Necker cube reversals during long-term EEG recordings: Sub-bands of alpha activity

International Journal of Psychophysiology 59 (2006) 179 – 189 www.elsevier.com/locate/ijpsycho

Necker cube reversals during long-term EEG recordings: Sub-bands of alpha activity ¨ mmu¨han I˙s¸og˘lu-Alkac¸ a,*, Daniel Stru¨ber b,c U b

a˙ Istanbul University, I˙stanbul Medical Faculty, Department of Physiology, 34390 C¸apa-I˙stanbul, Turkey University of Bremen, Institute of Psychology and Cognition Research and Center for Cognitive Sciences, Bremen, Germany c Hanse Institute for Advanced Study, Delmenhorst, Germany

Received 31 August 2004; accepted 7 May 2005 Available online 14 July 2005

Abstract Reversible figures, such as the Necker cube, make up a well-known class of visual phenomena in which an invariant stimulus pattern gives rise to at least two different perceptual interpretations. A better understanding of the neurophysiological processes underlying perceptual reversals might help to disentangle bottom-up from top-down influences on multistable perception. Recently, we reported alpha activity decrease during multistable visual perception. The aim of the present study was to define more specifically the functional roles of the EEG alpha band during the perception of Necker cube reversals by subdividing the extended alpha band into three sub-bands (lower-1 alpha, lower-2 alpha, upper alpha). We employed a long-term recording condition, during which 10 healthy participants observed the Necker cube for approximately 60 min and responded by pressing a button to any perceived reversal. The results showed a reversal induced alpha desynchronization for the lower alpha bands, with the lower-2 alpha desynchronization differing across the time course of the experiment. The upper alpha band demonstrated no reliable effects. It is concluded that the lower-1 alpha desynchronization reflects an automatic arousal reaction which triggers attentional processing in a bottom-up manner, whereas the lower-2 alpha desynchronization is related to attentional processes that are achieved by top-down control with limited resources. The lack of reliable effects in the upper alpha band is presumably due to the relatively low semantic task demands in figure reversal. D 2005 Elsevier B.V. All rights reserved. Keywords: Alpha band; Sub-bands of alpha frequency; Multistable perception; Reversible figures; Event-related potentials; Necker cube

1. Introduction Reversible or ambiguous figures, such as the Necker cube (Fig. 1), make up a well-known class of visual phenomena in which an invariant stimulus pattern gives rise to at least two different perceptual interpretations and, therefore, allows to dissociate perceptual from stimulusdriven mechanisms. Traditionally, there are two general approaches for the explanation of reversible figures. One favours the more passive or bottom-up processes of neural satiation or fatigue and recovery among cortical structures, originally

* Corresponding author. Tel.: +90 212 414 22 52; fax: +90 212 414 22 53. E-mail address: [email protected] (U¨. I˙s¸ og˘lu-Alkac¸). 0167-8760/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpsycho.2005.05.002

proposed by Ko¨hler (1940). The other approach stresses top-down influences on illusory reversals by means of cognitive processes like intention, attention, expectation, imagining, and the matching of memory representations (Rock et al., 1994). A better understanding of the neurophysiological processes underlying perceptual reversals might help to disentangle bottom-up from top-down influences on multistable perception. In this context, the timing of the involved neural mechanisms is of special importance, which can be analyzed by means of EEG or MEG providing high temporal resolution. Previous ERP studies on multistable perception found a late P300-like component induced by the figure reversal (O’Donnell et al., 1988; Basar-Eroglu et al., 1993; Isoglu-Alkac et al., 1998, 2000; Stru¨ber et al., 2001; Stru¨ber and Herrmann, 2002), which indicates the conscious

180

¨ . I˙s¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U

Fig. 1. The Necker cube.

recognition of perceptual change and, therefore, the closure of the reversal process. Recently, Kornmeier and Bach (2004) also reported a late positive component, but this positivity was preceded by an early negative deflection at posterior locations. Therefore, the authors interpreted the early negativity as a correlate of bottom-up processing during figure reversal. In the frequency domain, EEG studies of our research group on multistable perception were mainly concerned with the gamma and alpha band. Basar-Eroglu et al. (1996) analyzed the EEG gamma-band (30-50 Hz) during unstable (reversal phase) and stable (non-reversal phase) perceptual states while subjects continuously observed an ambiguous motion paradigm. The results demonstrated enhanced gamma-band activity compared to spontaneous EEG, especially at right frontal area during the reversal phase. A frontal topography of the gamma activity was replicated in two later studies (Stru¨ber et al., 2000, 2001), and interpreted as reflecting attentional top-down processes related to monitoring demands during multistable perception, i.e. attentively recognizing and reporting the reversals by a motor response. Concerning the alpha band, Isoglu-Alkac et al. (2000) reported a decrease of alpha activity in the EEG during perceptual reversals of the Necker cube. The timing of this alpha desynchronization coincided with the occurrence of a P300-like component, as described in the ERP studies above. In an MEG study, Stru¨ber and Herrmann (2002) also found an alpha activity decrease during figure reversal which started already in a relatively long time interval preceding the reversal. This finding was interpreted as a passive bottom-up process which initiates a reversal before it appears in visual awareness. Similarly, Mu¨ller et al. (1999) have shown that perceptual alternations were preceded by a decrease of the EEG theta and alpha power (reflecting a vigilance decrease), and followed by an increase of EEG frequencies (reflecting an arousal reaction). These results are in accordance with the interpretation that decreasing alpha activity reflects bottom-up

processing, while higher frequency bands signal attentional top-down processes. As demonstrated by these EEG and MEG studies, the analysis of different frequency bands seems to be a promising tool for studying the interrelation of top-down and bottom-up processes in multistable perception. The main purpose of the present study was to define more specifically the functional roles of the EEG alpha band during the perception of Necker cube reversals by subdividing the extended alpha band into three sub-bands. Additionally, we intended to study possible influences of an overall decrease in vigilance on the alpha desynchronization during perceptual reversals, which we have shown previously (Isoglu-Alkac et al., 2000). Therefore, we introduced a long-term recording condition, during which healthy participants observed the Necker cube for 66.6 min. The rationale of this experiment was, that if alpha desynchronization reflects a passive and automatic bottom-up processing, then it should not depend on large-scale changes in vigilance, which are expected to be induced by the 66.6 min observation period. When, on the other hand, vigilance or other attention-based processes play the dominant role in figure reversal, then we would expect a decreasing amount of alpha desynchronization during the long-term recording condition. However, when analyzing the alpha frequency range, one has to take into account that a variety of functional correlates have been found for the alpha band. According to Basar, the alpha rhythm is a prototype of a dynamic process which governs a large ensemble of integrative brain functions (Basar et al., 1992). Alpha patterns can be induced, evoked, and emitted (Galambos, 1992; Basar, 1980, 1999; Basar et al., 1997a,b). Also alpha desynchronization is not a unitary phenomenon. Event-related desynchronization or ERD, as coined by Pfurtscheller and Aranibar (1977), is described as the attenuation or blocking within the alpha and other frequency bands (Pfurtscheller and Klimesch, 1992; Pfurtscheller et al., 1996; Krause et al., 1997). ERD was observed during visual stimulation, voluntary movement, and cognitive activity by several investigators. Recent research on eventrelated changes in alpha band power suggests that desynchronization in the alpha frequency range is associated with active information processing, and that different alpha frequencies have quite different functions (Klimesch, 1999; Klimesch et al., 2000, 2004). Accordingly, we analyzed different alpha sub-bands in order to prove for possible differences in the patterns of desynchronization during figure reversal. As suggested by Klimesch (Klimesch et al., 1997a,b,c; Klimesch, 1999), we used the individual alpha frequency (IAF) as an anchor point and distinguished three alpha bands with a width of 2 Hz each: two lower alpha bands (below IAF) and one upper alpha band (above IAF). The comparison between these sub-bands allowed us to establish whether the alpha desynchronization was modulated differently in three alpha

¨ . I˙s¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U

frequencies during long-term EEG recording of Necker cube reversals.

2. Methods 2.1. Participants Ten healthy, right-handed volunteers (5 males) participated in this study and received financial remuneration. The mean age was 26 T 2.3 years. They were psychology students and members of Bremen University. In order to obtain compliance of the subjects during the long-term recording condition, we selected subjects high in motivation and experience with EEG measurements during visual tasks. All participants had normal or corrected-to-normal vision and none of them had neurological or psychiatric disorders and all gave written informed consent. They were instructed to keep their eyes open, to maintain fixation all the time, and to minimize blinking and eye movements. No breaks were given during the long-term recording. 2.2. Stimulus pattern We used the Necker cube as stimulus with 10 cm line length of the outline cube. At a viewing distance of 150 cm, the resulting visual angle was 3.8-. The stimulus was presented as white lines on a black background. The upper right corner of the cube served as fixation point (seen from the upright position of the cube) (see Fig. 1). 2.3. Experimental procedure The participants sat in a soundproof and echo-free room, which was dimly lit. They were informed about the reversibility of the Necker cube and instructed to look at the fixation point all the time and to press a button immediately following reversals. They indicated the time period of the perceptual reversal by shortly and slightly pressing down a button with their right index finger, thus breaking a contact impinging on one channel of the EEGwriter. An additional electromyographic (EMG) recording was carried out in order to estimate the contribution of the motor potentials to the main recording. 2.4. Electrophysiological recording The EEG was registered at recording sites Fz, Cz, Pz, and Oz according to the international 10/20 system. Linked earlobe electrodes served as reference. All electrode impedances were maintained at less than 5 kV. The electrooculogram (EOG) was registered from medial upper and lateral orbital rim of the right eye. The EEG was amplified by means of a Nihon Kohden (EEG 4421 G) apparatus with band limits at 0.1– 70 Hz (24 dB/octave). An additional notch filter (36 dB/octave) was also applied to remove the

181

main interferences. For the recording of EOG, the time constant was set at 0.3 s and a low pass filter at 70 Hz was applied. All channels were displayed on paper and on-line by monitor scope in order to observe both single trials and averaged trials. The EEG was digitized with a sampling rate of 256 points/s and stored on a computer disc memory. Each button press was also recorded as a simple on– off signal on one EEG channel. The recording session was made up of 2000 EEG epochs with a duration of 2000 ms each, resulting in an overall duration of 66.6 min per session and per subject. 2.5. Data analysis 2.5.1. Selective averaging and rejection of artifacts An automatic on-line artifact rejection procedure was used for the elimination of global artifactual EEG-epochs. These epochs contained movement artifacts, excessive muscle activities, and amplitudes exceeding 50 AV at any electrode. Additionally, in an off-line procedure, the on-line recorded, digitized, and stored single artifact-free EEG epochs were selected. The EOG channel was visually inspected for each trial and trials with eye movement or blink artifact were rejected. The mean number of artifact-free EEG epochs was between 900 and 1000 for each recording. Therefore, we equalized the number of artifact-free sweeps for each participant to 900. These artifact-free epochs were averaged time-locked to the onset of the button press response. Note that these 900 epochs represent a subgroup of the overall recorded 2000 epochs in which subjects had reported a reversal. That is, we analyzed only the reversal-related artifact-free EEG epochs. The number of the reversal-related epochs without artifact elimination, i.e. the individual reversal rates, had a range from 977 to 1820 during the 66.6 min recording time (mean: 1276.3, S.D.: 347.1). According to this relation, two third of the relevant data were artifact-free across subjects (900 epochs from 1276.3 on average). 2.5.2. Selection of time windows Since the perceived change of Necker cube is generated endogenously, response-locked analysis of the time interval preceding the button press (t = 0 ms) was applied in order to analyze the reversal-related activity. Data were recorded from 1000 ms before the button press (t = 1000 to 0 ms) until 1000 ms after (t = 0 ms to 1000 ms). In our prior study on ERPs during Necker cube reversals (Isoglu-Alkac et al., 1998), we showed that a P300-like wave was induced by figure reversals. As described previously (Isoglu-Alkac et al., 2000), we used the timing of this positive wave in order to define two time windows for analyzing alpha power changes related to the perceptual reversals. One time window corresponded to the mean peak latency of the positive wave, i.e. the mean time difference between button press (t = 0 ms) and peak amplitude (t = 256.7 T 60 ms; left side of Fig. 2). We defined the length of this time window as T 3 standard deviations around the mean peak latency (t2:

182

¨ . ˙Is¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U

Fig. 2. Grand average responses in the broad band (0.1 – 30 Hz), delta band (0 – 4 Hz), and extended alpha band (6 – 12 Hz) for Fz, Cz, Pz, and Oz channels. xaxis: time in ms (t = 0 ms refers to the button press); y-axis: amplitude in AV.

436 to 76 ms, Figs. 3 and 4). Alpha power in this time window was compared to another time window of the same length preceding the occurrence of the positive wave (t1: 796 to 436 ms, Figs. 2 and 3). We restricted our analysis to relatively short time intervals before the button press in order to avoid an overlap of reversal-related activity with motor activity following the button press. As mentioned above, the mean reversal rate of the subjects in this

Fig. 3. Temporal relation between the grand average delta (0 – 4 Hz) and extended alpha band (6 – 12 Hz) response for Cz location. x-axis: time in ms (t = 0 ms refers to the button press); y-axis: amplitude in +/ 5 AV for delta and +/ 2.5 AV for alpha band. t1 and t2 refer to the time windows in which the alpha desynchronization was measured (t1: 796 to 436 ms, t2: 436 to 76 ms). For definition of the time window length see text.

experiment was 1276.3 reversals during the 66.6 min recording, which corresponded to 19 reversals per minute or one reversal every three seconds. Due to this high mean reversal rate of the subjects, it was not feasible to reliably define a Fneutral_ time window as a control condition which contained neither perception-related nor response-related activity (see Basar-Eroglu et al., 1996 for a comparison of reversal-related activity with a non-reversal phase). 2.5.3. Selection of alpha frequency bands In order to analyze alpha power change in the described time windows, we computed root mean square (RMS) values from the averaged sweeps for three alpha frequency subbands. Each single RMS value represented the mean alpha amplitude of the data points for the selected time windows. Since the alpha frequency is subject to large interindividual differences, we used the peak frequency of the dominant EEG frequency in the alpha band, averaged over all recording sites, as an anchor point to adjust frequency bands individually for each subject (Klimesch et al., 1998; Klimesch, 1999). The frequency windows had a standard bandwidth of 2 Hz and were the same for all participants (Klimesch et al., 1998; Klimesch, 1999). Three alpha frequency sub-bands with a bandwidth of 2 Hz each were defined by using the individual alpha frequency (IAF) as the individual anchor point: (IAF 4) to (IAF 2), (IAF 2) to IAF and IAF to (IAF + 2), termed lower-1 alpha (6 –8 Hz), lower-2 alpha (8 –10 Hz), and upper-alpha (10 –12 Hz). Each IAF was determined on the basis of clear peaks in the EEG power spectra. Averaged over the entire sample, IAF was 10.37 Hz with a range from 8.98 Hz to 12.77 Hz. 2.5.4. Alpha activity during long-term recording As described above, alpha RMS values were calculated for three sub-bands in each of two time windows (t1, t2).

¨ . ˙Is¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U

183

Fig. 4. Grand average of event-related alpha oscillations in the three sub-bands for Fz, Cz, Pz, and Oz locations. x-axis: time in ms (t = 0 ms refers to the button press); y-axis: amplitude in AV.

This analysis was conducted across three different time sections (I, II, III) of the entire 66.6 min recording session. For analyzing possible effects related to the long-term condition, we compared grand averages of 300 epochs per participant, taken from the beginning (I), middle (II), and end (III) of the experiment. 2.6. Statistical analyses For analyzing differences between the alpha sub-bands, we conducted a three-way analysis of variance (ANOVA) for repeated measures with the factors time window (t1, t2), frequency (lower-1, 6– 8 Hz; lower-2, 8– 10 Hz, and upper, 10 –12 Hz), and channel (Fz, Cz, Pz, Oz). For analyzing effects of long-term recording, we conducted three-way ANOVAs for repeated measures with the factors time window (t1, t2), section (I: beginning, II: middle, III: end), and channel (Fz, Cz, Pz, Oz) separately for the three alpha sub-bands.

3. Results 3.1. ERPs and event-related alpha oscillations Fig. 2 shows the ERPs induced by figure reversals and the activity of the extended alpha band (6 –12 Hz) for the midline electrodes. As can be clearly seen in the 0.1 – 30 Hz frequency range and in the delta range (0 –4 Hz), a P300like positive wave occurred approximately 250 ms prior to the button press (t = 0), which fits to earlier results of our research group (Basar-Eroglu et al., 1993; Isoglu-Alkac et al., 1998). The extended alpha band (6– 12 Hz) desynchron-

ized during the transition from time window t1 to t2, most clearly at Fz and Cz electrodes. As presented in Fig. 3 for Cz location, the time course of the alpha desynchronization (6 –12 Hz) overlapped with the occurrence of the positive slow wave in the delta range (0– 4 Hz), which is also a replication of our previous results (Isoglu-Alkac et al., 2000). Fig. 4 shows the grand averages of the event-related alpha oscillations for each location and the three alpha sub-bands, respectively. As can be seen, the lower-1 alpha band (6 –8 Hz) desynchronized strongest at posterior sites (Cz, Pz), while the lower-2 alpha band (8 – 10 Hz) desynchronized strongest at anterior sites (Fz, Cz). For the upper alpha band (10 – 12 Hz), no clear modulation could be observed. 3.2. Comparison of alpha RMS values in the three sub-bands For statistical analyses of alpha activity differences in three sub-bands and two time windows, we used RMS values based on the average of 900 epochs per participant as obtained from the 66.6 min recording time (see Figs. 5a, 6a, and 7a). A three-factor (2 time windows  3 frequencies  4 channels) ANOVA revealed a significant main effect for the factors time window [ F(1,9) = 91.59, p < 0.0001] and frequency [ F(2,18) = 5.79, p < 0.01], indicating an overall decrease in alpha power from t1 to t2 (desynchronization), and a difference of alpha RMS values between sub-bands, with lowest alpha power in the upper alpha band. Furthermore, there were significant interaction effects for factors time window  frequency [ F(2,18) = 42.23, p < 0.0001], time window  channel [ F(3,27) =3.98,

¨ . I˙s¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U

a

b

AVERAGED RESPONSES (6-8 Hz)

RMS Values

8 t1

4

t2

0 Fz

Cz

Pz

AVERAGED RESPONSES-I (6-8)

RMS Values

184

8 t1

4

t2

0 Fz

Oz

d

8 t1

4

t2

0 Cz

Pz

Oz

AVERAGED RESPONSES-III (6-8 Hz)

RMS Values

RMS Values

AVERAGED RESPONSES-II (6-8 Hz)

Fz

Pz

Channels

Channels

c

Cz

8 t1

4

t2

0 Fz

Oz

Cz

Pz

Oz

Channels

Channels

Fig. 5. Averaged RMS values for the lower-1 alpha band. x-axis: channel; y-axis: amplitude in AV. Black bars show alpha RMS values in t1, white bars show alpha RMS values for t2. The difference between t1 and t2 reflects the amount of desynchronization. (a) Grand average RMS values from 900 epochs per participant. (b) RMS values based on the average of 300 epochs per participant taken from the first section (I, beginning). (c) RMS values based on the average of 300 epochs per participant taken from the second section (II, middle). (d) RMS values based on the average of 300 epochs per participant taken from the third section (III, end).

p < 0.01], frequency  channel [ F(6,54) = 4.4, p < 0.001], and time window  frequency  channel [ F(6,54) = 4.24, p < 0.001]. As suggested by this pattern of interactions, alpha sub-band activity differed across channels, and, most importantly, the amount of desynchronization differed with respect to alpha sub-band frequency and scalp distribution. In order to analyze each alpha sub-band in more detail, we conducted two-way ANOVAs with the factors time window and channel for each sub-band separately. For the lower-1 alpha band (6– 8 Hz), the ANOVA revealed a significant main effect for the factors time window [ F(1,9) = 59.96, p < 0.0001] and channel [ F(3,27) = 51.5,

b

AVERAGED RESPONSES (8-10Hz)

8 t1

4

t2

0 Fz

Cz

Pz

RMS Values

RMS Values

a

p < 0.0001], and a significant interaction time window  channel [ F(3,27) = 5.2, p < 0.01]. Similarly for the lower-2 alpha band (8 –10 Hz), there were significant main effects for time window [ F(1,9) = 244.84, p < 0.0001], channel [ F(3,27) = 4.37, p < 0.01], and a significant interaction time window  channel [ F(3,27) = 5.95, p < 0.003]. For the upper alpha band (10 – 12 Hz), however, the ANOVA did not yield any significant effects. This pattern of results indicates a pronounced alpha desynchronization (t1 > t2) with a centro-parietal maximum for the lower-1 alpha band (Fig. 5a), and a fronto-central maximum for the lower-2 alpha band (Fig. 6a), whereas the upper alpha band

AVERAGED RESPONSES-I (8-10 Hz) 8 t1

4

t2

0

Oz

Fz

Channels

4

t2

0 Channels

Oz

RMS Values

RMS Values

t1

Pz

Oz

d AVERAGED RESPONSES-III (8-10 Hz)

8

Cz

Pz

Channels

c AVERAGED RESPONSES-II (8-10 Hz)

Fz

Cz

8 t1

4

t2

0 Fz

Cz

Pz

Oz

Channels

Fig. 6. Averaged RMS values for the lower-2 alpha band. x-axis: channel; y-axis: amplitude in AV. Black bars show alpha RMS values in t1, white bars show alpha RMS values for t2. The difference between t1 and t2 reflects the amount of desynchronization. (a) Grand average RMS values from 900 epochs per participant. (b) RMS values based on the average of 300 epochs per participant taken from the first section (I, beginning). (c) RMS values based on the average of 300 epochs per participant taken from the second section (II, middle). (d) RMS values based on the average of 300 epochs per participant taken from the third section (III, end).

¨ . I˙s¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U

b

AVERAGED RESPONSES (10-12 Hz) 8 t1

4

t2

0 Fz

Cz

Pz

AVERAGED RESPONSES-I (10-12 Hz)

RMS Values

RMS Values

a

8 t1

4

t2

0

Oz

Fz

Channels AVERAGED RESPONSES-II (10-12 Hz) 8 t1

4

t2

0 Fz

Cz

Pz

Cz

Pz

Oz

Channels

d

AVERAGED RESPONSES-III (10-12 Hz)

RMS Values

RMS Values

c

185

8 t1

4

t2

0

Oz

Fz

Cz

Pz

Oz

Channels

Channels

Fig. 7. Averaged RMS values for the upper alpha band. x-axis: channel; y-axis: amplitude in AV. Black bars show alpha RMS values in t1, white bars show alpha RMS values for t2. The difference between t1 and t2 reflects the amount of desynchronization. (a) Grand average RMS values from 900 epochs per participant. (b) RMS values based on the average of 300 epochs per participant taken from the first section (I, beginning). (c) RMS values based on the average of 300 epochs per participant taken from the second section (II, middle). (d) RMS values based on the average of 300 epochs per participant taken from the third section (III, end).

was not modulated by time window and/or scalp distribution (Fig. 7a). 3.3. Comparison of alpha RMS values in the three sub-bands during long-term recording For analyzing possible effects of the long-term recording on alpha desynchronization of the sub-bands, we divided the 900 epochs of the entire recording session into three sections, each containing 300 epochs per participant, taken from the beginning (I), middle (II), and end (III) of the experiment. For section I, the grand average RMS values for t1 and t2 in the lower-1 alpha band are presented in Fig. 5b as a function of electrode site. Fig. 5c and d illustrate the grand average RMS values for sections II and III. Figs. 6b – d and 7b –d show the same information for the lower-2 and upper alpha band, respectively. Table 1 summarizes the results of the separate three-factor repeated measures analyses of variance performed on each alpha sub-band (2 time windows  3 sections  4 channels). The lower-1 alpha band demonstrated strong desynchronization for all sections, with a topography slightly changing from anterior to more posterior locations during

the experiment as indicated by the time window  channel interaction (as a trend) and the significant three-way interaction. For the lower-2 alpha band, there was also a strong desynchronization for all sections. However, the amount of desynchronization differed across sections as indicated by the interaction between time window and section, with comparable values for section I, II, and a decrease at section III. Topographically, desynchronization was larger at fronto-central location as indicated by the strong time window  channel interaction. The upper alpha band demonstrated no relevant effects, with only a significant main effect for channel unrelated to desynchronization and section.

4. Discussion The present study aimed at defining the functional roles of different alpha sub-bands for perceptual reversals during long-term EEG recordings with the Necker cube. The results demonstrated a marked alpha desynchronization induced by figure reversals for the lower-1 (6– 8 Hz) and lower-2 (8– 10 Hz) alpha band, which was absent for the upper alpha band

Table 1 Summary of F-ratios and probabilities from the 3-factor (2 time windows  3 sections  4 channels) ANOVA performed on the lower-1 alpha band, the lower-2 alpha band, and the upper alpha band Source (df)

Lower-1 alpha (6 – 8 Hz)

Lower-2 alpha (8 – 10 Hz)

Upper alpha (10 – 12 Hz)

F

p

F

p

F

p

Time window (1,9) Channel (3,27) Section (2,18) T  C (3,27) T  S (2,18) C  S (6,54) T  C  S (6,54)

59.96 0.59 0.84 2.38 0.27 0.93 2.50

0.0001 0.62 0.44 0.09 0.76 0.48 0.03

244.46 4.37 1.89 5.95 4.80 0.35 0.25

0.0001 0.01 0.17 0.003 0.02 0.90 0.95

1.70 6.54 0.48 2.39 0.85 0.24 0.45

0.22 0.002 0.62 0.09 0.44 0.95 0.84

186

¨ . I˙s¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U

(10 –12 Hz). A modulation of the amount of desynchronization during the 66.6 min long-term recording was obtained only for the lower-2 alpha band. These findings partly replicate earlier results of our research group (IsogluAlkac et al., 2000; Stru¨ber and Herrmann, 2002), showing that alpha activity decrease plays an important role for figure reversal. However, all previous studies analyzed an extended alpha band without further subdivisions. Thus, the present results demonstrate for the first time different responses of three alpha sub-bands during the perception of Necker cube reversals. Alpha activity in the human EEG can be classified into different types (Basar et al., 1997a,b). Most frequently, alpha activity is found as a spontaneous rhythm (Pfurtscheller et al., 1996). However, alpha activity can also be evoked by experimental stimuli (Isoglu-Alkac et al., 2004) and reflects cognitive processes (Basar et al., 1999, 2000; Klimesch, 1999; Klimesch et al., 2004). Whereas alpha synchronization occurs during alert wakefulness and marks cortical areas at rest or idling state, actual cognitive information processing is signaled by alpha desynchronization (Pfurtscheller et al., 1996; Klimesch, 1997; Klimesch et al., 1997a,b,c, 1998). Accordingly, the perceptual reversal induced alpha desynchronization of the present study probably reflects cognitive processes related to the recognition of a reversal and the need for a behavioral reaction. The system changes from a relaxed state of alert wakefulness preceding a reversal to an arousal reaction following a reversal (Mu¨ller et al., 1999). This interpretation fits also to the overlapping occurrence of a P300-like component in the time period following a perceptual reversal (t2), as illustrated in Figs. 2 and 3. Thus, the P300-like component and the accompanying alpha desynchronization might signal corresponding cognitive processes in the context of multistable perception. When taking the three alpha sub-bands into account, however, a more detailed distinction between different cognitive processes might evolve. According to Klimesch (1999), lower alpha desynchronization in the range of approximately 6 –10 Hz is obtained in response to a variety of non-stimulus specific factors which may be subsumed under the term Fattention_. It is topographically widespread over the entire scalp and probably reflects general task demands and attentional processes. Upper alpha desynchronization in the range of approximately 10 –12 Hz is topographically restricted and develops during the processing of semantic information (Klimesch, 1999). In a modified visual oddball task, Klimesch et al. (1998) demonstrated that the lower-1 alpha band reflected alertness (or arousal) by desynchronizing in response to a warning and imperative stimulus (targets and non-targets), whereas the lower-2 alpha band reflected expectancy and desynchronized already before the imperative stimulus occurs. The upper alpha band reflected semantic processes that were related to task performance and, therefore, showed maximal desynchronization in the late

poststimulus interval for targets only (Klimesch et al., 1998). Although the present study – due to the internal origin of perceptual reversals – cannot be directly related to results obtained with an oddball task, there were some similarities which might be useful for interpreting the data. It can be assumed that the perception of a reversal was comparable to the perception of a target in the oddball task and increased alertness (or arousal), as already mentioned above. It can be further assumed that the time preceding a reversal reflected a state of increased expectancy. This assumption is based on the often reported fact that, in spite of high interindividual variation, the withinsubject reversal rate is remarkably stable (Bergum and Bergum, 1980; Borsellino et al., 1972). Therefore, observers might have become respondent to the rhythmical structure of their reversal rate and, thus, expected the occurrence of a reversal. Similarly in the oddball task, after getting familiar with the (not completely random) sequence of targets and non-targets within a few trials, observers were able to predict the targets (Klimesch, 1999; Klimesch et al., 1998). While this interpretation seems plausible for the desynchronization in the lower-1 alpha band as reflecting increased arousal following a reversal, there is a problem in interpreting the desynchronization in the lower-2 alpha band as reflecting expectancy since expectancy-related changes in alpha activity have been shown to start as early as 1000 ms before the onset of a stimulus in the oddball task (Klimesch, 1999; Klimesch et al., 1998). Because the time window used in this study was too short for analyzing the time course of such long-lasting processes, the present results are not decisive in this respect. However, in this context, it must be noted that a recent MEG study by Stru¨ber and Herrmann (2002) demonstrated a steady decrease of alpha activity starting at approximately 1000 ms before the estimated time of reversal during the perception of a bistable apparent motion paradigm. Although the authors interpreted this finding as indicative for a bottom-up process in figure reversal and did not analyze sub-bands of alpha, it cannot be ruled out that part of the MEG alpha response may be related to expectancy. Regardless of the time course preceding a reversal, the desynchronization in the lower-2 alpha band of the present study could also reflect the expected task of pressing the response key following the perception of a reversal. This interpretation would correspond to the larger lower-2 alpha band desynchronization for targets (button press expected) than for non-targets (no expectation of button press) in the oddball task (Klimesch, 1999; Klimesch et al., 1998). Taken together, the above considerations suggest a possible role for the lower-1 alpha band as arousal reaction induced by a reversal, while the lower-2 alpha band might reflect the expectancy to perform still another task following the perception of a reversal, i.e. to press the response button. The similarity of the results for the lower-1 and lower-2

¨ . I˙s¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U

alpha band might reflect a large degree of overlapping of the proposed sub-processes. Concerning the upper alpha band, our results did not show any sign of desynchronization. This lack of alpha desynchronization is in contradiction to the results of the modified oddball paradigm used by Klimesch et al. (1998). The authors reported maximal desynchronization of the upper alpha band in the late post-stimulus interval for targets only. This finding was interpreted in that the upper alpha band reflected semantic processes that were related to task performance and, therefore, showed desynchronization during the second half of the poststimulus interval when the stimulus type is recognized and subjects started to count the target (Klimesch, 1999; Klimesch et al., 1998). In the present study, however, participants had not to differentiate targets from non-targets. Therefore, the amount of working memory load and semantic processing by identifying a target as meaningful and ignoring the nontargets is much larger in the oddball task than in the figure reversal task where subjects only had to recognize and report the occurrence of a reversal. Consequently, the upper alpha band in the present study demonstrated no reliable effects presumably due to relatively low task demands. Desynchronization in the lower (8– 10 Hz) and upper (10 –12 Hz) alpha bands has been related to the preparation of voluntary finger movements (Babiloni et al., 1999; Pfurtscheller et al., 2000). Since the present study involved a button-press response, the desynchronization in the lower2 and upper alpha band may signal motor preparation rather than cognitive processes related to multistable perception. Although we cannot rule out this possibility completely, there are several arguments which speak against an interpretation of the results solely based on motor preparation. First, the literature on alpha ERD during voluntary movement generally suggests a role for frequencies higher than 8 –10 Hz, i.e. the upper alpha and lower beta bands (see Pfurtscheller and Lopes da Silva, 1999 for a review). Studies comparing the lower and higher alpha bands found stronger effects for the upper than for the lower alpha band during movement preparation (Babiloni et al., 1999; Pfurtscheller et al., 2000). In the present study, however, we found the opposite: alpha ERD was strongest in the lower alpha bands but absent in the upper alpha band. Second, the topographical distribution of the lower-2 alpha band desynchronization in our study does not fit to movement-related processes. ERD was strongest at frontal location and comparable at parietal and occipital locations, while one would expect movement-related activity to be largest at central and parietal sites (Babiloni et al., 1999). Third, the alpha ERD in the lower bands was connected to the appearance of a P300-like wave, which is a cognitive component and not responsive to motor preparation (Picton, 1992). According to these arguments, it seems unlikely that the desynchronization in the lower alpha bands was substantially related to movement preparation due to the button-press response.

187

These interpretations of the obtained alpha sub-band activities, however, do not take into account large-scale effects according to the long-term recording session. It is reasonable to assume that the 66.6 min observation time could engage substantial processing requirements which might be reflected by different outcomes of the alpha subbands. While the desynchronization of the lower-1 alpha band demonstrated no long-term recording effects, the lower-2 alpha desynchronization was reduced at the end of the experiment. Given this outcome and the above interpretations, it follows that the lower-1 alpha desynchronization reflects an arousal reaction induced by perceptual reversals, which is unaffected by long-term recording. This finding would be consistent with the notion that the ability of arousal-inducing stimuli to trigger attentional processing is mediated in a bottom-up manner (Sarter et al., 2001). Accordingly, the lower-1 alpha band desynchronization could reflect the activity of a low-level mechanism leading to an automatic ‘‘detection’’ of percept change even without a change of stimulation. In contrast to that, the lower-2 alpha band desynchronization could be related to contexts in which attention is achieved through top-down processes. Top-down processes in this sense describe knowledge-based mechanisms designed to enhance the neuronal processing of relevant stimuli, to enhance the signal-to-noise ratio, and to bias the subject toward particular locations in which signals may appear (Kastner and Ungerleider, 2000). According to Sarter et al. (2001), in sustained attention performance, the subject knows where to expect what type of signal and how to respond in accordance with previously acquired response rules. Furthermore, the subject develops expectations concerning the probability for signals (Sarter et al., 2001). Obviously, this description of sustained attention (or vigilance) performance fits very well to the task demands of the long-term recording in the present study and to the interpretation of the lower-2 alpha sub-band as indexing expectancy-related attentional processes. Since the ability to sustain attention via top-down control has been considered to represent a limited resource, the decrease of lower-2 alpha desynchronization in the third section of the experiment may reflect a decrement in sustained attention performance over time-on-task. Furthermore, the interpretation of the lower-2 alpha band as indexing attentional top-down processes is corroborated by its frontal topography. Activation of the ‘‘anterior attention system’’ (Posner and Petersen, 1990) has been suggested to modulate the functions of posterior cortical areas in a top-down manner (Sarter et al., 2001). Conceptually, this interpretation of the lower-2 alpha band is similar to that given for the gamma band during multistable perception in former studies of our research group which also showed a frontal topography (Stru¨ber et al., 2000, 2001). Additionally, the relevance of gamma oscillations for attentional mechanisms was demonstrated for other visual classification tasks (Herrmann and Mecklinger, 1999). Therefore, it might be possible that attention-

188

¨ . ˙Is¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U

related top-down processes during multistable visual perception are indexed by a combined decrease of the lower-2 alpha band and an increase of the gamma activity. However, future research is needed in order to test this hypothesis, and to define the reversal-related time course of alpha sub-band and gamma activity. In conclusion, the results of this study demonstrate that differentiating three alpha sub-bands proved useful for a better distinction between different cognitive processes related to multistable perception. By employing long-term recording, the present findings suggest that the lower-1 alpha band desynchronization reflects an automatic arousal reaction which triggers attentional processing in a bottomup manner. In contrast, the findings for the lower-2 alpha band desynchronization can be related to attentional processes that are achieved by top-down control with limited resources. The upper alpha band demonstrated no reliable effects presumably due to the relatively low semantic task demands in figure reversal.

Acknowledgments This research was supported by The Scientific and Technical Research Council of Turkey (TUBITAK; B.02.1. BAK.0.09.00.00/832/285) and the Deutsche Forschungsge¨ R 112/12/04). meinschaft (DFG; GZ: 446TU

References Babiloni, C., Carducci, F., Cincotti, F., Rossini, P.M., Neuper, C., Pfurtscheller, G., Babiloni, F., 1999. Human movement-related potentials vs. desynchronization of EEG alpha rhythm: a high-resolution EEG study. NeuroImage 10, 658 – 665. Basar, E., 1980. EEG-Brain Dynamics. Relation between EEG and Brain Evoked Potentials. Elsevier, Amsterdam. Basar, E., 1999. Functional alphas selectively distributed in the brain—a theory. In: Basar, E. (Ed.), Brain Function and Oscillations II. Integrative Brain Function, Neurophysiology and Cognitive Processes. Springer, Berlin, pp. 331 – 351. Basar, E., Basar-Eroglu, C., Parnefjord, R., Rahn, E., Schu¨rmann, M., 1992. Evoked potentials: ensembles of brain induced rhythmicities in the alpha, theta and gamma ranges. In: Basar, E., Bullock, T.H. (Eds.), Induced Rhythms in the Brain. Birkha¨user, Boston, pp. 155 – 181. Basar, E., Schu¨rmann, M., Basar-Eroglu, C., Karakas, S., 1997a. Alpha oscillations in brain functioning: an integrative theory. Int. J. Psychophysiol. 26, 5 – 29. Basar, E., Hari, R., Lopes da Silva, F.H., Schu¨rmann, M., 1997b. Brain alpha activity—new aspects and functional correlates. Int. J. Psychophysiol. 26, 1 – 3. Basar, E., Basar-Eroglu, C., Karakas, S., Schu¨rmann, M., 1999. Oscillatory brain theory: a new trend in neuroscience. IEEE Eng. Med. Biol. 18, 56 – 66. Basar, E., Basar-Eroglu, C., Karakas, S., Schu¨rmann, M., 2000. Brain oscillations in perception and memory. Int. J. Psychophysiol. 26, 95 – 124. Basar-Eroglu, C., Stru¨ber, D., Stadler, M., Kruse, P., 1993. Multistable visual perception induces a slow positive EEG wave. Int. J. Neurosci. 73, 139 – 151.

Basar-Eroglu, C., Stru¨ber, D., Kruse, P., Basar, E., Stadler, M., 1996. Frontal gamma band enhancement during multistable visual perception. Int. J. Psychophysiol. 24, 113 – 125. Bergum, J.E., Bergum, B.O., 1980. Reliability of reversal rates as a measure of perceptual stability. Percept. Mot. Skills 50, 1039 – 1046. 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. Galambos, R., 1992. A comparison of certain gamma band (40-Hz) brain rhythms in cat and man. In: Basar, E., Bullock, T.H. (Eds.), Induced Rhythms in the Brain. Birkha¨user, Boston. Herrmann, C.S., Mecklinger, A., 1999. Gamma responses and ERPs in a visual classification task. Clin. Neurophysiol. 110, 636 – 642. ¨ ., Basar-Eroglu, C., Ademoglu, A., Demiralp, T., Miener, Isoglu-Alkac, U M., Stadler, M., 1998. Analysis of the EEG activity during the Necker cube reversals by means of the Wavelet Transform (WT). Biol. Cybern. 79, 437 – 442. ¨ ., Basar-Eroglu, C., Ademoglu, A., Demiralp, T., Miener, Isoglu-Alkac, 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. ¨ ., Keskindemirci, G., Karamu¨rsel, S., 2004. Auditory onIsoglu-Alkac, U and off-responses and alpha oscillations in the human EEG. Int. J. Neurosci. 114 (7), 879 – 906. Kastner, S., Ungerleider, L.G., 2000. Mechanisms of visual attention in the human cortex. Annu. Rev. Neurosci. 23, 315 – 341. Klimesch, W., 1997. EEG alpha rhythms and memory processes. In: Basar, E., Hari, R., Lopes da Silva, F.H., Schu¨rmann, M. (Eds.), Brain Alpha Activity—New Aspects and Functional Correlates, Int. J. Psychophysiol., vol. 26, pp. 319 – 340. Klimesch, W., 1999. EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain Res. Cogn. Brain Res. 29 (2 – 3), 169 – 195. Klimesch, W., Doppelmayer, M., Pachinger, T., Russegger, H., 1997a. Event-related desynchronization in the alpha band and the processing of semantic information. Brain Res. Cogn. Brain Res. 6 (2), 83 – 94. Klimesch, W., Doppelmayr, M., Schimke, H., Ripper, B., 1997b. Theta synchronization and alpha desynchronization in a memory task. Psychophysiology 34 (2), 169 – 176. Klimesch, W., Doppelmayr, M., Pachinger, T., Ripper, B., 1997c. Brain oscillations and human memory performance: EEG correlates in the upper alpha and theta bands. Neurosci. Lett. 238, 9 – 12. Klimesch, W., Doppelmayr, M., Russegger, H., Pachinger, T., Schwaiger, J., 1998. Induced alpha band power changes in the human EEG and attention. Neurosci. Lett. 244, 73 – 76. Klimesch, W., Doppelmayr, M., Ro¨hm, D., Po¨llhuber, D., Stadler, W., 2000. Simultaneous desynchronization and synchronization of different alpha responses in the human electroencephalograph: a neglected paradox? Neurosci. Lett. 284, 97 – 100. Klimesch, W., Schack, B., Schabus, M., Doppelmayr, M., Gruber, W., Sauseng, P., 2004. Phase-locked alpha and theta oscillations generate the P1-N1 complex and are related to memory performance. Brain Res. Cogn. Brain Res. 19, 302 – 316. Ko¨hler, W., 1940. Dynamics in Psychology. Liveright, New York. 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. Krause, C.M., Porn, B., Lang, A.H., Laine, M., 1997. Relative alpha desynchronization and synchronization during speech perception. Brain Res. Cogn. Brain Res. 5 (4), 295 – 299. Mu¨ller, T.J., Federspiel, A., Fallgatter, A.J., Strik, W.K., 1999. EEG signs of vigilance fluctuations preceding perceptual flips in multistable illusionary motion. NeuroReport 10 (16), 3423 – 3427. O’Donnell, B.F., Hendler, T., Squires, N.K., 1988. Visual evoked potentials to illusory reversals of the Necker cube. Psychophysiology 25, 137 – 143.

¨ . I˙s¸og˘lu-Alkac¸, D. Stru¨ber / International Journal of Psychophysiology 59 (2006) 179 – 189 U Pfurtscheller, G., Aranibar, A., 1977. Event-related cortical desynchronization detected by power measurements of scalp EEG. Electroencephalogr. Clin. Neurophysiol. 42, 817 – 826. Pfurtscheller, G., Klimesch, W., 1992. Event-related desynchronization and desynchronization of alpha and beta waves in a cognitive task. In: Basar, E., Bullock, T.H. (Eds.), Induced Rhythms in the Brain. Birkha¨user, Boston. Pfurtscheller, G., Lopes da Silva, F.H., 1999. Event-related EEG/MEG synchronization and desynchronization: basic principles. Clin. Neurophysiol. 110, 1842 – 1857. Pfurtscheller, G., Stancak, A., Neuper, C., 1996. Event-related synchronization (ERS) in the alpha band—an electrophysiological correlate of cortical idling: a review. Int. J. Psychophysiol. 24, 39 – 46. Pfurtscheller, G., Neuper, C., Krausz, G., 2000. Functional dissociation of lower and upper frequency mu rhythms in relation to voluntary limb movements. Clin. Neurophysiol. 111, 1873 – 1879. Picton, T.W., 1992. The P300 wave of the human event-related potential. J. Clin. Neurophysiol. 9, 456 – 479.

189

Posner, M.I., Petersen, S.E., 1990. The attention system of the human brain. Annu. Rev. Neurosci. 13, 25 – 42. Rock, I., Hall, S., Davis, J., 1994. Why do ambiguous figures reverse? Acta Psychol. 87, 33 – 59. Sarter, M., Givens, B., Bruno, J.P., 2001. The cognitive neuroscience of sustained attention: where top-down meets bottom-up. Brain Res. Rev. 35, 146 – 160. Stru¨ber, D., Herrmann, C.S., 2002. MEG alpha activity decrease reflects destabilization of multistable percepts. Brain Res. Cogn. Brain Res. 14, 370 – 382. Stru¨ber, 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 (3), 243 – 252. Stru¨ber, D., Basar-Eroglu, C., Miener, M., Stadler, M., 2001. EEG gammaband response during the perception of Necker cube reversals. Vis. Cogn. 8, 609 – 621.