Biol Cybern (2005) 93: 343–354 DOI 10.1007/s00422-005-0011-2
O R I G I N A L PA P E R
Hironori Nakatani · Cees van Leeuwen
Individual differences in perceptual switching rates; the role of occipital alpha and frontal theta band activity
Received: 1 November 2004 / Accepted: 18 July 2005 / Published online: 20 October 2005 © Springer-Verlag 2005
Abstract Prolonged presentation of visually ambiguous figures leads to perceptual switching. Individual switching rates show great variability. The present study compares individuals with high versus low switching rates by investigating human scalp electroencephalogram and blink rates. Eight subjects viewed the Necker cube continuously and responded to perceptual switching by pressing a button. Frequent switchers showed characteristic occipital alpha and frontal theta band activity prior to a switch, whereas infrequent switchers did not. The alpha activity was specific to switching, the theta activity was generic to perceptual processing conditions. A negative correlation was observed between perceptual switching and blink rates. These results suggest that the ability to concentrate attentional effort on the task is responsible for the differences in perceptual switching rates.
1 Introduction Figures such as Rubin’s vase/face or the Necker cube are ambiguous; they have at least two, frequent, alternative interpretations. When figure presentation lasts sufficiently long, these alternate spontaneously ( Attneave 1971; Einh¨auser et ˙so˘glu-Alka¸c et al. 1998, 2000; Leopold and Loal. 2004; I¸ gothetis 1999). Because this occurs without changes in the figure, perceptual switching phenomena are eminently suitable to study brain activity related to the intrinsic dynamics of visual perception. Perceptual switching rates show a great deal of individual variation ( Borsellino et al. 1972; Brown 1955; Str¨uber et al. 2000). We raise the question whether there is a systematic difference in brain activity between frequent and infrequent switchers. This relates to perceptual flexibility. Lesion H. Nakatani (B) · C. van Leeuwen Laboratory for Perceptual Dynamics, RIKEN Brain Science Institute. 2-1, Hirosawa, Wako-shi, Saitama 351-0198, Japan E-mail:
[email protected] Fax: +81-48-467-7236
studies ( Meenan and Miller 1994; Ricci and Blundo et al. 1990) have related perceptual flexibility to frontal brain areas; frontal damage patients show lower switching rates than normal subjects. We, therefore, expect individual differences in switching rate to be related to frontal activity. We will investigate whether there are distinctive characteristics in the human scalp electroencephalogram (EEG) in frontal areas and elsewhere. When distinctive characteristics can be found, we would be interested in whether these individuals differ on behavioral characteristics related to switching. In particular, we are interested in eye-blink rates. Ito et al. (2003) showed that eye blink frequency is reduced prior to the occurrence of a perceptual switch. Individuals who are more inclined to switching, therefore, are expected to show reduced blinking rates.
2 Method 2.1 Subjects Eight subjects, three male and five female, with normal or corrected-to-normal vision, participated in this study. Mean age was 26.8 years with a range of 20–35 years. The study was approved by RIKEN BSI Institutional Review Board No. 2 (Research Ethics Committee), and the experiments were undertaken with written informed consent.
2.2 Experimental design The experiment consisted of three conditions: a perceptual switching (PS) condition, a stimulus driven (SD) condition, and an eyes closed (EC) condition. Each experimental session consisted of these three conditions in this order (Subjects-E, G, H of Table 1 did not join in the experiment of the EC condition). Two or three sessions separated by a break were conducted for each subject.
344
H. Nakatani, C. van Leeuwen
(a)
(b)
Fig. 1 Visual stimulus in our experiment. The visual angle of the stimulus in both PS and SD conditions was 5.7◦ . Subjects were instructed to look at the center region of the stimulus, but were not required to fixate at a specific point. a In the PS condition, the Necker cube was presented as a white line drawing on a black ground. The stimulus was continuously presented for 4 min. b In the SD condition, two biased Necker cubes were presented alternately at the same location. Each figure was continuously presented for a variable time interval of 5.0–10.0 seconds (mean: 7.5 s)
2.2.1 Perceptual switching condition
2.2.2 Stimulus driven condition
Switching fails to occur if subjects know only one of the possible interpretations of an ambiguous figure ( Rock et al. 1994). To assure that all subjects started with equal information, they were advised before the experiment that the Necker cube has two alternative interpretations. During the experiment, a Necker cube (Fig. 1a) drawn in white lines on a black ground was continuously presented on a flat-panel monitor in a dark soundproof room. Subjects viewed the stimulus at a distance of 150 cm from the monitor. The size of the stimulus was 5.7◦ of visual angle. Subjects were instructed to look at the center of the stimulus, but were not forced to fixate their view. To report their percept, subjects pressed the appropriate button on a response box using their dominant hand. Two alternative buttons on the box corresponded to the alternative interpretations of the Necker cube. Subjects reported their new interpretation by pressing a button whenever they perceived the occurrence of a switch but not when their percept merely became inconsistent or vague. An experimental session, in which subjects viewed the Necker cube continuously and reported switches while EEG was recorded, took 4 min.
Two biased Necker cubes (Fig. 1b) were presented alternately at the same location. The visual angle of each cube was 5.7◦ . Each biased cube was continuously presented for a variable time interval of 5.0–10.0 s (mean : 7.5 s). When one figure was changed to another, subjects pressed the appropriate button using their dominant hand. An experimental session took 4 min. 2.2.3 Eyes closed condition Subjects closed their eyes and pressed two alternative buttons with their dominant hand. They were instructed to replicate intervals of button press in the PS condition. An experimental session took 4 min. 2.3 Grouping of subjects We divided subjects into two groups, frequent and infrequent switchers, based on their switching rates. The frequent
Individual differences in perceptual switching rates; the role of occipital alpha and frontal theta band activity
switching group consisted of subjects whose switching rates were higher than 20 times per minute. 2.4 Statistical properties of perceptual switching intervals It is known that intervals of perceptual switching follow a gamma distribution ( Borsellino et al. 1972; Leopold and Logothetis 1999; Murata et al. 2003) and lack short term temporal correlations ( Lehky 1995). We analyzed these statistical properties of our button press interval data, in order to evaluate the reliability of button presses in the PS condition as switch responses. 2.4.1 Gamma distribution The cumulative probability density function of a gamma distribution F (x) is defined as follows, x α β α−1 t exp (−βt)dt, (1) F (x) = (α) 0
where α and β are positive real numbers and (α) is the gamma function, ∞ (α) =
x α−1 exp (−x)dx.
(2)
0
345
amplifiers was 100 M. Gains were 94 dB for EEG signals and 80 dB for EOG signals, and cut-off frequencies were 0.08 and 100 Hz. Signals were digitized with a 12 bit analog-todigital converter at 500 Hz and stored on the hard disk of a PC. 2.6 Reduction of baseline fluctuation and artifacts We used a discrete wavelet transform (DWT) to reduce low frequency baseline fluctuation in our recordings. The hierarchical structure of the multiresolution analysis (MRA) for recordings was generated by the DWT ( Bartnik et al. 1992; Nakatani et al. 2001); the center frequency of the signal at the i-th scale is fs /2i+1 , where fs is 500 Hz. Recordings were decomposed into the 10th scale of the hierarchical structure of MRA and subsequently reconstructed with the discrete detail signals of the first to nineth scales. We used the compactly supported cardinal B-spline wavelets ( Chui ˙so˘glu-Alka¸c et al. 1998, 2000) as scaling and Wang 1992; I¸ functions. One of the merits of this method is that it causes no phase shift in the recordings. We used principal component analysis (PCA) to reduce blink and EOG artifacts from EEG recordings as follows: we applied PCA to fluctuation-reduced recordings u u= (uEEG−O1 (t), . . ., uEEG−F 4 (t), uV EOG (t), uH EOG (t)) ,
(4)
Least-squares estimators of α and β for each subject were obtained from the cumulative probability distribution of button press intervals from all experimental sessions. As each possible interpretation of the Necker cube has different values of α and β ( Murata et al. 2003), we obtained cumulative probability distributions for both percepts. To determine whether the button press intervals follow a gamma distribution, we used the Kolmogorov-Smirnov one-sample test ( Press et al. 1993; Zhou et al. 2004).
and obtained principal components. As EOG recordings usually have a larger amplitude than EEG recordings, most of the blink and EOG artifacts components tend to appear in the first principal component. By reconstructing the signal without this component, artifact-reduced EEG recordings were obtained.
2.4.2 Autocorrelation function
The scalogram, the amplitude of the EEG recordings in the time-frequency domain, was calculated by a continuous wavelet transform. The mother function of the wavelet transform was the complex Gabor function g(t), t2 1 (5) exp − 2 exp (j 2πt), g(t) = √ 4α 2 πα
The autocorrelation function ρ(τ ) is defined as follows, ρ(τ ) =
E[(x(n) − µ)(x(n + τ ) − µ)] , (E[(x(n) − µ)2 ])1/2 (E[(x(n + τ ) − µ)2 ])1/2
(3)
where x(n) is the length of the n-th interval, µ is mean of x(n), and E[·] represents expectation of ·. 2.5 EEG recording Plate electrodes were placed on O1, O2, Pz, P3, P4, Cz, C3, C4, Fz, F3, F4 recording sites in accordance with the international 10/20 system. Reference and ground electrodes were placed on nose and forehead, respectively. Vertical and horizontal electrooculogram (EOG) were also recorded. Signals were amplified and band-limited with differential amplifiers and first order band-pass filters. Input impedance of the
2.7 Continuous wavelet transform
where α = 0.5. When we take α = 0.5, the size of the mother function is about three cycles. Each sweep of data consisted of an episode preceding the button press (BP), which ran from −900 to 0 ms, where 0 ms corresponded to BP. 2.8 Parameter to characterize switching rate-dependent activity Our aim was to identify an observable parameter ϕr from the EEG recordings of an individual that has a systematic relationship to individual switching rate in the PS condition.
346
H. Nakatani, C. van Leeuwen
The parameter should be derived from a single recording site, such as O1 or Pz, represented by the subscript r. We identified ϕr as the location, or locations, where the average waveform µ in the time-frequency domain showed its local maximum within a broad range. ϕr = ϕr (µ) = (t0 , f0 ).
(6)
The next step in our investigation was to determine the recording sites and to identify the broad range in which ϕr was to be obtained. Based on our exploration of the data, we chose as potentially interesting the recording sites O1 and Fz and the range from (−900 ms, 4.0 Hz) to (0 ms, 10.0 Hz). We used the following conditions to identify the coordinates of ϕr , √ g(t0 , f0 ) = arg max g(t, f ) ≥ 0.4 (µV/ Hz) (t,f ) (7) ∂ 2 g(t,f ) | = 0 (t,f )=(t ,f ) 0 0 ∂t∂f where g(t, f ) is the amplitude of µ at (t, f ). If no point (t, f ) satisfied these conditions, ϕr took the null value. 2.9 Bootstrap resampling technique to estimate the variance of the estimator Variance occurs in the estimated location ϕˆr . To determine the variance, we considered our data as a random sample from an unknown probability distribution F ( Efron 1979; Efron and Tibshirani 1986), Xi = xi ,
Xi ∼ F
i = 1, 2, . . . , n
(8)
where the variable Xi represents the random sample, xi represents a sweep of data which is a realization of Xi , and n is the number of data sweeps. The true average waveform is given by µ = EF (X),
(9)
where the notation EF indicates the expectation. We obtain the average waveform of the experimental data as an unbiased estimator of µ, n
x¯ =
1 xi . n i=1
(10)
ϕˆr , an estimator of ϕr , ϕˆr = ϕr (x) ¯
(11)
is a function of x. ¯ According to the central limit theorem, ϕˆr will follow the normal distribution N(ϕr , σ 2 /n), where σ 2 is the asymptotic variance, when n is large. We used the bootstrap resampling technique ( Efron 1979; Efron and Tibshirani 1986) to estimate the variance of the estimator. This technique can be used, in principle, for individual as well as group data; here we are concerned with individual data samples only. We generated a bootstrap sample xi∗ (i = 1, . . . , n) by random sampling with replacement from the actual sweeps of EEG data xi (i = 1, . . . , n) of an individual subject, and obtained bootstrap parameter ϕˆr∗ = ϕr (x¯ ∗ )
(12)
from the average waveform x¯ ∗ of the bootstrap sample. For each individual subject, we repeated this procedure and obtained 1,024 bootstrap parameter values. The variance of ϕˆr∗ is the estimated sample variance of ϕˆr and the distribution of ϕˆr∗ specifies the margin of error for the estimated location of ϕr . As ϕˆr∗ also follows N(ϕr , σ 2 /n), the margin is given by the ellipse sf2 (t − t¯)2 − 2ctf (t − t¯)(f − f¯) + st2 (f − f¯)2 = 1, (13) 2 st2 sf2 − ctf where (t¯, f¯) is the mean of ϕˆr∗ , st and sf are standard deviations of the time and frequency components, respectively, of ϕˆr∗ , and ctf is the covariance between time and frequency components of ϕˆr∗ . This ellipse consists of points (t, f ) where the Mahalanobis distance from (t¯, f¯) is 1. Each ellipse provides us with the margin of error for the estimated location of ϕr for a frequent switching subject. In order to combine these data, we took the rectangular range of the time-frequency domain which contains these ellipses. If ϕr is a reliable criterion, none of the infrequent switchers should have any significant activity within this range. 2.10 Detection of blink timing from vertical EOG recordings In order to analyze the relationship between perceptual switching and eye blink rates, we obtained blink rates from the vertical EOG (VEOG) recordings. Blinks cause spike-like artifacts that have larger amplitude than the EOG activity in VEOG recordings. Blink timing could, therefore, be estimated using the threshold detection method ( Lewicki 1998).
3 Results Table 1 shows the perceptual switching rates of individual subjects in the PS condition. The differences in individual switching rates between subjects were large, whereas those between sessions of individual subjects were relatively small. Button press interval data of all subjects exhibited gamma distributions (the chosen level of significance : 0.05). The values of (α, β) of the gamma distribution are shown in Table 1. The autocorrelation function for all subjects except for G showed no short term temporal correlations, ρ(τ ) < 0.4 f or τ ≥ 1. These results are consistent with the ones previously published ( Borsellino et al. 1972; Lehky 1995; Leopold and Logothetis 1999; Murata et al. 2003). We, therefore, consider button press responses in the PS condition to be reliable indicators of perceptual switching. Figure 2 shows the average waveforms of each recording site of the two most frequently switching subjects. These two subjects show activity in O1 and Fz recording sites in the 6–9 Hz (theta and alpha band) region. We focused our exploration on this activity in these regions for other subjects as well. Figure 3 shows average waveforms of O1 and Fz recording sites of subjects with higher (A, B), middle-level (D, E),
Individual differences in perceptual switching rates; the role of occipital alpha and frontal theta band activity
347
Table 1 Switching rates of eight subjects in the PS condition
Group
Subject
Switching rate average (1/min) [first session, second session, (third session)] 22.4
269
: (2.83, 1.06)
Frequent
A
[25.0, 21.8, 20.5]
(three sessions)
: (5.50, 2.80)
20.9
251
: (6.62, 2.27)
Frequent
B
[20.8, 22.0, 20.0]
(three sessions)
: (5.02, 2.14)
20.8
249
: (2.03, 0.58)
Frequent
C
[19.8, 23.0, 19.5]
(three sessions)
: (2.66, 1.39)
15.9
191
: (2.24, 0.44)
Infrequent
D
[16.5, 16.3, 15.0]
(three sessions)
: (2.61, 1.01)
7.9
95
: (2.33, 0.32)
Infrequent
E
[7.0, 7.8, 9.0]
(three sessions)
: (1.48, 0.22)
7.9
95
: (2.74, 0.40)
Infrequent
F
[5.3, 6.3, 12.3]
(three sessions)
: (3.79, 0.59)
4.0
32
: (2.08, 0.14)
Infrequent
G
[5.5, 2.5]
(two sessions)
: (1.95, 0.19)
3.3
26
: (1.49, 0.05)
Infrequent
H
[3.8, 2.8]
(two sessions)
: (2.79, 0.10)
and lower (G, H) switching rates. Only subjects with higher switching rates showed activity at both the O1 site about 600 ms and at the Fz site about 450 ms before the switch response. Their frequency components corresponded to the theta and lower alpha bands. We calculated the parameters ϕˆO1 and ϕˆFz according to eq.(11) in the range from (−900 ms, 4.0 Hz) to (0 ms, 10.0 Hz). Table 2 shows the values of ϕˆO1 and ϕˆFz of each subject. This result indicates that non-zero values of ϕˆO1 and ϕˆFz were obtained consistently for subjects with higher switching rates. Figure 4 shows the distribution of the bootstrap parame∗ ∗ and ϕˆFz and the margins of the estimated locations ters ϕˆO1 of ϕO1 and ϕFz for each frequent switcher. Both for O1 and Fz, these elliptical margins were situated within a narrow rectangular range. Our proposed criterion for belonging to the frequent switching group is significant activity in these areas, that are located in the range from (−836 ms, 6.7 Hz) ∗ and from (−612 ms, 6.1 Hz) to to (−405 ms, 9.3 Hz) for ϕˆO1 ∗ (−254 ms, 7.9 Hz) for ϕˆFz . According to this criterion, none of subjects of the infrequent switching group should have any activity in these ranges. Figure 5 shows scatter plots of the ∗ ∗ and ϕˆFz of each probability that the bootstrap parameters ϕˆO1 subject appeared in these specific ranges. In Fig. 5a, not only the frequent switchers but also two infrequent switchers, subjects-G and H, showed a relatively high probability of activity in the broad range from (−900 ms, 4.0 Hz) to (0 ms, 10.0 Hz)
Switching number
(α, β) of the gamma distribution
of our initial exploration. However, as shown in Fig. 5b, only frequent switchers showed activity in the narrow ranges of Fig. 4b specified by our bootstrap resampling analysis. This result clearly indicates that the parameter ϕr defined by Eq. 6 and the specified ranges in Fig. 4b are specific to EEG activity of frequent switchers. Therefore, the combined presence of specific episode of theta and alpha band activity in occipital and frontal areas was specific to frequent switchers. Figure 6 shows average waveforms of O1 and Fz recording sites of six subjects in the SD condition and Fig. 7 shows average waveforms of two frequent switchers in the EC condition. The characteristic alpha band activity of frequent switchers in the PS condition did not appear in the SD condition. Subject-A showed alpha band activity in the SD condition. However, the frequency of this activity is higher than that of the alpha band activity in the PS condition. As this activity has similar frequency to that of the EC condition, it appears to be unrelated to perceptual processing. We concluded that the alpha band activity observed in the PS condition is specific to the perceptual switching processes of frequent switchers. As for theta band activity, frequent switchers also showed this activity in the SD condition. The activity occurred in a position in time-frequency domain similar to that in the PS condition. Infrequent switchers, who did not show this theta band activity in the PS condition, did not show it in the SD condition either. If theta band activity of
348
H. Nakatani, C. van Leeuwen
0.00 0.13 0.26 0.39 0.52 (uV)
Freq. (Hz)
(a)
15 F3
15 Fz
15 F4
10
10
10
5
5
5
Freq. (Hz)
-800
-800
10
10
5
5
5
-400 Time (ms)
BP
-800
-400 Time (ms)
BP
-800
15 P3
15 Pz
15 P4
10
10
10
5
5
5
-400 Time (ms)
Freq. (Hz) Freq. (Hz)
BP
10
BP
-800
-400 Time (ms)
15 O1
15 O2
10
10
5
5 -400 Time (ms)
BP
-800
BP
-800
-400 Time (ms)
15 F3
15 Fz
15 F4
10
10
10
5
5
5
-800
Freq. (Hz)
-400 Time (ms)
15 C4
-800
-400 Time (ms)
BP
-800
-400 Time (ms)
BP
-800
15 C3
15 Cz
15 C4
10
10
10
5
5
5
-800
Freq. (Hz)
-800
15 Cz
-800
(b)
BP
15 C3
-800
Freq. (Hz)
-400 Time (ms)
-400 Time (ms)
BP
-800
-400 Time (ms)
BP
-800
15 P3
15 Pz
15 P4
10
10
10
5
5
5
-400 Time (ms)
Freq. (Hz)
-800
BP
-800
-400 Time (ms)
15 O1
15 O2
10
10
5
5 -800
-400 Time (ms)
BP
-800
BP
-800
-400 Time (ms)
-400 Time (ms)
BP
-400 Time (ms)
BP
-400 Time (ms)
BP
BP
-400 Time (ms)
BP
-400 Time (ms)
BP
-400 Time (ms)
BP
BP
Fig. 2 Average waveform of all recording sites of two subjects in the PS condition, time-locked to the button press (BP). a Subject-A. b Subject-B
Individual differences in perceptual switching rates; the role of occipital alpha and frontal theta band activity
0.39
0.13
0.00
0.00
0.52
0.52
0.39
0.39
0.26
10 5
15
15
0.13
0.00
0.00
0.52
0.52
0.39
0.39
5
15
0.13
0.00
0.00
0.52
0.52
0.39
0.39 0.26
0.13
0.13
0.00
0.00
0.52
0.52
0.39
0.39
0.26
10
0.26
0.13
0.26
10
0.26
0.13
0.26
10
(uV)
15
0.26
0.13
0.26
0.13
5 -400 Time (ms)
(uV)
0.13
0.00 -800
(uV)
0.52
0.39
(uV)
0.52
(uV)
0.00
(uV)
0.00
(uV)
5
Lower
subject-G Freq. (Hz)
0.13
0.26
10
0.26
0.13
(uV)
subject-E Freq. (Hz)
subject-D Freq. (Hz)
subject-B Freq. (Hz)
Higher
15
0.39
(uV)
0.26
(uV)
0.39
10
5
subject-H Freq. (Hz)
0.52
(uV)
15
5
Middle-level
Fz
0.52
(uV)
subject-A Freq. (Hz)
O1
349
0.00
BP
-800
-400 Time (ms)
BP
Fig. 3 Average waveforms of O1 and Fz recording sites of higher (subjects-A, B), middle-level (D, E), lower (G, H) switching rates subjects in the PS condition
frequent switchers would be related to motor functions for button pressing, it should have appeared also in the EC condition. However, no theta band activity occurred in this condition. Apparently, the theta band activity is neither specific to perceptual switching nor related to motor functions, but appears to be a generic, characteristic of frequent switchers’ perceptual processing. A hint to the possible function of the frontal theta band activity may be obtained by comparing the blink rates of frequent switchers to those of infrequent switchers. Figure 8a shows blink rates (1/s) during the experiments on the PS condition. There was a negative correlation between perceptual switching and blink rates. Figure 8b shows the normalized number of blinks (1/BP) before the button press responses. The number of blinks that appeared in a specified
time window were normalized by the number of button press responses (Top of Fig. 8b). The normalized numbers of blinks in the PS condition are shown in the middle part of Fig. 8b. Table 2 The values of ϕˆO1 and ϕˆFz of each subject Subject
ϕˆO1
ϕˆFz
A B C D E F G H
(−482 ms, 8.6 Hz) (−614 ms, 7.0 Hz) (−746 ms, 8.8 Hz) Null Null Null Null (−464 ms, 7.8 Hz)
(−478 ms, 7.6 Hz) (−358 ms, 6.2 Hz) (−478 ms, 6.8 Hz) Null Null Null (−752 ms, 9.2 Hz) Null
350
H. Nakatani, C. van Leeuwen
0.08
Freq. (Hz)
0.06 0.04 0.02 5
0.00
10
0.08 0.06
Freq. (Hz)
subject-A subject-B
Fz
O1
(a) 10
0.04 0.02 5
0.00 0.08 0.06
Freq. (Hz)
subject-C
10
0.04 0.02 5
0.00
(b) 10 Freq. (Hz)
9.3 Hz 7.9 Hz 6.7 Hz -405 ms
-836 ms
-612 ms
5 -800
-400 Time (ms)
BP -800 subject-A subject-B subject-C
6.1 Hz -254 ms
-400 Time (ms)
BP
∗ ∗ and and the margin of error for the estimated locations of ϕO1 and ϕFz in frequent Fig. 4 The distribution of bootstrap parameters ϕˆO1 and ϕˆFz ∗ ∗ and ϕˆFz of each switchers (subjects A-C). Left hand side: O1 recordings; right hand side: Fz recordings. a Colors indicate the probability that ϕˆO1 frequent switcher appeared in the range from (−900 ms, 4.0 Hz) to (0 ms, 10.0 Hz) in the time-frequency domain. b The elliptical margins were obtained from Eq. 10 as the margin of the estimated location of ϕO1 and ϕFz of each frequent switcher. Orange rectangular ranges surrounding the elliptical margins were used to distinguish frequent from infrequent switchers in Fig. 5b
B
0.8 A 0.6 0.4 H E 0.0 0.0
D
0.2
0.4 0.6 0.8 Probability - O1
1.0
B
C
0.8 0.6
A
0.4 0.2 D E F G
0.2 F
1.0
C
G
Probability - Fz
(b)
1.0
Probability - Fz
(a)
0.0 0.0
0.2
H 0.4 0.6 0.8 Probability - O1
1.0
Fig. 5 Scatter plots of the probability that the bootstrap parameter of each subject appeared in the specified range. The values of horizontal and ∗ ∗ ∗ ∗ and ϕˆFz , respectively. a The range was from (−900 ms, 4.0 Hz) to (0 ms, 10.0 Hz) for both ϕˆO1 and ϕˆFz . vertical axes represent the probability of ϕˆO1 ∗ ∗ and from (−612 ms, 6.1 Hz) to (−254 ms, 7.9 Hz) for ϕˆFz b The range was specified in Fig. 4b, from (−836 ms, 6.7 Hz) to (−405 ms, 9.3 Hz) for ϕˆO1
Individual differences in perceptual switching rates; the role of occipital alpha and frontal theta band activity
0.45
0.32
15
0.15
0.00
0.00
0.52
0.52
0.39
0.39
0.26
10 5
15
0.26
0.13
0.13
0.00
0.00
0.52
0.52
0.39
0.39
0.26
10
(uV)
5
0.30
0.16
(uV)
10
0.26
0.13
0.13
0.00
0.00
* 0.80
0.52
0.60
0.39
15
0.40
10 5
15
0.13
0.00
0.00
0.52
0.52
0.39
0.39
0.26
10
0.26
0.20
0.26
0.13
5 -400 Time (ms)
BP
(uV)
0.13
0.00 -800
(uV)
0.48
15
(uV)
0.00 * 0.60
(uV)
0.00 * 0.64
(uV)
0.17
(uV)
subject-D Freq. (Hz)
subject-B Freq. (Hz)
Higher subject-E Freq. (Hz)
0.34
0.25
Lower
subject-G Freq. (Hz)
0.51
(uV)
0.50
(uV)
0.75
10
5
subject-H Freq. (Hz)
* 0.68
(uV)
15
5
Middle-level
Fz
* 1.00
(uV)
subject-A Freq. (Hz)
O1
351
0.00 -800
-400 Time (ms)
BP
Fig. 6 Average waveforms of O1 and Fz recording sites of subjects-A, B, D, E, G, H in the SD condition. Subjects-A, B showed higher, D, E middle-level, and G, H lower switching rates, in the PS condition. A blue asterisk on the color scale indicates that the scale value is larger than that (0.52 µV ) in Fig. 3
Frequent switchers showed lower numbers of blinks than infrequent switchers both 1 s and just before the responses. Both frontal theta band activity and inhibition of blinks have been related to concentration on a task (Yamada 1998). If frequent switchers are able to concentrate better on the task than infrequent switchers, frequent switchers also would show inhibition of blinks just before the responses in the SD condition. As shown in the bottom of Fig. 8b, inhibition of blinks occurred in frequent switchers during the time in which the figure was changed and a response was required, whereas the number of blinks of frequent and infrequent switchers were similar when no response was required. We conclude that the difference in switching rates can be associated with subjects’ ability to concentrate on the task.
4 Discussion We investigated EEG activity in perceptual switching and found that alpha and theta band activity in occipital and frontal areas was dependent on individual switching rates. Using a bootstrap resampling technique, the timing and frequency component of occipital and frontal activity were estimated as ranging from (−836 ms, 6.7 Hz) to (−405 ms, 9.3 Hz) and from (−612 ms, 6.1 Hz) to (−254 ms, 7.9 Hz), respectively. The precision of the estimated range depends on the number of the data for analysis; larger numbers of data enable more precise estimation. These, however, will be within the range we specified. Comparing Fig. 3 with Figs. 6 and 7, we infer that occipital alpha band activity in perceptual switching
352
H. Nakatani, C. van Leeuwen
* 0.76
0.19
0.00
0.00
* 1.00
* 0.64
0.75
0.48
Higher
0.50
10 5 -400 Time (ms)
0.38
0.30
15
-800
0.57
BP
0.32
0.25
0.16
0.00
0.00 -800
-400 Time (ms)
(uV)
0.60
(uV)
0.90
10
(uV)
15
5
subject-B Freq. (Hz)
Fz
* 1.20
(uV)
subject-A Freq. (Hz)
O1
BP
Fig. 7 Average waveforms of O1 and Fz recording sites of subjects-A, B in the EC condition. Subjects-A, B showed higher switching rates in the PS condition. A blue asterisk on the color the scale value indicates that scale is larger than that (0.52 µV ) in Fig. 3
is specific to this process in frequent switchers. By contrast, frontal theta band activity is not specific to perceptual switching. Frequent switchers show this activity also in the stimulus driven condition. Inhibition of blinks in frequent switchers is observed in both the perceptual switching and the stimulus driven conditions. Recently, Leopold et al. (2002) have reported that intermittent presentation of an ambiguous figure causes lower rates of perceptual switching than continuous presentation. They concluded that continuous viewing of an ambiguous figure was a requirement for the initiation of the perceptual switching processes. Blinks during continuous presentation can be considered similar to intermittent presentation. Thus, our findings in Fig. 8a are consistent with their report. Horlitz and O’Leary (1993) have reported that attention to ambiguous figures enhances perceptual switching rates, and Stern et al. (1993) have reported that endogenous blinking is inhibited by attentional processes. Inhibition of blinks in our study was related to concentration on the tasks. Frequent switchers pay more attention to the visual stimulus. They do so, however, only when a response is required. In the stimulus driven condition, no such inhibition is observed for frequent switchers when no response was required. Therefore, as shown in Fig. 8b, frequent switchers showed smaller number of blinks both just before and 1 s before the responses in the perceptual switching condition and when the response was required in the stimulus driven condition. These findings contribute to our conclusion that attentional effort causes the lower blink rates that provide enough continuous viewing time to initiate perceptual switching process, and this leads to higher perceptual switching rates. If perceptual switching rates are controlled by attentional effort, several distant brain areas would be related to the switching, because widely separated cortical areas is related to attentional processes ( Corbetta and Shulman 2002). This view is consistent with Str¨uber et al. (2000)’s report of switching rates dependent gamma band activity as well as several functional magnetic resonance imaging (fMRI)
studies ( Kleinschmidt et al. 1998; Lumer et al. 1998). One may wonder if any of these areas controls the perceptual switching rate. One candidate area would be frontal cortex, because frontal lobe lesions result in low switching rates ( Meenan and Miller 1994; Ricci and Blundo et al. 1990). Furthermore, it has been reported that damage to the right frontal cortex causes difficulty to sustain attention to a stimulus ( Meenan and Miller 1994). We found a difference in frontal activity between frequent and infrequent switchers. Frequent switchers showed high amplitude theta band activity around the frontal midline region. This theta band activity, Fm-theta, is known to be related to a variety of cognitive processes ( Inanaga 1998; Jensen and Tesche 2002; Kahana et al. 2001; Mizuhara et al. 2004). It has been reported that Fm-theta activity and inhibition of blink is related to concentration on a task ( Yamada 1998). Fm-theta tends to show maximal amplitude at Fz site even if sources of the activity are localized a part of frontal areas because of the summation of the scattered and widely distributed extracellular electrical currents ( Sasaki et al. 1996). Fm-theta of frequent switchers was related to sustaining attention to the task, whereas brain areas related to sustaining attention localize in right frontal cortex as suggested by the lesion study ( Meenan and Miller 1994). Accordingly, this activity appeared not only in the perceptual switching but also in the stimulus driven conditions. As for the occipital alpha band activity, Yamagishi et al. (2003) reported that an increase in lower alpha band (7–10 Hz) activity within the calcarine cortex is related to attention to the visual stimulus. One possible role of alpha band activity in frequent switchers is to enhance the efficiency of processing information about visual representations. Kornmeier and Bach ( 2004; 2005) reported positive and negative components of occipital activity which suggest a crucial role of visual areas in perceptual switching processes. The alpha band activity in occipital areas we observed in frequent switchers prior to the switch was different from theirs, because if this activity were essential for switching, it should
Individual differences in perceptual switching rates; the role of occipital alpha and frontal theta band activity
353
(a) 0.6 In the PS condition
Blink rates (1/s)
0.5 0.4 0.3 0.2 0.1 0.0
(b)
A,B,C Frequent Blink
D,E,F G,H Infrequent
Blink
BP
Blink
T (s) normalized number of blinks = number of blinks in T / number of BPs (1/BP) In the PS condition
-1.7 s
BP
-1.0 s
Freq. 0.082 (1/BP) Infreq. 0.256 (1/BP)
-0.7 s
0.0 s
0.065 (1/BP) 0.262 (1/BP)
In the SD condition reaction time was about 0.7 s presented figure BP was changed -1.7 s
-1.0 s
Freq. 0.162 (1/BP) Infreq. 0.177 (1/BP)
-0.7 s
0.0 s
0.067 (1/BP) 0.172 (1/BP)
Fig. 8 Comparison of blink rates between frequent and infrequent switchers. a Blink rates (1/s) during the experiments of the PS condition. Average blink rates of three groups, (subjects-A, B, C) (D, E, F) (G, H), are presented. b Normalized number of blinks (1/BP) before the button press responses. Top The definition of normalized numbers of blinks. The number of blinks that appeared in a specified time window, T, indicated by a red arrow were normalized by the number of button press responses. Middle Normalized numbers of blinks in the PS condition. Bottom Normalized numbers of blinks in the SD condition
appear in infrequent switchers as well. On the other hand, we inferred that the alpha activity merely facilitates perceptual switching. 5 Conclusion One of the differences between frequent and infrequent switchers appears to be that perceptual switching process is governed by attentional processes distributed over several distant brain areas. In frequent switchers, attentional modulations were
observed as sequential occipital alpha and frontal theta band activity. These did not occur in infrequent switchers. Attentional control of perceptual switching process, therefore, was shifted in frequent switchers from occipital to frontal areas, resulting in facilitation of switching behavior. Acknowledgements We thank Dr.Andrey R. Nikolaev of RIKEN Brain Science Institute for comments on an early manuscript, Dr. Tsutomu Murata of National Institute of Information and Communications Technology for the discussion, and Dr. Hiroaki Mizuhara of RIKEN Brain Science Institute for the advice for the revision of the manuscript.
354
References Attneave F (1971) Multistability in perception. Sci Am 225:63–71 Bartnik EA, Blinowska KJ, Durka PJ (1992) Single evoked potential reconstruction by means of wavelet transform. Biol Cybern 67:175– 181 Borsellino A, De Marco A, Allazetta A, Rinesi S, Bartolini B (1972) Reversal time distribution in the perception of visual ambiguous stimuli. Kybernetik 10:139–144 Brown KT (1955) Rate of apparent change in a dynamic ambiguous figure as a function of observation-time. Am J Psychol 68:358–371 Chui CK, Wang JZ (1992) On compactly supported spline wavelets and a duality principle. Trans Amer Math Soc 330:903–915 Corbetta M, Shulman GL (2002) Control of goal-directed and stimulus-driven attention in the brain. Nat Rev Neurosci 3:201–215. Efron B (1979) Bootstrap methods: another look at the jackknife. The Annals of Statistics 7:1–26 Efron B, Tibshirani R (1986) Bootstrap methods for standard errors, confidence intervals, and the other measures of statistical accuracy. Statistical Science 1:54–77 Einh¨auser W, Martin KA, K¨onig P (2004) Are switches in perception of the Necker cube related to eye position? Eur J Neurosci 20:2811– 2818 Horlitz KL, O’Leary A (1993) Satiation or availability? Effects of attention, memory, and imagery on the perception of ambiguous figures. Percept Psychophys 53:668–681. Inanaga K (1998) Frontal midline theta rhythm and mental activity. Psychiatry Clin Neurosci 52:555–566 ¨ Ba¸sar-Ero˘glu C, Ademo˘glu A, Demiralp T, Miener M, ˙so˘glu-Alka¸c U, I¸ Stadler M (1998) Analysis of the electroencephalographic activity during the Necker cube reversals by means of the wavelet transform. Biol Cybern 79:437–42 ¨ Basar-Eroglu C, Ademoglu A, Demiralp T, Miener Isoglu-Alka¸c U, M, Stadler M (2000) Alpha activity decreases during the perception of Necker cube reversals: an application of wavelet transform. Biol Cybern 82:313–320 Ito J, Nikolaev AR, Luman M, Aukes MF, Nakatani C, van Leeuwen C (2003) Perceptual switching, eye movements, and the bus paradox. Perception 32:681–698 Jensen O, Tesche CD (2002) Frontal theta activity in humans increases with memory load in a working memory task. Eur J Neurosci 15:1395–1399 Kahana MJ, Seelig D, Madsen JR (2001) Theta returns. Curr Opin Neurobiol 11:739–744 Kleinschmidt A, Buchel C, Zeki S, Frackowiak RS (1998) Human brain activity during spontaneously reversing perception of ambiguous figures. Proc R Soc Lond B Biol Sci 265:2427–2433 Kornmeier J, Bach M (2004) Early neural activity in Necker-cube reversal: evidence for low-level processing of a gestalt phenomenon. Psychophysiology 41:1–8
H. Nakatani, C. van Leeuwen
Kornmeier J, Bach M (2005) The Necker cube-an ambiguous figure disambiguated in early visual processing. Vision Res 45:955–960 Lehky SR (1995) Binocular rivalry is not chaotic. Proc R Soc Lond B Biol Sci 259:71–76 Leopold DA Logothetis NK (1999) Multistable phenomena: changing views in perception. Trends Cogn Sci 3:254–264 Leopold DA, Wilke M, Maier A, Logothetis NK (2002) Stable perception of visually ambiguous patterns. Nat Neurosci 5:605–609 Lewicki MS (1998) A review of methods for spike sorting: the detection and classification of neural action potentials. Network 9:R53–R78 Lumer ED, Friston KJ, Rees G (1998) Neural correlates of perceptual rivalry in the human brain. Science 280:1930–1934 Meenan JP, Miller LA (1994) Perceptual flexibility after frontal or temporal lobectomy. Neuropsychologia 32:1145–1149 Mizuhara H, Wang LQ, Kobayashi K, Yamaguchi Y (2004) A longrange cortical network emerging with theta oscillation in a mental task. Neuroreport 15:1233–1238 Murata T, Matsui N, Miyauchi S, Kakita Y, Yanagida T (2003) Discrete stochastic process underlying perceptual rivalry. Neuroreport 14:1347–1352 Nakatani H, Watanabe T, Hoshimiya N (2001) Detection of nerve action potentials under low signal-to-noise ratio condition. IEEE Trans Biomed Eng 48:845–9 Press WH, Teukolshy SA, Vetterling WT, Flannery BP (1993) Numerical Recipes in C (Japanese translation). Gijutsu Hyoron Sha Ricci C, Blundo C (1990) Perception of ambiguous figures after focal brain lesions. Neuropsychologia 28:1163–1173 Rock I, Hall S, Davis J (1994) Why do ambiguous figures reverse. Acta Psychol (Amst) 87:33–59 Sasaki K, Tsujimoto T, Nishikawa S, Nishitani N, Ishihara T (1996) Frontal mental theta wave recorded simultaneously with magnetoencephalography and electroencephalography. Neurosci Res 26:79–81 Stern JA, Walrath LC, Goldstein R (1993) The endogenous eyeblink. Psychophysiology 21:22–33 Str¨uber D, Basar-Eroglu C, Hoff E, Stadler M (2000) Reversal-rate dependent differences in the EEG gamma-band during multistable visual perception. Int J Psychophysiol 38:243–252 Yamada F (1998) Frontal midline theta rhythm and eyeblinking activity during a VDT task and a video game: useful tools for psychophysiology in ergonomics. Ergonomics 41:678–688 Yamagishi N, Callan DE, Goda N, Anderson SJ, Yoshida Y, Kawato M (2003) Attentional modulation of oscillatory activity in human visual cortex. Neuroimage 20:98–113. Zhou YH, Gao JB, White KD, Merk I, Yao K (2004) Perceptual dominance time distributions in multistable visual perception. Biol Cybern 90:256–263
