated that the apparent across-arms jump
can be fully accounted for by response bias alone. Thus,
the findings of the present study are in line with earlier
observations by Geldard and colleagues that sensory
saltation does not produce illusory shifts of tactile events
across the body midline. In other words, the cutaneous
rabbit remained unwilling to jump across arms. How-
ever, a reliable across-arms pull remained present in Ex-
periment 2 above and beyond any effects of response
bias triggered by perceived motion. This finding that the
location of a tactile event delivered to one arm is shifted
toward a subsequent tactile event presented to the other
arm is important, since it demonstrates that this illusion
can occur even when successive tactile events are ini-
tially represented in different hemispheres. This obser-
vation appears to rule out the primary somatosensory
cortex (S1) as the basis of sensory saltation, since S1 ex-
clusively represents the contralateral body side and has
few commissural connections to the opposite hemisphere
(Powell, 1977). Unlike S1, higher order somatosensory
areas (secondary and posterior parietal somatosensory
cortex) contain a large number of neurons with bilateral
receptive fields (Iwamura, Iriki, & Tanaka, 1994). These
areas may have mediated the tactile localization bias on
one arm produced by a subsequent tactile event on the
other arm.
In summary, the present experiments have demon-
strated the utility of a new objective measure of cuta-
neous saltation and have challenged previous theoretical
accounts of this illusion. On the methodological level,
the similarity of perceptual reports in Experiments 1 and
2 observed despite differences in the T1–T2 ISI, and de-
spite the fact that ISI conditions were blocked in Exper-
iment 2 and randomized within blocks in Experiment 1,
underlines the reliability of perceptual reports as mea-
sures of cutaneous saltation. With respect to the validity
of this measure, Experiment 2 revealed that response
bias effects related to the difference between stationary
and moving tactile stimulation sequences represent a po-
tential confounding factor. This factor needs to be con-
trolled in order to be able to interpret perceptual reports
as valid indicators of the saltation illusion.
On the theoretical level, the present findings, and es-
pecially the results of Experiment 3, suggest that cuta-
neous saltation is sufficiently powerful to produce illu-
sory percepts that are highly similar to, or perhaps even
phenomenologically indistinguishable from, veridical
tactile sensations. Although the saltation illusion clearly
cannot be fully explained in terms of top-down factors
such as attention and expectation (Kilgard & Merzenich,
1995), the fact that a tap on one arm can affect the lo-
calization of a previous tap on the other arm is also in-
consistent with the view that tactile saltation originates
at early somatosensory processing stages (Geldard, 1975,
1982) and suggests that this illusion is either generated
in higher order somatosensory areas or is a postpercep-
tual phenomenon.
REFERENCES
Cholewiak, R. W. (1976). Satiation in cutaneous saltation. Sensory
Processes, 1, 163-175.
Cholewiak, R. W. (1999). The perception of tactile distance: Influ-
ences of body site, space, and time. Perception, 28, 851-875.
Cholewiak, R. W., & Collins, A. A. (2000). The generation of vi-
brotactile patterns on a linear array: Influences of body site, time, and
presentation mode. Perception & Psychophysics, 62, 1220-1235.
Dennett, D. C. (1991). Consciousness explained. Boston: Little, Brown.
Geldard, F. A. (1975). Sensory saltation: Metastability in the percep-
tual world. Hillsdale, NJ: Erlbaum.
Geldard, F. A. (1982). Saltation in somesthesis. Psychological Bul-
letin, 92, 136-175.
Geldard, F. A., & Sherrick, C. E. (1972). The cutaneous “rabbit”: A
perceptual illusion. Science, 178, 178-179.
Geldard, F. A., & Sherrick, C. E. (1983). The cutaneous saltatory
area and its presumed neural basis. Perception & Psychophysics, 33,
299-304.
Helson, H., & King, S. M. (1931). The tau effect: An example of psy-
chological relativity. Journal of Experimental Psychology, 14, 202-
217.
Iwamura, Y., Iriki, A., & Tanaka, M. (1994). Bilateral hand repre-
sentation in the postcentral somatosensory cortex. Nature, 369, 554-
556.
Kilgard, M. P., & Merzenich, M. M. (1995). Anticipated stimuli
across skin. Nature, 373, 663.
Powell, T. P. S. (1977). The somatic sensory cortex. British Medical
Bulletin, 33, 129-136.
Sherrick, C. E., & Rogers, R. (1966). Apparent haptic movement. Per-
ception & Psychophysics, 1, 175-180.
NOTES
1. Although the use of equal ISIs separating T1, T2, and T3 is advis-
able in order to counteract any impact of the tau effect on perceptual re-
ports, this manipulation can inadvertently give rise to undesirable ap-
parent motion effects, especially when ISIs are short. This issue will be
further explored in Experiment 2.
2. We thank David Shore for suggesting this manipulation to control
for response bias effects.
3. Our thanks to an anonymous reviewer for making this important
point and for suggesting the two-alternative forced choice procedure as
a tool to find out whether subjective percepts are equivalent on tap and
rabbit trials.
4. It might be possible to develop an alternative attentional explana-
tion of the saltation illusion in terms of the involuntary (exogenous)
capture of attention by T3 at L1 or L3, even though attention is en-
dogenously focused at L2. However, such an account would clearly be
distinct from the explanation of saltation in terms of endogenously me-
diated states of attention and expectation proposed by Kilgard and
Merzenich (1995).
(Manuscript received October 22, 2003;
revision accepted for publication June 25, 2004.)
