Manipulating the Experienced Onset of Intention - 麻省理工学院人工智能研究专业

Manipulating the Experienced Onset of Intention
after Action Execution

Hakwan C. Lau1,2, Robert D. Rogers2, and Richard E. Passingham1,2

抽象的

& Using transcranial magnetic stimulation (TMS), 我们有
tested the time needed for the perceived onset of spontaneous
motor intention to be fully determined. We found that TMS
applied over the presupplementary motor area after the exe-
cution of a simple spontaneous action shifted the perceived
onset of the motor intention backward in time, and shifted
the perceived time of action execution forward in time. 这
size of the effect was similar regardless of whether TMS was
applied immediately after the action or 200 msec after. 这

results of three control studies suggest that this effect is time-
limited, specific to modality, and also specific to the anatom-
ical site of stimulation. We conclude that the perceived onset
of intention depends, at least in part, on neural activity that
takes place after the execution of action. A model, 这是
based on the mechanism of cue integration under the pres-
ence of noise, is offered to explain the results. The implica-
tions for the conscious control of spontaneous actions are
discussed. &

介绍

We experience a strong sense of conscious control when
generating spontaneous, or self-paced, 运动动作.
This experience has been challenged as ‘‘illusory’’: Wegner
(2002, 2003) has argued that although we perceive our
motor intentions to arise before the execution of actions,
we cannot confidently conclude that the former is caus-
ing the latter; there might be a mere temporal correla-
tion between the two events. 尽管如此, this argument
alone is insufficient to establish that the conscious will is
illusory, as it does not show that motor intentions are in
fact not causing the actions. One strong demonstration
for the case of illusory conscious control would be that
our perceived temporal order of intentions and actions
是, 实际上, 错误的. If intentions, 实际上, arise after actions,
they could not, 原则, be causing the actions.

Using a cross-modal timing method, Libet (1985) 和
Libet, Gleason, 赖特, and Pearl (1983) have reported
that participants start to experience their motor in-
tention at about 200 msec before making a spontane-
ous finger movement. 然而, critics have suggested
that the reported timings given by participants in the
Libet clock paradigm might not be accurate ( Joordens,
van Duijn, & Spalek, 2002; 克莱因, 2002; Trevena &
磨坊主, 2002; Gomes, 1998; Dennett & Kinsbourne,
1995; Dennett, 1991; Libet, 1985). In the context of the
debate of whether conscious intentions cause actions, A
critical possibility is that the reported onset of intentions
is determined by neural activity that takes place after
action execution. This has not been tested before.

1伦敦大学学院, 2牛津大学

Despite its counterintuitiveness, this possibility is
supported by psychophysical research on conscious
意识. Phenomena, such as backward masking
(Breitmeyer 1984) and postdiction (Alais & Burr, 2003;
Eagleman & Sejnowski, 2000), highlight the retrospec-
tive nature of perceptual experience, 那是, the inten-
sity and content of an experience can depend on
information that only becomes available after the sub-
jective time of perception. 尤其,
it has been
shown that the temporal extent of this retrospective
effect can last for as long as 200 毫秒, 正如所证明的
in the auditory modality (Alais & Burr, 2003).

We have therefore tested whether transcranial mag-
netic stimulation (TMS), applied after action execution,
has any effect on the reported onset of intention as
reported by the participants using the method of Libet
等人. (1983). Due to concerns about the absolute
accuracy of the reported timings, 然而, 我们只
assessed the shifts in the reported timings due to TMS,
but not the reported absolute measures. 还, 因为
TMS was only applied in half of the trials in a random
fashion together with sham TMS, participants could not
predict if they were going to be stimulated in a particular
trial until the action was executed.

We have previously performed a functional magnetic
resonance imaging (功能磁共振成像) experiment using Libet et al.’s
范例 (Lau, 罗杰斯, Haggard, & Passingham, 2004),
and we found that when participants were required to
estimate the onset of their intentions, activations were
found in areas that are known to be involved in motor
preparation and attention to action, which suggests that,
in this paradigm, the participants were in fact trying to

D 2007 麻省理工学院

认知神经科学杂志 19:1, PP. 81–90

D
哦
w
n
我
哦
A
d
e
d

F
r
哦
米

我

/
j

F
/

t
t

我
t
.

:
/
/

D
H
哦
t
w
t
p
n
:
哦
/
A
/
d
米
e
我
d
t
F
r
p
哦
r
米
C
.
H
s
我
p
我
v
d
e
我
r
r
e
C
C
H
t
.
米
A
我
r
e
.
d
C
你
哦
米
哦
/
C
n
j
哦
A
C
r
t
n
我
C
/
e
A
–
r
p
t
d
我
C
1
我
9
e
1
–
8
p
1
d
F
1
/
9
1
3
6
9
0
/
6
1
1
/
8
哦
1
C
/
n
1
2
7
0
5
0
6
7
4
1
1
9
4
/
1
j
8
哦
1
C
p
n
d
.
2
乙
0
y
0
G
7
你
.
e
1
s
9
t
.
哦
1
n
.
0
8
8
1
S
.
p
e
d
p
F
e
米
乙
乙
y
e
r
G
2
你
0
e
2
s
3
t

/
j

哦
n

1
8

中号
A
y

2
0
2
1

access information related to the generation of action. 在
addition to activations in the dorsal prefrontal and
parietal cortices that are commonly found in attention-
ally demanding tasks, we found activation in the pre-
supplementary motor area (pre-SMA), which we argued
is likely to reflect the representation of intention. 这
idea that the pre-SMA is particularly important for
spontaneous intention is also supported by previous
学习. 第一的, it has been reported that electrical stimu-
lation of the medial frontal cortex elicits the feeling of
an urge to move (Fried et al., 1991). 第二, 病变
of the medial frontal cortex abolish self-initiated move-
ments in macaque monkeys (Thaler, 陈, Nixon, Stern,
& Passingham, 1995). 最后, we have previously re-
ported activity in the pre-SMA when participants gener-
ate actions of their own free choice (Lau, 罗杰斯, &
Passingham, 2006). The pre-SMA was therefore chosen
as the main site of interest in this study (数字 2). 全部
stimulations conducted in Experiments 1, 2, 和 3 是
targeted at this anatomical region. In Experiment 1, 我们
found that TMS after action execution induces shifts in
the perceived onset of both intentions and movements.
实验 2, 3, 和 4 were set up to rule out alter-
native interpretations.

EXPERIMENT 1: MAIN EXPERIMENT

方法

Six male and four female healthy participants were
tested in this experiment. The task instructions were
explained to them verbally before they received safety
screening and gave informed consent.

The psychological tasks were based on Libet et al.’s
(1983) clock paradigm (数字 1), and the detailed
procedures were similar to those used in a previous

实验 (Lau, 罗杰斯, Haggard, 等人。, 2004). Partic-
ipants rested their head on a chin rest, and in every trial,
they watched a red dot revolving around a clock face
(diameter (西德:1) 38) on a computer screen placed at about
50 cm away from the chin rest. There was a fixation cross
presented in the center of the clock face, and partic-
ipants were required to maintain fixation while the
red dot was moving. The dot revolved at a speed of
2560 msec per cycle, and after the first revolution in
每次试验, the participants were required to press a
computer mouse button using their left index finger,
at a random time point of their own choice. After the
button press, the dot kept on moving for a period of
1280–2560 msec. There was then a delay of 2000 毫秒,
after which the red dot reappeared at the center of the
clock face, and participants were required to control the
dot as a cursor using another computer mouse held in
their right hand.

In the intention condition, they were required to
move the cursor to where the dot was when they first
felt their intention to press the button. In the movement
状况, they were required to move the cursor to
where the dot was when they actually pressed the
button with their left index finger. After they selected
the location, they clicked the mouse button with their
right hand to finish the trial. The next trial began after an
interval of 1000 毫秒. In half of the trials, TMS was
applied over the pre-SMA. In the remaining trials, 那里
was a sham TMS triggered by another TMS machine with
a coil placed near the back of the head of the partic-
爱普茨, but directed away from the cortex. The stimula-
的 (real or sham) was either presented immediately
after action execution or 200 msec afterwards. 那里
were a total of 240 试验. The two main task conditions
were organized into eight 30-trial blocks. Real and sham
TMS were randomly allocated, and so were early and late

数字 1. Libet’s clock
范例. Participants made a
spontaneous finger movement
while watching a red dot
revolving around a clock face
(左边). After the action, the dot
kept on going for 1/2–1 cycle
before it disappeared. 然后
there was a delay, 后
which the dot reappeared
at the center of the clock.
Participants used a computer
mouse to control the dot
as a cursor and moved it to
the location where the dot
was when they first felt the
intention to move in the
intention condition, 和
to the location where the
dot was when they actually
made the movement in the
movement condition.

认知神经科学杂志

体积 19, 数字 1

D
哦
w
n
我
哦
A
d
e
d

F
r
哦
米

我

/
j

t
t

F
/

我
t
.

:
/
/

/
j

哦
n

1
8

中号
A
y

2
0
2
1

near the midline. The sham TMS pulses were triggered
by a similar TMS machine, and the intensity was adjusted
individually for each participant to a level that produced
an auditory ‘‘click’’ sound of a volume that was reported
to be similar to that of the real TMS.

实验 2, 3, 和 4 were set up, after obtaining
the results in this experiment, to exclude alternative
解释. 所以, the procedures were similar
to those used in this experiment.

结果

When only sham TMS trials were considered, the group
mean for the judged onset of intention was (西德:2)148 毫秒
(标准差= 103 毫秒) relative to the time of the recorded
button press. The group mean for the judgment of
movement was (西德:2)50 毫秒 (标准差= 42 毫秒). As reported
in previous studies (Lau, 罗杰斯, Haggard, 等人。, 2004;
Sirigu et al., 2004; Haggard & Eimer, 1999), the two
measures differed significantly [one-tailed t test was
used because the test was motivated by evidence ob-
served in previous studies, t(9) = 2.899, p = .009].

The effect of TMS for each individual was assessed by
subtracting the median judgment value for the sham
TMS trials from that of the TMS trials. The group mean
for the TMS effect was (西德:2)9 毫秒 (标准差= 32 毫秒) 为了
the intention condition with 0 msec delayed TMS,
(西德:2)16 毫秒 (标准差= 32 毫秒) for the intention condition
和 200 msec delayed TMS, 14 毫秒 (标准差= 27 毫秒) 为了
the movement condition with 0 msec delayed TMS, 和
9 毫秒 (标准差= 16 毫秒) for the movement condition.
These data are plotted in Figure 3. They were entered
into an analysis of variance (ANOVA) with task (Inten-
tion vs. Movement) and time (0 msec or 200 毫秒)
considered as experimental factors, and it was found
that task was a significant factor [F(1,9) = 24.089,
p = .001] but time was not [F(1,9) = 0.485, p = .504).
There was no significant interaction between time and
任务 [F(1,9) = 0.009, p = .926].

When the data for the different time points of TMS
were considered together, it could be shown that the
effect of TMS on the judgment of intention is signifi-
cantly negative to zero[two-tailed t test, t(9) = (西德:2)2.6,
p = .029] and the effect on the judgment of movement
是, although weaker, significantly positive to zero [二-
tailed t test, t(9) = 2.35, p = .045].

EXPERIMENT 2: TIME SPECIFICITY

方法

Seven male and three female healthy participants were
tested in this experiment where the effects of TMS
applied at other time points were evaluated. The meth-
odology and procedures were similar to those used in
实验 1, except that the time points for TMS were
不同的. Instead of applying a delay of either 0 毫秒

Lau, 罗杰斯, and Passingham

数字 2. The pre-SMA. This area was the target of stimulation in
实验 1 到 3. The coordinates (X, y, z = 2, 4, 54) were obtained
from a previous fMRI study (Lau, 罗杰斯, Haggard, 等人。, 2004), 和
figure is also adapted from the report of that study. The VCA line
vertically passes through the anterior commissure, and is the border
between the pre-SMA and the SMA proper.

stimulations. 所以, from the point of view of the
参与者, they did not know whether they would be
stimulated, or at which point would the pulse be
triggered, until after they have pressed the button.

The anatomical site of stimulation was the pre-SMA
(数字 2). The localization for stimulation here was
based on the coordinates obtained from the imaging
study described in the previous section (X, y, z, = 2, 4,
54). The relationship between the brain of each indi-
vidual and the standard Talairach space was computed
by applying spatial normalization (FLIRT version 5.0,
www.fmrib.ox.ac.uk/fsl/flirt/index.html) to the previous-
ly acquired MRI scans of each participant, using the
Montreal Neurological Institute canonical single-subject,
high-resolution MRI image as the template as in the
previous fMRI study. Using the Brainsight Frameless
系统 (version 1.5B3, www.rogue-research.com/), 这
location of the pre-SMA was then marked on each in-
dividual’s MRI scan, which was then registered with
the actual brain in 3-D space. Because the location
of the intended site of stimulation is quite far away from
the surface of the scalp, a double cone coil (Magstim
公司, Whitland, South West Wales, 英国) was used in
this experiment. The hotspot of the coil was placed
directly above the marked position for stimulation on
the MRI scan as presented on a computer screen by the
Brainsight software. The stimulation pulses were trig-
gered by a Magstim Rapid Rate TMS machine (Magstim
公司), and the intensity of stimulation was set at 5%
above the active motor threshold for a noticeable foot
twitch, tested over the foot area of the motor cortex

D
哦
w
n
我
哦
A
d
e
d

F
r
哦
米

我

/
j

t
t

F
/

我
t
.

:
/
/

/
j

哦
n

1
8

中号
A
y

2
0
2
1

D
哦
w
n
我
哦
A
d
e
d

F
r
哦
米

我

/
j

t
t

F
/

我
t
.

:
/
/

/
j

哦
n

1
8

中号
A
y

2
0
2
1

数字 3. The effects of TMS on the perceived times. TMS produced a forward shift in the reported onset of intention and a backward shift
in the reported onset of movement, regardless of whether TMS was administered immediately after the action (button press) 或者 200 毫秒
afterwards (实验 1, top left). This effect of exaggerating the difference between the two reported onsets was significant ( p < .001, main effect of task in the ANOVA; see also Figure 4). Experiments 2, 3, and 4 were control studies set up to rule out alternative explanations. In Experiment 2 (top right), TMS was administered either 500 msec after the action or immediately before the subjects were required to report the experienced onsets. In Experiment 3 (bottom left), the subjects were required to judge the onset and offset of a slowly ramping-up tactile stimuli (see Figure 5), instead of the onsets of intention and movements; they were not required to make a spontaneous action. Experiment 4 was identical to Experiment 1, except that the motor cortex, instead of the pre-SMA, was targeted in TMS. None of the control studies showed any significant result in the statistical analyses, and the patterns of the effects were clearly different from those observed in Experiment 1. The error bars represent standard errors across participants. or 200 msec, the delay in this experiment was either 500 msec, or between 3280 and 4560 msec, so that the TMS pulse (shame or real) was triggered at the point when the cursor appeared at the middle of the clock face, prompting the participants to report the estimated tim- ings. These times were chosen to test whether the effect obtained in Experiment 1 was actually due to memory or responding, rather than the experienced onset itself. Results When only the sham TMS trials were considered, on average, the participants judged the onset of intention to be (cid:2)110 msec (SD = 82 msec) relative to the times for recorded button press. The average for the judgment of movement was (cid:2)7 msec (SD = 69 msec). The two measures differed significantly [one-tailed t test was used as in Experiment 1 because the test was motivated by evidence observed in previous studies, t(9) = 4.964, p = .0005]. The group mean for the TMS effect was 9 msec (SD = 28 msec) for the intention condition with 500 msec delayed TMS, 0 msec (SD = 26 msec) for the inten- tion condition with 3280–4560 msec delayed TMS, 5 msec (SD = 21 msec) for the movement condition with 500 msec delayed TMS, and 5 msec (SD = 32 msec) for the movement condition with 3280–4560 msec de- layed TMS. These data are plotted in Figure 3. As in Experiment 1, the effects of TMS for each task condition were entered into an ANOVA with task (In- tention vs. Movement) and time (500 msec or 3280– 4560 msec) considered as experimental factors, and it was found that neither task nor time was a significant factor [F(1,9) = 0.001, p = .974 and F(1,9) = 0.208, p = .659, respectively]. There was also no significant interaction between time and task [F(1,9) = 0.549, p = .478]. The main effect of task was of special interest as this was found to be significant in Experiment 1. However, from the p value it is clear that this effect would not have been significant even if a one-tailed t test had been applied. The size of this effect (i.e., the ef- fect of TMS for intention minus the effect of TMS for movement) is plotted in Figure 4 to allow a comparison across all experiments reported here. Also, the effect of 84 Journal of Cognitive Neuroscience Volume 19, Number 1 up that is characteristic of the electrophysiological measures of the readiness potential (Deecke, Scheid, & Kornhuber, 1969; Kornhuber & Deecke, 1965). In- stead of estimating the onset of intention and move- ment, participants were required to estimate the timing of either the onset or the peak of the tactile stimulus using the Libet clock. A small tactile stimulator (100-(cid:1) bone conduction vibrators; Oticon, Hamilton, Scotland) was connected to the audio output from the sound card of a PC, where a wave file containing white noise was played, to give the tactile sensation. The intensity of the stimulation was set at a low level, only slightly beyond detection threshold for each participant, such that the earliest part of the stimulation would not be felt. The tactile stimulator and the left index finger of the partic- ipants were inserted into a piece-insulating sponge so as to minimize the auditory noise associated with the tactile stimulus. TMS was triggered at either 0 msec or 200 msec after the offset of the tactile stimulus. Results When only the sham TMS trials were considered, on average, the participants judged the onset of the tactile stimulus to be (cid:2)107 msec (SD = 95 msec) relative to the offset of the stimulus. The average for the judgment of the peak of the stimulus was 26 msec (SD = 92 msec) relative to the actual offset of the stimulus. The two measures differed significantly [one-tailed t test used as in Experiment 1, t(9) = 4.764, p = .0005], and the magnitude of this difference was similar to the difference between the intention judgment and movement judg- ment in Experiment 1 (see Figure 5). The group mean for the TMS effect was (cid:2)18 msec (SD = 46 msec) for the onset timing condition with 0 msec delayed TMS, (cid:2)7 msec (SD = 41 msec) for the onset timing condition with 200 msec delayed TMS, (cid:2)17 msec (SD = 31 msec) for the peak timing condition with 0 msec delayed TMS, and 2 msec (SD = 22 msec) for the peak timing condition with 200 msec delayed TMS. These data are plotted in Figure 3. As in Experiment 1, the effects of TMS for each task condition were entered into an ANOVA with task (On- set Timing vs. Offset Timing) and time (0 msec or 200 msec) considered as experimental factors, and it was found that neither task nor time was a significant factor [F(1,9) = 0.222, p = .649 and F(1,9) = 1.716, p = .223, respectively]. There was also no significant interaction between the two factors [F(1,9) = 0.159, p = .699]. The main effect of task was of special inter- est as this was found to be significant in Experiment 1. However, from the p value it is clear that this effect would not have been significant even if a one-tailed t test was applied. The size of this effect (the effect of TMS for onset timing minus the effect of TMS for peak timing) is plotted in Figure 4 to allow a comparison across all experiments reported here. Lau, Rogers, and Passingham 85 Figure 4. The exaggeration of the difference between the two reported onsets as induced by TMS. Plotted on the vertical scale is the size of the main effect of task, that is, the difference between the effects of TMS on the reported onset of intention and the reported onset of movement (for Experiments 1, 2, and 4) or between the effects of TMS on the reported onset of the tactile stimulus and the reported offset of the stimulus (for Experiment 3). This effect was clearly not present for the control experiments, Experiments 2, 3, and 4 ( p = .974, .649, and .493, respectively). The error bars represent standard errors across participants. TMS (TMS vs. sham) was not significant on either the intention task [two-tailed t test, t(9) = 0.720, p = .435] or the movement task alone [two-tailed t test, t(9) = 0.818, p = .490]. The effect of task was also assessed when the trials for the different time points of TMS were considered separately. One-tailed t test was used based on the results obtained in Experiment 1 to maximize statistical power. The effect of task was not significant for either the 500 msec trials [direction of effect opposite to prediction based on Experiment 1, one-tailed t test, t(9) = (cid:2)0.694, p = ns] or the 3280–4560 msec trials [one-tailed t test, t(9) = 0.331, p = .374]. EXPERIMENT 3: MODALITY SPECIFICITY Methods Eight male and two female healthy participants were tested in this experiment, which was set up to test if the effect in Experiment 1 was actually due to the general mechanism of cross-modal timing using the clock face. In particular, we wanted to know whether TMS simply exaggerates the perceived temporal difference between early and late events. This was tested by presenting a tactile stimulus to the subjects and requiring them to judge the timing of this stimulus using the clock. The methodology and procedures were similar to those used in Experiment 1. Instead of requiring the participants to make a self-paced left index finger movement, a slowly ramping up tactile stimulus (duration = 600 msec) was applied to the tip of this finger at a random time point (2960–7980 msec) in every trial. The tactile stimulus used here had the same property of slowly ramping D o w n l o a d e d f r o m l l / / / / / j t t f / i t . : / / D h o t w t p n : o / a / d m e i d t f r p o r m c . h s i p l v d e i r r e c c h t . m a i r e . d c u o m o / c n j o a c r t n i c / e a - r p t d i c 1 l 9 e 1 - 8 p 1 d f 1 / 9 1 3 6 9 0 / 6 1 1 / 8 o 1 c / n 1 2 7 0 5 0 6 7 4 1 1 9 4 / 1 j 8 o 1 c p n d . 2 b 0 y 0 g 7 u . e 1 s 9 t . o 1 n . 0 8 8 1 S . p e d p f e m b b y e r g 2 u 0 e 2 s 3 t / j . t . . f . . o n 1 8 M a y 2 0 2 1 D o w n l o a d e d f r o m l l / / / / / j f / t t i t . : / / D h o t w t p n : o / a / d m e i d t f r p o r m c . h s i p l v d e i r r e c c h t . m a i r e . d c u o m o / c n j o a c r t n i c / e a - r p t d i c 1 l 9 e 1 - 8 p 1 d f 1 / 9 1 3 6 9 0 / 6 1 1 / 8 o 1 c / n 1 2 7 0 5 0 6 7 4 1 1 9 4 / 1 j 8 o 1 c p n d . 2 b 0 y 0 g 7 u . e 1 s 9 t . o 1 n . 0 8 8 1 S . p e d p f e m b b y e r g 2 u 0 e 2 s 3 t / j . . . . . t f o n 1 8 M a y 2 0 2 1 Figure 5. The temporal profile of the tactile stimulus used in Experiment 3. The tactile stimulus used in Experiment 3 took a slowly ramping-up form (left), and was presented at a low intensity such that the early part of the stimulus was not felt. This was set up to mimic the form of the readiness potential preceding spontaneous actions (Deecke et al., 1969; Kornhuber & Deecke, 1965). The magnitude of the difference between the perceived onset of the tactile stimulus and the perceived time of the peak of the stimulus in Experiment 3 was very similar to the magnitude of the difference between the perceived onset of intention and the perceived time of movement in Experiment 1 (right). Also, the effect of TMS (TMS vs. sham) was not sig- nificant on either the onset judgments [two-tailed t test, t(9) = (cid:2)1.197, p = .262] or the offset judgments alone [two-tailed t test, t(9) = (cid:2)1.185, p = .266]. Despite the fact that task was not found to be a sig- nificant factor for the effect of TMS in this experiment, the magnitude of the effect of TMS on the onset judg- ments was somewhat similar to that of the effect of TMS on intention in Experiment 1. However, this nonsignif- icant effect was not specific to onset judgment alone but was also there in offset judgments, which means that, unlike in Experiment 1, this effect was not specific to task. It is likely that because TMS produces a tactile sensation to the scalp, it produced a nonspecific inter- ference with the tactile judgments in general. That would fit with the fact that this effect was stronger for the TMS applied closer to the stimuli. In any case, it is clear that TMS did not exaggerate the perceived tempo- ral difference between early and late events as observed in Experiment 1. EXPERIMENT 4: ANATOMICAL SPECIFICITY Methods Seven male and three female healthy participants were tested in this experiment in which we try to assess if the effect obtained in the experiment was simply because TMS added noise to the motor system. The methodol- ogy and procedures were similar to those used in Experiment 1, except that the site of TMS was the right primary motor cortex instead of the pre-SMA. Because of the ease of accessibility of the motor cortex as shown in previous studies, a normal figure-of-eight shaped coil (Double 70mm Coil, Magstim, www.magstim.com/ Products.html) was used. The localization of the primary motor cortex was done by checking the intensity of TMS-induced motor twitch for the first dorsal interos- seus in the right hand of the participants while moving the coil around the expected region of the scalp. Once the most sensitive spot was identified, the threshold for stimulation was set at 5% above the active motor threshold for a noticeable hand twitch. Results When only the sham TMS trials were considered, the group mean for the judged onset of intention was (cid:2)99 msec (SD = 79 msec) relative to the time of the recorded button press. The group mean for the judg- ment of movement was (cid:2)13 msec (SD = 76 msec). The two measures differed significantly [one-tailed t test as in Experiment 1 because the test was motivated by evidence observed in previous studies, t(9) = 3.733, p = .002]. The group mean for the TMS effect was (cid:2)8 msec (SD = 30 msec) for the intention condition with 0 msec delayed TMS, 16 msec (SD = 28 msec) for the intention condition with 200 msec delayed TMS, (cid:2)4 msec (SD = 86 Journal of Cognitive Neuroscience Volume 19, Number 1 23 msec) the movement condition with 0 msec delayed TMS, and (cid:2)2 msec (SD = 26 msec) for the movement condition with 200 msec delayed TMS. These data are plotted in Figure 3. As in Experiment 1, the effects of TMS for each task condition were entered into an ANOVA with task (Inten- tion vs. Movement) and time (0 msec or 200 msec) con- sidered as experimental factors, and it was found that neither task nor time was a significant factor [F(1,9) = 0.510, p = .493 and F(1,9) = 2.418, p = .154, respective- ly]. There was also no significant interaction between time and task [F(1,9) = 2.052, p = .186]. The main effect of task was of special interest as this was found to be significant in Experiment 1. However, from the p value it is clear that this effect would not have been signifi- cant even if a one-tailed t test was applied. The size of this effect (i.e., the effect of TMS for intention minus the ef- fect of TMS for movement) is plotted in Figure 4 to allow a comparison across all experiments reported here. Also, the effect of TMS (TMS vs. sham) was not significant on either the intention task [two-tailed t test, t(9) = 0.686, p = .510] or the movement task alone [two-tailed t test, t(9) = (cid:2)0.456, p = .659]. DISCUSSION Summary of Findings To summarize the results, Experiment 1 showed that there was a retrospective effect for TMS over the pre- SMA on the perceived onset of intention as well as for perceived timing of the movement itself. TMS shifted the perceived onset of intention backward in time and shifted the perceived timing of the movement forward in time. The main effect of task, that is, the differential effect on intention and movement, is about 24 msec. This difference is not driven by the TMS effect on the movement condition alone, as the effect of TMS in the intention condition is also significant on its own. This differential effect was found for both stimulations ap- plied at 0 msec and 200 msec after action execution. One could argue that the TMS effect we have observed was not genuinely retrospective, but rather, it simply af- fected the memory of timing, or the reporting of the timing, which took place after the TMS. However, Exper- iment 2 showed that the effect observed in Experiment 1, that is, the exaggeration of the difference of the judg- ments for the onsets of intention and movement, was not found when TMS was applied at 500 msec after the action or right before the participants made their reports of their time estimates. One could argue that this only shows that there is a critical time window within which memory is susceptible to manipulation. However, this interpretation would require the ad hoc assumption that an experience could be first generated and then misre- membered as something else, unbeknownst to the sub- ject. If this assumption is allowed, one could as well argue against the existence of well-established phenom- ena, such as backward masking (Breitmeyer, 1984) and flash-lag (Nijhawan, 1994, 2002; Eagleman & Sejnowski, 2000), by reinterpreting them as results of failure or change of memory. Although philosophers have pointed out that we cannot unequivocally reject these alternative interpretations (Dennett & Kinsbourne, 1995; Dennett, 1991), one generally prefers parsimonious explanations that do not require ad hoc assumptions. Neither did we observe the same effect when we tested this for the tactile modality, as shown in Experiment 3 when the same timing method was used. This suggests that TMS did not simply affect the visual processing re- quired to make accurate judgments in the Libet task. In particular, TMS did not simply exaggerate the perceived temporal difference between early and late events. The tactile task in Experiment 3 was similar in structure to the task in Experiment 1. When only the control sham TMS trials where considered, the differences between the perceived onset of the tactile stimulus and the perceived time for its peak were very similar to the differences between the perceived onset of intention and perceived time of movement (see Figure 4). Yet, no significant effect was observed in Experiment 3. The targeted anatomical site of stimulation of Experi- ments 1 to 3 was the pre-SMA. This anatomical localiza- tion was based on the fMRI results of our previous study (Lau, Rogers, Haggard, et al., 2004), and each individual participant’s structural MRI scan. However, we could not be sure that only the pre-SMA was stimulated, as it was not clear if TMS afforded such high spatial specificity, especially given the potential spread of the effect and interaction between different areas. Also, the lateral premotor areas might have been stimulated by the magnetic fields generated by the wings of the double- cone coil used in this experiment, too. Nonetheless, the conclusion of the study does not depend on the precision of the stimulation, as the main question is about whether the experience of intention is fully determined before action execution, but not about where it is represented in the brain. Experiment 4 further showed that the effect observed in Experiment 1 is, in fact, reasonably specific, in that even when another area in the motor system (the primary motor cortex) is stimulated, the same effect is not observed. This suggests that the effect was not simply due to the fact that TMS added noise to the motor system or that participants were simply startled. Effect Size and Reliability of the Data One could argue that the temporal shifts produced by TMS are small, and thus, unimpressive. Due to consid- erations about the safety and comfort of the partici- pants, we did not apply the stimulations at a higher intensity or in a rapid-rate, repetitive fashion. This might be one reason why the size of the effect did not appear to be very large. However, the size of the temporal shifts Lau, Rogers, and Passingham 87 D o w n l o a d e d f r o m l l / / / / / j t t f / i t . : / / D h o t w t p n : o / a / d m e i d t f r p o r m c . h s i p l v d e i r r e c c h t . m a i r e . d c u o m o / c n j o a c r t n i c / e a - r p t d i c 1 l 9 e 1 - 8 p 1 d f 1 / 9 1 3 6 9 0 / 6 1 1 / 8 o 1 c / n 1 2 7 0 5 0 6 7 4 1 1 9 4 / 1 j 8 o 1 c p n d . 2 b 0 y 0 g 7 u . e 1 s 9 t . o 1 n . 0 8 8 1 S . p e d p f e m b b y e r g 2 u 0 e 2 s 3 t / j . . t . f . . o n 1 8 M a y 2 0 2 1 cannot be directly compared with the results obtained in most other cognitive TMS experiments ( Walsh & Pascual-Leone, 2003; Stewart, Ellison, Walsh, & Cowey, 2001), as the common measures in previous experi- ments are normally a change in reaction times or error rates, which are very different from subjective onset estimates. One study (Haggard, Clark, & Kalogeras, 2002), which used the same paradigm to investigate the retrospective effect of external stimuli on the per- ceived onset of movement (but not intention), re- vealed a temporal shift of only 15 msec. Another study (Haggard & Magno, 1999) used the same paradigm to investigate the prospective effect of TMS on the per- ceived onset of movement (but not intention), and one of the targeted areas of stimulation was close to the pre- SMA (FCz). It was found that TMS applied to this area prior to action execution can only shift the perceived onset of movement by 54 msec, even when the actual shift of movement execution was as large as 113 msec. Taken together, it seems that experimentally induced temporal shifts as measured by the Libet method are, in general, not very large. Nonetheless, this does not mean that the measures are uninformative, as the principled way to assess the significance of an effect and compare it with other studies that use different measures is to characterize it in terms of the variance of the data (i.e., to look at the statistical effect size). The eta-squared value, which is a standard estimate of statistical effect size, is .728 for the main effect of task in Experiment 1; the corresponding p value is as small as .001. This is a relatively high level of significance as far as cognitive TMS experiments are concerned. Despite this high level of statistical significance, and the fact that each experiment involved 10 subjects (even for each control studies), there might still be concerns about the reliability of the presented data. In particular, the baseline values of the reported onsets (i.e., values for sham TMS trials) fluctuated across Experiment 1 and the control studies. This could be due to the fact that some of the control studies differed from Experiment 1 in various aspects: Experiment 2 presented sham TMS at later time points; Experiment 3 required subjects to judge the timing of a tactile stimulus instead of a self- generated action; Experiment 4 presented the sham TMS using a different coil in order to match for the sound of the TMS administered in that experiment. These suggest that subtle factors, such as the auditory level and the timing of the sham TMS, could, in fact, affect the reported onsets in this difficult timing task. Also, the variability of the effect of TMS was also a source of concern. Although the statistical analyses revealed no significant effect in all control studies, there were sometimes weak trends observed in unpredicted directions, as shown in Figure 3. We emphasize that TMS exaggerated the difference between the reported onsets of intention and movement in a robust fashion, and this effect was clearly not present in the control studies, as shown in Figure 4. This is the main effect of interest in the article. Nonetheless, one major limitation could be that the control studies involved different experimental sessions and did not always involve the same subjects. Future studies would benefit from replicating these results in the same sessions by, for instance, testing more different time points of TMS in each session, thereby combining and replicating the results of Experi- ments 1 and 2, and possibly exploring other critical time points such as those before action execution as well. Direction of Temporal Shifts In Experiment 1, it was found that retrospective TMS shifted the perceived onset of intention backward in time, and shifted the perceived timing of action execu- tion forward in time. The directions of these effects were not predicted a priori. Here we offer a tentative model to account for this pattern of temporal shifts, which we hope could be tested in future experiments. When performing difficult perceptual tasks, we often try to combine information from different sources, such as cues from different sensory modalities. A natural and optimal method to combine these different sources of information is to perform Bayesian cue integration (Knill & Pouget, 2004). Assuming that the noise or uncertainty in the information can be characterized as Gaussian in form, this method amounts to taking a weighted average of the information from the different sources, where the weights are inversely proportional to the amount of noise in each respective source. The readiness potential (Deecke et al., 1969; Kornhuber & Deecke, 1965), measurable by EEG from the scalp, reflects the slowly ramping-up neural activity that pre- cedes spontaneous movements. It has been argued that the readiness potential comprises different components that are functionally distinct (Deecke, 1987). The earliest onset of the readiness potential is around 1 to 1.5 sec. Despite the lack of an uncontroversial method to mea- sure accurately the onset of the experience of intention, it seems plausible that the experienced onset is not as early as the onset of the readiness potential. One possibility is that due to the difficult nature of the task to estimate the onset of intention, the brain combines the information reflected by both the early and the late components of the readiness potential. The late compo- nents of the readiness potential might indicate that the onset of intention is much closer to the time of action execution, and because the signal of the early compo- nent of the readiness potential is weak, the weighting for the representation reflected by early component is likely to be low. That might explain why we typically experience the onset of the intention as much later than the onset of the readiness potential. It has been suggested that the best way to character- ize the effect of TMS is to view it as adding neural noise to the targeted cortical area ( Walsh & Pascual-Leone, 88 Journal of Cognitive Neuroscience Volume 19, Number 1 D o w n l o a d e d f r o m l l / / / / / j f / t t i t . : / / D h o t w t p n : o / a / d m e i d t f r p o r m c . h s i p l v d e i r r e c c h t . m a i r e . d c u o m o / c n j o a c r t n i c / e a - r p t d i c 1 l 9 e 1 - 8 p 1 d f 1 / 9 1 3 6 9 0 / 6 1 1 / 8 o 1 c / n 1 2 7 0 5 0 6 7 4 1 1 9 4 / 1 j 8 o 1 c p n d . 2 b 0 y 0 g 7 u . e 1 s 9 t . o 1 n . 0 8 8 1 S . p e d p f e m b b y e r g 2 u 0 e 2 s 3 t / j t . . f . . . o n 1 8 M a y 2 0 2 1 2003). The pre-SMA is likely to be one of the sources of the readiness potential (Cunnington, Windischberger, Deecke, & Moser, 2002). Because TMS is applied after action execution in the experiments described in this article, it is reasonable to assume that noise is mostly added to the late components of the readiness potential, but not the early ones. If the brain combines the timing information of intention in a way that is similar to optimal Bayesian cue integration, increasing the uncer- tainty of the representations reflected by the late com- ponents of the readiness potential would decrease the weight assign to these representations, and thus, in- crease the relative weight for the early components. The result would be that we estimate the onset of intention to be earlier. Similarly, when estimating the timing of action execu- tion, the brain might combine the information from proprioceptive sensory feedback as well as from the representations reflected by the late components of the readiness potential. The late components of the readiness potential may indicate an earlier time of action execution than does the sensory feedback, as the former is a preparatory signal. According to the model, adding noise to the late component would lead to a forward temporal shift of the resulted estimation, as the relative weight for the sensory feedback is increased. To explain the present data, the model need not be fully Bayesian in nature, in the sense that it explicitly represents the probability distribution functions for the different cues. Neither does it require the assumption that noise has a Gaussian structure or is normally dis- tributed. The foregoing explanation could be derived so long as the system weighs the different cues accord- ing to the noise associated with each cue, so that the higher the noise, the smaller the weight would be ascribed to that cue. Conclusions The main conclusion of this article does not depend on the aforementioned model. Regardless of what mecha- nism is used in the brain, the results suggest that the perceived onset of intention depends on neural activity that can be manipulated by TMS at as late as 200 msec after the execution of a spontaneous action. The mech- anism for this retrospective effect is unclear. One possi- bility is that TMS over the medial frontal area interferes with feedback processes that confirm the execution of the action. The present data also do not determine whether neural activity that takes place before the execution of the action is sufficient to yield some form of experience of intention. It could be the case that some weaker form of experience of intention is suffi- ciently determined by neural activity that takes place before the execution of the action, and such experience might have some causal impact on the control of the action. One way to examine this possibility would be to apply TMS before the execution of action, as suggested above, and compare the effect with those obtained in this study. Although the onsets of spontaneous actions cannot be easily predicted, TMS could be delivered at random time points. Given sufficient number of trials, one could identify, after the experiment, the trials where TMS are delivered just before action onsets, and perform the analysis on these trials. Before this experiment is conducted, however, one cannot draw the strong con- clusion that the experience of having conscious control of a simple motor action is entirely illusory. Nonetheless, the current results throw doubt on the commonsensical view that the experience of intention, including the experienced onset, is completely deter- mined before an action. The commonsensical view is attractive when we assume that the main function of experience of intention is for the conscious control of action, but it cannot account for the data presented here. The data suggest that the perceived onset of inten- tion depends at least in part on neural activity that takes place after the execution of action, which could not, in principle, have any causal impact on the action itself. An alternative view that is compatible with the data is that one function of the experience of intention might be to help clarify the ownership of actions ( Wegner, 2002, 2003), which can help to guide future actions. This process could take place immediately after action execution. Acknowledgments This work was supported by the Wellcome Trust (R. E. P.) and a Rhodes Scholarship (H. C. L.). We thank Jon Driver for his critical comments on part of the results represented. Reprint requests should be sent to Hakwan Lau, Functional Im- aging Laboratory, 12 Queen Square, London, WC1N 3BG UK, or via e-mail: h.lau@fil.ion.ucl.ac.uk. REFERENCES Alais, D., & Burr, D. (2003). The ‘‘Flash-Lag’’ effect occurs in audition and cross-modally. Current Biology, 13, 59–63. Breitmeyer, B. G. (1984). Visual masking: An integrative approach. Oxford: Oxford University Press. Cunnington, R., Windischberger, C., Deecke, L., & Moser, E. (2002). The preparation and execution of self-initiated and externally-triggered movement: A study of event-related fMRI. Neuroimage, 15, 373–385. Deecke, L. (1987). Bereitschaftspotential as an indicator of movement preparation in supplementary motor area and motor cortex. Ciba Foundation Symposium, 132, 231–250. Deecke, L., Scheid, P., & Kornhuber, H. H. (1969). Distribution of readiness potential, pre-motion positivity, and motor potential of the human cerebral cortex preceding voluntary finger movements. Experimental Brain Research, 7, 158–168. Dennett, D. C. (1991). Consciousness explained (1st ed.). Boston: Little Brown and Co. Dennett, D. C., & Kinsbourne, M. (1995). Time and the Lau, Rogers, and Passingham 89 D o w n l o a d e d f r o m l l / / / / / j t t f / i t . : / / D h o t w t p n : o / a / d m e i d t f r p o r m c . h s i p l v d e i r r e c c h t . m a i r e . d c u o m o / c n j o a c r t n i c / e a - r p t d i c 1 l 9 e 1 - 8 p 1 d f 1 / 9 1 3 6 9 0 / 6 1 1 / 8 o 1 c / n 1 2 7 0 5 0 6 7 4 1 1 9 4 / 1 j 8 o 1 c p n d . 2 b 0 y 0 g 7 u . e 1 s 9 t . o 1 n . 0 8 8 1 S . p e d p f e m b b y e r g 2 u 0 e 2 s 3 t / j . . . f t . . o n 1 8 M a y 2 0 2 1 observer: The where and when of consciousness in the brain. Behavioral and Brain Sciences, 15, 183–247. Eagleman, D. M., & Sejnowski, T. J. (2000). Motion integration and postdiction in visual awareness. Science, 287, 2036–2038. Fried, I., Katz, A., McCarthy, G., Sass, K. J., Williamson, P., Spencer, S. S., et al. (1991). Functional organization of human supplementary motor cortex studied by electrical stimulation. Journal of Neuroscience, 11, 3656–3666. Gomes, G. (1998). The timing of conscious experience: A critical review and reinterpretation of Libet’s research. Consciousness and Cognition, 7, 559–595. Haggard, P., Clark, S., & Kalogeras, J. (2002). Voluntary action and conscious awareness. Nature Neuroscience, 5, 382–385. Haggard, P., & Eimer, M. (1999). On the relation between brain potentials and the awareness of voluntary movements. Experimental Brain Research, 126, 128–133. Haggard, P., & Magno, E. (1999). Localising awareness of action with transcranial magnetic stimulation. Experimental Brain Research, 127, 102–107. Joordens, S., van Duijn, M., & Spalek, T. M. (2002). When timing the mind one should also mind the timing: Biases in the measurement of voluntary actions. Consciousness and Cognition, 11, 231–240. Klein, S. (2002). Libet’s research on the timing of conscious intention to act: A commentary. Consciousness and Cognition, 11, 273–279. Knill, D. C., & Pouget, A. (2004). The Bayesian brain: The role of uncertainty in neural coding and computation. Trends in Neurosciences, 27, 712–719. Kornhuber, H. H., & Deecke, L. (1965). Changes in the brain potential in voluntary movements and passive movements in man: Readiness potential and reafferent potentials. Pflu¨gers Archiv fu¨r die Gesamte Physiologie des Menschen und der Tiere, 284, 1–17. Lau, H. C., Rogers, R. D., Haggard, P., & Passingham, R. E. (2004). Attention to intention. Science, 303, 1208–1210. Lau, H., Rogers, R. D., & Passingham, R. E. (2006). Dissociating response selection and conflict in the medial frontal surface. Neuroimage, 29, 446–451. Libet, B. (1985). Unconscious cerebral initiative and the role of conscious will in voluntary action. Behavioral and Brain Sciences, 8, 529–566. Libet, B., Gleason, C. A., Wright, E. W., & Pearl, D. K. (1983). Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential). The unconscious initiation of a freely voluntary act. Brain, 106, 623–642. Nijhawan, R. (1994). Motion extrapolation in catching. Nature, 370, 256–257. Nijhawan, R. (2002). Neural delays, visual motion and the flash-lag effect. Trends in Cognitive Sciences, 6, 387. Sirigu, A., Daprati, E., Ciancia, S., Giraux, P., Nighoghossian, N., Posada, A., et al. (2004). Altered awareness of voluntary action after damage to the parietal cortex. Nature Neuroscience, 7, 80–84. Stewart, L., Ellison, A., Walsh, V., & Cowey, A. (2001). The role of transcranial magnetic stimulation (TMS) in studies of vision, attention and cognition. Acta Psychologica, 107, 275–291. Thaler, D., Chen, Y. C., Nixon, P. D., Stern, C. E., & Passingham, R. E. (1995). The functions of the medial premotor cortex: I. Simple learned movements. Experimental Brain Research, 102, 445–460. Trevena, J. A., & Miller, J. (2002). Cortical movement preparation before and after a conscious decision to move. Consciousness and Cognition, 11, 162–190. Walsh, V., & Pascual-Leone, A. (2003). Transcranial magnetic stimulation: A neurochronometrics of mind. Cambridge: MIT Press. Wegner, D. M. (2002). The illusion of conscious will. Cambridge: MIT Press. Wegner, D. M. (2003). The mind’s best trick: How we experience conscious will. Trends in Cognitive Sciences, 7, 65–69. D o w n l o a d e d f r o m l l / / / / / j t t f / i t . : / / D h o t w t p n : o / a / d m e i d t f r p o r m c . h s i p l v d e i r r e c c h t . m a i r e . d c u o m o / c n j o a c r t n i c / e a - r p t d i c 1 l 9 e 1 - 8 p 1 d f 1 / 9 1 3 6 9 0 / 6 1 1 / 8 o 1 c / n 1 2 7 0 5 0 6 7 4 1 1 9 4 / 1 j 8 o 1 c p n d . 2 b 0 y 0 g 7 u . e 1 s 9 t . o 1 n . 0 8 8 1 S . p e d p f e m b b y e r g 2 u 0 e 2 s 3 t / j . . . . t f . o n 1 8 M a y 2 0 2 1 90 Journal of Cognitive Neuroscience Volume 19, Number 1 Manipulating the Experienced Onset of Intention image

下载pdf