memory.pdf

(355 KB) Pobierz
49039809 UNPDF
Exp Brain Res (2008) 185:623–633
DOI 10.1007/s00221-007-1190-y
RESEARCH ARTICLE
Effect of tactile stimulus frequency on time perception:
the role of working memory
Mohammad Ali Khoshnoodi Æ
Rouzbeh Motiei-Langroudi Æ Mohsen Omrani Æ
Mathew E. Diamond Æ Abdol Hossein Abbassian
Received: 28 January 2007 / Accepted: 18 October 2007 / Published online: 8 November 2007
Springer-Verlag 2007
Abstract In most models of interval timing, there is a
central clock, which is considered to be highly protected
from the effects of external stimuli. However, many studies
have reported such effects and different theories are pro-
posed to explain the observations. These include the effect
of arousal, attention sharing, memory load and information
processing on central clock as well as change in the speed
of the pacemaker. In this study, we used regular vibro-
tactile stimuli with different frequencies in a ‘‘duration
reproduction task’’ to investigate the effect of stimulus
content on interval timing. Results showed that subjects
overestimated the duration as a function of test stimulus
frequency. A significant correlation between increasing the
test frequency and overestimation of subjective time was
observed. We further investigated the effect of blank and
filled gaps with various durations on time estimation.
Analysis revealed that regardless of gap duration, sub-
jective time increased in the filled gap condition, compared
to the blank gap. This effect was independent from con-
textual stimuli and correlated to the mean number of
stimuli during the temporal interval rather than rate of
stimulus presentation.
Keywords Time perception Stimulus content
Central clock Gap Working memory
Introduction
Most models of time perception are based on a specialized
chronometric mechanism (Creelman 1962 ; Fraisse 1984 ;
Frankenhauser 1959 ; Gibbon et al. 1984 ; Killeen and Fet-
terman 1988 ; Treisman 1963 ). The basic structure of these
models includes five major components: (1) time base, (2)
gate, (3) counter, (4) reference memory, and (5) compar-
ator. In some models, the mean number of events in a given
time interval specifies the time base (Frankenhauser 1959 )
while in others there is an internal pacemaker, which emits
pulses (Gibbon and Church 1984 ; Treisman 1963 ). In the
scalar expectancy theory, a variant of internal pacemaker
models, the pacemaker generates pulses, which are trans-
ferred to the counter through a switch. The comparator
continuously compares the number of pulses gathered in
the counter with the reference number stored in memory.
Reaching the criterion number, response is elicited (Cata-
nia 1970 ; Church et al. 1991 ; Gibbon and Church 1990 ;
Gibbon 1977 ).
In most clock-counter based models of interval timing,
the clock is internal, receiving no input from the outside
world. Thus, to explain the reported effects of external
stimuli on time perception, some theories modulated the
clock and its components. Treisman et al. ( 1990 ) suggested
that the pacemaker consists of an oscillator coupled to a
calibrator unit, which is modulated by arousal level. They
suggested two mechanisms for the effects of external
stimuli on time perception. Frequencies near to the oscil-
lator’s fundamental frequency or its harmonics would drive
the oscillator and figuratively, change the clock speed. On
M. A. Khoshnoodi ( & ) R. Motiei-Langroudi M. Omrani
A. H. Abbassian
School of Cognitive Sciences,
Institute for Studies in Theoretical Physics and Mathematics
(IPM), Niavaran, Tehran, Iran
e-mail: khoshnoodi@ipm.ir
M. E. Diamond
Cognitive Neuroscience Sector,
International School for Advanced Studies (SISSA),
Trieste, Italy
123
49039809.001.png
624
Exp Brain Res (2008) 185:623–633
the other hand, the external stimulus may possess an
arousing effect on the calibrator, increasing the number of
emitted pulses. Other models, which explicitly account for
the effect of non-temporal attributes of the signal, consider
time as an emergent property of information processing.
Grondin ( 2001 ) suggested that the evidence for influence of
non-temporal stimulus characteristics on duration dis-
crimination might question the existence of a central clock.
Ornstein ( 1969 ) rejected the idea of a specialized chrono-
metric mechanism, suggesting that perception of temporal
duration depends on the amount of information available
during the time interval and the load of perceptual and
memory processing.
Another suggested explanation for the effects of external
stimuli on time perception within clock-counter based
models of interval timing, is attention sharing theory. It is
proposed that there are two processor systems involved in
obtaining temporal information which recruit common
attentional resources: a temporal processor and a general
processor (Block and Zakay 1997 ; Cantor and Thomas
1977 ; Fortin and Masse 2000 ; Gibbon and Church 1984 ;
Hicks et al. 1976 ; Eagleman et al. 2005 ; Thomas and
Cantor 1978 ; Thomas and Weaver 1975 ; Zakay 2000 ;
Zakay and Tsal 1989 ). In Gibbon and Church’s ( 1984 )
information processing model of timing, the proposed
switch located between the pacemaker and accumulator
controls the transfer of pulses- which are generated con-
tinuously in the pacemaker- to accumulator. It has been
proposed that attention controls activation of the switch.
Diversion of attention from timing processor opens the
switch resulting in underestimation of the subjective time
(Meck 1984 ; Tremblay and Fortin 2003 ). The gap para-
digm has been frequently used to examine the predictions
of attention sharing theory. Fortin and Masse ( 2000 )
showed that human subjects use stop strategy during the
gap while the produced intervals lengthened proportionally
with increasing the pre-gap duration. They further showed
that increasing the uncertainty and level of expectancy of
the gap results in more underestimation of pre-gap duration
(Tremblay and Fortin 2003 ). Buhusi and Meck ( 2000 )
showed that using empty gaps prompts rats to stop timing
while using filled gaps prompts rats to reset the entire
timing process, possibly due to allocation of more atten-
tional resources to the general processing of filled gaps.
Making the gap content similar to the inter-trial interval
leads to resetting the time processor, presumably due to
allocation of fewer resources to timing in the inter-trial
interval (Buhusi and Meck 2002 ; Buhusi et al. 2002 ).
It has been suggested that there may be different
implementations for memory-based and clock-counter
attentional models of interval timing. Block and Zakay
( 1997 ) suggested that in prospective timing paradigms (in
which subjects are explicitly aware that they have to
estimate the interval duration), attentional models are
needed to explain temporal judgments. In retrospective
paradigms (in which subjects need to judge a duration after
it has ended), however, changes in time perception are
described better by memory-based models.
Here, we designed a prospective time reproduction task,
using vibrotactile stimuli as the timing target. In this kind
of stimuli, the content can be modulated by changing the
frequency without altering its structure or complexity.
First, we studied the effect of stimulus frequency on time
perception. We then used the gap paradigm, both empty
and filled with a vibrotactile stimulus to further investigate
probable sources of stimulus frequency effect. According
to the attention sharing theory, filling the gap should result
in shifting attention away from the time processor, pro-
longing the reproduced interval. On the contrary, if filling
the gap would cause reproduction of shorter durations, it
would be better explained with memory-based theories or
change in the clock speed. Next, inserting a filled/empty
gap and changing the context of target stimulus, we tried to
examine the influence of clock speed and arousal level on
the observed effect. If changing the clock speed was the
attributing factor for the effect, it would respectively vary
based on the context frequency. Finally, we evaluated the
effect of mean number of events on time perception.
Material and methods
Participants
Participants were students of Tehran University of Medical
Sciences, who voluntarily participated in the present study.
All participants were na¨ve to the purpose of the experi-
ment. No participant took part in more than one
experiment. All subjects were right handed, with a mean
age of 22.6 ± 1.8 (range, 19–30). Seven (6 males and 1
female), five (2 males and 3 females), five (2 males and 3
females), eight (5 males and 3 females) six (all males) and
seven (all males) attended experiments 1, 2, the first and
second parts of experiment 3, 4, and 5, respectively.
Recruitment of the subjects and experimental procedures
were conducted in accordance with Declaration of
Helsinki.
Apparatus
The vibrotactile stimuli were produced using piezoelectric
wafers (length, 38 mm; width, 19 mm; thickness, 0.5 mm;
Morgan Matroc, Bedford, OH) individually driven by 80 V
pulses from custom-built amplifiers controlled by a com-
puter running MATLAB software (version 6.5). A flat
123
Exp Brain Res (2008) 185:623–633
625
rubber pad (length, 19 mm; width, 19 mm; thickness,
2 mm) was glued on the top face at the outer end of each
wafer. Wafer positions adjusted to accommodate the
lengths of each subject’s fingers. A padded wooden block
was used as wrist support.
were provided with feedback after each test trial. This was
done to ensure that each subject had learned the duration of
the timing stimulus precisely. In different experiments,
variations were applied to the test phase of the main blocks.
Only the main blocks were used in data acquisition and
analysis. In each trial, time reproduction error (TRE) was
defined as: reproduced duration—learned duration. In each
subject, the data from different main blocks of each exper-
iment was pooled and a mean TRE was calculated.
Stimuli
Vibrations were composed of a sequence of square wave
deflections where each edge of the square wave was felt as
a discrete event (Harris et al. 2001 ). Two kinds of stimuli
were used: The timing stimulus, a periodic train of
deflections and the distractor stimulus, an irregular pattern
of deflections that consisted of 4–16 deflections per trial
with random deflection times.
Experiment 1: effect of test stimulus frequency on time
reproduction
This experiment was designed to find out whether changing
the frequency of test trials would affect time reproduction.
Hence, the frequency of the tactile timing stimulus in the
learning phase was 12 Hz in half of the main blocks and
24 Hz in the remaining blocks. In the test phase, the fre-
quency of the tactile stimulus was either 12 or 24 Hz,
which was randomly distributed in each block. The main
blocks were also arranged randomly. The experimental
design resulted in four pairs of learning-test frequencies:
12–12, 12–24, 24–12 and 24–24 (e.g., 12–12 means that
the frequency of the stimulus was 12 Hz in the learning
phase and 12 Hz in the test phase). To control the effect of
the distractor stimulus, four additional blocks were added
to the experiment. Here, the distractor stimulus was also
presented in the test phase, and the subjects were asked to
count and report the number of vibrations while simulta-
neously reproducing the learned duration. 2-way ANOVA
was used to compare the TREs.
Procedure
We used the peak interval procedure and its variant, the
gap paradigm, which are used primarily as the standard
procedure in non-human interval timing experiments
(Gibbon and Church 1990 ). Human performance with peak
procedure is comparable to that of laboratory animals
(Rakitin et al. 1998 ).
The experiment was conducted in two practice blocks
and 10 main blocks. Each main block consisted of two
phases: the learning phase and the test phase. The learning
phase consisted of 15 trials; in each trial, a vibrotactile
timing stimulus was presented to the subjects’ right index
finger. The subjects were asked to learn the duration of the
timing stimulus. To force the subjects to learn duration by a
time measurement rather than by a counting process (Rak-
itin et al. 1998 ), we discouraged counting during learning
phase by presenting an irregular distractor stimulus simul-
taneously to the subjects’ right middle finger; they were
asked to count the number of the deflections and report it
verbally at the end of the trial. The duration of the timing
stimulus was 8 s in every trial of the learning phase while
the subjects were unaware of the exact duration. Learning
trials were separated by random intervals ranging 1–3 s.
Test phase consisted of 15 trials in which the stimulus was
presented to the subject’s right index finger, and they were
asked to reproduce the learned duration by pressing the
space bar when they estimated that the duration of the
running stimulus equaled the learned duration. If the subject
failed to press the space bar, the tactile stimulus would stop
in a time double the criterion duration. The distractor
stimulus was played in the test phase of four additional
blocks in experiment 1 that was designed to control the
effect of distractor. In the practice blocks, the stimuli in both
learning and test phases were exactly the same and subjects
Experiment 2: correlation between test stimulus
frequency and time reproduction error
In this experiment, we tried to find out the correlation
between frequency of test phase stimulus and change in
time perception. Here, the frequency of the learning phase
stimulus was 12 Hz in all blocks, but four different fre-
quencies (6, 12, 24 and 48 Hz) were applied in the test
phase of main blocks. This resulted in four pairs of learn-
ing-test frequencies (12–6, 12–12, 12–24 and 12–48).
Correlation analysis was used to show the correlation
between frequency of the test phase stimulus and TRE.
Experiment 3: effect of inserting filled/empty gaps on
time reproduction
In the first part of this experiment we inserted an empty gap
in the middle of the test stimulus. Frequency of the stimulus
123
626
Exp Brain Res (2008) 185:623–633
was 12 Hz in both learning and test phases in all blocks. A
‘‘gap’’ of either 0, 2, 4 or 6 s duration was randomly inserted
in the middle of the test trials (i.e., 4 s after test trial onset).
Gap was defined as a pause in stimulus presentation and
gaps with zero second duration were test trials without a gap
and served as control trials. The subjects were asked to
ignore the duration of the gap and press the space key when
the sum of the pre- and post-gap durations equaled the
learned duration. Repeated measure ANOVA was used to
compare the effect of gap duration on TRE. Afterwards, we
investigated the effect of filling the gap with stimulus. We
applied two types of gaps in the middle of the test trials:
empty gaps (the same as used in the first step) and filled
gaps, during which instead of pausing the stimulus, its
vibration frequency was increased from 12 to 24 Hz
(Fig. 3 a). The difference between 12 and 24 Hz was per-
fectly discriminable to the subjects (Harris et al. 2001 ). The
gap durations were 2, 4, 6 and 8 s. Different gap durations
and gap types were randomly distributed through blocks.
Repeated measure ANOVA was used to compare the effect
of gap duration and gap content on TRE.
of deflections from frequency of stimulus and examine the
result. Experiment 5 was designed to investigate the effect
of mean frequency of the test stimulus on TRE while the
rate of deflections changes during the stimulus presenta-
tion. The subjects learned the duration of a 12 Hz stimulus.
The test stimuli were split into two uninterrupted halves,
each with different frequencies. Six pairs of stimuli were
produced (6–12, 12–6, 6–48, 48–6, 12–48 and 48–12) and
were compared with the control stimulus in which both
halves were 12 Hz (Fig. 5 a). The correlation between TRE
and mean test frequency was analyzed. The mean test
frequency was calculated from the average frequency in the
whole test trial (e.g., in both 6–12 and 12–6 test trials, the
average frequency would be nine, and so on).
Results
Results of experiment 1 showed that subjects reproduced a
shorter duration of elapsed time when the test trials con-
tained a vibrotactile frequency of 24 Hz compared to test
frequency of 12 Hz, regardless of vibrotactile frequency
presented during learning phase (2 9 2 repeated measure
ANOVA; factor learning frequency: F(1,6) = 0.169,
p = 0.699; factor test frequency: F(1,6) = 18.937,
p = 0.005) (Fig. 1 ). We further compared the resulted TREs
to zero, which was only statistically significant in 24–12
condition (t = 2.550,df =6, p = 0.043; for the other
conditions (12–12, 12–24, 24–24), t was equal to 1.540,
-0.613, and -0.540 with df = 6 in all conditions, and
p value equaled to 0.174, 0.562, and 0.608, respectively). In
blocks in which both learning and test phases contained the
distractor stimulus, no effect of learning or test frequency
was observed (2 9 2 repeated measure ANOVA; factor
learning frequency: F(1,6) = 2.02, p = 0.205; factor test
frequency: F(1,6) = 4.17, p = 0.087).
Result of experiment 2 showed a significant negative
correlation between the vibrotactile frequency presented
during the test trials and TRE (Pearson correlation
r = -0.734, F(1,18) = 20.97, p \ 0.001; Fig. 2 ); in other
words, when the test frequency was increased, subjects
pressed the response key earlier. This indicates that the
finding of experiment 1—that the presence of 24 Hz stimuli
during the test phase led to a shorter reproduced duration—
is in fact a systematic effect of the test phase frequency on
TRE. To evaluate the correlation between the magnitude of
the effect and the frequency of test stimulus, the TRE of
blocks with 12 Hz test frequency was considered as an
internal error of subject’s performance and the measure was
subtracted from TREs of other experiments for each
subject; then, the results were analyzed with Linear
Regression analysis. Again a significant negative correla-
tion was observed (Pearson correlation r = -0.817,
Experiment 4: interaction of peri-gap stimulus
and gap content
In the previous experiment, the frequency of the vibrations
in the filled gap period was 24 Hz and the peri-gap fre-
quency was 12 Hz. As 24 Hz frequency has more
deflections in an objective time unit, it is possible that it
would speed up the clock or increase the arousal level of
the subjects (Treisman et al. 1990 ). Experiment 4 was
designed to investigate whether the effect of gap content on
TRE is dependent on the contextual difference of fre-
quencies between the gap itself and the pre- and post-gap
intervals. In this experiment, the effect of a 4 s gap, either
blank or filled with the frequency of 24 Hz was compared
in blocks with learning and test frequencies of either 12 or
36 Hz. These blocks were arranged and presented ran-
domly. Repeated measure ANOVA was applied to assess
the effect of gap content and context frequency on TRE.
Experiment 5: correlation between mean frequency
of split test stimuli and time reproduction error
In the current experiment, it is possible that time perception
is affected by change in clock speed, arousal level or the
mean number of deflections. According to Triesman the-
ory, increasing the stimulus frequency may drive the clock
to higher speed or may increase the arousal level, so we
could not discriminate the possible source of observed
effect. So we tried to dissociate the change in mean number
123
Exp Brain Res (2008) 185:623–633
627
Fig. 2 Correlation between TRE and mean frequency of test stimuli
in experiment 2
Fig. 1 a A schematic illustration of the stimuli used in experiment 1.
b Results of experiment 1. The first and second number in each pair
illustrated in horizontal axis represents the frequencies of the learning
and test phase stimuli, respectively. The bars represent S.E.M
for gap duration (gap content: F(1,7) = 29.348, p = 0.001;
gap duration: F(3,21) = 2.321, p = 0.105; gap content 9
gap duration: F(3,21) = 0.253, p = 0.858). This means that
regardless of gap duration, filling the gap with a 24 Hz
frequency led to an increase in subjective perception of
time; therefore, the response of the subjects appeared ear-
lier, compared to empty gaps (Fig. 3 b). We further
compared the TREs in filled gap condition to zero, to
examine the role of performance error in observed effect.
No significant difference was observed in either of the
groups (t = 0.265, p = 0.798 for 2 s gap; t = -0.042,
p = 0.968 for 4 s gap; t = -0.416, p = 0.690 for 6 s gap;
t = 0.003, p = 0.997 for 8 s gap). To assess the effect of
gap content and duration on time reproduction reliability of
subjects, we calculated the coefficient of variance (CV) of
reproduced durations. We used 2 (gap content) 9 4 (gap
type) ANOVA to analyze the effect of gap on CV. Neither
gap content (F(1,7) = 1.043,p = 0.431) nor gap duration
(F(3,21) = 0.788, p = 0.674) had significant effect on CV.
Experiment 4 revealed that a 24 Hz filled gap affects
time reproduction regardless of the vibration frequency
presented during learning and test (Fig. 4 ). 2 (gap con-
tent) 9 2 (learning and test frequency) repeated measure
ANOVA analysis showed a significant effect of gap con-
tent on TRE while contextual frequency had no significant
effect (gap content: F(1,5) = 9.912, p = 0.025; contextual
frequency: F(1,5) = 0.220, p = 0.659; gap content 9 con-
textual frequency: F(1,5) = 3.848, p = 0.107; Fig. 4 ). This
result suggests that the effect of gap content in this
experiment is independent of the contextual frequency.
In experiment 5, we first analyzed the effect of transition
from high-to-low or low-to-high frequency on TRE. We
used a 2 (transition) 9 3 (pairs, 6–12, 6–48, 12–48)
ANOVA to examine the effect. No significant effect for
F(1.18) = 25.154, p \ 0.001) and a high correlation
between the frequency of test stimulus and the amount of
change in subjective time was demonstrated, as described
by the equation: magnitude of change in TRE = -
0.038 9 test stimulus frequency + 0.42 (t values and p
values for the slope and constant factors were t = 4.136,
p \ 0.0001 and t = -5.015, p = 0.026, respectively). We
also analyzed the effect of stimulus frequency on coefficient
of variance of reproduced durations. One-way ANOVA
showed no significant effect of frequency on variability of
subjects response (F(1, 3) = 0.240, p = 0.487).
Results of experiment 3 showed that neither inserting a
blank gap nor the gap duration had a significant effect on
TRE (Repeated measure ANOVA; F (3,12) = 2.715,
p = 0.167). Thus subjects maintained memory of pre-gap
duration during the gap period, linked the pre-gap duration
to the post-gap duration, and then reproduced the learned
duration with an added shift equal to the duration of the
gap. In the second part of experiment 3, the effect of gap
content and duration on TRE were evaluated by inserting
either blank or filled gaps of four different durations in test
trials. 2 (gap content) 9 4 (gap duration) repeated measure
ANOVA showed a significant effect for gap content but not
123
49039809.002.png 49039809.003.png 49039809.004.png
Zgłoś jeśli naruszono regulamin