CC BY
24
ОРИГИНАЛЬНЫЕ СТАТЬИ УДК 612.825
THE DYNAMICS OF GAMMA BAND OSCILLATORY ACTIVITY DURING PERCEPTION OF SUBJECTIVE CONTOURS
Belova E.
Rostov State Medical University 29 Nakhichevansky st., Rostov-on-Don, 344022, Russia [email protected]
Abstract
Purpose. The purpose of the current study was to investigate the functional roles of early and late gamma band responses to subjective contours.
Materials and methods. Variations of gamma band activity on the human scalp were measured during different task paradigms: a passive perception task, a simple reaction task, and a choice reaction task. The experiment was designed to compare four different types of perception of the same subjective contour, in this case, an illusory square. We used a time-frequency wavelet decomposition of electroencephalogram (EEG) signals to quantify the changes in gamma band oscillatory activity.
Results. Gamma band oscillatory activity rose in the first 150 ms after stimulus onset in all paradigms. In contrast to several previous studies, the latency of the early gamma responses varied with task relevance. The probability of the early gamma band responses to target and non-target stimuli tended to be higher at frontal, left occipital and parietal electrodes. This probability increased at occipital electrodes during the 200-250 ms time interval only in response to the target stimulus. For central electrodes, the probability of gamma responses increased within 250-350 ms in the case of the simple reaction task and after 350 ms of stimulus onset in the choice task condition.
Conclusion. We suggest that, in the choice paradigm, the early gamma band responses reflect the comparison of working memory contents with stimulus-related information, whereas the late gamma activity reflects the conscious stimulus representation and its utilization for motor reactions.
Keywords: gamma band responses; visual perception; bottom-up and top-down factors; subjective contours; wavelet; EEG.
ДИНАМИКА ГАММА-ЧАСТОТНОЙ ОСЦИЛЛЯТОРНОЙ АКТИВНОСТИ ПРИ ВОСПРИЯТИИ ИЛЛЮЗОРНЫХ КОНТУРОВ
Белова Е.И.
Ростовский государственный медицинский университет Россия, 344022, Ростов-на-Дону, пер. Нахичеванский, 29 [email protected]
Цель. Исследовать функциональную роль ранних и поздних компонентов гамма-частотного ответа в процессе восприятия человеком иллюзорных контуров.
Материалы и методы. Гамма-частотные ответы в ЭЭГ человека регистрировали при восприятии иллюзорных стимулов в разных парадигмах деятельности: пассивное восприятие, простая сенсомоторная реакция и реакция выбора. Дизайн эксперимента был построен таким образом, что при восприятии одного и того же иллюзорного контура (квадрата) менялась его функциональная значимость, однако не менялись физические параметры. При обработке результатов эксперимента использовали вейвлет преобразование ЭЭГ.
Результаты. Гамма-частотная осцилляторная активность повышалась в первые 150 мс после начала стимула во всех парадигмах деятельности. При этом латентный период раннего ответа варьировал в зависимости от значимости стимула. Вероятность появления ранних гамма-ответов увеличивалась во фронтальных, левом затылочном и теменном отведениях в ответ на целевой и нецелевой стимулы. На целевой стимул
во временном интервале 200-250 мс вероятность гамма-ответов увеличивалась в затылочных отведениях. В центральных отведениях вероятность гамма-ответов увеличивалась во временном интервале 250-350 мс в ситуации простой сенсомоторной реакции, при реализации реакции выбора — после 350 мс от начала стимула.
Заключение. Результаты позволяют предположить, что в ситуации выбора повышение вероятности ранних гамма-ответов отражает сравнение информации о предъявляемом стимуле с содержанием рабочей памяти, в то время как поздние гамма-ответы — осознанное представление о стимуле и его использования для формирование моторной реакции.
Ключевые слова: гамма-частотные ответы, зрительное восприятие, иллюзорные контуры вейвлет-анализ, электроэнцефалограмма.
Introduction
Findings from many studies, including recording of neuronal activity, electroencephalogram (EEG) and magnetoencephalogram (MEG), show that high-frequency (y) oscillations are functionally relevant to information processing in the brain [1-7]. Gamma band activity has been observed during a variety of behavioral states in both human and animal studies. Based on the assumption that synchronized neuronal activity is related to visual feature binding [1, 8], Y-oscillations are considered as a neurophysiological signature of object representation [for review, see 9, 10]. Some research suggests that this fast oscillatory activity is associated with top-down processes including attention [3, 11, 12], memory [7, 13] and multisensory integration [14, 15]. Studies of the memory processes have supported the notion that Y-band activity reflects the cell assembly synchronization during working memory maintenance when sustained activation of the neural representation is required [16]. Together, these findings suggest that these high-frequency oscillations underlie the activation of the neural representation of the object being represented [3, 9, 10, 17]. Such gamma oscillations are neither phase-locked nor time-locked to stimulus onset; thus, they are termed induced gamma band responses [3, 18] rather than stimulus-locked evoked Y-responses, such as the event evoked potential. Evoked Y-band responses can be isolated and analyzed using the time-frequency energy of the averaged evoked potential, whereas the induced responses disappear when averaging because of latency jitters across trials [3, 4]. The evoked responses have a latency of <200 ms (early component), whereas the induced responses appear 200 ms after stimulus onset (late component). The evoked gamma activity is insensitive to stimulus type and is not likely to reflect sensory binding mechanisms [3, 4], whereas the induced gamma activity is related to high-level cognitive processes and correlates closely with object representation, attention, object recognition and sensory motor integration [4, 10, 18, 19]. Conversely, some authors [20, 21] have indicated that early phase-locked Y-responses are related to stimulus processing and are sensitive to task difficulty.
Despite the growing number of publications
about gamma band activity, the functional roles of the early and late high-frequency responses are not fully understood. In the present investigation, we examined the possible functional roles of the early and late Y-band responses to visual stimuli. The subjective contours were presented under various task relevance conditions, and the time course of the gamma response probability for each task condition was investigated.
Materials and Methods
Subjects. 18 right-handed participants (men students on average 25 years old had normal or corrected-to-normal vision) took part in the present study as volunteers. Before the experiment, each participant signed a consent form. The current study and the consent form conformed to The Code of Ethics of the World Medical Association (Declaration of Helsinki) and were approved by the Southern Federal University Ethical Committee.
Stimuli and tasks. The following three subjective (illusory) contours were used as test stimuli: triangle, square and star. These contours subtended a visual angle of 70 at a viewing distance of 1.5 m. Stimuli were delivered for 200 ms on a video display on a light gray background in randomized order and were synchronized with EEG recoding equipment. The interval between stimuli was randomized from 2 to 5 seconds. The participants were instructed to fixate on a small, centrally placed cross.
The recording conditions included: (1) a passive perception task (PP), in which only illusory squares were presented, and the subjects were required to watch the stimulus without any movement. In this session, the illusory square served as an irrelevant stimulus; (2) a simple reaction task (SR), in which the same stimulus was presented, and participants were required to press a button as soon as possible. In this task, the illusory square was defined as a relevant stimulus; (3) a choice-reaction task, in which the illusory triangle and star were added as test stimuli. The motor reaction (button pressing) was required only when the target was presented (GO); the participants were required to ignore the nontarget stimuli (NOGO). For each PP and SR task, 60 stimuli (illusory squares) were presented. For the
third task, each test stimulus was presented 60 times in randomized order. The test stimulus was located in the center of the monitor, and the sample target was located in the right top corner of the monitor. After presenting the test stimuli 16 times, a new target was set and the procedure was repeated. In the third session, only trials with the illusory square as target or non-target stimuli were used for subsequent analysis. Participants received task training for a few minutes before the experiment began.
Electrophysiological recording and analysis. EEG activity was recorded from 9 electrodes with linked mastoids as a reference. Their locations according to the international 10-20 system were: F3, F4, C3, C4, P3, P4, O1, O2 and Cz. The EEG was digitized at 1000 Hz, and the initial bandpass recording filter was set at 0.3-280 Hz. Analysis epochs contained 1024 ms before and 1024 ms after stimulus onset. Epochs with artifacts (EEG>100 |V or EOG>threshold) were rejected and 40-60 artifact-free sweeps of each stimulus type were used for further analysis.
To quantify the changes in y-band oscillatory activity due to the different stimulus types, we used a time-frequency wavelet decomposition of the EEG signal (complex of Morlet's wavelets). The wavelet window (T) was equal to 6 periods of predetermined frequency. The sine and cosine components of the wavelet were used. The calculations were performed
for the frequency band from 30 to 45 Hz with a step size of 1 Hz. Each analysis epoch was scanned by the chosen window in the time interval from T/2 to L-T/2 (in which L is an epoch duration). For each sample, we calculated the convolution of the sine and cosine components of the wavelet with time sequence f(t):
Wf(x, a) = - I i|f * f—') f{t)dt
where: y — some function; y* — complex conjugation of y; t — corresponds to time scale and is named as a parameter of location; a — prescribes process of scaling and is named as a parameter of stretching.
Then the module of the resultant vector was calculated:
M
V С 2 — S2
where: C — cosine component; S — sine component of wavelet.
The wavelet transformation was applied to the average evoked potential and provided information about the phase-locked oscillatory activity. We also applied this method to single trials to analyze nonphase-locked high-frequency components
(Fig. 1).
2 3
I I I—I I I I
200 400 600 800 1000 ms
Fig. 1. Event-related ^-responses from one subject.
(1) Native EEG recorded at P3. (2) The result of a time-frequency wavelet decomposition of a single EEG trial at 40 Hz frequency.
The presence of envelope peaks indicates the occurrence of y bursts. (3) The time-frequency energy calculated at 40 Hz and averaged over single trials. The arrow points to stimulus onset.
The latencies and amplitudes of the envelope peaks were calculated at 35 Hz for the first 500 ms after stimulation. These were measured to determine whether they exceeded the baseline level. The baseline level was calculated as a sum of the averaged time-frequency energy in the pre-stimulus period (-200 ms) and the mean root square deviation. Gamma band responses to all four stimulus types (irrelevant in PP, relevant in SR, target and non-target in GO/NOGO) were compared. Then the post-stimulus periods were divided into ten 50 ms time windows for which the
probability of y-responses was calculated over single trials for each subject, stimulus type and electrode. The probability values were assigned to line plots.
A MANOVA was used to analyze the envelope peak amplitudes. The distribution of the peak latencies was far from Gaussian; therefore, we performed the nonparametric Wilcoxon test. The differences between samples were considered statistically significant at the 5% level. In the case of 0.05<p<0.1, the effect was considered nonsignificant, but a difference was noted at the trend line level.
Results
Behavioral results showed that all subjects performed all of the tasks successfully. They identified the illusory square in the third condition (GO/NOGO) with high accuracy (96%). The subjects' reaction time was significantly shorter (F(1, 17)=871.4; p<0.001) for the simple sensory motor reaction (286 ms) than for the choice reaction (496 ms).
Gamma band responses were observed during the post-stimulus period for all conditions (PP, SR, GO, NOGO). One component of the high-frequency activity, phase-locking to the stimulus onset, was identified by the wavelet time-frequency analysis of the averaged evoked potential at the occipital and parietal electrodes. The phase-locking was identified in both the PP condition (for 10 of the 18 total subjects) and the SR condition (for 14 of the 18 total subjects) conditions. Phase-locking was identified in the GO/ NOGO condition for only 3 subjects. For this reason, in subsequent analyses we calculated time-frequency energy of single trials, with latency and amplitude of single envelope picks being measured.
The average amplitude of significant Y-band responses was from 5.5 to 8.7 |V. For the frontal electrodes, the amplitude differences reached significance only when
non-target stimuli were compared with irrelevant stimuli. Single response amplitudes were larger for non-target stimuli at F3, but they only showed a nonsignificant trend at the trend line level at F4 (F(1, 17)=5.6; p=0.03; F(1, 17)=3.4; p=0.08, respectively).
The latency analysis of the single-sweep envelopes showed that the time course of Y-band responses to illusory contours depended on the task paradigm. For all stimulus conditions, the highest probability of Y-band responses was observed within the first 150 ms after stimulus onset. The Y-band response appeared at all electrode locations, but was maximal at the occipital and parietal locations. After 150 ms, a decrease was observed in Y-band activity to the irrelevant stimulus in the PP condition and to the non-target stimuli in the choice condition (Fig. 2A).
At O1 and O2 (p=0.09 and 0.05, respectively), only the target stimuli tended to increase the Y-band response probability within the 200-250 as shown at Fig. 3A. At Cz (p=0.02) and C3 (p=0.09), the target stimuli induced the same increase in Y-band response probability after 350 ms (Fig. 3B). Compared with irrelevant stimuli, gamma responses to relevant stimuli in the SR condition occurred more frequently within the 250-350 ms time window. These responses were significant at Cz (p=0.04).
0,135-,
0,130-
0,125-
0,120-
0,115-
0,110-
& 0,105 -
n 0,100-
о 0,095 -
a. 0,090-
0,085 -
0,080 -
0,075 -
0,070 -
0,065-
A
■ irrelevant •— non-target
0,120 0,115 0,110 0,105 0,100 0,095 0,090 0,085 0,080 0,075
100 200 300 400 500 IDS
time
Fig. 2. The probability distribution of y band responses to irrelevant stimuli in the PP condition and to the non-target in the choice condition at P3 (A) and F3 (B) within the post-stimulus period.
* indicates a statistically significant difference.
Fig. 3. The probability distribution of y band responses to relevant stimuli in the SR condition and to the target in the choice condition at O1 (A) and C3 (B) within the post-stimulus period.
Within the first 150 ms, the target and non-target stimuli provided a higher probability of Y-band responses at O1, P3 and both frontal electrodes than the relevant (SR) and irrelevant (PP) stimuli did. The Wilcoxon test showed significant differences at F3 (p=0.04) and F4 (p=0.03) for non-target stimuli (Fig. 2B). Target stimulation revealed a higher probability of Y-band responses at F3 (p=0.09) within the 100150 ms interval, whereas the probability effects of the different conditions reversed within the 50-100 ms interval. That is, Y-band responses to the illusory square appeared more rarely in the GO condition than in the SR condition (p=0.03). At P3 (p=0.08)
and O1 (p=0.07), gamma band activity occurred more frequently within the 50-100 ms time window for non-target stimuli and within 100-150 ms for target stimuli (Fig. 2A, 3A).
A comparison between the NOGO and GO conditions indicates that the probability of Y-band responses was higher within the 50-100 ms interval to the non-target stimuli at F3 (p=0.02), P3 (p=0.04) and O1 (p=0.07). Fig. 4 illustrates this for P3. A comparison between the SR and PP conditions indicates that there were no significantly different Y-band responses to relevant and irrelevant stimuli within the first 150-ms interval.
-■-non-target • target
--1-1-1-1-1-1-1-1-1-1-1
о 100 200 300 400 500 m s
time
Fig. 4. The probability distribution of y band responses to non-target and target stimuli in the choice condition at P3 within the post-stimulus period.
* indicates p<0.05.
Discussion
The present experiments showed an increase in the probability of Y-responses at the parietal and occipital electrodes within the first 150 ms after illusory square presentation for all experimental paradigms. Unlike previous data [3], the early Y-band responses varied with task relevance and did not always appear with the same latency. As a result, we identified the phase-locked to stimulus onset component for only 3 subjects in the GO/NOGO paradigm. We suppose that the early gamma activity at the occipital and parietal electrodes could be related to bottom-up stimulus processing in all conditions. Moreover, we assume that more detailed stimulus processing was required in the GO/NOGO condition than in the PP and SR conditions. For this reason, the relevant stimuli in the SR condition elicited Y responses more frequently within the 50-100 ms interval, whereas the target stimuli elicited Y responses more frequently within the 100-150 ms interval. The illusory square was the
only stimulus presented in the passive perception and simple sensory motor conditions, and it was delivered 60 times in succession. For these conditions, only this stimulus presentation was essential for the task performance. Conversely, successfully categorizing the stimulus as the target or non-target in the GO/NOGO condition required binding the parts that induce illusory contours together to the coherent percept. Consistent with the hypothesis that an object elicits a coherent percept by the oscillatory synchronization of activity of neurons encoding the different parts and features of this object [1, 5, 22, 23], feature binding could be the source of the enhanced gamma activity observed at P3 and O1 in the GO/NOGO condition compared with the PP and SR conditions. Due to the design of our experiments, we were unable to exclude the possible effects of attention on the increased early gamma activity in the posterior leads in the GO/ NOGO paradigm. When subjects were required to attend to the target stimulus, its neural representation was built up.
Our suggestion about the role of the early y-responses supports the classical event related potentials experimental data [24], which shows that the visual system can perform the stimuli categorization task before 150 ms. The psychophysical observations with the ERP recordings also demonstrated that estimation of the stimulus relevance could be performed at an unconscious level within 100-150 ms [25, 26]. This fast feed-forward (bottom-up) mechanism provides a rough identification of the specific features of the stimulus and is necessary for more adaptive behaviors [10]. Top-down influences from high-level brain areas are required for the categorization of stimuli that possess complex physical or semantic features. As Buschman and Miller's [27] work shows, that prefrontal neurons were activated first during top-down attention, when monkey identified target location. Parietal neurons were activated first during bottom-up attention.
We posit that fast bottom-up processing of illusory squares occurs in the PP and SR conditions, whereas top-down working memory modulation of early gamma activity is required in the GO/NOGO condition. The y-response probability increased simultaneously over the frontal and parieto-occipital cortices within the first 50-100 ms to non-target stimuli and within 100-150 ms to target stimuli in the choice condition but not the other conditions. This finding partially conflicts with previous data [3, 19]. According to the hypothesis formulated by Tallon-Baudry and Bertrand [4], gamma activity that is induced later (>200 ms) is specifically related to the integration of bottom-up and top-down factors. Our results showed that such matching could be much faster. Moreover, integration of bottom-up processing and memory content in response to non-target stimuli was shortened compared with the response to target stimuli; the maximum y-response probability occurred within the 50-100 ms after stimulus presentation for the non-target stimuli, whereas this integration fully developed and occurred within 100-150 ms in response to the target stimuli.
Thus, both bottom-up and top-down factors can influence the early y-band activity. The current results support the evidence that task difficulty modulates the neuronal firing rate at the earliest stages of cortical processing of visual stimuli (area V1) in monkeys
[28]. An ingenious model has been proposed [29] to simulate aspects of the coupling between bottom-up preattentive processing and the task-directed top-down attentive processes in the visual system. It is believed that the visual cortex is not merely a bottom-up filtering, but that neurons in 2/3 of the layers in VI comprise connections between bottom-up filtering, horizontal grouping and top-down influence
[29]. Herrmann's 'match-and-utilization-model' [30] also attempts to explain the early y-band activity in
terms of matching stimulus-related information with the memory contents, and the late y-band responses as a readout and utilization of the resultant information.
Our results support the above-mentioned models concerning the roles of both early and late y band activity. We observed an increase in the y-band response probability within the 200-250 ms interval at occipital electrodes only for the target stimuli. This increase is likely to reflect the construction of a detailed conscious representation of the presented stimulus that will be used for organizing the subsequent motor response. The doubled latencies in the choice motor reaction task compared with those in the simple reaction task imply that the increase in y-band response probability at C3 and Cz that occurred within 250350 ms in the SR condition and within 350-500 ms in the GO condition is related to the organization of the motor reaction. Conversely, in the NOGO condition, a decrease in gamma activity occurred after 150 ms of stimulus onset. This decrease is likely due to the inhibition of processing in the visual cortex when the non-target signal was presented.
Conclusion
The data summarized here showed that both bottom-up and top-down factors can influence the y-band activity in parieto-occipital areas as early as 150 ms after stimulus presentation. The early y-band responses were observed for all experimental conditions (PP, SR, GO, NOGO). However, these responses were not always synchronized with the stimulus onset. In the choice condition target and non-target stimuli provided a higher probability of y-band response in both the frontal and left parieto-occipital areas before 150 ms after stimulus onset. The relevant stimuli in the SR condition elicited y responses more frequently within the 50-100 ms interval, whereas the target stimuli elicited y responses more frequently within the 100-150 ms interval. We suggest that early (50-150ms) y-frequency increase in choice condition was caused by the combining of stimulus-related information and working memory content.
Only the target stimuli provided the increase the y-band response probability within the 200-250 ms at O1 and O2. We suggest that more detailed stimulus processing was required in the GO condition, and this late gamma activity in response to target stimulus reflects the conscious stimulus representation in visual cortex that can be used for subsequent analyses.
The relevant and target stimuli provided increase the y-band response probability at C3 within the 200-250 and after 350 ms, respectively. These late y-band responses likely reflect the correction of processing results and the use of this information for the organization of the subsequent motor reaction.
REFERENCES
1. Gray CM, Konig P, Engel AK, Singer W. Oscillatory responses in cat visual cortex exhibit inter-columnar synchronization which reflects global stimulus properties. Nature. 1989; 338(6213): 334-7.
2. Kiroy VN, Belova EI. Mechanisms of formation and role of the oscillatory activity of neuronal populations in brain state and information processing. Zhurnal Vysshei Nervnoi Deoatelnosti IP Pavlova. 2000; 50:179-91 (in Russian).
3. Tallon-Baudry C, Bertrand O, Delpuech C, Pernier J. Oscillatory gamma-band (30-70 Hz) activity induced by a visual search task in humans. J Neurosci 1997; 17(2):722-34.
4. Tallon-Baudry C, Bertrand O. Oscillatory gamma activity in humans and its role in object representation. Trends Cogn Sci. 1999; 3(14):151-62.
5. Esfahani ZG, Gollo LL, Valizadeh A. Stimulus-dependent synchronization in delayed-coupled neuronal networks. Sci Rep. 2016; 6: 23471. doi: 10.1038/srep23471.
6. Canavier CC. Phase-resetting as a tool of information transmission. Curr Opin Neurobiol. 2015 Apr; 31:206-13. doi: 10.1016/j.conb.2014.12.003.;31:206-13.
7. Kaiser J. Dynamics of auditory working memory. Front Psychol. 2015; 11:6:613. doi: 10.3389/fpsyg.2015.00613
8. Eckhorn R, Bauer R., Jordan W., Brosch M, W. Kruse, M. Munk et. al., Coherent oscillations: a mechanism of feature linking in the visual cortex? Multiple electrode and correlation analyses in the cat, Biol Cybern. 1988; 60:121-30.
9. Singer W. Consciousness and the binding problem Annu Rev Neurosci. 2001; 929:123-46.
10. Tallon-Baudry C. Oscillatory synchrony and human visual cognition. J Physiology - Paris. 2003; 97(2-3):355-63.
11. Womelsdorf T, Fries P. The role of neuronal synchronization in selective attention. Curr Opin Neurobiol. 2007; 17(2):154-60.
12. Lakatos P, Karmos G, Mehta AD, Ulbert I, Schroeder C.E. Entrainment of neuronal oscillations as a mechanism of attentional selection. Science. 2008; Apr 4;320(5872):110-3. doi: 10.1126/science.1154735.
13. Pipa G, Stadtler ES, Rodrigues EF, Waltz JA, Muckli LF, Singer W. et al. Performance- and stimulus-dependent oscillations in monkey prefrontal cortex during short-term memory. Frontiers in Integrative Neuroscience. 2009; 3:1-13.
14. Lakatos P, Chen CM, O'Connell MN, Mills A, Schroeder CE. Neuronal oscillations and multisensory interaction in primary auditory cortex. Neuron. 2007; Jan 18;53(2):279-92.
15. Meehan SK, Staines WR. Task-relevance and temporal synchrony between tactile and visual stimuli modulates cortical activity and motor performance during sensory-guided movement. Hum Brain Mapp. 2009;30(2):484-96.
16. Pesaran B, Pesaris JS, Sahani M, Mitra PP, Andersen RA. Temporal structure in neuronal activity during working memory in macaque parietal cortex. Nat Neurosci. 2002; Aug;5(8): 805-11.
17. Castelhano J, Rebola J, Leitao B, Rodriguez E, Castelo-Branco M. To perceive or not perceive: the role of gamma-band activity in signaling object percepts. To perceive or not perceive: the role of gamma-band activity in signaling object percepts. PLoS ONE. 2013; 8(6): e66363. doi:10.1371/journal.pone.0066363.
18. Kaiser J1, Lutzenberger W. Induced gamma-band activity and human brain function. Neuroscientist. 2003;9(6):475-84.
19. Kranczioch C, Debener S, Herrmann CS, Engel AK. EEG gamma-band activity in rapid serial visual presentation. Exp Brain Res. 2006 Feb;169(2):246-54.
20. Senkowski D, Herrmann CS. Effects of task difficulty on evoked gamma activity and ERPs in a visual discrimination task. Clin Neurophysiol. 2002 Nov;113(11):1742-53.
21. Stefanics G, Jakab A, Bernath L, Kellenyi L, Hernadi I. EEG early evoked gamma-band synchronization reflects object recognition in visual oddball task. Brain Topogr. 2004; 16(4):261-4.
22. Singer W, Gray CM. Visual features integration and the temporal correlation hypothesis. Annu Rev Neurosci. 1995; 18:555-86.
23. Fries P, Nikolic D, Singer W. The gamma cycle. Trends Neurosci. 2007 Jul; 30(7):309-16.
24. Thorpe S, Fize D, Marlot K. Speed of processing in the human visual system. Nature. 1996 Jun 6;381(6582):520-2.
25. Ivanitsky AM. Brain basis of subjective experience: information synthesis hypothesis. Zhurnal Vysshei Nervnoi Deoatelnosti IP Pavlova. 1996; 46: 241-52 (in Russian).
26. Kostandov E.A. Conscious and unconscious forms of human higher nervous activity, in: Mechanisms of Human Brain Functioning. Part One. Human Neurophysiology. Nauka Pablishing House. Leningrad. 1988. pp. 677 (in Russian).
27. Buschman TJ, Miller EK. Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. Science. 2007 Mar 30; 315(5820):1860-2.
28. Chen Y, Martinez-Conde S, Macknik SL, Bereshpolova Y, Swadlow H, Alonso J. Task difficulty modulates the activity of specific neuronal populations in primary visual cortex. Nat Neurosci. 2008. Aug; 11(8):974-82. doi: 10.1038/nn.2147.
29. Grossberg S. How does the cerebral cortex work? Development, learning, attention and 3-D vision by laminar circuits of visual cortex. Behav and Cogn Neurosc Reviews. 2003 Mar;2(1):47-76.
30. Herrmann CS, Munk MHJ, Engel A.K. Cognitive functions of gamma-band activity: memory match and utilization. Trends Cogn Sci. 2004 Aug; 8(8):347-55.
