Scholarly article on topic 'Frequency-dependent spatiotemporal profiles of visual responses recorded with subdural ECoG electrodes in awake monkeys: Differences between high- and low-frequency activity'

Frequency-dependent spatiotemporal profiles of visual responses recorded with subdural ECoG electrodes in awake monkeys: Differences between high- and low-frequency activity Academic research paper on "Clinical medicine"

CC BY-NC-ND
0
0
Share paper
Academic journal
NeuroImage
OECD Field of science
Keywords
{Electrocorticography / "Frequency band" / "Visual response" / "Spatiotemporal profile" / Monkey}

Abstract of research paper on Clinical medicine, author of scientific article — Kana Takaura, Naotsugu Tsuchiya, Naotaka Fujii

Abstract Electrocorticography (ECoG) constitutes a powerful and promising neural recording modality in humans and animals. ECoG signals are often decomposed into several frequency bands, among which the so-called high-gamma band (80–250Hz) has been proposed to reflect local cortical functions near the cortical surface below the ECoG electrodes. It is typically assumed that the lower the frequency bands, the lower the spatial resolution of the signals; thus, there is not much to gain by analyzing the event-related changes of the ECoG signals in the lower-frequency bands. However, differences across frequency bands have not been systematically investigated. To address this issue, we recorded ECoG activity from two awake monkeys performing a retinotopic mapping task. We characterized the spatiotemporal profiles of the visual responses in the time-frequency domain. We defined the preferred spatial position, receptive field (RF), and response latencies of band-limited power (BLP) (i.e., alpha [3.9–11.7Hz], beta [15.6–23.4Hz], low [30–80Hz] and high [80–250Hz] gamma) for each electrode and compared them across bands and time-domain visual evoked potentials (VEPs). At the population level, we found that the spatial preferences were comparable across bands and VEPs. The high-gamma power showed a smaller RF than the other bands and VEPs. The response latencies for the alpha band were always longer than the latencies for the other bands and fastest in VEPs. Comparing the response profiles in both space and time for each cortical region (V1, V4+, and TEO/TE) revealed regional idiosyncrasies. Although the latencies of visual responses in the beta, low-, and high-gamma bands were almost identical in V1 and V4+, beta and low-gamma BLP occurred about 17ms earlier than high-gamma power in TEO/TE. Furthermore, TEO/TE exhibited a unique pattern in the spatial response profile: the alpha and high-gamma responses tended to prefer the foveal regions, whereas the beta and low-gamma responses preferred the peripheral visual fields with larger RFs. This suggests that neurons in TEO/TE first receive less selective spatial information via beta and low-gamma BLP but later receive more fine-tuned spatial foveal information via high-gamma power. This result is consistent with a hypothesis previously proposed by Nakamura et al. (1993) that states that visual processing in TEO/TE starts with coarse-grained information, which primes subsequent fine-grained information. Collectively, our results demonstrate that ECoG can be a potent tool for investigating the nature of the neural computations in each cortical region that cannot be fully understood by measuring only the spiking activity, through the incorporation of the knowledge of the spatiotemporal characteristics across all frequency bands.

Academic research paper on topic "Frequency-dependent spatiotemporal profiles of visual responses recorded with subdural ECoG electrodes in awake monkeys: Differences between high- and low-frequency activity"

Frequency-dependent spatiotemporal profiles of visual responses recorded with subdural ECoG electrodes in awake monkeys: Differences between high- and low-frequency activity

Kana Takaura a, Naotsugu Tsuchiya b,c,d, Naotaka Fujiia'*

a Laboratory for Adaptive Intelligence, RIKEN Brain Science Institute, 2-1 Hirosawa, Wako-shi, Saitama, 351-0198, Japan b School of Psychological Sciences, Faculty of Biomedical and Psychological Sciences, Monash University, Melbourne, VIC 3800, Australia c Decoding and Controlling Brain Information, Japan Science and Technology Agency, Chiyoda-ku, Tokyo 102-8266, Japan d Monash Institute of Cognitive and Clinical Neurosciences, Monash University, Melbourne, VIC 3800, Australia

ARTICLE INFO ABSTRACT

Electrocorticography (ECoG) constitutes a powerful and promising neural recording modality in humans and animals. ECoG signals are often decomposed into several frequency bands, among which the so-called high-gamma band (80-250 Hz) has been proposed to reflect local cortical functions near the cortical surface below the ECoG electrodes. It is typically assumed that the lower the frequency bands, the lower the spatial resolution of the signals; thus, there is not much to gain by analyzing the event-related changes of the ECoG signals in the lower-frequency bands. However, differences across frequency bands have not been systematically investigated. To address this issue, we recorded ECoG activity from two awake monkeys performing a retinotopic mapping task. We characterized the spatiotemporal profiles of the visual responses in the time-frequency domain. We defined the preferred spatial position, receptive field (RF), and response latencies of band-limited power (BLP) (i.e., alpha [3.9-11.7 Hz], beta [15.6-23.4 Hz], low [30-80 Hz] and high [80-250 Hz] gamma) for each electrode and compared them across bands and time-domain visual evoked potentials (VEPs). At the population level, we found that the spatial preferences were comparable across bands and VEPs. The high-gamma power showed a smaller RF than the other bands and VEPs. The response latencies for the alpha band were always longer than the latencies for the other bands and fastest in VEPs. Comparing the response profiles in both space and time for each cortical region (V1, V4+, and TEO/TE) revealed regional idiosyncrasies. Although the latencies of visual responses in the beta, low-, and high-gamma bands were almost identical in V1 and V4+, beta and low-gamma BLP occurred about 17 ms earlier than high-gamma power in TEO/TE. Furthermore, TEO/TE exhibited a unique pattern in the spatial response profile: the alpha and high-gamma responses tended to prefer the foveal regions, whereas the beta and low-gamma responses preferred the peripheral visual fields with larger RFs. This suggests that neurons in TEO/TE first receive less selective spatial information via beta and low-gamma BLP but later receive more fine-tuned spatial foveal information via high-gamma power. This result is consistent with a hypothesis previously proposed by Nakamura et al. (1993) that states that visual processing in TEO/TE starts with coarse-grained information, which primes subsequent fine-grained information. Collectively, our results demonstrate that ECoG can be a potent tool for investigating the nature of the neural computations in each cortical region that cannot be fully understood by measuring only the spiking activity, through the incorporation of the knowledge of the spatiotemporal characteristics across all frequency bands.

© 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/).

|Wj| CrossMark

Article history: Received 4 March 2015 Accepted 3 September 2015 Available online 10 September 2015

Keywords:

Electrocorticography Frequency band Visual response Spatiotemporal profile Monkey

Abbreviations: BLP, band-limited power; ECOG, electrocorticography; FDR, false discovery rate; FP, fixation point; HGP, high-gamma power; MRI, magnetic resonance image; ND, normalized difference; PP, preferred position; RF, receptive field; RL, response latency; SRF, size of the receptive field; VEP, visual evoked potential.

* Corresponding author.

E-mail address: na@brain.riken.jp (N. Fujii).

1. Introduction

Electrocorticography (ECoG) is a technique that uses subdural electrodes to record neural activity directly from the cortical surface and has been widely used in patients with epilepsy to localize the origin of epileptic seizures (Palmini et al., 1995; Zumsteg and Wieser, 2000). Recently, ECoG has gained attention as a tool for electrophysiological research in animals because it offers several advantages, including large spatial coverage, fine spatiotemporal resolution, and stable recordings

http: //dx.doi.org/10.1016/j.neuroimage.2015.09.007

1053-8119/© 2015 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

over weeks to months, which are impossible to attain simultaneously with other recording techniques (Bosman et al., 2012; Chao et al., 2010; Matsuo et al., 2011; Rubehn et al., 2009; Shimoda et al., 2012; Viventi et al., 2011). These advantages make ECoG a promising technique for neuroengineering applications, including brain machine interfaces (Graimann et al., 2004; Schalk et al., 2008).

Typically, ECoG signals are analyzed using the short-time window Fourier transform or its related spectral analysis techniques. The spectral decomposition of ECoG is critical, as it can isolate a putatively stable and informative aspect of the signal: the high-gamma power (HGP, 80-200 Hz). ECoG recordings in patients with epilepsy indicate that, compared to low-frequency band-limited power (BLP), HGP can better determine the timing and localization of motor-, sensory-, and task-related changes in neural activity (Canolty et al., 2007; Crone et al., 1998a, 1998b, 2001 ; Edwards et al., 2009, 2010; Kawasaki et al., 2012; Miller et al., 2007,2010; Pei et al., 2011; Tsuchiya et al., 2008). Although HGP almost invariably shows event-related increments (but see Foster et al., 2012), event-related low-frequency BLP exhibits both increments and decrements in complicated temporal patterns over a few seconds that vary across subjects and recording sites (Canolty et al., 2007; Crone et al., 1998b; Fukuda et al., 2010; Harvey et al., 2013; Ohara et al., 2000). These observations have resulted in a view that, as far as event-related changes are concerned, the HGP reliably reflects local cortical functions at the cortical surface below the ECoG electrodes (Crone et al., 1998a, 2006; Jerbi et al., 2009), while it remains unclear what the lower-frequency BLP reflects, leading to less attention to event-related changes in the low-frequency BLP. Note that the functional roles of the low-frequency synchronized oscillations in steady states are a major issue in systems neuroscience (Engel and Fries, 2010; Ward, 2003).

However, detailed functional characterization of event-related changes in the low-frequency BLP might provide additional insights into local cortical processing. Although it is often assumed that electrical signals spatially propagate at different rates in the brain depending on their frequencies, the actual impedance spectra of cortical tissue are independent of frequency (Logothetis et al., 2007; Rank, 1963). This suggests that, in principle, low-frequency BLP might also reflect localized neural events. If this is true, it would be possible to learn the nature of the inputs and outputs of a given cortical area by analyzing both the HGP and low-frequency BLP, as output spikes and input synaptic activity are mainly correlated with HGP and low-frequency BLP, respectively (Bartos et al., 2007; Buzsâki et al., 2012; Logothetis, 2008). Furthermore, low-frequency BLP can reflect activity below the spiking threshold, which may not be reflected in HGP. To test this hypothesis, it is necessary to systematically investigate the differences between HGP and low-frequency BLP. In doing so, critical knowledge regarding what kind of information can be extracted from the different frequency bands can be obtained, which in turn will be useful for future ECoG applications in both basic neuroscience and applied neuroengineering.

In this study, we characterized the spatiotemporal profiles of visually driven ECoG responses across frequency bands using a retinotopic mapping paradigm, with a special focus on electrodes in the occipitotemporal cortex. Retinotopic responses have been investigated extensively for all known visual areas with various recording methods, the results of which facilitate the quantitative assessment and interpretation of our own results (Boussaoud et al., 1991; Brewer et al., 2002; Fize et al., 2003; Gattass et al., 1981,1988; Van Essen and Zeki, 1978; Van Essen et al., 1984). Specifically, in two awake monkeys, we defined the spatial preference, size of the receptive field (RF), and response latency (RL) for each subdural ECoG electrode in the time-frequency domain, as well as time-domain visual evoked potentials (VEPs).

As to the response frequencies, we took an approach that is agnostic about the presence or absence of "oscillatory" responses (For this controversial issue in the field, see Miller et al., 2009 and Gaona et al., 2011). Following previous studies that identified functionally different subcomponents in the event-related changes of lower-frequency BLP (Canolty et al., 2007; Crone et al., 1998a, 1998b; Edwards et al., 2005;

Harvey et al., 2013), we decomposed the spectral power change into the high-gamma band (80-250 Hz) and the alpha (3.9-11.7 Hz), beta (15.6-23.4 Hz), and low-gamma (30-80 Hz) bands. As we employed transient visual flash stimuli, which are typically used in a retinotopic mapping paradigm, we do not expect that changes in the power of these frequency band to reflect steady-state oscillations, which requires a longer duration of stimulation. Instead, we simply report the event-related changes in spectral power in different frequency bands as they are without giving too much interpretation. With these divisions of spectral power, we found similarities and striking differences in the spatial preference, RF size, and RL for HGP, each BLP, and VEPs, in the early (V1), middle (including V4), and high-level (TEO/TE) visual areas. Our findings constrain the models of how the measured ECoG responses are generated by a population of neurons, not necessarily from a perspective that presupposes oscillatory mechanisms in the brain.

2. Materials and methods

All experimental procedures were performed in accordance with the experimental protocols of the RIKEN Ethics Committee and the recommendations of the Weatherall report, "The use of non-human primates in research." All procedures were approved by the Committee for Animal Experiment at RIKEN (No. H24-2-2-3 (4)).

2. 1. Subjects and set-up for fixation training

Two macaque monkeys identified as Q(male, 8.1 kg) and B (male, 7.0 kg) were used in the experiments after brain magnetic resonance images (MRIs) were acquired. Before the monkeys were implanted with subdural ECoG electrodes, they were familiarized with the experimental settings and trained with a fixation task. During the fixation task, they sat in a primate chair with their head in a fixed position using a custom-made helmet for each monkey. Throughout the training and experiments, we used the same display and a custom-built computer control system (LabVIEW, National Instruments, Austin, TX) and measured horizontal and vertical eye positions at 500 Hz using an infrared video-based eye tracker (iView XTM HiSpeed Primate, SMI). For visual stimuli, we positioned a liquid crystal display (Eizo, Japan) 30 cm from the eyes. We used MATLAB (Mathworks, Natick, MA) and Psychophysics Toolbox (Brainard, 1997; Kleiner et al., 2007; Pelli, 1997) to draw visual stimuli on the display. We wrote custom codes in LabVIEW running on a realtime PXI platform that controlled the flow of the experiment and synchronized the eye tracker, MATLAB, and other equipment (e.g., reward delivery).

2.2. Electrode implantation

Subdural electrodes were surgically implanted after the monkeys completed fixation training. To anesthetize the monkeys, we administered ketamine (5 mg/kg, intramuscular), atropine (0.05 mg/kg), and pentobarbital (20 mg/kg, intravenous). Throughout surgery, we continuously monitored their heart rate and sometimes checked their reflexive responses to noxious stimulation, adjusting the dose of pentobarbital accordingly. In the subdural space, we chronically implanted a customized multichannel ECoG electrode array (Unique Medical, Japan; Nagasaka et al., 2011) embedded with 2.1-mm diameter platinum electrodes (1-mm diameter exposed from a silicone sheet). The center-to-center inter-electrode distance was 5 mm. Both monkeys were implanted with 128 ECoG electrodes, a reference electrode in the subdural space, and a ground electrode in the epidural space above the right hemisphere (the reference and ground electrodes were 5x10 mm rectangular platinum plates). To localize the electrodes, we acquired post-operative X-ray images and co-registered them with the MRIs (Fig. 1A). We manually identified the locations of each electrode by projecting the electrodes in the X-ray images onto the cortical surface reconstructed from the MRIs. In Fig. 1A, we depicted some

Fig. 1. Experimental design. (A) Locations of the 128 electrocorticography electrodes in each monkey. In monkey Q, 10 electrodes were located under the orbitofrontal cortex. The right panel for each monkey shows the electrode locations on the visual cortex where we investigated the spatiotemporal profiles of the visual responses. Colored lines indicate key anatomical landmarks. (B) The task schema (left) and spatial configuration of the mapping stimuli (right). As the monkeys held their gaze on the fixation point (FP), a series of brief flashes (mapping stimuli) were presented. The stimuli positions were randomly selected from 61 positions (shown in the right panel).

electrodes on the dorsal and ventral surfaces with tilted orientations to indicate the cortical curvature of those areas. For monkey Q, 10 electrodes are shown beneath the orbitofrontal cortex outside the cortical boundary.

2.3. Retinotopic mapping task and data acquisition

In the main retinotopic mapping task (Fig. 1B), the monkeys were required to maintain fixation as in the pre-surgical fixation task. The fixation duration varied from 2100 to 2500 ms for monkey Qand from 1800 to 2300 ms for monkey B. To investigate the visual responses in each ECoG channel, the monkeys were presented with mapping stimuli that consisted of a series of 50-ms white flashes (208.6 cd/m2) presented on a gray (4.6 cd/m2) background. The intervals between the offset of one flash and the onset of the next were randomized from 100 to 300 ms. The stimulus position was randomized across the display at 61 positions, which consisted of the central fixation location and 5 levels of eccentricity (e.g., visual angles of 0.7°-2.5°, 2.5°-5°, 5°-10°, 10°-15°, and 15°-20°) and 12 polar angles, as depicted in Fig. 1B (right panel), similar to those employed in previous studies (Hansen et al., 2004, Henriksson et al, 2012). A sequence of mapping stimuli appeared up to 7 (for monkey Q) or 4 (for monkey B) times during a single fixation. The onset timing of the mapping stimulus was defined by the response of a photodiode attached to the display. The photodiode output was recorded simultaneously with the 128 ECoG signals at 1 kHz using a Cerebus Data Acquisition System (Blackrock Microsystems, Salt Lake City, UT, USA).

2.4. Data analysis

2.4.1. Pre-processing

We removed 50-Hz line noise from the continuous ECoG data in each channel using the multi-taper method implemented as a MATLAB function (rmlinesc.m) in an open-source Chronux Toolbox (http://chronux.org) (Mitra and Bokil, 2008). From the cleaned continuous data, we extracted 380-ms epochs beginning 130 ms before stimulus onset and defined each 380-ms epoch as one trial.

After extracting the single-trial data, we excluded the trials in which the eye position deviated from the fixation point (FP) by more than 0.7°. We also excluded trials if the duration between the offset time and the onset time of the following trial was less than 200 ms. Because of this trial removal procedure, the number of trials analyzed for each stimulus position varied, ranging from 46 to 88 in monkey Qand from 45 to 89 in

monkey B. These data sets were obtained from experiments conducted over 4 and 5 days for monkey Qand B, respectively. We merged the data recorded in the daily sessions into a single data set for each subject.

2.42. Spatiotemporal characteristics

For our analysis, we first examined whether the ECoG signals recorded by each electrode showed significant changes in spectral power after stimulus onset. These data were examined for each stimulus position and each frequency band. We also examined the stability of the evoked power changes separately for increments and decrements. As shown in Fig. 3, we found that the increments in power were more stable across trials, thus all subsequent reports focus on the increments. We then characterized the spatial profile of each increment; for each electrode and frequency band, we determined the most preferred stimulus position and the extent of activity at the responsive stimulus positions (e.g., RF). In addition, we characterized the temporal profile of each increment by determining the RL for each electrode, frequency, and stimulus position. Finally, we compared these response characteristics among frequency bands and electrodes implanted on different cortical areas.

2.4.3. Test for the evoked power changes

Throughout the analysis, we used a double-resampling method (Fig. 2A). We used the following variables to represent the key aspects of the ECoG responses:

i electrode (i = 1 ... 128)

f frequencies (f = a [alpha, 3.9-11.7 Hz], b [beta, 15.6-23.4 Hz],

l [low-gamma, 30-80 Hz], and h [high-gamma, 80-250 Hz]) j stimulus position (j = 1 ... 61)

k kth trial in the 50 resampled trials (k = 1 ... 50, the 1st

resampling step) allowing replacement m mth repeat in the 100 repeats (m = 1 ... 100, the 2nd

resampling step) t stimulus period, t = before or after stimulus onset

In this manuscript, we represent these variables as abbreviations written in lower and uppercase letters (e.g., blp and BLP for band-limited power) to denote the variable estimated from a single-trial and the median across 50 samples, respectively (the 1st resampling step). We use upper case letters with a bar (e.g., BLP) to denote the mean, median, or mode across 100 repeats (the 2nd resampling step).

Resampled dataset

1st repeat

4 bands

Preferred Position (m)

Size of RF (m) x Electrodes x 4 bands Response Latency (j,m)

Preferred Position (100) -> Size of RF (100) x Electrodes x 4 bands Response Latency (j,100)

Most preferred position Size of RF Response latencies

x Electrodes x 4 bands

m th repeat

Spectrogram

Band limited power

* .....

★............ 'A

RUi-m)

"I 5 (dB)

v BLPi,hT(i>m) RL|hij,m)

v. BLP|3(j,m) RL|B(j,m)

,BLP|„(j,m)

p th repeat q th repeat

Shuffled data

p th repeat q th repeat

0 5 10 15 20

p th (dB)

r = -0.010

Cû 15

-»ST" »0

cq 5 o 0 °

0 5 10 15 20

p th (dB)

0 100 200 (ms)

Preferred BLPf (m) Position (m) RF (m)

r = 0.975

RLi,T(j,m)

Time (ms)

Fig. 2. Flow of the analysis procedure with representative data. (A) Summary of our analysis procedure. We conducted a double-step resampling analysis. First, we resampled 50 trials with replacement (k = 1 ... 50) to estimate the variable of interest (e.g., preferred position, size of the receptive field [RF], and response latency [RL]) for a single electrode i at frequency f for stimulus position j. We repeated this procedure 100 times (m = 1... 100) and averaged the data from the 100 repeats to estimate the variable of interest (for details, see Section 2.4.3). (B) Left column (middle panel): An example of the mean spectrogram of the mth resampled data set at stimulus position j (top panel) for electrode i, whose location is indicated in the bottom panel. Right column: The corresponding band-limited power (BLP). The dashed vertical line represents the stimulus onset, the colored vertical lines represent the detected response onsets (RIf, m)), and the stars represent the maximum evoked power in each time series (BLPfj, m)). The shading represents the confidence interval (see Section 2). (C) The corresponding response map for electrode i in the high-gamma band (BLPih) in thepth and qth resampled data sets and scatter plots of BLPih used to evaluate the stability of the spatial selectivity (see Section 2). The correlation coefficients between BLPy,(p) and BLPy,(q) across stimulus positionj's are indicated above the scatter plots. The top row is from the real data, whereas the bottom row is generated from the shuffled data (see Section 2). For an exemplar response with decrement, see Fig. 3G. (D) The response map obtained from the mth repeat (left), from which we derived the preferred position (middle) and the RF (right).

To determine whether an ECoG signal was driven by visual stimuli in each frequency band and each stimulus position, we first randomly sampled 50 trials and examined whether the visual stimuli evoked a significant change in spectral power. Throughout the analysis, we always resampled the data with replacement. For electrode i, in the kth trial at position j, we estimated the spectrogram using the multi-taper method, where a half-bandwidth (W) is defined by W = (tp + 1 )/2 T, with tp and T being the number of tapers and the size of the time window in seconds, respectively (Mitra and Bokil, 2008; Tsuchiya et al., 2008). We used tp = 1 and T = 0.128 s, resulting in a half-bandwidth of 7.8 Hz, which is sufficient to estimate the power examined in this analysis (alpha, beta, low-, and high-gamma). We confirmed that the time window of 0.128 s was sufficient to resolve the BLP in the low-frequency bands by using an artificially generated signal (Supplementary Fig. 1). Note that we denote the frequency range as ^ — F2 Hz] including the error margin of the half-bandwidth. For example, we analyzed the BLP at the center frequency of 7.8 Hz as the alpha (F1 = 7.8 — W/2 = 3.9 and F2 = 7.8 + W/2 = 11.7 Hz). We obtained a single-trial estimate of the power spectrogram, pf,t(j, m, k), where f spans from 0 to 250 Hz. We then transformed the power into a log-scale and converted it into an evoked power spectrogram using the following equation:

mj' ,t U, m, k) = 20loS10(pi,f, t(j> m, k)) —220log10(p;_ f, baseline (j; m, k))

where t is the latest point in the time window (e.g., the time window ranged from t — 128 ms to t), and baseline is the period between -130 ms and + 20 ms from stimulus onset. To obtain the time course of the BLP, we averaged the evoked power within each frequency band:

Mpij,t(j,m,k) = -^Y,epi,j,t(j; m>k) (2)

where njis the number of frequency points j) for the frequency bandf. We then defined the mean band-limited power, BLPi<it(j, m), by averaging blpi<it(/, m, k) over trials k = 1 ... 50. The spectrogram and BLP averaged across 50 trials are shown in Fig. 2B.

Over k =1 ... 50 sampled trials, we estimated the confidence intervals of the BLP at each time step, as follows:

Lower limit for BLPi, f, t(j,m) = q2 —1.57(q3—q1)/n1/2

Upper limit for BLPi; f, t(j, m) = q2 + 1.57(q3—q1)/n1/2 ( )

where q1, q2, and q3 are the 25th, 50th, and 75th percentiles of BLPift(j, m), respectively, among k =1 ... 50 trials (McGill et al., 1978). If we observed the lower limit for BLPij;t(j, m) to be larger than the upper limit of the confidence interval estimated for BLP^toO', m), when t0 = 0.02 s (e.g., -108 to +20 ms from the stimulus onset), we considered it a significant evoked response at time t. We repeated this procedure with a time step of 3 ms from t = 20 to t = 250 ms after

stimulus onset. When significant changes were observed at five consecutive points (or 15 ms), we defined the first significant time point as the RL, RLy(j, m) (Fig. 2B, right). When the onset was not detected within t = 20 to 250 ms, RL was not defined.

We repeated this procedure for m = 1 ... 100 repeats. If RLyO, m) was defined for more than 50% of the repeats, we defined the RLi; f (j) as the median across the subset of resampled data sets in which the latency was defined. When RLi;f (j) was defined, we regarded electrode i as showing a significant increment in frequency band f at position j. We used RLf in the analysis for the temporal profiles of visual responses.

We also defined RLi;f (j) for decrements by comparing the upper limit for BLPyt(j, m) to the lower limit of the confidence interval estimated for BLPyt0(j, m). To quantify the strength of the evoked power change, we defined an index for the strength of the increment, incBLPy(j, m) as the maximum value of BLPyt(j, m) (Fig. 2B, right). We also defined an index for the strength of the decrement, decBLP,f(j, m) as the minimum value of BLPi<it(j, m).

2.4.4. Stability of increments and decrements

To evaluate the consistency of a response pattern across all tested stimulus positions, we created a spatial response stability index. The index reflects the correlation of the mean spatial response pattern across resampled data sets. The stability reaches 1 when the mean evoked changes are the same across resampled data sets (m = 1 ... 100) in all the stimulus positions (j = 1 ... 61) and approaches 0 for random data sets. Specifically, we defined the spatial response stability as the mean correlation coefficient of incBLPf among all combinations of the resampled repeats (e.g., 100 x 99/2 combinations). In Fig. 2C, we demonstrate how we computed the stability index. For a given combination of pth and qth repeats, we computed a correlation coefficient among 61 stimulus positions. When we calculated the stability index for a null distribution created by shuffling stimulus positions (the bottom row in Fig. 2C), the index was close to 0. We also repeated the procedures for the decrement using decBLP.

2.4.5. Preferred position

We defined the most preferred stimulus position using incBLPyO, m). For each mth repeat, we determined position j that elicited the maximal incBLPyO, m) as a preferred position (PP, Fig. 2D). Thus, PPy(m) takes a value that represents stimulus position j = 1 ... 61. We defined the most preferred position as position j that was most frequently selected as the preferred position over m = 1 ... 100 repeats. Note that we defined the most preferred position for all electrodes in each frequency band from the same set of trials.

We evaluated the differences in the most preferred position among frequency bands separately for the polar angle and eccentricity. At a given electrode, we calculated the difference in the most preferred polar angle (or in the most preferred eccentricity) between the frequency bands f1 and f2 (f1 < f2). The difference in the most preferred angle becomes positive when the higher-frequency band (f2) prefers the upper visual field relative to the low-frequency band (f1) (see the inset at the bottom of Supplementary Fig. 4). To determine whether the difference was significant, we performed permutation tests. We generated a null distribution for electrode i by randomly swapping the frequency labels and computed the angle,' and eccentricity,' (the mean over m = 1 ... 100) 1000 times. We regarded the observed difference in anglei and eccentricityi to be significant if it was in the top 2.5% or the bottom 2.5% of the null distribution (two-sided test, a = 0.05). When we tested the difference in the preferred angles, we removed the electrodes representing the fovea in either band.

To test if there is a reliable difference in the most preferred angle or eccentricity between frequency bands, we calculated the median of the differences for all of the electrodes within the occipitotemporal cortex (see Fig. 1A, right panel for each monkey) and performed a non-

parametric two-sided sign test contrasting ^anglei and ^eccentricity; against 0 (a = 0.05). We corrected for multiple comparisons with the false discovery rate (FDR) at q = 0.05 (Benjamini and Hochberg, 1995).

We also examined the differences in the preferred eccentricity across bands within each brain region. We classified the electrodes into three groups: V1, V4+, and TEO/TE (see Fig. 6A), based on the electrode locations relative to the lunate sulcus, the inferior occipital sulcus, and the superior temporal sulcus (Fig. 1A). We excluded some electrodes because they were positioned over sulci or on a boundary between the regions. We labeled the electrodes represented by triangles in Fig. 6A as V4+, as most of V2 and V3 are buried within the lunate and inferior occipital sulci, but they do partially extend onto the surface.

We tested the significance of the effect of frequency bands on the preferred eccentricity using Friedman's test (two-sided test, a = 0.05) and a post hoc two-sided sign test with FDR correction (q = 0.05). We used the same statistical approaches for the other response characteristics (i.e., RF size and response latencies).

We defined the preferred stimulus position in each resampled data set because we observed that the spatial selectivity was unstable, especially in the lower-frequency bands (see Figs. 3G and J). As an alternative, we defined the most preferred position as the stimulus position where the mean evoked power across 100 resampled data sets was the largest among all of the positions. The results for these two methods were highly similar (data not shown).

2.4.6. RF size

We defined the RF size for each electrode with the double-step resampling method and using incBLPyO', m). For a given mth repeat, we regarded stimulus position j to be inside the RF if incBLPyO, m) was above the half-maximum of incBLPyO, m) (Fig. 2D). We excluded stimulus position j if it was not spatially contiguous with other positions in the same RF. The size of RF (SRF) was defined as

SRFi,f(m) ^Ytj^j(m)areaofp°sition j (4)

The area of position j is the physical area of the mapping stimulus (Fig. 1B, right panel). We repeated this procedure for m = 1 ... 100 repeats and defined the SRF as the mean of SRFy(m) over m. Because the area of each visual flash at the periphery was larger than the area near the fovea (Fig. 1), this measure can overestimate the SRF if it includes the stimulus position at the periphery. Thus, we performed an alternative analysis by regarding the area to be the same across the stimulus positions (i.e., simply counting the number of stimulus positions). We found that the results did not differ qualitatively between the two methods (data not shown).

In order to compare the RF sizes among the bands recorded with the same electrode (Figs. 7B-D), we defined the normalized difference (ND) for electrode i as

NDi = (SRFi,f2-SRFi,f 1)/(SRFf + SRF^ f 1) (5)

where f1 and f2 represent different frequency bands (f1 < f2). We performed permutation tests to determine whether the difference was significant. We generated null distributions by randomly swapping the frequency labels and computed the ND of the shuffled trials 1000 times. We regarded the difference as significant if the original and observed ND; ranked in the top 2.5% or the bottom 2.5% of the null distribution (two-sided test, a = 0.05).

Similar to the analysis for the most preferred position, we evaluated the significance of the population difference in the visual cortex using non-parametric two-sided tests that contrasted NDi against 0.

C Visual cortex

A Visual cortex

B Outside Visual cortex

O 0.4 t o a.

a p y hy VEP I I Increment I I Decrement

a ß y hY VEP

ß Y hY VEP

D Outside Visual cortex

Y hY VEP

50 100 150 200 250

Latency

50 100 150 200 250

Latency

□ Increment Decrement VEP

Visual Outside Shuffle cortex Visual cortex

Example

p th repeat q th repeat

r = 0.483 o

-3 -2 -1 0

p th (dB)

monk Q

monk B

monk Q

monk B

J Visual cortex

\a <>.5 .2 w

a p y hy VEP I | Increment I I Decrement

Fig. 3. Incremental visual responses in the electrocorticography are observed more frequently, with faster latencies and higher stability compared with decremental responses. (A and B) For a given electrode in a given frequency band, the latency was defined for a stimulus position if the stimulus-evoked significant incremental and decremental responses (see Section 2). The values represented here are the mean proportion of stimulus positions that showed significant responses, averaged over the electrodes on the visual cortex (A) or over the other electrodes outside the visual cortex (B), separately for each frequency band, as well as for the visual evoked potential (VEP). (C and D) Box plots for the response latencies for the increment (white boxes) and decrement (gray boxes) for each band and the VEP for the electrodes on the visual cortex (C) or for the electrodes outside the visual cortex (D). (E) Box plots for the stability indices (see Section 2) for the increment (white boxes) and decrement (gray boxes) for all frequency bands and the VEP (dark gray boxes) for the electrodes on the visual cortex (left) and the electrodes outside the visual cortex (middle). A box at the right edge represents the indices for the shuffled data sets (see Section 2) for the electrodes on the visual cortex. (F) An exemplar response map of decBLPf in the pth and qth resampled data sets (left panel), and for the scatter plots of decBLPf obtained in the procedure to evaluate the stability of the spatial selectivity (right panels, see Section 2). (G-I) Spatial distribution of the stability indices on the cortical surface of each monkey for the increment (G) and decrement (H) for all BLPs and VEPs (I). J Stability indices for the BLP (white bars, increment; gray bars, decrement) and the VEP. BLP, band-limited power.

2.4.7. RL

To compare the RL among the bands recorded with the same electrode, we defined the difference in RL among the frequency bands recorded at each stimulus position as:

dRLi(j) = RLi, f 2( j)-RLi, f 1 (j) (6)

We assessed the significance of the differences with permutation tests using the procedure described above (i.e., constructing a null distribution by swapping the frequency band labels, and determining significance with two-sided tests, a = 0.05). Then we took the median of significant dRLs across positions as a representative value of the difference in RL for each electrode in each paired comparison. When we computed the median of all the dRLs, including the non-significant dRLs, the absolute value of the median dRL was reduced, but the conclusions from the statistical tests were not different (data not shown).

In the population analysis across regions or within a region, we exploited the advantage of simultaneous recordings with many ECoG channels. Specifically, we compared the relevant RLs that were estimated with the same set of trials. For each frequency band f and position j, we identified a set of electrodes where RL was defined within region A (A = V1, V4+, or TEO/TE). We call this set EAfj (e.g., if i £ EVhaj, we defined the RL for electrode i, which is located within V1, in the alpha band for stimulus position j). Then the RL at frequency f for position j and region A is

RLaJ (j) = nf RLif(j) (7)

where nAfj represents the number of electrodes in the set Ef

To compare RLs across regions within each frequency band in a fair manner (Fig. 9A, black), we fixed the stimulus position, j, and selected

a subset of electrodes, whose RL could be defined at stimulus position j. We denote the average RL for those electrodes as a population-latency RLf) for frequency band f in cortical region A (A = V1, V4+, or TEO/ TE) at stimulus position j. In some stimulus positions (), there were no electrodes for which we could define RLs for all cortical regions. Thus, the number of stimulus positions j for which RLf) was defined varied for each monkey and frequency band (35,49,35, and 31 stimulus positions in monkey Q, and 49,55,27, and 33 positions in monkey B, for the alpha, beta, low-, and high-gamma bands, respectively). For the analysis reported in Fig. 9A, we further averaged RLf) across stimulus positions j to obtain RLAf which can be meaningfully compared across all frequency bands. In other words, our procedure removes the effects of stimulus positions and electrodes, allowing us to compare the RLs, which were computed from the same set of the trials, across cortical regions in a fair manner. We also computed RLs at the most preferred position for each electrode (Fig. 9A, gray), which were obviously computed from different sets of trials depending on the region and frequency band.

Similarly, we compared RLs across frequency bands within a region in a fair manner (Fig. 9B); we fixed the stimulus position and selected a subset of electrodes whose RLs could be defined for all frequency bands. For some stimulus positions, we could not define the RL for any of the frequency bands or electrodes. Thus, the number of stimulus positions for which RLAfj) was defined varied for each monkey and region (22, 26, and 16 in monkey Q, and 24, 38, and 11 in monkey B, for V1, V4+, and TEO/TE, respectively). For the analysis reported in Fig. 9B, we further averaged RLAfj) across stimulus positions to obtain RLAf for comparing the RLs across all frequency bands, thus removing the effects of stimulus position and electrode, based on the same set of trials.

We examined the possibility that analytic artifacts affected the estimate of RLs by using artificially generated signals. This analysis confirmed that our findings regarding the difference in RLs across frequency bands could not be attributed to analytic artifacts. For additional details, please see the section "Simulation of response latencies" in the Supplementary Materials and Supplementary Fig. 2.

2.4.8. Analysis ofVEPs

To examine the relationship between the BLP changes and VEPs, we defined the preferred stimulus position, RF size, and RL from the VEP for each electrode, using the same double-resampling procedure as described for the BLP. For the VEPs, we corrected the baseline by subtracting the mean voltage in the time window of 150 ms, starting from 130 ms before stimulus onset in each trial; then we averaged the signals across the trials to obtain the VEP. For all analyses, we disregarded the polarity of the VEP. For the PP and RF size, we used the maximum absolute value of the VEP. We defined the earliest significant increase or decrease from the baseline as the RL. The rest of the analysis procedures were identical to those for the BLP.

3. Results

3.1. Visual stimuli evoked an increase in power across frequency bands

Fig. 3 summarizes the proportion of significant responses (A and B), RLs (C and D), and response stability (E-J) for the visually evoked incremental and decremental spectral power and potentials. Increments were more frequent than decrements across all frequency bands. Furthermore, in the visual cortex, an increment typically reached significance much earlier than a decrement, except in the alpha band (Fig. 3C). We also observed that decrements were less stable in spatial selectivity, especially in the occipitotemporal cortex (Figs. 3E, H, andJ).

This may seem inconsistent with the findings of previous studies reporting event-related decrements in the low-frequency activity, including the alpha and beta bands (Aoki et al., 1999; Crone et al., 1998a,1998b; Edwards et al., 2009; Hermes et al., 2012; Miller et al., 2007, 2009; Ohara et al., 2000). This apparent difference could be due

to our focus on the transient component of responses (50-250 ms) that occurred just after stimulus onset, whereas the previous studies focused more on the sustained component of responses in a later time window. For example, a study in the somatosensory cortex reported that, in the alpha and beta bands, the response initially increased, but then decreased after median nerve stimulation (Fukuda et al., 2010). This is consistent with our present finding that decrements occurred later than increments.

Hereafter, we focus on the response increment in the electrodes on the occipitotemporal cortex (see Fig. 1A, right panel for each monkey). Within the occipitotemporal cortex, we examined the recording stability across days by comparing the waveforms of the visual responses. Our analysis confirmed that the waveforms in all bands as well as VEP are highly stable in the most preferred stimulus position, with correlation coefficients raging from 0.87-0.98 over 4 (monkey Q) or 5 days (monkey B). Our results extend the stability of ECoG recording reported by Rubehn et al. (2009) and Chao et al. (2010). For details, please see the section "Validation of the recording stability" in the Supplementary Materials and Supplementary Fig. 3.

3.2. The most preferred positions were similar among frequency bands in V1 and V4 +

We defined the most preferred position for each electrode and frequency band (Figs. 4 and 5A) and then compared them among bands recorded from the same electrode. We analyzed the polar angle (Supplementary Fig. 4) and eccentricity (Figs. 5B, C, and D) of the most preferred position, respectively.

In general, retinotopic organization in the early visual areas (V1 and V4+) was similar across frequency bands (Figs. 4 and 5A). In terms of the polar angles, the most preferred positions gradually shifted from the lower to the upper quadrants of the contralateral visual field (from green to pink in Fig. 4) as the electrodes go from the dorsal to the ventral surface. This trend was similar across all of the bands. In terms of the eccentricity (Fig. 5A), the foveal representation was observed around the inferior occipital sulcus, and the preferred eccentricity increased as the distance from the sulcus increased. We could not identify an ordered structure in the temporal cortex that was consistent between monkeys.

The preferred angles among the frequency bands were similar, which was confirmed by pair-wise comparisons between the frequency bands (Supplementary Fig. 4A-C). As a population, only a few electrodes showed significant differences between bands (Supplementary Fig. 4A, left panel), and they failed to show any notable spatial patterns that were consistent between the two monkeys (Supplementary Fig. 4B and C). The differences were distributed around 0 (Supplementary Fig. 4A, right panel), and the overall differences were not significantly different from 0 for any of the frequency band pairs (two-sided sign test, p > 0.1 for all comparisons).

In the population analysis, the preferred eccentricities were comparable among the bands (Fig. 5B), except between the beta and high-gamma bands (red asterisk in Fig. 5B, right panel. p = 0.0016, which survived FDR correction q = 0.05). This difference was mainly due to the difference in preferred eccentricity of the electrodes around the inferior occipital sulcus and temporal cortex; these electrodes showed preference toward the fovea in their HGP and preference toward the periphery in their beta-band power (blue circles in panel v in Figs. 5C and D), and preference toward the periphery for the low-gamma band than for the alpha band (red circles in panel ii in Figs. 5C and D). This is more discernible in monkey Q, but a similar trend was observed in monkey B. By grouping electrodes within each cortical area (V1, V4+, TEO/TE), we quantified the above observation; the preferred eccentricities were similar across the frequency bands in V1 andV4+ (p > 0.05), but differed in TE/TEO (Friedman's test, p = 0.0291, Fig. 6B). Post hoc comparisons confirmed that in TEO/TE the beta band preferred more peripheral positions than the alpha band (p = 0.049) and that the low-gamma

Fig. 4. The most preferred angles were similar among the frequency bands (see also Supplementary Fig. 4). Spatial distribution of the most preferred angles. Upper row, monkey Q lower row, monkey B. The color represents the polar angle of the most preferred position for each electrode, and the sizes indicate the stability of the spatial selectivity (see Fig. 3). Each column corresponds to each frequency band. The electrodes representing the fovea are represented by yellow stars.

band preferred more peripheral positions than the high-gamma band (p = 0.0352).

Overall, the electrodes over V1 and V4+ showed similar spatial preferences, for polar angles and eccentricities across all frequency bands. The electrodes over TEO/TE showed a trend indicating that the beta

and gamma bands prefer the peripheral visual field while the alpha and high-gamma bands prefer the foveal region, an observation that we will revisit in Section 4. These results imply that our systematic across-band comparisons of visual responses may be particularly sensitive for revealing the properties of TEO/TE.

Q_ 0.25 0

— 0 □c-10 -20

a vs ß a vs y a vs hy ß vs y ß vs hy yvs hy

s ß a vs y a vs hy ß vs y ß vs hy yvs hy

¡i OT 10

L<H œ

Fig. 5. The most preferred eccentricities were comparable across the bands, with some region-specific differences. (A) Spatial distribution of the most preferred eccentricities. Upper row, monkey Q lower row, monkey B. (B) Population analysis for differences between the bands. Proportion of electrodes with significant differences (left panel) and the distribution of differences (right panel) in each pair of frequency bands. The yellow circles and red asterisk in the right panel represent the median values among all electrodes. The red asterisk indicates that the median of the overall differences was significantly different from zero. (C andD) Spatial distribution of the significant electrodes in each pair of frequency bands in monkey Q(C) and monkey B (D). Electrodes with significant differences are shown with colored circles. Red and blue circles represent electrodes with positive and negative differences in eccentricity, respectively. The size of the circles represents the average of the stability indices (see Fig. 3), which were defined separately for the lower- and higher-frequency bands within each electrode. Black rectangles in panel (ii) and (v) indicate the cluster of electrodes mentioned in the main text.

3.3. The RFsize was smaller in the high-gamma band, but not necessarily larger in the alpha band

We examined the extent of the responsive spatial locations for each electrode in each frequency band as a proxy of the population RF size (Fig. 7A, the circle color indicates the square root of the RF sizes; exemplar RFs are shown in Supplementary Fig. 5). We compared the RF sizes among the bands recorded on the same electrode (Figs. 7B-D).

Previous human ECoG studies suggested that high-gamma activity reflects local processing (Crone et al., 1998a, 2006; Jerbi et al., 2009), which indicates smaller RF sizes. Consistent with the findings of previous studies, at a population level, the RF sizes for the high-gamma band in monkey ECoGs were significantly smaller than the RF sizes for the other bands (Fig. 7B, two-sided sign test, p < 0.005 for all comparisons). On the other hand, the RF sizes for the alpha, beta, and low-gamma bands did not differ significantly (all p > 0.1), indicating that the RF sizes were not necessarily larger for the lower-frequency bands (see examples in Supplementary Fig. 5).

The above results were obtained by pooling data from all electrodes across all regions. By assigning electrodes into V1, V4+, or TEO/TE, we found that the relationship between the RF sizes and the frequency bands was highly dependent on the regions (Fig. 6C). In V1, the RF sizes decreased as the frequency bands increased (Friedman's test, p < 0.0005). Post hoc paired tests revealed that the high-gamma band defined smaller RFs than the alpha (p < 0.001) and beta (p < 0.0001) bands. The low-gamma band also defined smaller RFs than the beta band (p < 0.01). In V4+, the frequency band had only a weak effect on the RF sizes (Friedman's test, p = 0.0426. No significant post hoc comparisons). Finally, TEO/TE showed a strong non-linear dependency with the frequency band (Friedman's test, p < 0.0005). In fact, the pattern of frequency-dependency for RF sizes is similar to that for the preferred eccentricities (see Figs. 6B and C, right panels), though it is statistically much more reliable for the RF sizes. In TEO/TE, the RF sizes were larger in the beta and low-gamma bands than in the alpha and high-gamma bands, as confirmed by post hoc comparisons (p < 0.005

alpha band vs. beta band, beta band vs. high-gamma, band, and low-gamma band vs. high-gamma band).

Because the area of each visual flash at the periphery was larger than the area near the fovea (Fig. 1), our estimates of the RF sizes and the preferred eccentricities were not independent. However, when we simply counted the number of stimulus positions that elicited more than half of the maximum response, we found a similar dependency on frequency bands and cortical regions (data not shown, see Section 2.4.6). Note that because all of the analyses were performed uniformly across the frequency bands, any frequency-dependent differences in retinotopic organization among the cortical areas cannot be attributed to artifacts related to our stimulus design or analyses.

As a population, the high-gamma responses showed smaller RFs than the other frequency bands. There is a widespread assumption that low-frequency signals show poorer stimulus specificity because the brain tissue acts as a low-pass capacitive filter (Bedard et al., 2004, 2006). While this notion is consistent with our findings using the stability index (Fig. 3J), it does not entirely explain the findings. For V4+, the RF sizes for the alpha, beta, and low-gamma bands were similar. For TEO/TE, the RF sizes for the alpha band were much smaller than the RF sizes for the beta and low-gamma bands. These results are difficult to explain if one assumes a lower spatial specificity for lower-frequency activity and revised the assumption of the poor stimulus specificity for the low-frequency signal. We will elaborate on these points in Section 4.

3.4. RLs respect the visual hierarchy and are much slower in the alpha band

Finally, we examined the RLs. We defined the latency for each stimulus position in each frequency band for each electrode recording. Fig. 8A shows the latencies for the most preferred position of each electrode, which was determined separately for each band. In the paired comparisons (Figs. 8B-D), we evaluated significant differences between bands using the same set of trials for each stimulus position and utilized the median of the significant differences across the stimulus positions as

monk Q

monk B

1 V4+ TEO/TE

a p Y hY V1

a p y hY

a p Y hY V4+

a p Y hY

TEO/TE

a p Y hY TEO/TE

a p Y hY

Fig. 6. V1 and TEO/TE revealed idiosyncratic relationships with the receptive field (RF) sizes among the frequency bands. (A) Electrode classification. The electrodes in V1, V4+, and TEO/TE are represented by white circles, black triangles, and gray squares, respectively. Colored lines indicate key anatomical landmarks (yellow: lunate sulcus, red: inferior occipital sulcus, blue: superior occipital sulcus). (B) Comparison of the preferred eccentricities (Ecc) within each region. (C) Comparison of RF sizes within each region. The frequency bands had a significant effect on the RF sizes in V1 and TEO/TE ( p < 0.0005, Friedman's test). Horizontal bars represent pairs showing significant differences in the post hoc comparison. Vertical bars represent the 95% confidence intervals. *p < 0.01, **p < 0.005, ***p < 0.001.

monk Q

monk B

o i» a

O 05 i o 0.1 y

L<H L>H

a vs p a vs y a vs hy p vs y p vs hy yvs hy

............a a = |

L>H a vs p a vs y a vs hy p vs y p vs hy yvs hy

monk Q

a vs p

monk B

¡ a vs p

L>H L<H

D¡ff Idx

Fig. 7. The receptive field (RF) sizes were smaller in the high-gamma band but were comparable among the other low-frequency bands. (A) Spatial distribution of RF sizes. (B) Population analysis for differences among the bands. Proportion of electrodes with significant differences (left panel) and the distribution of differences (Diff Idx; right panel) in each pair of frequency bands. (C and D) Spatial distribution of the significant electrodes in each comparison in monkey Q (C) and in monkey B (D). The format is the same as Fig. 5.

a representative value for each electrode (see Section 2). When using the median of all the differences, including non-significant differences, the absolute values of the RLs were reduced, but our conclusions did not change.

For most of the electrodes, the RLs of the alpha band were slower than those of the other three bands (Fig. 8B, left), which was confirmed by a statistical test at the population level (Fig. 8B, right panel, all p < 0.0001). We also observed that the beta and low-gamma bands responded earlier than the high-gamma band (p < 0.005 for both comparisons). In both monkeys, the earlier responses in the beta band than in the high-gamma band were observed in the electrodes around the inferior occipital sulcus and on the temporal cortex (black rectangles in Figs. 8C and D, panel v). The earlier response in the low-gamma band than in the high-gamma band were primarily attributed to monkey Q. These results are demonstrated in the time-frequency plots for two representative electrodes in Supplementary Fig. 6. We also note that these latency differences did not arise due to an artifact of our analysis procedure (Supplementary Fig. 2).

Utilizing the advantage that simultaneous ECoG recordings offer, we further compared the RLs within each frequency band (Fig. 9A) or within each visual area (Fig. 9B) using the same set of trials (see Section 2). Within each frequency band (Fig. 9A), the RLs increased in the same order as the cortical hierarchy (Friedman's test: p < 0.0005 for all frequency bands). The latency differences between the lowest-level (V1)

and highest-level (TEO/TE) regions in the hierarchy were significant (post hoc two-sided test, p < 0.0005 in all bands). The paired differences between V1 and V4+ reached significance for the low- and high-gamma BLP (p < 0.01), as did those between V4+ and TEO/TE for all frequencies (p < 0.005), although the low-gamma band was less significant than the other bands (p = 0.0151).

Fig. 9 (gray circles) depicts the averaged latencies in the most preferred positions for each band and region (also see Table 1). Note that these analyses were performed based on different sets of trials, which were optimized for each band and each channel, in a similar manner as the RL analysis in single unit studies.

Within each region (Fig. 9B), the RLs were highly dependent on the frequency bands (Friedman's test, p < 0.0005 for all regions). Furthermore, in all regions, the RLs for the alpha band were slower than the RLs for the other bands (post hoc two-sided test, p < 0.0005). The frequency dependence of the RLs was similar between V1 and V4+. On the other hand, TEO/TE exhibited a rather unique pattern. In addition to the slower latencies in the alpha band, the latencies for the high-gamma band were also slower than the latencies for the beta and low-gamma bands (p < 0.005). The difference with the beta band was 17 ms (median, 25-75 percentiles: 7-35 ms), and the difference with the low-gamma band was 17 ms (25-75 percentiles: 7-22 ms). The unique pattern in TEO/TE occurred for both the temporal RLs and the spatial selectivity (Figs. 6B and C).

Since high-gamma BLP is typically regarded as an indicator of spiking activity (Kayser et al., 2007; Ray et al., 2008, 2011), the latencies of the high-gamma BLP in V1 might seem too long (see Table 1, 83 ms as a population even in the most preferred position). However, studies with single unit recordings have reported variability in the unit latencies (sometimes beyond 75 ms, as in Maunsell and Gibson, 1992, or in the range of 34 to 97 ms, as in Schmolesky et al., 1998) and increases in unit latencies from layer 4 to the cortical surface (Maunsell and Gibson, 1992). Furthermore, we did not optimize the stimuli for all neurons that contributed to the ECoG recording, which would delay the estimates (as can be inferred from the differences in the latencies for the most preferred [gray circles in Fig. 9A] and all available positions [white circles in Fig. 9A]). Given that the ECoG signal better correlates with activity in the superficial layers than activity in the deeper layers (Watanabe et al., 2012), the latencies of the high-gamma responses we observed seem reasonable.

In summary, we found that the RLs in the alpha band were significantly slower than the RLs in the other bands, and that the RLs within each frequency band increased with the cortical hierarchy, from V1 to TEO/TE. Notably, we identified a unique pattern of frequency-dependent response latencies in TEO/TE, which mirrored the unique frequency-dependent profiles of spatial selectivity in TEO/TE (Figs. 6B

and C). Although the responses in V1 and V4+ for the beta, low-, and high-gamma bands occurred almost simultaneously, the responses in TEO/TE in the beta and low-gamma bands preceded the responses in the high-gamma band by ~17 ms.

3.5. Comparison between BLP and VEPs

To examine the relationships between BLP and VEPs, we defined the PP, RF size, and RL from the VEP for each electrode and compared them to those estimated from the BLP.

The preferred angles and eccentricities were comparable between the VEP and the BLP in all frequency bands (Supplementary Fig. 7). Most of the electrodes did not show significant differences between the VEP and BLP, and the difference as a population was not different from zero (two-sided sign test with FDR correction, q = 0.05; Supplementary Fig. 7).

As a population, the RF sizes estimated from the VEP were larger than the RF sizes of the BLP, in all frequency bands (two-sided sign test, p < 0.0001 for all comparisons; Fig. 10A, bottom panel). Although some electrodes showed smaller RFs in the VEP than the BLP (red circles in Fig. 10B), their spatial distribution was not consistent across subjects.

monk Q i

monk B

O io t

O 0.5 g

• 0.1 gy

CL 0.25 0

L<H L>H

1» 60 E

~ 0 Í -60 -120

a vs p a vs y a vs hy p vs y p vs hy y vs hy

Ml . 1

1 11 s * * ■ 1

a vs p a vs y a vs hy p vs y p vs hy yvs hy

monk Q

i a vs p

monk B

i a vs p

vi a vs hy v p vs hy iv y vs hy vi a vs hy v p vs hy iv y vs hy

Not defined

Fig. 8. Slower response latencies in the alpha band. (A) Spatial distribution of the response latencies in the most preferred positions defined for each electrode and each frequency band. (B) Population analysis for the differences among the bands. (C and D) Spatial distribution of the differences. Red and blue circles represent the electrodes in which the mean latency for the lower-frequency band was earlier or later, respectively. The electrodes where the latencies were not defined for both bands are not depicted. Black rectangles in panel (v) indicate the cluster of electrodes we mentioned in the main text. The format is the same as that in Fig. 5.

The RLs of the VEP were shorter than the RLs of the BLP, for all frequency bands (two-sided sign test, p < 0.0001 for all comparisons; Fig. 11A, bottom panel) in almost all electrodes (Fig. 11B).

In sum, the VEP and the BLP showed similar spatial preferences; the VEP defined larger RF sizes and faster RLs than almost all the BLPs across all areas. In general, these results can be explained by the fact that the VEP reflects the phase-locked components of the spectral power across all frequencies, and as evoked power changes in any frequency can cumulatively contribute to the VEP, the VEP may show less spatial selectivity and faster latencies than the BLP.

4. Discussion

In the present study, we characterized the responses evoked by small visual flashes in ECoG data recorded from two monkeys with implanted subdural electrodes using a retinotopic mapping paradigm. By comparing the responses across frequency bands as a population, we observed highly consistent spatial preferences across bands and found that the smallest RF was in the high-gamma band, while the longest RLs were in the alpha band. Furthermore, our novel findings emerged from the comparison of the BLP across the regions. We found that only V1 showed a decrease in RF size as the frequency increased and that TE/TEO showed unique patterns in the spatiotemporal profiles, which has not been reported in previously. Below, we discuss our results with regards to previous findings, and deliberate on the neural underpinning of our findings. Finally, we speculate on the functional implications of our observations.

4.1. Similar spatial preferences across bands

For a given electrode, we found that spectral power in all bands and the VEP demonstrated similar spatial preferences, which we examined with the most preferred polar angles and eccentricities (Figs. 4 and 5, also see Supplementary Figs. 4 and 7). The same results were observed across the visual areas. It is reassuring that the most basic property of the visual response, that is the most preferred stimulus position, is preserved across frequencies, as well as the VEP, and potentially even with single unit activity and functional MRI blood-oxygen-level dependent

Table 1

Response latencies in the most preferred position. Values are the medians with the minimums and maximums in parentheses. Units are in milliseconds.

alpha beta Low-gamma High-gamma

V1 113(102-125) 98 (84-104) 85(80-92) 83(80-90)

V4+ 125(108-139) 107 (101-117) 100(84-112) 95 (84-120)

TEO/TE 142 (135-158) 115(80-131) 119 (108-125) 125(116-131)

signals, when using a retinotopic mapping paradigm, one of the most extensively studied paradigms in vision science (Wandell and Winawer, 2011). However, we point out a limitation of our study: our retinotopic mapping stimuli were rather coarse. Future studies using finer mapping stimuli might reveal finer differences in the most preferred positions across frequency bands for a given ECoG electrode. In addition, we cannot reject the possibility of distinct spatial preferences across frequencies when using more local measurements, such as recordings with laminar electrodes (Chen et al., 2007).

4.2. Smaller RFfor the high-gamma band than for the other bands

Previous studies showed a tight-relationship between spiking and HGP in local field potentials recorded by a single electrodes inserted into the cortex (Kayser et al., 2007; Ray and Maunsell, 2011, Ray et al., 2008). In the present study, we found that, as a population, the HGP defined the smallest RF compared to the BLP in the other bands (Fig. 7). This finding has an important implication as to how HGP and other BLP relate to the underlying neuronal spiking and subthreshold membrane potentials, especially when considered together with our present findings regarding the RLs across frequencies in each visual area.

Note that the RF sizes we estimated were much larger compared to the RFs described in previous studies. It was possible that this difference was related to our stimulus design since our stimuli were not optimized to estimate the RF for each electrode recording or to fit two-dimensional Gaussian functions (Yoshor et al., 2007). However, all frequency bands were analyzed in the same way. Thus, our finding of different RF sizes across bands cannot be explained by the features of our stimulus design.

® 120 to

200 160 120 80

200 160 120 80

200 160 120 80

V1 V4+ TEO/TE

V1 V4+ TEO/TE

V1 V4+ TEO/TE V1 V4+ TEO/TE

Latency in the most preferred position

Fig. 9. (A) Response latencies increased along with the cortical hierarchy across the alpha, beta, low-, and high-gamma bands (from left to right). White circles connected by black lines represent the median latencies across electrodes within a region. Gray circles represent the median latencies defined at the most preferred position (see Table 1). (B) The temporal profile across bands is dependent on the brain regions. In V1 (left) and V4+ (middle), the alpha-band response occurred later than the other bands, while the responses in the other bands occurred almost simultaneously. TEO/TE (right) showed a unique profile, in which the beta and gamma responses preceded the high-gamma response, followed by the alpha response. Horizontal bars represent the pairs with significant differences in the post hoc comparison. Vertical bars represent the 95% confidence intervals. *p < 0.05, **p < 0.005, ***p < 0.0005.

Fig. 10. Receptive field (RF) sizes estimated from the visual evoked potential (VEP) were larger than the RF sizes from the band-limited power (BLP). (A) Spatial distribution of RF sizes estimated from the VEP (top) and the population analysis for the differences within each frequency band (middle and bottom). In the bottom panel, red asterisk indicates that the median of the overall differences was significantly different from zero. (B) Distribution of the electrodes with significant difference between the VEP and the BLP. Top and bottom panels represent the results from monkeys Qand B, respectively. Blue and red circles represent the electrodes showing larger or smaller RFs in the VEP than the paired BLP, respectively. The sizes of the circles represent the mean of the stability indices estimated for the VEP and the paired BLP.

As we noted in Section 1, we focused on the event-related change of band-limited powers using transient visual flash stimuli, which induce neural activity related to onsets and offsets of the stimuli in close temporal proximity. While our latency analysis does isolate the effects of stimulus onset given that it reflects the earliest response component, other spatial aspects of the stimulus selectivity are likely to reflect the onset and offset responses. While we are not aware of any receptive field mapping study that showed separate spatial selectivity for the onset and offset, it is plausible that different frequency bands in BLPs may show such dependencies as well as some non-linear interactions between the onset and offset responses. Future studies with stimuli of a longer duration would address such possibility, characterizing the spatial selectivity of steady-state oscillations, as well as the onset and offset response, and their interactions.

4.3. The RF sizes in each region as a function of frequency band

We found that the RF sizes in each region were related to the frequency bands (Fig. 6C). In V1, the RF sizes decreased as the frequency increased, while in V4+, there was no clear relationship. In TEO/TE, the RF sizes showed a non-linear inverted U-shape similar to the pattern in the preferred eccentricities across bands (Fig. 6B). Taken together, the results for the RF sizes and preferred eccentricities imply that a given electrode in TEO/TE typically prefers the small foveal region in the alpha and high-gamma bands, and simultaneously, more peripheral regions in the beta and low-gamma bands.

The dependency in V1 is consistent with a commonly held assumption that cortical tissue serves as a low-pass capacitive filter (Bédard et al., 2004, 2006). In this model, electrical activity in the lower

Fig. 11. Response latencies (RLs) for the visual evoked potential (VEP) were much faster than the RLs for the band-limited power (BLP). (A) Spatial distribution oftheRLsin the most preferred position (top) and of the population analysis (middle and bottom). (B) Spatial distribution of the differences in RL between the VEP and the BLP. For each electrode, we calculated the mean difference in RLs across the positions. Blue and red circles represent electrodes with slower or faster latencies for the VEP than for the BLP, respectively. The format is same as that used in Fig. 10.

frequency spreads across distant areas presumably due to the lower impedance of cortical tissue, whereas electrical activity in the higher frequency remains local due to a higher cortical impedance. However, recent studies demonstrated that the impedance of the cortex is independent of frequency (Kajikawa and Schroeder, 2011; Logothetis et al., 2007). Indeed, our results in V4+ and TEO/TE are not compatible with the capacitive filter model. Rather, the very small RF sizes observed in the alpha band for TEO/TE suggest that alpha responses can reflect local computations, as local as the high-gamma band. Our findings imply there may be a great deal that can be learned from the low-frequency activity about the nature of "local" computations.

4.4. Longest RLs in the alpha band

One of the most surprising results relates to the latencies of the BLP in the alpha band. Across all areas, the RLs in the alpha band were much slower than the RLs in the other bands (Figs. 8 and 9), and this finding was not due to artifacts of our analyses (Supplementary Fig. 2). Note that we estimated these latencies based on the same set of trials, which were presented in the same location and visual space. Even when we defined the RLs based on the most preferred stimulus positions, the alpha band latencies were still the slowest in each area (Fig. 9A, gray circles and Table 1).

Typically, low-frequency activity is thought to reflect fluctuations in the membrane potential caused by synaptic inputs (Bartos et al., 2007; Buzsaki et al., 1983; Buzsaki et al., 2012; Logothetis, 2008; Ray et al., 2008, 2011). In our retinotopic paradigm, which emphasized event-related responses, we might expect to see the synaptic inputs in the low-frequency bands, including the alpha band, before the spiking outputs and the HGP (Kayser et al., 2007; Ray et al., 2008,2011). In fact, several studies using event-related potentials reported very early RLs, which might reflect synaptic inputs rather than spiking activity (Chen et al., 2007; Kirchner et al., 2009). Indeed, our VEP analyses replicated these findings (see Fig. 11).

It is plausible that the initial cortical inputs that are directly driven by visual stimuli may consist of a signal above 12 Hz that spares the narrow alpha band (3.9-11.7 Hz). The reason why it takes an additional 25-50 ms for the alpha power to increase compared to other bands (Fig. 9) is unclear. One possibility is that it may reflect the time needed for local computations to settle in a steady state. Another possibility is that the alpha band in the occipital lobe primarily reflects the ongoing spontaneous states of the cortex (Klimesch et al., 2007; Mathewson et al., 2011). If so, it may take some time to entrain the ongoing alpha activity with external stimuli, which would result in slower response latencies compared to those in the other bands. Yet another possibility is that the alpha responses reflect synchronized synaptic activity triggered by feedback signals from the other regions. Given the latencies, it is likely the responses are triggered by feedback from other areas rather than by an initial sweep of the direct feedforward inputs. Indeed, recent studies showed that feedback signals could drive the local circuits that generate the oscillations in the alpha-band range (Bollimunta et al., 2008, 2011). Such a late oscillatory signal might be reflected by alpha-band responses in the ECoG signal. It is an open question whether or not the evoked change of the BLP in the alpha band, which we studied here, shares common neural mechanisms with the sustained oscillatory activity in the alpha band, which was examined previously by Bollimunta et al. (2008, 2011).

In contrast to the alpha band, the beta and low-gamma bands responded almost simultaneously (in V1 and V4+) or earlier (in TEO/ TE) than the high-gamma band. Thus, the beta and low-gamma BLP might reflect the initial sweep into each region, rather than the feedback inputs or oscillatory modulation.

As for the neural sources of the evoked BLP in the different frequency bands, ECoG recording alone does not let us draw strong conclusions due to its limited spatial resolution. In our study, each electrode sampled the electrical signal generated by the neuronal populations below

and nearby the electrode surface whose diameter is 1 mm. Further, ECoG recordings heavily reflect the neural activity in the superficial layers of the cortex (Watanabe et al., 2012). Thus, to understand how the BLP in different frequencies across the depth of cortical columns contribute to the surface ECoG recordings requires additional studies that combine ECoG with other recording techniques that achieve finer depth and lateral resolution, such as laminar electrodes (Watanabe et al., 2012). Such studies are more feasible in animal models than in human epilepsy patients. Our technique of using ECoG in monkeys will help determine the physiological source of ECoG signals in different frequency bands. Our current study strongly suggests that the neural sources for different frequency bands are distinct.

4.5. The unique patterns of the spatiotemporal profile in TEO/TE and its potential neural mechanisms

The spatiotemporal characteristics of TEO/TE revealed a non-linear dependency on frequency (Figs. 6 and 9B). While the beta and low-gamma bands responded early (~130 ms) and showed bigger RFs (~10°) with a trend to prefer the periphery (~3° from the fovea), the alpha and high-gamma bands responded later (~175 and ~145 ms, respectively) and showed smaller RFs that preferred the foveal representation.

What neural mechanisms can explain the unique profiles ofTEO/TE? Although several explanations are possible, the general response pattern in TEO/TE is consistent with a previous hypothesis that visual processing in the temporal lobe begins with coarse-grained information, with the details being processed later (Nakamura et al., 1993; Ungerleider et al., 2008). This hypothesis was originally inspired by the anatomical finding that neurons in the ventral stream have diverse connections, including many shortcuts from the low- to high-level areas. For example, a well-established shortcut exists between V2 and TEO. Such shortcut pathways might provide "a means for coarse-grained information to rapidly arrive in the temporal lobe" (Ungerleider et al., 2008). At this point, we are not certain which pathways are crucial for the initial beta and low-gamma band visual responses in TEO/TE since neuroanatomical studies have demonstrated several shortcut connections (Builler and Kennedy, 1983; Fries, 1981; Nakamura et al., 1993; Rodman et al., 2001; Ungerleider et al., 2008; Webster et al., 1993; Yukie and Iwai, 1981, 1985; Zeki, 1978). Nevertheless, the basic idea of "coarse-to-fine" processing in the ventral visual stream is attractive for explaining our results.

In TEO/TE, the beta and low-gamma bands responded significantly faster than the high-gamma band (note that the results differ when we analyzed only the most preferred positions, as shown in Table 1). The 17-ms difference in latency implies that the beta and low-gamma activity in TEO/TE may not be the direct cause of spiking activity or the associated HGP in TEO/TE. In fact, given the differences in RF size and the most preferred eccentricities, there seems to be some degree of independence between the beta and low-gamma activity and the HGP (as well as spikes). In fact, the idea that "coarse information primes the processing of the fine" seems consistent with the finding that the beta and gamma bands tended to prefer peripheral visual fields, which was followed by the foveal HGP. Obviously, further work is necessary to precisely identify the neural mechanisms underlying the distinctive frequency bands, especially in the high-level visual areas including TEO/TE.

5. Conclusions

In conclusion, our results indicate that low-frequency BLP is more informative for learning about the local cortical mechanisms than previously thought. The low-frequency BLP does not always suffer from an unspecific spread of electrical activity across neighboring cortical areas. We also revealed unexpected frequency-dependent spatiotemporal profiles, especially upon close examination within each visual

area. In particular, we found that the high-level visual area TEO/TE is unique in its spatiotemporal profiles, implicating the involvement of the neural mechanisms that allow "coarse-to-fine" processing for object recognition in the ventral pathway. These interesting results would not have been found if we had only focused on the HGP of ECoG signals. Likewise, future studies that integrate the information gained from all frequency bands would provide a fuller picture of how the brain works, and in particular, the neural basis of ECoG. Such a basic understanding is necessary if ECoG is to be used in the future as a promising recording technique for basic neuroscience and other applications such as developing brain machine interfaces.

Acknowledgments

We are grateful to Naomi Hasegawa and Tomonori Notoya for providing animal care and technical support. This work was supported byJST PRESTO (No. 3630), ARC Future Fellowship (No. FT120100619) to NT, ARC Discovery Project to KT and NT (No. DP130100194), and MEXT KAKENHI to KT (No. 24700269). National BioResource Project "Japanese monkey" provided monkey B. The authors declare no competing financial interests.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.neuroimage.2015.09.007.

References

Aoki, F., Fetz, E.E., Shupe, L., Lettich, E., Ojemann, Ga, 1999. Increased gamma-range activity in human sensorimotor cortex during performance of visuomotor tasks. Clin. Neurophysiol. 110, 524-537. Bartos, M., Vida, I., Jonas, P., 2007. Synaptic mechanisms of synchronized gamma oscillations in inhibitory interneuron networks. Nat. Rev. Neurosci. 8,45-56. Bedard, C., Kröger, H., Destexhe, A., 2004. Modeling extracellular field potentials and the

frequency-filtering properties of extracellular space. Biophys. J. 86,1829-1842. Bedard, C., Kröger, H., Destexhe, A., 2006. Model of low-pass filtering of local field potentials in brain tissue. Phys. Rev. E. Stat. Nonlin. Soft Matter Phys. 73, 051911. Benjamini, Y., Hochberg, Y., 1995. Controlling the false discovery rate: a practical and

powerful approach to multiple testing. J. R Statist. Soc. B 57,289-300. Bollimunta, A., Chen, Y., Schroeder, C.E., Ding, M., 2008. Neuronal mechanisms of cortical

alpha oscillations in awake-behaving macaques. J. Neurosci. 28, 9976-9988. Bollimunta, A., Mo, J., Schroeder, C.E., Ding, M., 2011. Neuronal mechanisms and atten-

tional modulation of corticothalamic alpha oscillations. J. Neurosci. 31,4935-4943. Bosman, C.A., Schoffelen, J.M., Brunet, N., Oostenveld, R., Bastos, A.M., Womelsdorf, T., Rubehn, B., Stieglitz, T., De Weerd, P., Fries, P., 2012. Attentional stimulus selection through selective synchronization between monkey visual areas. Neuron 75, 875-888.

Boussaoud, D., Desimone, R, Ungerleider, L.G., 1991. Visual topography of area TEO in the

macaque. J. Comp. Neurol. 306, 554-575. Brainard, D.H., 1997. The Psychophysics Toolbox. Spat. Vis. 10,433-436. Brewer, A.A., Press, W.A., Logothetis, N.K., Wandell, B.A., 2002. Visual areas in macaque cortex measured using functional magnetic resonance imaging. J. Neurosci. 22, 10416-10426.

Builler, J., Kennedy, H., 1983. Projection of the lateral geniculate nucleus onto cortical area

V2 in the macaque monkey. Exp. Brain Res. 53,168-172. Buzsäki, G., Leung, L.S., Vanderwolf, C.H., 1983. Cellular bases of hippocampal EEG in the

behaving rat. Brain Res. 287,139-171. Buzsäki, G., Anastassiou, C.A., Koch, C., 2012. The origin of extracellular fields and

currents—EEG, ECoG, LFP and spikes. Nat. Rev. Neurosci. 13,407-420. Canolty, R.T., Soltani, M., Dalal, S.S., Edwards, E., Dronkers, N.F., Nagarajan, S.S., Kirsch, H.E., Barbaro, N.M., Knight, R.T., 2007. Spatiotemporal dynamics of word processing in the human brain. Front. Neurosci. 1,185-196. Chao, Z.C., Nagasaka, Y., Fujii, N., 2010. Long-term asynchronous decoding of arm motion

using electrocortigographic signals in monkeys. Front. Neuroeng. 3, 3. Chen, C.M., Lakatos, P., Shah, A., Mehta, A.D., Givre, S.J., Javitt, D.C., Schroeder, C.E., 2007. Functional anatomy and interaction of fast and slow visual pathways in macaque monkeys. Cereb. Cortex 17,1561-1569. Crone, N.E., Miglioretti, D.L., Gordon, B., Lesser, R.P., 1998a. Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. II. Event-related synchronization in the gamma band. Brain 121, 2301-2315. Crone, N.E., Miglioretti, D.L., Gordon, B., Sieracki, J.M., Wilson, M.T., Uematsu, S., Lesser, R.P., 1998b. Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. I. Alpha and beta event-related desynchronization. Brain 121, 2271-2299.

Crone, N.E., Boatman, D., Gordon, B., Hao, L., 2001. Induced electrocorticographic gamma activity during auditory perception. Clin. Neurophysiol. 112,565-582.

Crone, N.E., Sinai, A., Korzeniewska, A., 2006. High-frequency gamma oscillations and human brain mapping with electrocorticography. Prog. Brain Res. 159,275-295.

Edwards, E., Soltani, M., Deouell, L.Y., Berger, M.S., Knight, R.T., 2005. High gamma activity in response to deviant auditory stimuli recorded directly from human cortex. J. Neurophysiol. 94, 4269-4280.

Edwards, E., Soltani, M., Kim, W., Dalal, S.S., Nagarajan, S.S., Berger, M.S., Knight, R.T., 2009. Comparison of time-frequency response and the event-related potential to auditory speech stimuli in human cortex. J. Neurophysiol. 102,377-386.

Edwards, E., Nagarajan, S.S., Dalal, S.S., Canolty, R.T., Kirsch, H.E., Barbaro, N.M., Knight, R.T., 2010. Spatiotemporal imaging of cortical activation during verb generation and picture naming. Neuroimage 50, 291-301.

Engel, A.K., Fries, P., 2010. Beta-band oscillations—signaling the status quo? Curr. Opin. Neurobiol. 20,156-165.

Fize, D., Vanduffel, W., Nelissen, K., Denys, K., Chef d'Hotel, C., Faugeras, O., Orban, G.A., 2003. The retinotopic organization of primate dorsal V4 and surrounding areas. J. Neurosci. 23, 7395-7406.

Foster, B.L., Dastjerdi, M., Parvizi, J., 2012. Neural populations in human posteromedial cortex display opposing responses during memory and numerical processing. Proc. Natl. Acad. Sci. U. S. A. 109,15514-15519.

Fries, W., 1981. The projection from the lateral geniculate nucleus to the prestriate cortex of the macaque monkey. Proc. R. Soc. Lond. B Biol. Sci. 213, 73-80.

Fukuda, M., Juhâsz, C., Hoechstetter, K., Sood, S., Asano, E., 2010. Somatosensory-related gamma-, beta- and alpha-augmentation precedes alpha- and beta-attenuation in humans. Clin. Neurophysiol. 121,366-375.

Gaona, C.M., Sharma, M., Freudenburg, Z.V., Breshears, J.D., Bundy, D.T., Roland, J., Barbour, D.L., Shall;, G., Leuthardt, E.C., 2011. Nonuniform high-gamma (60-500 Hz) power changes dissociate cognitive task and anatomy in human cortex. J. Neurosci. 31, 2091-2100.

Gattass, R., Gross, C.G., Sandell, J.H., 1981. Visual topography of V2 in the macaque. J. Comp. Neurol. 201, 519-539.

Gattass, R., Sousa, A.P., Gross, C.G., 1988. Visuotopic organization and extent of V3 and V4 of the macaque. J. Neurosci. 8,1831-1845.

Graimann, B., Huggins, J.E., Levine, S.P., Pflurtscheller, G., 2004. Toward a direct brain interface based on human subdural recordings and wavelet-packed analysis. IEEE Trans. Biomed. Eng. 51, 954-962.

Hansen, K.A., David, S.V., Gallant, J.L., 2004. Parametric reverse correlation reveals spatial linearity of retinotopic human V1 BOLD response. Neurolmage 23,233-241.

Harvey, B.M., Vansteensel, M.J., Ferrier, C.H., Petridou, N., Zuiderbaan, W., Aarnoutse, E.J., Bleichner, M.G., Dijkerman, H.C., van Zandvoort, M.J., Leijten, F.S., Ramsey, N.F., Dumoulin, S.O., 2013. Frequency specific spatial interactions in human electrocorticography: V1 alpha oscillations reflect surround suppression. Neurolmage 65, 424-432.

Henriksson, L., Karvonen, J., Salminen-Vaparanta, N., Ralio, H., Vanni, S., 2012. Retinotopic maps, spatial tuning, and locations of human visual areas in surface coordinates characterized with multifocal and blocked fMRI designs. PLoS One 7, e36859.

Hermes, D., Miller, K.J., Vansteensel, M.J., Aarnoutse, E.J., Leijten, F.S., Ramsey, N.F., 2012. Neurophysiologic correlates of fMRI in human motor cortex. Hum. Brain Mapp. 33, 1689-1699.

Jerbi, K., Ossandôn, T., Hamamé, C.M., Senova, S., Dalal, S.S., Jung, J., Minotti, L., Bertrand, O., Berthoz, A., Kahane, P., Lachaux, J.P., 2009. Task-related gamma-band dynamics from an intracerebral perspective: review and implications for surface EeG and MEG. Hum. Brain Mapp. 30,1758-1771.

Kajikawa, Y., Schroeder, C.E., 2011. How local is the local field potential? Neuron 72, 847-858.

Kawasaki, H., Tsuchiya, N., Kovach, C.K., Nourski, K.V., Oya, H., Howard, M.A., Adolphs, R., 2012. Processing of facial emotion in the human fusiform gyrus. J. Cogn. Neurosci. 24, 1358-1370.

Kayser, C., Petkov, C.I., Logothetis, N.K., 2007. Tuning to sound frequency in auditory field potentials. J. Neurophyisol. 98,1806-1809.

Kirchner, H., Barbeau, E.J., Thorpe, S.J., Régis, J., Liégeois-Chauvel, C., 2009. Ultra-rapid sensory responses in the human frontal eye field region. J. Neurosci. 29, 7599-7606.

Kleiner, M., Brainard, D., Pelli, D., 2007. What's new in Psychtoolbox-3? Perception 36ECVP Abstract Supplement

Klimesch, W., Sauseng, P., Hanslymayr, S., 2007. EEG alpha oscillations: the inhibition-timing hypothetis. Brain Res. Rev. 53, 63-88.

Logothetis, N.K., 2008. What we can do and what we cannot do with fMRI. Nature 453, 869-878.

Logothetis, N.K., Kayser, C., Oeltermann, A., 2007. In vivo measurement of cortical impedance spectrum in monkeys: implications for signal propagation. Neuron 55, 809-823.

Mathewson, K.E., Lleras, A., Beck D.M., Fabiani, M., Ro, T., Gratton, G., 2011. Pulsed out of awareness: EEG alpha oscillations represent a pulsed-inhibition of ongoing cortical processing. Front. Psychol. 2, 99.

Matsuo, T., Kawasaki, K., Osada, T., Sawahata, H., Suzuki, T., Shibata, M., Miyakawa, N., Nakahara, K., Iijima, A., Sato, N., Kawai, K., Saito, N., Hasegawa, I., 2011. Intrasulcal electrocorticography in macaque monkeys with minimally invasive neurosurgical protocols. Front. Syst. Neurosci. 5, 34.

Maunsell, J.H., Gibson, H., 1992. Visual response latencies in striate cortex of the macaque monkey. J. Neurophyisol. 68,1332-1344.

McGill, R., Tukey, J.W., Larsen, W.A., 1978. Variations of box plots. Am. Stat. 32,12-16.

Miller, K.J., Leuthardt, E.C., Schalk, G., Rao, R.P.N., Anderson, N.R., Moran, D.W., Miller, J.W., Ojemann, J.G., 2007. Spectral changes in cortical surface potentials during motor movement. J. Neurosci. 27, 2424-2432.

Miller, K.J., Zanos, S., Fetz, E.E., den Nijs, M., Ojemann, J.G., 2009. Decoupling the cortical power spectrum reveals real-time representation of individual finger movements in humans. J. Neurosci. 29, 3132-3137.

Miller, K.J., Schalk, G., Fetz, E.E., den Nijs, M., Ojemann, J.G., Rao, R.P., 2010. Cortical activity during motor execution, motor imagery, and imagery-based online feedback Proc. Natl. Acad. Sci. U. S. A. 107,4430-4435.

Mitra, P., Bokil, H., 2008. Observed Brain Dynamics. Oxford University Press, New York

Nagasaka, Y., Shimoda, K., Fujii, N., 2011. Multidimensional recording (MDR) and data sharing: an ecological open research and educational platform for neuroscience. PLoS One 6, e22561.

Nakamura, H., Gattass, R., Desimone, R., Ungerleider, L.G., 1993. The modular organization of projections from areas V1 and V2 to areas V4 and TEO in macaque. J. Comp. Neurol. 13, 3681-3691.

Ohara, S., Ikeda, A., Kunieda, T., Yazawa, S., Baba, K., Nagamine, T., Taki, W., Hashimoto, N., Mihara, T., Shibasaki, H., 2000. Movement-related change of electrocorticographic activity in human supplementary motor area proper. Brain 123,1203-1215.

Palmini, A., Gambardella, A., Andermann, F., Dubeau, F., da Costa, J.C., Olivier, A., Tampieri, D., Gloor, P., Quesney, F., Andermann, E., Paglioli, E., Paglioli-Neto, E., Coutinho, L., Leblanc, R., Lim, H.I., 1995. Intrinsic epileptogenicity of human dysplastic cortex as suggested by corticography and surgical results. Ann. Neurol. 37,476-487.

Pei, X., Leuthardt, E.C., Gaona, C.M., Brunner, P., Wolpaw, J.R., Schalk G., 2011. Spatiotemporal dynamics of electrocorticographic high gamma activity during overt and covert word repetition. NeuroImage 54, 2960-2972.

Pelli, D.G., 1997. The VideoToolbox software for visual psychophysics: transforming numbers into movies. Spat. Vis. 10,437-442.

Rank Jr., J.B., 1963. Specific impedance of rabbit cerebral cortex. Exp. Neurol. 7,144-152.

Ray, S., Maunsell, J.H., 2011. Different origins of gamma rhythm and high-gamma activity in macaque visual cortex. PLoS Biol. 9, e1000610.

Ray, S., Crone, N.E., Niebur, E., Franaszczuk P.J., Hsiao, S.S., 2008. Neural correlates of high-gamma oscillations (60-200 Hz) in macaque local field potentials and their potential implications in electrocorticography. J. Neurosci. 28,11526-11536.

Rodman, H.R., Sorenson, K.M., Shim, A.J., Hexter, D.P., 2001. Calbindin immunoreactivity in the genicul-extrastriate system of the macaque: implications for heterogeneity in the koniocellular pathway and recovery from cortical damage. J. Comp. Neurol. 431, 168-181.

Rubehn, B., Bosman, C., Oostenveld, R., Fries, P., Stieglitz, T., 2009. A MEMS-based flexible multichannel ECoG-electrode array. J. Neural Eng. 6, 036003.

Schalk, G., Miller, K.J., Anderson, N.R., Wilson, J.A., Smyth, M.D., Ojemann, J.G., Moran, D.W., Wolpaw, J.R., Leuthardt, E.C., 2008. Two-dimensional movement control using electrocorticographic signals in humans. J. Neural Eng. 5, 75-84.

Schmolesky, M.T., Wang, Y., Hanes, D.P., Thompson, K.G., Leutgeb, S., Schall, J.D., Leventhal, A.G., 1998. Signal timing across the macaque visual system. J. Neurophysiol. 79, 3272-3278.

Shimoda, K., Nagasaka, Y., Chao, Z.C., Fujii, N., 2012. Decoding continuous three-dimensional hand trajectories from epidural electrocorticographic signals in Japanese macaques. J. Neural Eng. 9, 036015.

Tsuchiya, N., Kawasaki, H., Oya, H., Howard 3rd, M.A., Adolphs, R., 2008. Decoding face information in time, frequency and space from direct intracranial recordings of the human brain. PLoS One 3, e38692.

Ungerleider, L.G., Galkin, T.W., Desimone, R., Gattass, R., 2008. Cortical connections of area V4 in the macaque. Cereb. Cortex 18,477-499.

Van Essen, D.C., Zeki, S.M., 1978. The topographic organization of rhesus monkey prestriate cortex. J. Physiol. 277, 193-226.

Van Essen, D.C., Newsome, W.T., Maunsell, J.H., 1984. The visual field representation in striate cortex of the macaque monkey: asymmetries, anisotropies, and individual variability. Vision Res. 24, 429-448.

Viventi, J., Kim, D.H., Vigeland, L., Frechette, E.S., Blanco, J.A., Kim, Y.S., Avrin, A.E., Tiruvadi, V.R., Hwang, S.W., Vanleer, A.C., Wulsin, D.F., Davis, K., Gelber, C.E., Palmer, L., Van der Spiegel, J., Wu, J., Xiao, J., Huang, Y., Contreras, D., Rogers, J.A., Litt, B., 2011. Flexible, foldable, actively multiplexed high-density electrode array for mapping brain activity in vivo. Nat. Neurosci. 14,1599-1605.

Wandell, B.A., Winawer, J., 2011. Imaging retionotopic maps in the human brain. Vision Res. 51, 718-737.

Ward, L.M., 2003. Synchronous neural oscillations and cognitive processes. Trends Cogn. Sci. 7, 553-559.

Watanabe, H., Sato, M.A., Suzuki, T., Nambu, A., Nishimura, Y., Kawato, M., Isa, T., 2012. Reconstuction of movement-related intracortical activity from micro-electrocorticogram array signals in monkey primary motor cortex. J. Neural Eng. 9, 036006.

Webster, M.J., Bachevalier, H., Ungerleider, L.G., 1993. Subcortical connections of inferior temporal areas TE and TEO in macaque monkeys. J. Comp. Neurol. 335, 73-91.

Yoshor, D., Bosking, W.H., Ghose, G.M., Maunsell, J.H., 2007. Receptive fields in human visual cortex mapped with the surface electrodes. Cereb. Cortex 17, 2293-2302.

Yukie, M., Iwai, E., 1981. Direct projection from the dorsal lateral geniculate nucleus to the prestriate cortex in macaque monkeys. J. Comp. Neurol. 201, 81-97.

Yukie, M., Iwai, E., 1985. Laminar origin of direct projection from cortex area V1 to V4 in the rhesus monkey. Brain Res. 346, 383-386.

Zeki, S.M., 1978. Uniformity and diversity of structure and functions in rhesus monkey prestriate visual cortex. J. Phyisol. 277, 273-290.

Zumsteg, D., Wieser, H.G., 2000. Presurgical evaluation: current role of invasive EEG. Epilepsia 41 (Suppl. 3), S55-S60.