Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 Nov 21.
Published in final edited form as: Neuron. 2018 Oct 11;100(4):953–963.e3. doi: 10.1016/j.neuron.2018.09.019

Gamma synchronization between V1 and V4 improves behavioral performance

Gustavo Rohenkohl 1, Conrado Arturo Bosman 2,3, Pascal Fries 1,2,4,*
PMCID: PMC6250574  NIHMSID: NIHMS1508921  PMID: 30318415

SUMMARY

Behavior is often driven by visual stimuli, relying on feedforward communication from lower to higher visual areas. Effective communication depends on enhanced interareal coherence, yet it remains unclear, whether this coherence occurs at an optimal phase relation that actually improves stimulus transmission to behavioral report. We recorded local field potentials from V1 and V4 of macaques performing an attention task, during which they reported changes of the attended stimulus. V1-V4 gamma synchronization immediately preceding the stimulus change partly predicted subsequent reaction times (RT). RTs slowed systematically as trial-by-trial interareal gamma phase relations deviated from the phase relation at which V1 and V4 synchronized on average. V1-V4 gamma phase relations accounted for RT differences of 13-31 ms. Effects were specific to the attended stimulus and not explained by local power or phase. Thus, interareal gamma synchronization occurs at the optimal phase relation for transmission of sensory inputs to motor responses.

eTOC Blurb:

Rohenkohl et al. show that visually induced interareal gamma synchronization between primary and higher visual cortex occurs at the phase relation that optimally subserves stimulus transmission. This directly links interareal synchronization to behavior, strongly supporting the Communication-through-Coherence hypothesis.

INTRODUCTION

At the heart of many cognitive functions is the dynamic modulation of effective connectivity, i.e. the context-dependent modulation of the postsynaptic impact of a given neuronal group. The impact of a group of neurons can be enhanced when they engage in gamma-band synchronization, because this renders their inputs to postsynaptic target neurons coincident in time (Azouz and Gray, 2003; Salinas and Sejnowski, 2000). Indeed, groups of visual cortical neurons show enhanced local gamma-band synchronization when they process attended as compared to ignored stimuli (Bichot et al., 2005; Fries et al., 2001; Taylor et al., 2005), and this enhancement predicts reaction times (Womelsdorf et al., 2006). Thus, effective connectivity depends on synchronization among presynaptic neurons.

Importantly, effective connectivity also depends on the synchronization between pre- and postsynaptic neurons. When postsynaptic neurons are synchronized, this modulates their gain rhythmically, and inputs are most effective when they consistently arrive during high gain. Thus, good neuronal communication requires coherence between pre- and postsynaptic neurons, a concept referred to as Communication-trough-Coherence (CTC) (Akam and Kullmann, 2010; Börgers and Kopell, 2008; Buehlmann and Deco, 2010; Fries, 2015; Palmigiano et al., 2017). Indeed, gamma phase relations between neuronal groups affect their power-power correlation (Womelsdorf et al., 2007) and their transfer entropy (Besserve et al., 2015). Importantly, this mechanism might serve a functional role for selective attention. When two visual stimuli induce two local gamma rhythms in macaque V1, only the gamma rhythm induced by the attended stimulus establishes coherence to V4 (Bosman et al., 2012; Grothe et al., 2012).

Whether interareal coherence can in fact affect interareal communication according to the CTC mechanism, depends on whether the postsynaptic gamma rhythm modulates the gain of spike responses. One study investigated whether a spike in V1 is followed by a spike in V2, and found this spike transmission modulated by the gamma phase in V2 (Jia et al., 2013). Another recent study showed that visually induced gamma in V4 rhythmically modulates the gain of spike responses and also behavioral reaction times (Ni et al., 2016). Similar effects have been described for gamma-frequency rhythms that were optogenetically entrained in rodent somatosensory cortex (Cardin et al., 2009; Siegle et al., 2014). Thus, several studies have shown that postsynaptic gamma rhythmically modulates gain.

However, it remains to be shown that pre- and postsynaptic neurons engage in coherence at the phase relation that actually improves communication. The mentioned studies on gamma-rhythmic gain modulation do not show that presynaptic activity is coherent at such a phase relation that inputs are timed to moments of maximal postsynaptic gain. The mentioned studies on selective attention effects report enhanced coherence, i.e. an enhanced consistency of phase relations. Yet, while consistent phase relations are necessary for good communication, they are not sufficient. A consistent phase relation could as well consistently time inputs to postsynaptic phases of low gain. For CTC to mediate selective attention, phase relations need to be not merely consistent but they need to be consistent at the optimal phase relation (Akam and Kullmann, 2010; Fries, 2005). Thus, one of the most important requirements of CTC has yet to receive empirical support.

Therefore, we set out to test whether interareal gamma synchronization actually occurs at the optimal phase relation for communication and thereby improves behavioral performance. We investigated simultaneous V1-V4 recordings in macaques performing a selective visual attention task to test whether the transmission of a stimulus change to a behavioral response depends on the V1-V4 phase relation. As the stimulus change is timed randomly, it is independent of ongoing neuronal activity and provides an ideal probe for the efficiency of transmission. Furthermore, the analysis relates interareal synchronization directly to behavioral responses, thereby directly probing its behavioral relevance. We first describe that fast responses are preceded by enhanced V1-V4 gamma-band synchronization. We then show that interareal gamma phase relations just before the stimulus change predict the speed at which the change is transformed into a behavioral response. Most importantly, the analysis shows that the phase relation, at which V1 and V4 synchronize, is optimal for stimulus transmission: Synchronization entails that the distribution of phase relations is non-uniform and centered around a mean phase relation, i.e. the mean phase relation is the phase relation of synchronization; We show that when the phase relation on a given trial deviates from this phase relation of synchronization, the subsequent behavioral response is systematically slowed.

RESULTS

We intended to investigate the relation between interareal synchronization and behavior. To this end, we focused on neuronal activity preceding the behaviorally relevant stimulus event, and related it to the subsequent reaction time in response to the event. The relevant stimulus event was a shape change of a cued drifting grating. The cued and an un-cued grating were simultaneously presented in opposite hemifields, while monkeys kept fixation and attentionally monitored the cued grating (Figure 1). During stimulus presentation and task performance, we recorded neuronal activity simultaneously from areas V1 and V4. Both areas were covered with an electrocorticographic (ECoG) grid (see Methods for details). Trials with attention to the stimulus contralateral to the ECoG are referred to as “attend-IN”, and trials with attention ipsilateral to the ECoG as “attend-OUT”. Unless otherwise noted, the analysis focuses on the 200 ms immediately preceding the change of the attended stimulus and uses the corresponding reaction times (RTs).

Figure 1. Stimuli and behavioral task.

Figure 1.

Open in a new tab

Macaques were trained to touch a bar, which triggered the appearance of a central fixation point. Two stimuli were presented, one in the RF of the recording sites (illustrated as dashed circle, not visible to the monkey) and one in the opposite quadrant. Blue and yellow tints were randomly assigned to the two stimuli. The fixation point assumed the color of one of the stimuli, cueing that stimulus as the behaviorally relevant, i.e. attended, one. Either one of the stimuli could undergo a bend at an unpredictable time. If the change occurred in the attended stimulus, and the bar was released within a short time window thereafter (illustrated in red), a reward was given. The analysis focused on a 200 ms time window immediately before the stimulus change (illustrated in green). Full details of the stimuli and the task are described in the Methods section.

Visually induced gamma-band activities in V1 and V4 and their interareal synchronization.

Visual stimulation induced gamma-band activities in retinotopically corresponding parts of V1 and V4 (Figure 2A,B)(Lewis et al., 2016b). For further analysis, we selected per monkey the visually driven recording sites, which were defined as the top third of sites with the strongest visually induced gamma-band activity (see Methods for details). The spectra of stimulation-induced local field potential (LFP) power changes showed clear gamma-band peaks for both V1 (Figure 2C,D) and V4 (Figure 2E,F), with particularly strong power increases in V1. The gamma rhythms in V1 and V4 showed interareal synchronization, as evidenced by the spectra of pairwise phase consistency (PPC, Figure 2G,H; in this and the following analyses of interareal coherence, N=140 interareal site pairs and N=2550 trials).

Figure 2. ECoG coverage of areas V1 and V4, and spectra of visually induced power changes and of interareal synchronization.

Figure 2.

Open in a new tab

(A) Dots represent positions of ECoG recording sites on V1 and V4 in monkey K, projected onto a standard brain. Overlaid color map illustrates visually induced gamma-band activity in areas V1 and V4. Note that the color range is scaled separately per area, because visually induced gamma was substantially stronger in V1. (B) Same as A, but for monkey P. (C) Spectrum of visually induced power changes in V1, quantified as percent change relative to pre-stimulus baseline. (D) Same as C, but for monkey P. (E,F) Same as C,D, but for area V4. (G) Spectrum of V1-V4 synchronization in monkey K, as quantified by the pairwise phase consistency (PPC). (H) Same as G, but for monkey P. In C-F, for area V1 and area V4 separately, data were averaged over the top third of recording sites with the strongest visually induced gamma-band activity, because those sites were used for further analysis. Shaded areas around the curves show ±1 SEM across trials, averaged over monkeys.

For further analysis, we determined the individual gamma peak frequency for each monkey separately (see Methods for details). Spectra of local power changes and of interareal coherence agreed largely but not perfectly in peak frequency, as has been observed in previous studies of interareal synchronization (Bosman et al., 2012; Gregoriou et al., 2009). This corresponds to the predictions of the theory of weakly coupled oscillators, which also applies to the synchronization among cortical gamma rhythms (Lowet et al., 2017). As our focus was on investigating whether interareal synchronization is predictive of behavior, we determined the individual gamma peak frequency by fitting a Gaussian to the PPC spectrum. The gamma peak frequency was 74 Hz in monkey K and 63 Hz in monkey P. This is in agreement with previous studies showing inter-individual variability in gamma peak frequency (van Pelt et al., 2012; Vinck et al., 2010a). Note that the V1-V4 PPC spectra also showed peaks in the beta band (16 Hz in monkey K and 14 Hz in monkey P). We will investigate those beta peaks separately below.

Short reaction times are preceded by strong interareal gamma-band synchronization.

As a first approach to test for a putative relation between interareal gamma-band synchronization and behavior, we performed a median split of the trials according to RTs. Note that RTs did not differ between attend-IN and attend-OUT (p=0.18, Figure S1). For fast and slow trials separately, we calculated the V1-V4 PPC in a 200 ms window immediately preceding the stimulus change. To combine results from both monkeys, PPC spectra were aligned to the individual gamma peak frequencies. V1-V4 PPC was stronger during the attend-IN as compared to the attend-OUT condition (compare Figure 3A,B; p < 0.05, 2×2 ANOVA with factors attend-IN/attend-OUT and RT-fast/RT-slow). This is consistent with previous findings of increased interareal coherence with attention (Bosman et al., 2012; Gregoriou et al., 2009; Grothe et al., 2012; Richter et al., 2017).

Figure 3. Short reaction times are preceded by strong interareal gamma-band synchronization.

Figure 3.

Open in a new tab

(A) V1-V4 PPC during attend-IN as a function of frequency relative to the individual gamma peak frequency, calculated for a 200 ms window immediately preceding the onset of stimulus change. PPC spectra from trials with short (long) RTs are shown as solid (dashed) lines. (B) Same as A, but for attend-OUT. (C) V1-V4 PPC during attend-IN at the individual gamma peak frequency, as a function of time relative to the onset of stimulus change (vertical dashed line). (D) Same as C, but for attend-OUT. Shaded areas around the curves show ±1 SEM across site pairs, averaged over monkeys. Note that statistical inferences are based on a non-parametric randomization test including correction for the multiple comparisons. Black horizontal bars indicate statistical significance (p<0.05). See also Figure S2 and S3.

Importantly, V1-V4 gamma PPC was 77% stronger preceding short as compared to long RTs (Figure 3A, p<0.05, non-parametric randomization test with correction for multiple comparisons across frequencies). When RTs were split into quartiles, this revealed a systematic increase in V1-V4 gamma PPC, with a 125% increase between longest and shortest RTs (Figure S2A). These effects were absent for the attend-OUT condition (Figure 3B, Figure S2B), when monkeys reported changes in the ipsilateral stimulus, and thereby the observed RTs related to the stimulus not processed by the recorded neurons. Thus, the RT-related effect of gamma synchronization is specific to the V1-V4 area pair that is actually involved in the communication of the behaviorally relevant stimulus, and the effect is not due to fluctuations in overall arousal. The RT-related effect was furthermore specific in time to the period just before and around the stimulus change. The time-resolved PPC analysis during attend-IN revealed a start of the effect around 300 ms before the stimulus change and an end around 100 ms after the stimulus change (Figure 3C). The absence of an effect during attend-OUT was confirmed in the time-resolved analysis (Figure 3D).

As noted above, V1-V4 PPC spectra also showed a peak in the beta band. Therefore, we compared interareal beta phase locking between trials with fast and slow RTs (after alignment to the individual beta peak frequencies), yet this did not reveal any significant differences (Figure S3, non-parametric randomization test with correction for multiple comparisons across frequencies).

Interareal gamma-band phase relations predict reaction times.

To test whether interareal gamma-band synchronization is predictive of behavior, we determined the gamma-band phase relation immediately preceding the stimulus change, and investigated whether it systematically related to the subsequent behavioral reaction time. We calculated the V1-V4 gamma phase relation for the 200 ms window immediately preceding each stimulus change. Based on those phase relations, we sorted trials into 36 bins and averaged RTs per bin. Figure 4A shows the result of this analysis for an example site pair during attend-IN, and suggests a systematic dependence. By contrast, Figure 4B shows the analysis for the same pair during attend-OUT, and suggests a much weaker or no effect.

Figure 4. Interareal gamma-band phase relations predict reaction times.

Figure 4.

Open in a new tab

(A) Polar plot showing reaction time as a function of V1-V4 gamma phase relation during attend-IN, for one example pair of recording sites. First, trials were binned into 36 non-overlapping bins with the same number of trials per bin. Subsequently, each bin was symmetrically expanded to contain 25% of all available trials, thereby smoothing the data. (B) Same as A, but during attend-OUT. (C) Circular-linear correlation between gamma phase relation and RT, averaged over all V1-V4 site pairs and both monkeys. Data from attend-IN (attend-OUT) are shown in red (blue). The dashed lines indicate the significance thresholds (p<0.05) obtained with a non-parametric randomization test including correction for multiple comparisons.

To quantify the effect, we calculated the circular-linear correlation (Berens, 2009) between interareal phase relations and RTs, across trials and without binning, thereby reflecting the true trial-by-trial correlation coefficient (Richter et al., 2015, 2017). The red line in Figure 4C shows the correlation coefficient averaged over all interareal site pairs and both monkeys during attend-IN as a function of frequency, and reveals a significant correlation in the gamma-band range (p<0.05, non-parametric randomization test corrected for multiple comparisons across frequencies). When we repeated the same analysis for the attend-OUT condition, the correlation was absent.

Trial-by-trial deviation from mean interareal gamma-band phase relation predicts reaction time.

The correlation analysis captures any linear correlation between the phase-relation and RT, irrespective of the actual phase relations related to minimal and maximal RTs, respectively. Thus, the significant correlation reveals that gamma phase relations are predictive of RTs in the average across interareal site pairs. But the phase relations leading to minimal RTs could differ across site pairs, and even if they were consistent, the phase relation leading to minimal RTs could take some arbitrary and hard-to-interpret value. Therefore, as a next step, we aimed at testing the specific hypothesis that the phase-relation, at which interareal site pairs synchronized, is followed by the shortest RTs. Synchronization entails that site pairs spend relatively more time in a particular phase relation (Lowet et al., 2017). Therefore, the phase-relation of synchronization is the mean phase relation. Thus, we hypothesized that the mean phase relation is followed by the shortest RTs, and that deviations from this mean phase relation are followed by longer RTs. Note that the raw phase relation between a given V1-V4 site pair cannot be directly interpreted, because the absolute LFP phase depends on numerous accidental factors like the geometric relationship between source and electrode, and additionally the bipolar derivation used to remove the common recording reference incurs arbitrary phase rotations. Yet, irrespective of this, the mean phase relation reflects the phase relation of synchronization, and the phase relation in each trial can be expressed in terms of its deviation from this mean phase relation.

Figure 5A-D illustrates for the example site pair from Figure 4A,B how we tested the hypothesis. We calculated the mean gamma phase relation over trials (weighted by the power per trial) and named it the “good” phase relation (indicated by the yellow line in Figure 5A,C). RTs on a given trial should be predicted by the degree of deviation from this good phase relation. We rotated all phase relations by a fixed phase, such that the mean phase relation was at zero (Figure 5A was rotated into Figure 5C). We binned trials according to their phase relation and averaged RTs per bin (Figure 5B). After applying the same rotation (Figure 5B was rotated into Figure 5D), this revealed that phase relations close to the mean were indeed followed by particularly short RTs. We applied the same analysis steps to all interareal site pairs, and this confirmed the effect in the population of site pairs during attend-IN (Figure 5E, and red line in Figure 5F). During attend-OUT, this relation was much weaker or absent (blue line in Figure 5F).

Figure 5. Trial-by-trial deviation from mean interareal gamma-band phase relation predicts reaction time.

Figure 5.

Open in a new tab

Panels (A-D) illustrate, using the example pair of Figure 4A during attend-IN, that all phase relations of a given pair were rotated such that the mean phase relation was at zero. This rotation was applied to allow pooling of site pairs, and to investigate whether RTs depended systematically on the deviation from the mean phase relation. (A) Each dot represents the gamma phase relation between V1 and V4 in one trial. The polar histogram in red shows the corresponding distribution. The yellow bar represents the mean gamma phase relation, which corresponds to the phase relation at which gamma-band synchronization occurred. (B) Polar plot showing reaction time as a function of the V1-V4 gamma phase relation (copy of Figure 4A). (C,D) Same as A,B, after rotating all phase relations such that the mean phase relation is at zero. Note that the rotation is based on the mean phase relation shown in A, and it reveals that RTs systematically decrease with decreasing deviation from the mean phase relation. (E) Same as D, but averaged over all V1-V4 site pairs, after normalizing RTs per monkey. Color code reflects the cosine of the deviation from the mean phase relation, referred to as “goodness of phase relation” (GPR). (F) Same as E but in Cartesian coordinates, and showing both the attend-IN (red) and the attend-OUT (blue) condition. (G) Spectrum of correlation coefficients between the GPR and RT, separately for attend-IN (red) and attend-OUT (blue). The dashed lines indicate the significance thresholds (p<0.05) obtained with a non-parametric randomization test including correction for multiple comparisons. See also Figure S4 and S6.

We quantified the observed relation between the deviation from the mean phase relation and RT. We took for each trial the cosine of the phase-relation deviation as a metric of that trial’s “goodness of phase relation” (GPR). A GPR of one (minus one) indicates a good (bad) gamma phase relation (Figure 5E). The GPR is a linear (rather than circular) metric, which allowed us to calculate a linear correlation between single-trial GPR and RT values. We found a negative correlation between GPR and RTs, showing that good phase relations were followed by short RTs. This negative correlation occurred specifically in the gamma band during the attend-IN condition, but not the attend-OUT condition (Figure 5G, p<0.05, non-parametric randomization test corrected for multiple comparisons across frequencies).

The average correlation coefficient between GPR and RT was significant, yet relatively small, with a peak at r=−0.027. Note that this average is composed of correlation values calculated across single trials and for individual V1-V4 site pairs, i.e. between relatively noisy GPR and RT estimates. We found that across V1-V4 site pairs, the correlation between GPRs and RTs was itself correlated with the strength of V1-V4 PPC (r=−0.225, p<0.004; calculated at gamma peak frequency for attend-IN; negative value reflects the negative GPRxRT correlation). This suggests that the trial-by-trial gamma phase relations of the interareal site pairs showing strong coherence for a given stimulus have a strong influence on the communication of that stimulus and thereby explain a substantial fraction of RT variability; The gamma phase relations of pairs with weaker coherence contribute less to communication or potentially show less coherence and RT modulation due to lower signal-to-noise recordings.

To reduce the effect of noise, we first averaged GPR values over all site pairs per trial. This strengthened the average correlation between GPR and RT from −0.027 to −0.081, i.e. by a factor of three (Figure S4). To further reduce the effect of noise, we binned GPRs, averaged GPRs and corresponding RTs per bin, and then calculated the correlation between GPR and RT across bins (Figure S4). With 25 bins, the correlation strengthened to a value of −0.33, i.e. by a factor of «12 from the original value. Note that this latter approach of averaging per bin removes noise irrespective of whether this noise is accidental, like measurement noise, or whether it is genuine, reflecting a genuinely weak correlation (Richter et al., 2015, 2017).

The calculation of the correlation coefficient entails division by the variances of the correlated variables. Thereby, the correlation values are strongly affected by noise in either one of the two variables. This source of underestimation should be absent for the size of the RT modulation, i.e. the RT difference between the best and the worst gamma phase relation. To estimate this RT modulation, we fitted RT as a function of GPR (both estimated per trial) with a linear regression. The resulting slope revealed that best GPRs are followed by «13 ms shorter RTs than worst GPRs (p=0.002).

Conventional linear regression underestimates the slope, because it minimizes residuals only for the dependent variable, while assuming that the independent variable is observed or controlled without noise. In our case, the independent variable is the GPR, and its measurement is certainly noisy. This can be addressed by using a so-called model-2 standardized major axis (SMA) linear regression (Smith, 2009; Warton et al., 2006). SMA linear regression minimizes residuals for both the dependent and the independent variable. For the SMA regression, we binned the data per monkey (into 25 equally spaced non-overlapping bins, ensuring uniform distribution of the independent variable) and averaged data over monkeys per bin. We confirmed on those binned data that the conventional linear regression remained significant and had a similar slope as before binning, giving an RT-difference estimate of ≈13 ms (solid line in Figure 6A, p=0.033). The significance of the conventional linear regression justified the fitting of the SMA linear regression, which revealed an RT difference between best and worst GPR of ≈31 ms (dashed line in Figure 6A).

Figure 6. Reaction time difference between good and bad interareal gamma-band phase relation.

Figure 6.

Open in a new tab

(A) Scatter plot of RT differences from mean RT as a function of “goodness of phase relation” (GPR) in the gamma band for the attend-IN condition. Per monkey, the mean RT and the trial-wise RT differences from the mean were calculated. Trials were binned according to GPR into 25 equally spaced non-overlapping bins. Per bin, RT differences were first averaged over trials and then over monkeys. The solid line shows the conventional linear regression fit (p=0.033). The significant conventional linear regression justified the fitting of an SMA linear regression, which is shown as dashed line. The vertical lines to the right of the plot indicate the RT differences between best and worst GPRs, with a solid line for the conventional linear regression and a dashed line for the SMA linear regression. (B) Difference between RTs in trials with the top versus the bottom 5% GPR values, in the attend-IN (red) and attend-OUT (blue) condition. RTs were significantly faster following good versus bad gamma phase relations for the attend-IN (p<0.007), but not attend-OUT (p=0.503) condition. Error bars show ±1 SEM across trials, averaged over monkeys. Note that statistical inferences are based on non-parametric permutation tests.

Finally, in a model-free approach, we compared RTs between trials with the top versus the bottom 5% GPR values and found a difference of ≈24 ms (Figure 6B, p=0.007). These effects were absent in the attend-OUT condition: GPR was not correlated with subsequent RTs (p=0.878), which correspondingly did not allow the fitting of an SMR linear regression, and RTs did not differ between trials with top versus bottom GPRs (Figure 6B, p=0.503).

RT modulation by interareal phase-relation is not explained by effects of local power or phase.

Finally, we investigated to which degree the RT-predictive effect is specific to the V1-V4 gamma phase relation, rather than the gamma power in V1 or V4. Previous studies have shown that the gamma rhythm in V1 entrains the gamma rhythm in V4 in a feedforward manner (Bastos et al., 2015b; Bosman et al., 2012; Michalareas et al., 2016; van Kerkoerle et al., 2014). Thus, good gamma phase relations between V1 and V4 might be driven by strong gamma-band activity inside V1, and thereby the RT-predictive effect of the interareal phase relation might reduce to an effect of V1 (and/or V4) power.

We first repeated the above analysis steps for V1 and V4 gamma power. The median split analysis revealed that fast RTs were preceded by stronger V1 gamma power, both during attend-IN and attend-OUT (Figure S5A, B). Correspondingly, RTs showed a negative correlation with V1 power in the gamma band in both conditions (Figure S5C). In V4, the median split did not reveal any significant differences. Yet, there were trends for fast RTs compared to slow RTs to be preceded by stronger gamma during attend-IN and weaker gamma during attend-OUT (Figure S4D,E). The corresponding correlation analysis found no significant effect during attend-OUT and a significant negative relation during attend-IN (Figure S4F). The trends in the median-split analysis and the significant correlation during attend-IN are in agreement with a previous study in V4 (Womelsdorf et al., 2006).

We next investigated whether the RT-predictive effects of interareal phase locking held, when we controlled for power. We performed a linear regression with RT as dependent variable and the GPR as independent variable (Figure S6A). We then included the power in V1 and V4 as additional independent variables, and this left the results qualitatively unchanged (Figure S6B). We repeated this control analysis, but adding as independent variable not the frequency-wise power but the power value at its area-wise peak frequency, which again did not change results (Figure S6C).

The control analysis using multiple linear regression relies on the assumption that the relations between RT and the different neuronal metrics are linear. We therefore performed an additional analysis that avoids this assumption by using stratification for gamma power. First, we performed a median split of the trials according to GPR. We refer to these halves as “GPR conditions”. Then we sorted trials, separately for each GPR condition, according to gamma power, into nine bins, stratified conditions per bin for gamma power, and compared RTs between conditions. This revealed that RTs were significantly shorter for high than low GPR, even after eliminating potential influences of power. This held after both, stratification for gamma power in V1 and V4 (both p<0.001; non-parametric randomization test).

The stratification approach allowed us to apply the same control for the gamma phase in V1 or V4. A previous study had shown that the gamma phase in V4 around the time of a sudden change in stimulus color is predictive of reaction times (Ni et al., 2016). Note that we estimated the gamma phase for a Hann window end-aligned to the moment, when a transient and smooth stimulus deformation started, and the deformation lasted 0.15 s. We did not find any significant correlation between either the V1 or the V4 phase and the V1-V4 phase relation, for any one of the investigated frequencies. Nevertheless, to test whether eventual effects of the local phases in V1 or V4 explain the effect of the V1-V4 phase relation, we applied phase stratification. We used the same GPR conditions as for power stratification. Then, as described above for gamma power, we stratified for gamma phase and compared RTs between conditions. This again confirmed that RTs were significantly shorter for high than low GPR after both, stratifying for V1 gamma phase (p<0.001) and V4 gamma phase (p<0.001).

DISCUSSION

In summary, the results suggest that gamma synchronization between V1 and V4 improves behavioral performance. When trials were median split by their reaction times, faster RTs were preceded by stronger interareal synchronization in the gamma band. In fact, the trial-by-trial V1-V4 gamma phase relation just before the behaviorally relevant stimulus change correlated with the behavioral RTs. Most importantly, stimulus changes preceded by gamma phase relations close to the phase relation, at which synchronization occurred on average, were followed by the shortest RTs. RTs slowed systematically, when phase relations deviated from the mean. This suggests that the gamma phase relation at which V1 and V4 synchronize is optimal for stimulus transmission to motor output. These effects occurred only when the investigated gamma-band synchronization was induced by the attended stimulus, i.e. the stimulus that also triggered the behavioral response. This demonstrates that RTs depend specifically on the interareal synchronization involved in the transmission of the behaviorally relevant stimulus, rather than on fluctuations in overall arousal. Our results show that interareal communication depends on interareal coherence, directly supporting the central prediction of the CTC hypothesis.

We investigated the strength of the relation between V1-V4 gamma-band GPRs and behavioral RTs. This strength is difficult to quantify, because it will be underestimated if the GPR estimates and/or the RT measurements contain noise. We can safely assume that they do contain noise, but we can only partly separate accidental noise, like measurement noise, from noise that reflects a genuinely weak correlation. We were able to reduce noise across inter-areal site pairs, by first averaging over site pairs per trial and then calculating the GPR-RT correlation, which increased the correlation by a factor of three. Note that V1-V4 gamma phase relations are expected to explain only part of the variance in behavioral RTs, because the V1-V4 communication represents only one step in the sensory-motor transduction cascade. Importantly, we were able to provide an estimate of effect size that is less affected by noise than the correlation values. The slope of the regression between GPR and RT revealed that optimal, compared to worst, V1-V4 gamma phase relations are followed by RTs that are ≈13 ms faster. When we used a more sensitive regression method that reduces the influence of noise in the independent variable, this RT difference increased to ≈31 ms. When we simply compared RTs between trials with the best and worst 5% of GPRs, this revealed an RT difference of ≈24 ms. These effect sizes are in the same range as RT benefits of attention, as reported in human and non-human primate studies (Cutrell and Marrocco, 2002; Posner et al., 1980).

Granger causality (GC) analyses have shown that the gamma rhythm in V1 entrains the gamma in V4 much more than vice versa (Bosman et al., 2012; Richter et al., 2017; van Kerkoerle et al., 2014). Across many pairs of visual areas, GC analyses show that interareal gamma entrainment generally proceeds along anatomical feedforward projections (Bastos et al., 2015b; Michalareas et al., 2016; van Kerkoerle et al., 2014). Additionally, the available literature suggests that gamma-band synchronization among early visual areas V1, V2 and V4 is linked to the patterns of feedforward projections (Roberts et al., 2013; van Kerkoerle et al., 2014; Zandvakili and Kohn, 2015). Thus, the observed effect of V1-V4 gamma coherence on RT is most likely an effect of feedforward-directed influences from V1 to V4, i.e. a consequence of how accurately the gamma in V1 entrains the gamma in V4 at any given moment. Thereby, the V1-V4 gamma phase relation is partly dependent on the strength of the driving local gamma-band rhythm in V1. We investigated whether the relation between V1-V4 synchronization and RTs can be explained by local activity inside V1 or V4. Control analyses left the effect of interareal synchronization on behavior largely unchanged. As V1-V4 gamma synchronization is likely due to V1 gamma entraining V4 gamma through the respective monosynaptic feedforward projections, any explanation that invokes additional areas as common drivers is highly unlikely.

Several previous studies have shown that RTs can be predicted by local neuronal activity in individual visual areas. RTs in response to behaviorally relevant events can be predicted by neuronal firing rates evoked by these events in primary (Lee et al., 2010) and higher (Galashan et al., 2013; Womelsdorf et al., 2006) visual areas. Furthermore, RTs can be predicted by the strength or the absolute phase of local stimulus-induced gamma-band activity in visual areas of humans (Hoogenboom et al., 2010) and non-human primates (Ni et al., 2016; Womelsdorf et al., 2006). At the same time, numerous studies have shown that many cognitive functions, including perceptual organization, attention, working memory, and long-term memory encoding, do not only rely on local neuronal activity, but often specifically on interareal synchronization (Bosman et al., 2012; Buschman and Miller, 2007; Fell et al., 2001; Gregoriou et al., 2009; Grothe et al., 2012; Liebe et al., 2012; Salazar et al., 2012; van Kerkoerle et al., 2014). These findings have inspired the CTC hypothesis, stating that efficient interareal communication depends on interareal synchronization. Many recent studies have confirmed different predictions of the CTC hypothesis, such as the rhythmic gain modulation (Ni et al., 2016), and the selective interareal synchronization for attended information (Bosman et al., 2012; Grothe et al., 2012). The current study shows directly that an interareal phase relation that is close to the mean phase relation is optimal for information transmission as reflected in short RTs.

We found that behavioral response times are only predicted by interareal gamma phase relations, when gamma rhythms are induced by the attended, i.e. behaviorally relevant, stimulus. This supports the hypothesis that attentional selection is implemented by increased interareal communication (Fries, 2015). This notion has been supported by several studies using numerous different approaches. Firing rate recordings in higher visual areas combined with modeling have revealed that attentional effects on firing rates can be explained if attention enhances interareal gain (Reynolds et al., 1999; Ruff and Cohen, 2017). This was confirmed experimentally, when gain was directly quantified by the response of a higher area to the electrical microstimulation of a lower area. Selective attention enhances both the gain of LGN input to V1 (Briggs et al., 2013) and the gain of V1 input to MT (Ruff and Cohen, 2017). Finally, coherence and Granger causality between neuronal groups in different areas is enhanced, when they process attended stimuli (Bosman et al., 2012; Gregoriou et al., 2012; Gregoriou et al., 2009; Grothe et al., 2012; Richter et al., 2017; Saalmann et al., 2012; Zhou et al., 2016).

An interesting avenue for future research will be to investigate further interareal links on the way from V1 to motor cortex. The dependence of RTs on interareal synchronization that we demonstrate here for the V1-V4 link might hold along the entire way, or it might be transformed into an effect of firing rates or of synchronization in different frequency bands. Also, it will be important to expand the investigation to other metrics of behavioral performance. A recent study investigated neuronal activity in V1, V4 and frontal cortex, while macaques reported perception of weak visual stimuli (van Vugt et al., 2018). Under these conditions, conscious stimulus report can be analyzed as resulting from the interplay between response bias and perceptual sensitivity. Intriguingly, perceptual sensitivity depended on the success of interareal stimulus transmission, whereas response bias depended on the pre-stimulus brain state. In our experiment, the stimulus is present and inducing neuronal activity for a prolonged period of time, and the behaviorally reported event is an unpredictable change of this stimulus. We find that stimulus-induced V1-V4 gamma synchronization is partly predictive of RTs and interpret this as an effect on stimulus transmission. As van Vugt et al. found stimulus transmission to predict perceptual stimulus sensitivity, this suggests that gamma synchronization might also predict perceptual stimulus sensitivity. This suggestion can be tested in future experiments with near-threshold stimuli that are missed on a sufficient fraction of trials.

Future experiments should also investigate whether the effects described here generalize to gamma induced by natural stimuli, ideally under natural viewing conditions. A previous study failed to find clear gamma in human ECoG recordings for 44% of the employed naturalistic gray-scale photos (Hermes et al., 2015b). By contrast, another study found clear gamma in macaque ECoG for all employed naturalistic gray-scale and color photos, while animals freely viewed the images (Brunet et al., 2015). The discrepancy between those studies has been discussed (Brunet et al., 2014a; Hermes et al., 2015a). In any case, it will be interesting to investigate, whether the gamma induced in visual cortex during natural viewing is predictive of behavioral performance.

Another important task for future studies will be to provide further evidence for a causal relevance of the observed relationship. Optogenetics affords the opportunity to generate gamma band activity in visual areas in-vivo (Ni et al., 2016), and to modulate the phase and frequency of induced gamma (Akam et al., 2012). With these tools, it should be possible to generate interareal synchronization at pre-specified phase relations. The results presented here directly lead to the prediction that optogenetically controlled interareal phase relations should modulate the efficiency of interareal communication and ultimately affect behavior.

STAR ★ METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER
Experimental Models: Organisms/Strains
Macaque monkeys German Primate Center N/A
Software and Algorithms
Stimulus control NIMH CORTEX software dally.nimh.nih.gov/index.html
CARET (Van Essen, 2012) brainvis.wustl.edu/wiki/index.php/Caret:About
MATLAB MathWorks www.mathworks.com
FiledTrip Toolbox (Oostenveld et al., 2011) www.fieldtriptoolbox.org
CircStat Toolbox (Berens, 2009) github.com/circstat
Other
ECoG Grid (Rubehn et al., 2009) N/A
Digital Lynx system Neuralynx www.neuralynx.com
Headstage Amplifier Plexon www.plexon.com
Eye Tracker Thomas Recording www.thomasrecording.com
Open in a new tab

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources should be directed to and will be fulfilled by the Lead Contact, Pascal Fries (pascal.fries@esi-frankfurt.de).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

All experimental procedures were approved by the ethics committee of Radboud University Nijmegen (Nijmegen, The Netherlands). Data from two adult male macaque monkeys (macaca mulatta) were collected for this study. Parts of the data have been used in other publications (Bastos et al., 2015a; Bastos et al., 2015b; Bosman et al., 2012; Brunet et al., 2015; Brunet et al., 2014a; Brunet et al., 2014b; Hindriks et al., 2018; Lewis et al., 2016a; Lewis et al., 2016b; Pinotsis et al., 2014; Richter et al., 2015, 2017; Rubehn et al., 2009; Spyropoulos et al., 2018; Vinck et al., 2015).

METHOD DETAILS

Electrophysiological Recording and Signal Processing.

LFP recordings were made via a 252-channel electrocorticographic grid (ECoG) subdurally implanted over the left hemisphere (Bastos et al., 2015b; Rubehn et al., 2009). Signals were filtered with a passband of 0.159-8000 Hz and sampled at approximately 32 kHz (Neuralynx Digital Lynx system). Offline, signals were low-pass filtered at 250 Hz and downsampled to 1 kHz. The electrodes were distributed over eight 32-channel headstages, and referenced against a silver wire implanted onto the dura overlying the opposite hemisphere. The electrodes were re-referenced via a bipolar scheme to improve signal localization, cancel the common reference, and reject noise specific to the headstage. The bipolar derivation scheme subtracted the recordings from neighboring electrodes (spaced 2.5 mm) that shared a headstage, resulting in 218 bipolar derivations, referred to as “recording sites” or just “sites” (see (Bastos et al., 2015b) for a detailed description of the re-referencing scheme). Areas V1 and V4 used here were defined based on comparison of the electrode locations (coregistered to each monkey’s anatomical MRI and warped to the F99 template brain in CARET (Van Essen, 2012)) with multiple cortical atlases of the macaque (see (Bastos et al., 2015b) for a detailed description). Recording sites were coregistered to a common template (INIA19(Rohlfing et al., 2012)), as were the area definitions based on multiple cortical atlases. Based on these area definitions, 77 recording sites were localized to area V1 (monkey K: 29; monkey P: 48), 31 to area V4 (monkey K: 17; monkey P: 14). All analyses performed in this study were done on the sites that showed highest (top 1/3) visually induced gamma-band activity within areas V1 and V4 (Figure 2A,B), resulting in a subset of 26 recording sites from area V1 (monkey K: 10; monkey P: 16), and 11 from area V4 (monkey K: 6; monkey P: 5).

All signal processing was conducted in MATLAB (MathWorks, USA) using the FieldTrip toolbox (http://www.fieldtriptoolbox.org/) (Oostenveld et al., 2011). Line noise was removed by subtracting the 50, 100, and 150 Hz components estimated through a discrete Fourier transform. Trial epochs for each site were demeaned by subtracting the mean over all time points in the epoch. Epochs with any site having a variance of greater than 5 times the variance based on all data from that same site in the same session were rejected. In addition, epochs were manually inspected, and epochs with artifacts were rejected. Subsequently, per recording site, the signals from all remaining epochs of a given session were divided by the standard deviation of the signal across all those remaining epochs, and the resulting z-transformed signals were combined across sessions.

Stimuli and Task.

Stimuli were presented on a CRT monitor (120 Hz non-interlaced) in a dimly lit booth and controlled by CORTEX software (https://www.nimh.nih.gov/labs-at-nimh/research-areas/clinics-and-labs/ln/shn/software-proiects.shtml). The paradigm is illustrated in Figure 1. Upon touching a bar, a fixation point was presented, and the monkey’s gaze was required to remain inside the fixation window throughout the trial (monkey K: 0.85 deg radius, monkey P: 1 deg radius). Otherwise the trial would be terminated and a new trial initiated. Upon the acquisition of central fixation, and after an 0.8 s pre-stimulus interval had elapsed, two isoluminant and isoeccentric drifting sinusoidal gratings were presented, one in each visual hemifield (diameter: 3 deg, spatial frequency: ≈1 cycle/deg, drift velocity: ≈1 deg/s, resulting temporal frequency: ≈1 cycle/s, contrast: 100%). Blue and yellow tints were randomly assigned to each of the gratings on each trial. Following a random delay interval (monkey K: 1 - 1.5 s; monkey P : 0.8 - 1.3 s), the color of the central fixation point changed to match one of the drifting gratings, which indicated that the matching grating was the behaviorally relevant or “target” stimulus. Thus, fixation point color acted as the attentional cue. When attention was directed to the stimulus in the visual hemifield contralateral (ipsilateral) to the recorded left hemisphere, this is addressed as the attend-IN (attend-OUT) condition. In each trial, two times were randomly drawn according to a slowly rising hazard rate to fall into an interval of 0.75-5 s (monkey K), and 0.75-4 s (monkey P) after stimulus onset. These two times were randomly assigned to be the change times of the target and the distracter. Bar releases 0.15 - 0.5 s after target changes were rewarded. Overall, 94% and 84% of all target changes were correctly reported by monkey K and monkey P, respectively (excluding fixation breaks). On half of the trials, the target changed first, and a corresponding behavioral response terminated the trial. Only those trials are analyzed here. The stimulus changes were subtle changes in shape consisting of a transient bending of the bars of the respective grating (0.15 s duration of the full bending cycle). All analyses presented here were performed on an epoch of 200 ms immediately preceding stimulus change (unless specified in text). We excluded trials in which the attention cue was presented less than one second before the target change, to ensure that cue processing was fully completed and attention deployed by the time of the target change. A total of 2550 trials (monkey K: 1149; monkey P: 1401) were used in this study.

Spectral analysis.

Each 200 ms epoch was multiplied with a Hann taper, zero padded to 1 s, and Fourier transformed, resulting in an FFT spectrum with a frequency resolution of 1 Hz. In the following, we will denote the arrays of Fourier spectra for all Kepochs (k = 1,…, K) from the i-th site in V1 and the j-th site in V4 as Fik (f) and Fjk (f), respectively, with f being the frequency in Hz. Spectral power was derived as the squared magnitude of the complex Fourier spectra. The relative power change shown in Figure 2C-F was computed as percent change relative to a baseline period of 200 ms before stimulus onset. Cross-spectral densities (CSDs) were estimated for each pair ij of V1 and V4 recording sites, for each frequency f, and each epoch k as

Sijk(f)=Fik(f)Fjk¯(f)

where Fjk¯(f) is the conjugate Fourier spectrum for V4. The magnitude of the CSD, |Sijk (f)| reflects the product of spectral energies in V1 and V4, and the argument (angle) of the CSD reflects the phase relations between V1 and V4.

Phase locking was quantified by deriving from the CSD the pairwise phase consistency (PPC) metric, a phase locking metric that is not biased by the number of epochs (Vinck et al., 2010b).

Phase Relation Analysis.

We aimed at quantifying whether behavioral RTs were related on a trial-by-trial basis to the V1-V4 phase relation. We hypothesized that the mean V1-V4 phase relation is optimal for interareal communication and is therefore followed by the shortest RTs. Note that the raw phase relation between a given V1-V4 site pair cannot be directly interpreted, because the absolute LFP phase depends on numerous accidental factors like the geometric relationship between source and electrode, and additionally the bipolar derivation used to remove the common recording reference incurs arbitrary phase rotations. Yet, irrespective of this, the mean phase relation reflects the phase relation of synchronization, and the phase relation in each trial can be expressed in terms of its deviation from this mean phase relation. Therefore, for each interareal site pair, we calculated the mean phase relation and subtracted it from each epoch’s phase relation. The mean phase relation was determined by computing the mean resultant vector across epochs, defined as

ρij(f)=Σk=1kSijk(f)∣Σk=1kSijk(f)∣

For each epoch, we then determined the phase deviation from the mean phase relation by computing the “rotated CSD”, by multiplying with the conjugate of the mean resultant vector,

Sijkrot(f)=Sijk(f)ρij¯(f)

The argument (angle) of the rotated CSD, Arg Sijkrot(f), then corresponds to the phase difference between the CSD in the k-th trial and the mean phase relation. According to our hypothesis, any deviation from the average phase relation results in sub-optimal interareal communication. After rotating the average phase relation into zero, we could quantify the deviation of an epoch’s phase relation from the mean phase relation by taking the cosine of the angle of the rotated CSD. We refer to this metric as the goodness of phase relation ( GPR):

GPRijk(f)=cos(Arg(Sijkrot(f))

According to the hypothesis, a GPR value of 1 corresponds to a good phase relation, and a GPR value of −1 corresponds to a bad phase relation. Note that the GPR ignores the direction, in which a given epoch’s phase relation deviates from the average (see Figure 5E).

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical tests were based on the combined data from both animals, constituting a fixed-effect analysis that results in inferences limited to the investigated sample of two animals. To lend equal weight to each animal, data were first combined within each animal (across sites, site pairs, trials) and subsequently averaged over the two animals. All significance thresholds were computed using non-parametric permutation statistics (Maris and Oostenveld, 2007). For the median split analysis, the observed PPC spectra were derived by 1) calculating PPC spectra across all epochs of a given condition in a given animal, separately for all V1-V4 site pairs, 2) averaging PPC spectra across those site pairs, and 3) averaging across the two animals after aligning the peaks of the respective frequency bands. Surrogate distributions were then generated by randomly distributing epochs in two conditions, maintaining the sample sizes of the original conditions (short RTs versus long RTs) (Maris et al., 2007). Then, the same steps as for the observed PPC spectra were followed. For each of 1000 randomizations, the maximal absolute difference value across all frequencies was retained and placed into the randomization distribution. The observed differences were compared against the distribution of maximal absolute differences. This procedure corrects for multiple comparisons across frequencies. A similar approach was used in the phase relation analysis. A surrogate distribution was generated under the hypothesis that there is no correlation between RTs and the preceding phase relation between V1 and V4. So, in each randomization, the correlation coefficients were calculated on randomly shuffled RTs. The observed coefficients were then compared against the surrogate distribution of maximal absolute coefficients.

Supplementary Material

1

  • V1-V4 gamma coherence before stimulus change predicts speed of change detection

  • Deviations from the phase relation of gamma synchronization increase reaction times

  • V1-V4 gamma phase relation explains reaction time differences of 13 to 31 ms

  • Effects are specific to the attended stimulus and not due to local phase or power

ACKNOWLEDGEMENTS

We thank Craig Richter for preprocessing the data, Julien Vezoli for help with assigning electrodes and sites to brain areas, and Martin Vinck for helpful advice on data analysis. PF acknowledges grant support by DFG (SPP 1665, FOR 1847, FR2557/5-1-CORNET, FR2557/6-1-NeuroTMR), EU (HEALTH-F2-2008-200728-BrainSynch, FP7-604102-HBP, FP7-600730-Magnetrodes), a European Young Investigator Award, NIH (1U54MH091657-WU-Minn-Consortium-HCP), and LOEWE (NeFF).

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

DECLARATION OF INTERESTS

The authors declare no competing interests.

REFERENCES

  1. Akam T, and Kullmann DM (2010). Oscillations and filtering networks support flexible routing of information. Neuron 67, 308–320. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Akam T, Oren I, Mantoan L, Ferenczi E, and Kullmann DM (2012). Oscillatory dynamics in the hippocampus support dentate gyrus-CA3 coupling. Nature neuroscience 15, 763–768. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Azouz R, and Gray CM (2003). Adaptive coincidence detection and dynamic gain control in visual cortical neurons in vivo. Neuron 37, 513–523. [DOI] [PubMed] [Google Scholar]
  4. Bastos AM, Litvak V, Moran R, Bosman CA, Fries P, and Friston KJ (2015a). A DCM study of spectral asymmetries in feedforward and feedback connections between visual areas V1 and V4 in the monkey. NeuroImage 108, 460–475. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Bastos AM, Vezoli J, Bosman CA, Schoffelen JM, Oostenveld R, Dowdall JR, De Weerd P, Kennedy H, and Fries P (2015b). Visual areas exert feedforward and feedback influences through distinct frequency channels. Neuron 85, 390–401. [DOI] [PubMed] [Google Scholar]
  6. Berens P (2009). CircStat: A MATLAB Toolbox for Circular Statistics. J Stat Softw 31, 1–21. [Google Scholar]
  7. Besserve M, Lowe SC, Logothetis NK, Schölkopf B, and Panzeri S (2015). Shifts of Gamma Phase across Primary Visual Cortical Sites Reflect Dynamic Stimulus-Modulated Information Transfer. PLoS biology 13, e1002257. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bichot NP, Rossi AF, and Desimone R (2005). Parallel and serial neural mechanisms for visual search in macaque area V4. Science 308, 529–534. [DOI] [PubMed] [Google Scholar]
  9. Börgers C, and Kopell NJ (2008). Gamma oscillations and stimulus selection. Neural computation 20, 383–414. [DOI] [PubMed] [Google Scholar]
  10. Bosman CA, Schoffelen JM, Brunet N, Oostenveld R, Bastos AM, Womelsdorf T, Rubehn B, Stieglitz T, De Weerd P, and Fries P (2012). Attentional stimulus selection through selective synchronization between monkey visual areas. Neuron 75, 875–888. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Briggs F, Mangun GR, and Usrey WM (2013). Attention enhances synaptic efficacy and the signal-to-noise ratio in neural circuits. Nature 499, 476–480. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Brunet N, Bosman CA, Roberts M, Oostenveld R, Womelsdorf T, De Weerd P, and Fries P (2015). Visual cortical gamma-band activity during free viewing of natural images. Cereb Cortex 25, 918–926. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Brunet N, Vinck M, Bosman CA, Singer W, and Fries P (2014a). Gamma or no gamma, that is the question. Trends in cognitive sciences 18, 507–509. [DOI] [PubMed] [Google Scholar]
  14. Brunet NM, Bosman CA, Vinck M, Roberts M, Oostenveld R, Desimone R, De Weerd P, and Fries P (2014b). Stimulus repetition modulates gamma-band synchronization in primate visual cortex. Proceedings of the National Academy of Sciences of the United States of America 111, 3626–3631. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Buehlmann A, and Deco G (2010). Optimal information transfer in the cortex through synchronization. PLoS computational biology 6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Buschman TJ, and Miller EK (2007). Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. Science 315, 1860–1862. [DOI] [PubMed] [Google Scholar]
  17. Cardin JA, Carlén M, Meletis K, Knoblich U, Zhang F, Deisseroth K, Tsai LH, and Moore CI (2009). Driving fast-spiking cells induces gamma rhythm and controls sensory responses. Nature 459, 663–667. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Cutrell EB, and Marrocco RT (2002). Electrical microstimulation of primate posterior parietal cortex initiates orienting and alerting components of covert attention. Experimental brain research 144, 103–113. [DOI] [PubMed] [Google Scholar]
  19. Fell J, Klaver P, Lehnertz K, Grunwald T, Schaller C, Elger CE, and Fernández G (2001). Human memory formation is accompanied by rhinal-hippocampal coupling and decoupling. Nature neuroscience 4, 1259–1264. [DOI] [PubMed] [Google Scholar]
  20. Fries P (2005). A mechanism for cognitive dynamics: neuronal communication through neuronal coherence. Trends in cognitive sciences 9, 474–480. [DOI] [PubMed] [Google Scholar]
  21. Fries P (2015). Rhythms for Cognition: Communication through Coherence. Neuron 88, 220–235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Fries P, Reynolds JH, Rorie AE, and Desimone R (2001). Modulation of oscillatory neuronal synchronization by selective visual attention. Science 291, 1560–1563. [DOI] [PubMed] [Google Scholar]
  23. Galashan FO, Sassen HC, Kreiter AK, and Wegener D (2013). Monkey area MT latencies to speed changes depend on attention and correlate with behavioral reaction times. Neuron 78, 740–750. [DOI] [PubMed] [Google Scholar]
  24. Gregoriou GG, Gotts SJ, and Desimone R (2012). Cell-type-specific synchronization of neural activity in FEF with V4 during attention. Neuron 73, 581–594. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Gregoriou GG, Gotts SJ, Zhou H, and Desimone R (2009). High-frequency, long-range coupling between prefrontal and visual cortex during attention. Science 324, 1207–1210. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Grothe I, Neitzel SD, Mandon S, and Kreiter AK (2012). Switching neuronal inputs by differential modulations of gamma-band phase-coherence. The Journal of neuroscience : the official journal of the Society for Neuroscience 32, 16172–16180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Hermes D, Miller KJ, Wandell BA, and Winawer J (2015a). Gamma oscillations in visual cortex: the stimulus matters. Trends in cognitive sciences 19, 57–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Hermes D, Miller KJ, Wandell BA, and Winawer J (2015b). Stimulus Dependence of Gamma Oscillations in Human Visual Cortex. Cereb Cortex 25, 2951–2959. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Hindriks R, Micheli C, Bosman CA, Oostenveld R, Lewis C, Mantini D, Fries P, and Deco G (2018). Source-reconstruction of the sensorimotor network from resting-state macaque electrocorticography. NeuroImage 181, 347–358. [DOI] [PubMed] [Google Scholar]
  30. Hoogenboom N, Schoffelen JM, Oostenveld R, and Fries P (2010). Visually induced gamma-band activity predicts speed of change detection in humans. Neuroimage 51, 1162–1167. [DOI] [PubMed] [Google Scholar]
  31. Jia X, Tanabe S, and Kohn A (2013). Gamma and the coordination of spiking activity in early visual cortex. Neuron 77, 762–774. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Lee J, Kim HR, and Lee C (2010). Trial-to-trial variability of spike response of V1 and saccadic response time. Journal of neurophysiology 104, 2556–2572. [DOI] [PubMed] [Google Scholar]
  33. Lewis CM, Bosman CA, Brunet NM, Lima B, Roberts MJ, Womelsdorf T, de Weerd P, Neuenschwander S, Singer W, and Fries P (2016a). Two frequency bands contain the most stimulus-related information in visual cortex. bioRxiv. [Google Scholar]
  34. Lewis CM, Bosman CA, Womelsdorf T, and Fries P (2016b). Stimulus-induced visual cortical networks are recapitulated by spontaneous local and interareal synchronization. Proceedings of the National Academy of Sciences of the United States of America 113, E606–615. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Liebe S, Hoerzer GM, Logothetis NK, and Rainer G (2012). Theta coupling between V4 and prefrontal cortex predicts visual short-term memory performance. Nature neuroscience 15, 456–462, S451–452. [DOI] [PubMed] [Google Scholar]
  36. Lowet E, Roberts MJ, Peter A, Gips B, and De Weerd P (2017). A quantitative theory of gamma synchronization in macaque V1. eLife 6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Maris E, and Oostenveld R (2007). Nonparametric statistical testing of EEG- and MEG-data. Journal of neuroscience methods 164, 177–190. [DOI] [PubMed] [Google Scholar]
  38. Maris E, Schoffelen JM, and Fries P (2007). Nonparametric statistical testing of coherence differences. Journal of neuroscience methods 163, 161–175. [DOI] [PubMed] [Google Scholar]
  39. Michalareas G, Vezoli J, van Pelt S, Schoffelen JM, Kennedy H, and Fries P (2016). Alpha-Beta and Gamma Rhythms Subserve Feedback and Feedforward Influences among Human Visual Cortical Areas. Neuron 89, 384–397. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Ni J, Wunderle T, Lewis CM, Desimone R, Diester I, and Fries P (2016). Gamma-Rhythmic Gain Modulation. Neuron 92, 240–251. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Oostenveld R, Fries P, Maris E, and Schoffelen JM (2011). FieldTrip: Open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Computational intelligence and neuroscience 2011, 156869. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Palmigiano A, Geisel T, Wolf F, and Battaglia D (2017). Flexible information routing by transient synchrony. Nature neuroscience 20, 1014–1022. [DOI] [PubMed] [Google Scholar]
  43. Pinotsis DA, Brunet N, Bastos A, Bosman CA, Litvak V, Fries P, and Friston KJ (2014). Contrast gain control and horizontal interactions in V1: a DCM study. NeuroImage 92, 143–155. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Posner MI, Snyder CR, and Davidson BJ (1980). Attention and the detection of signals. Journal of experimental psychology 109, 160–174. [PubMed] [Google Scholar]
  45. Reynolds JH, Chelazzi L, and Desimone R (1999). Competitive mechanisms subserve attention in macaque areas V2 and V4. The Journal of neuroscience : the official journal of the Society for Neuroscience 19, 1736–1753. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Richter CG, Thompson WH, Bosman CA, and Fries P (2015). A jackknife approach to quantifying single-trial correlation between covariance-based metrics undefined on a single-trial basis. NeuroImage 114, 57–70. [DOI] [PubMed] [Google Scholar]
  47. Richter CG, Thompson WH, Bosman CA, and Fries P (2017). Top-Down Beta Enhances Bottom-Up Gamma. The Journal of neuroscience : the official journal of the Society for Neuroscience 37, 6698–6711. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Roberts MJ, Lowet E, Brunet NM, Ter Wal M, Tiesinga P, Fries P, and De Weerd P (2013). Robust gamma coherence between macaque V1 and V2 by dynamic frequency matching. Neuron 78, 523–536. [DOI] [PubMed] [Google Scholar]
  49. Rohlfing T, Kroenke CD, Sullivan EV, Dubach MF, Bowden DM, Grant KA, and Pfefferbaum A (2012). The INIA19 Template and NeuroMaps Atlas for Primate Brain Image Parcellation and Spatial Normalization. Frontiers in neuroinformatics 6, 27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Rubehn B, Bosman C, Oostenveld R, Fries P, and Stieglitz T (2009). A MEMS-based flexible multichannel ECoG-electrode array. Journal of neural engineering 6, 036003. [DOI] [PubMed] [Google Scholar]
  51. Ruff DA, and Cohen MR (2017). A normalization model suggests that attention changes the weighting of inputs between visual areas. Proceedings of the National Academy of Sciences of the United States of America 114, E4085–E4094. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Saalmann YB, Pinsk MA, Wang L, Li X, and Kastner S (2012). The pulvinar regulates information transmission between cortical areas based on attention demands. Science 337, 753–756. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Salazar RF, Dotson NM, Bressler SL, and Gray CM (2012). Content-specific fronto-parietal synchronization during visual working memory. Science 338, 1097–1100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Salinas E, and Sejnowski TJ (2000). Impact of correlated synaptic input on output firing rate and variability in simple neuronal models. The Journal of neuroscience : the official journal of the Society for Neuroscience 20, 6193–6209. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Siegle JH, Pritchett DL, and Moore CI (2014). Gamma-range synchronization of fast-spiking interneurons can enhance detection of tactile stimuli. Nature neuroscience. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Smith RJ (2009). Use and misuse of the reduced major axis for line-fitting. American journal of physical anthropology 140, 476–486. [DOI] [PubMed] [Google Scholar]
  57. Spyropoulos G, Bosman CA, and Fries P (2018). A theta rhythm in macaque visual cortex and its attentional modulation. Proceedings of the National Academy of Sciences of the United States of America. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Taylor K, Mandon S, Freiwald WA, and Kreiter AK (2005). Coherent oscillatory activity in monkey area v4 predicts successful allocation of attention. Cereb Cortex 15, 1424–1437. [DOI] [PubMed] [Google Scholar]
  59. Van Essen DC (2012). Cortical cartography and Caret software. NeuroImage 62, 757–764. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. van Kerkoerle T, Self MW, Dagnino B, Gariel-Mathis MA, Poort J, van der Togt C, and Roelfsema PR (2014). Alpha and gamma oscillations characterize feedback and feedforward processing in monkey visual cortex. Proceedings of the National Academy of Sciences of the United States of America. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. van Pelt S, Boomsma DI, and Fries P (2012). Magnetoencephalography in twins reveals a strong genetic determination of the peak frequency of visually induced gamma-band synchronization. The Journal of neuroscience : the official journal of the Society for Neuroscience 32, 3388–3392. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. van Vugt B, Dagnino B, Vartak D, Safaai H, Panzeri S, Dehaene S, and Roelfsema PR (2018). The threshold for conscious report: Signal loss and response bias in visual and frontal cortex. Science 360, 537–542. [DOI] [PubMed] [Google Scholar]
  63. Vinck M, Huurdeman L, Bosman CA, Fries P, Battaglia FP, Pennartz CM, and Tiesinga PH (2015). How to detect the Granger-causal flow direction in the presence of additive noise? NeuroImage 108, 301–318. [DOI] [PubMed] [Google Scholar]
  64. Vinck M, Lima B, Womelsdorf T, Oostenveld R, Singer W, Neuenschwander S, and Fries P (2010a). Gamma-phase shifting in awake monkey visual cortex. The Journal of neuroscience : the official journal of the Society for Neuroscience 30, 1250–1257. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Vinck M, van Wingerden M, Womelsdorf T, Fries P, and Pennartz CM (2010b). The pairwise phase consistency: A bias-free measure of rhythmic neuronal synchronization. NeuroImage 51, 112–122. [DOI] [PubMed] [Google Scholar]
  66. Warton DI, Wright IJ, Falster DS, and Westoby M (2006). Bivariate line-fitting methods for allometry. Biological reviews of the Cambridge Philosophical Society 81, 259–291. [DOI] [PubMed] [Google Scholar]
  67. Womelsdorf T, Fries P, Mitra PP, and Desimone R (2006). Gamma-band synchronization in visual cortex predicts speed of change detection. Nature 439, 733–736. [DOI] [PubMed] [Google Scholar]
  68. Womelsdorf T, Schoffelen JM, Oostenveld R, Singer W, Desimone R, Engel AK, and Fries P (2007). Modulation of neuronal interactions through neuronal synchronization. Science 316, 1609–1612. [DOI] [PubMed] [Google Scholar]
  69. Zandvakili A, and Kohn A (2015). Coordinated Neuronal Activity Enhances Corticocortical Communication. Neuron 87, 827–839. [DOI] [PMC free article] [PubMed] [Google Scholar]
  70. Zhou H, Schafer RJ, and Desimone R (2016). Pulvinar-Cortex Interactions in Vision and Attention. Neuron 89, 209–220. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1508921-supplement-1.pdf (532.5KB, pdf)

RESOURCES