Download Alpha-beta and Gamma Rhythms Subserve Feedback and

This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas.
Georgios Michalareas1, Julien Vezoli1, Stan van Pelt1,2, Jan-Mathijs Schoffelen2,3, Henry
Kennedy4,5 and Pascal Fries1,2
1
Ernst Strüngmann Institute (ESI) for Neuroscience in Cooperation with Max Planck Society,
Deutschordenstraße 46, 60528 Frankfurt, Germany.
2
Donders Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen, Kapittelweg 29,
6525 EN Nijmegen, Netherlands.
3
Max Planck Institute for Psycholinguistics, Wundtlaan 1, 6525 XD Nijmegen, Netherlands.
4
Stem Cell and Brain Research Institute, INSERM U846, 18 avenue Doyen Lépine, 69675 Bron, France.
5
Université de Lyon, 37 rue du Repos, 69361 Lyon, France.
Correspondence should be addressed to G.M. ([email protected]) and P.F.
([email protected])
Highlights:
•
•
•
•
Gamma mediates forward and alpha-beta feedback influences among human visual areas
Human inter-areal directed influences correlate with macaque laminar connectivity
Rhythmic inter-areal influences establish a hierarchy of 26 human visual areas
Alpha-beta influences differentially affect ventral- and dorsal-stream visual areas
In Brief:
Michalareas et al. show that in human visual cortex influences along feedforward projections
predominate in the gamma band, whereas influences along feedback projections
predominate in the alpha-beta band. These influences constrain a functional hierarchy in
agreement with macaque anatomical hierarchy.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
1
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
Summary
Primate visual cortex is hierarchically organized. Bottom-up and top-down influences are exerted
through distinct frequency channels, as was recently revealed in macaques by correlating inter-areal
influences with laminar anatomical projection patterns. Because this anatomical data cannot be
obtained in human subjects, we selected seven homologous macaque and human visual areas, and
correlated the macaque laminar projection patterns to human inter-areal directed influences as
measured with magnetoencephalography. We show that influences along feedforward projections
predominate in the gamma band, whereas influences along feedback projections predominate in the
alpha-beta band. Rhythmic inter-areal influences constrain a functional hierarchy of the seven
homologous human visual areas that is in close agreement with the respective macaque anatomical
hierarchy. Rhythmic influences allow an extension of the hierarchy to 26 human visual areas
including uniquely human brain areas. Hierarchical levels of ventral and dorsal stream visual areas
are differentially affected by inter-areal influences in the alpha-beta band.
Introduction
Non-human primate visual cortical areas are organized in a hierarchy with characteristic laminar
patterns of feedforward and feedback projections (Barone et al., 2000; Felleman and Essen, 1991;
Markov et al., 2014). Feedforward projections typically target layer 4. They originate predominantly
from superficial layers, and this predominance increases with the number of hierarchical levels
bridged by the projection. By contrast, feedback projections typically target layers 1 and 6. They
originate predominantly from infragranular layers, and this predominance increases with the number
of hierarchical levels bridged by the projection. In agreement with this, contextual feedback signals
activate predominantly superficial layers in human V1 (Muckli et al., 2015; Olman et al., 2012).
Cortical layers differ not only in their projection patterns, but also with regard to local rhythmic
synchronization. Synchronization in the gamma-frequency band is strongest in superficial layers,
whereas synchronization in the alpha-beta frequency band is strongest in infragranular layers
(Buffalo et al., 2011; Xing et al., 2012). These results taken together with the laminar projection
patterns suggest that gamma might subserve feedforward and alpha-beta feedback signaling (Fries,
2015; Lee et al., 2013). This is supported by patterns of inter-laminar delay and causality (Bollimunta
et al., 2008; Livingstone, 1996; Plomp et al., 2014). Further, electrical stimulation of V1 induces
enhanced gamma-band activity in V4, whereas area V4 stimulation induces enhanced alpha-betaband activity in V1 (van Kerkoerle et al., 2014).
When directed inter-areal influences across 28 area pairs are assessed through large-scale highdensity electrocorticography in rhesus macaques and correlated with anatomical projection patterns,
gamma is found systematically stronger in the feedforward direction and beta in the feedback
direction (Bastos et al., 2015). The high consistency of this effect made it possible to construct a
hierarchy of visual areas based on directed influences alone. These findings raise the intriguing
possibility that a similar analysis in human cortex would reveal the functional hierarchy of visual
areas. In human subjects, anatomical tract tracing methods, requiring active axonal transport, are
neither possible in the living nor in the post-mortem brain. Further, tractographic analysis of
diffusion MRI does not indicate the directionality of pathways and therefore cannot explore the
hierarchical organization of inter-areal pathways.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
2
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
In the present study, we have first selected seven human areas showing strong homology to
macaque areas. This enabled us to infer the structural hierarchy in the human cortex and to examine
its functional hierarchy using magnetoencephalography. This shows that causal interactions along
the feedforward and feedback pathways are exerted in distinct frequency bands. Based on these
functional markers, we were able to derive the full hierarchy of 26 human visual areas based on the
asymmetry of their feedforward and feedback interactions.
Results
Visually induced responses
Forty-three human subjects were instructed to attentively monitor a visual stimulus for
unpredictable changes, in order to engage both bottom-up and top-down influences. This paradigm
reliably induces gamma-band activity in visual cortex (Hoogenboom et al., 2006). While subjects
fixated centrally, a large circular sine-wave grating contracting towards the fixation point was
presented. The stimulus could change speed at any time between 0.75 and 3 s after onset, which had
to be reported within 0.5 s in order to obtain positive feedback (Figure 1A). Visual stimulation and
task performance induced grand-average power reductions in the alpha and beta bands and
enhancements in the gamma band in many sensors (Figure 1B), consistent with previous reports
(Hoogenboom et al., 2006; van Pelt et al., 2012). Gamma power enhancements were strongest over
occipital and parietal MEG sensors (Figure 1C). The sensors marked in Figure 1C show stimulusinduced gamma power enhancements that start around 0.1 s after stimulus onset and are sustained
for the duration of visual stimulation and task performance (Figure 1D). Alpha and beta power shows
sustained reductions at a slightly longer latency. Between 0.05 and 0.35 s, there is an additional theta
(3-8 Hz) power enhancement, most likely reflecting the event-related field (ERF) (Jutai et al., 1984).
To avoid the ERF and the effect of its non-stationarity on Granger Causality (GC) estimation (Wang et
al., 2008), further analyses use the data from 0.365 s after stimulus onset up to the stimulus change,
segmented into non-overlapping 0.365 s epochs (see Experimental Procedures).
Induced gamma power change estimated at the level of the cortical sheet revealed an occipital peak
extending into parietal and temporal regions (Figure 2A). In order to investigate GC among human
visual areas and to compare it to macaque laminar connectivity, we selected seven visual areas, for
which anatomical retrograde tracing data were available, and for which there is evidence of interspecies homology (see Experimental procedures). These areas and their homologies are illustrated in
Figure 2B. Areas V1 and V2 are substantially larger than any one of the other areas (Figure 2B) and
exhibit the largest gamma power enhancements (Figure 2A). To render V1 and V2 more comparable
in size to the other areas, and to focus on their visually activated subregions, we used the vertices
inside V1 and V2 that were within 2 cm from the local gamma maxima. These vertices and the
vertices of the remaining areas are shown in Figure 2C.
Granger Causality – Field spread effect
MEG sensors pick up mixtures of signals from multiple sources in the brain. This mixing can be partly
deciphered by projecting MEG data into source space. However, the closer two estimated sources
are, the larger the remaining mixing, an effect referred to as “field spread” (Schoffelen and Gross,
2009). Field spread leads to artifactual correlation or coherence for source signals from proximate
locations. Because field spread decays with distance between sources, the dominant spatial pattern
of coherence to a given source (black parcel in Figure 3) is a smooth decay with distance from that
source (Figure 3A, 3B). By contrast, the spatial pattern of GC exerted by that source does not show
3
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
this smooth decay, but rather shows spatial peaks that have their maximum not at the source or in
the immediate vicinity of the source (Figure 3C, 3D). This is likely due to the fact that field spread is
instantaneous. GC eliminates instantaneous correlations, because they are not indicative of a causal
influence.
Granger Causality in feedforward and feedback directions
GC was computed between all possible pairs of the seven selected visual areas. In order to estimate
bias due to the GC metric used and the finite sample size, data epochs were randomized at each
source location and GC computation was repeated. This bias estimation also entails an estimate of
residual stimulus-locked components remaining after exclusion of the first 0.365 s after stimulus
onset. The two hemispheres were analyzed independently.
We averaged all inter-areal GC spectra and compared them to the average bias estimate in order to
obtain a first estimate of overall inter-areal GC, in particular for the main spectral peaks, (Figure 4A).
Average inter-areal GC exceeded the bias estimate across the spectrum and showed two clear peaks,
one in the alpha-beta range, peaking at 11 Hz, and one in the gamma band, peaking around 60 Hz.
Previous studies have shown clear differences between alpha and beta rhythms (Bressler and
Richter, 2014; Gregoriou et al., 2015; Haegens et al., 2014; Wang, 2010), however, the present
analysis revealed one peak spanning across alpha and beta frequency ranges (similar to (Buffalo et
al., 2011)), which we therefore refer to as the alpha-beta band. Peak frequencies were essentially
identical in the two hemispheres.
We used macaque laminar connectivity data that quantifies the degree to which a given inter-areal
projection has a feedforward or feedback character in order to investigate, whether GC differed
between the feedforward and feedback directions. Retrograde tracer injection into a target area
leads to labeling of projection neurons in the source areas. The more pronounced the feedforward
(feedback) character of the anatomical projection, the higher (lower) is the proportion of
supragranular projection neurons (Barone et al., 2000; Markov et al., 2014). This anatomical
signature of the feedforward or feedback character of a projection is captured by the SLN index, i.e.
the number of supragranular labeled neurons divided by the sum of supragranular and infragranular
labeled neurons.
The SLN value for all projections among the 7 selected macaque visual areas was calculated (Markov
et al., 2014). If, between area A and area B, the SLN for the A-to-B projection was higher than the SLN
for the B-to-A projection, then A-to-B was considered a feedforward and B-to-A a feedback
projection. GC values were averaged independently for both directions in each area pair (Figure 4B,
4C for left and right hemisphere, respectively). The same was done with the surrogate data in order
to estimate bias. GC significantly exceeded the bias estimate across the entire spectrum. GC in the
alpha-beta band was stronger in feedback than in feedforward directions. In contrast, GC in the
gamma band was substantially stronger in feedforward than in feedback directions. The statistical
comparison revealed significant differences exclusively in these two frequency bands. Results are
very similar in both hemispheres. The alpha-beta band cluster (P<0.05) ranged from 7 to 17 Hz in
both hemispheres. The gamma-band cluster (P<0.05) ranged from 42 to 93 Hz in the left hemisphere
and from 41 to 84 Hz in the right hemisphere.
Figure 4D shows the GC in feedforward and feedback directions for all area pairs. GC peaks in the
gamma and alpha-beta bands for all area pairs of both hemispheres. For all area pairs, GC exceeded
the bias estimate for frequencies up to the upper end of the gamma band. When GC is compared
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
4
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
between feedforward and feedback directions, most area pairs confirm the above pattern. The
results for the alpha-beta band are as follows: In the left hemisphere, alpha-beta GC was stronger in
the feedback than in the feedforward direction in 13 out of 21 area pairs (61.91%), weaker in 1 area
pair (MT-TEO; 4.76%) and not significantly different in another 7 (33.33%), i.e. it was more frequently
significantly stronger in the feedback than in the feedforward direction (p=0.0018, sign test here and
for the following tests). In the right hemisphere, alpha-beta GC was stronger in the feedback than in
the feedforward direction in 13 out of 21 area pairs (61.91%), weaker in 2 area pairs (MT-TEO and
MT-DP; 9.52%) and not significantly different in another 6 (28.57%) (p=0.0074). The results for the
gamma band are as follows: In the left hemisphere, gamma GC was stronger in the feedforward than
in the feedback direction in 17 out of 21 area pairs (80.95%), weaker in 1 area pair (DP-7A; 4.76%)
and not significantly different in 3 other pairs (14.29%), i.e. it was more frequently significantly
stronger in the feedforward than in the feedback direction (p=0.00015). In the right hemisphere,
gamma GC was stronger in the feedforward than in the feedback direction in 16 out of 21 area pairs
(76.19%), weaker in 1 area pair (DP-7A; 4.76%) and not significantly different in 4 (19.05%)
(p=0.00028).
The peak in the GC spectra that we have identified as the alpha-beta peak spans the classical alphaand beta-frequency bands. Alpha- and beta-band rhythms vary in frequency across subjects and can
be influenced by stimulus and task conditions (Haegens et al., 2014). We calculated the average
inter-areal GC among the seven visual areas separately per subject. When Gaussians were fitted to
these averages in the 4-20 Hz range, most alpha-beta peaks were well approximated by a single
Gaussian and the resulting peak frequencies extended from 7 to 19 Hz (Figure S1A, B; mean±SD =
11.02±2.45 Hz). Likewise, when Gaussians were fitted to the GC averages in the 30-100 Hz range, the
resulting gamma peak frequencies extended from 52 to 69 Hz (Figure S1C, D; mean±SD =
59.16±4.16 Hz). Similar cross-subject variability has been described in previous studies using the
same stimulus and task as used here and has been related to genetic factors (Hoogenboom et al.,
2006; van Pelt et al., 2012). To account for this inter-subject variability in peak frequencies, we
repeated the above GC analyses after aligning alpha-beta and gamma peak frequencies, respectively,
across subjects (Figure 5). After alignment, GC values tended to increase, and the above differences
between GC in the feedforward and feedback direction were confirmed. Significant differences
remained identical except in a few cases, the most notable being V1-V2 in the right hemisphere,
changing from an insignificant difference to a stronger alpha-beta GC in the feedforward than in the
feedback direction.
Because the peak-aligned analysis accounts for inter-subject variability, it was used for all further
analyses (results without peak alignment are provided as supplemental information and are
qualitatively identical).
Correlation between human inter-areal Granger Causality and macaque anatomical
projections
So far, we have used the SLN metric to decide for each area pair, which one of the two reciprocal
projections is feedforward and which feedback. As explained above, the SLN of an anatomical
projection is the number of supragranular neurons normalized by the number of supragranular and
infragranular neurons that give rise to the projection. Thus, the SLN quantifies the feedforward or
feedback character of a projection in a graded metric. We next wanted to correlate this graded
anatomical metric with a similarly graded functional metric, which captures the above mentioned GC
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
5
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
asymmetries. Therefore, we defined as in (Bastos et al., 2015) the Directed influence Asymmetry
Index (DAI) between areas A and B as
DAI(A-to-B) = (GC(A-to-B) – GC(B-to-A)) / (GC(A-to-B) + GC(B-to-A))
The numerator captures the predominant or net direction of GC, while the denominator provides a
normalization factor such that the asymmetry in causal interactions can be compared between
different area pairs and between different frequencies, for which raw GC values can differ
substantially. Note that DAI is defined per frequency of the GC spectrum.
To test whether the GC asymmetry is systematically related to the feedforward-feedback character
of the anatomical projections, we calculated the Spearman correlation coefficient between DAI
values (from human MEG data) and SLN values (from macaque retrograde tracing data) across all
visual area pairs. The seven visual areas under consideration give rise to 42 pairs, of which 38 had
anatomical data that reached threshold (i.e. more than 10 labeled neurons). The DAI-SLN correlation
was determined per GC frequency and per subject. Figure 6 shows the mean and SEM across the 43
subjects. The results for the left (Figure 6A) and right (Figure 6B) hemisphere were qualitatively
identical. Also, the results remained qualitatively unchanged when the analysis was repeated without
alignment to the subject-wise alpha-beta- and gamma-band peak frequencies (Figure S2).
Feedforward (feedback) anatomical projections predominantly originate in supragranular
(infragranular) layers and are therefore characterized by a large (small) SLN (supragranular labeled
neuron proportion), typically above (below) 0.5. Correspondingly, when the correlation is calculated
between SLN values and DAI values, a positive DAI-SLN correlation for a given frequency range
indicates that this frequency is systematically stronger (weaker) in the direction of feedforward
(feedback) projections. Thus, the significant positive DAI-SLN correlation for the gamma band
indicates that the net influence in this frequency is feedforward (left (right) hemisphere: mean=0.465
(0.467), p=0.0019 (0.0019)). The significant negative correlation for the alpha-beta band indicates
that the net influence is feedback (left (right) hemisphere: mean=-0.226 (-0.178), p=0.0159 (0.0139)).
A Granger-Causality-based hierarchy correlates with the anatomical hierarchy
SLN values have been used to establish an anatomical hierarchy of primate visual areas (Markov et
al., 2014). In this hierarchy, the feedforward-feedback character of most inter-areal projections is in
conformity with a global hierarchy, in which each area occupies a particular hierarchical level. For
example, if the A-to-B projection is feedforward and the B-to-C projection is feedforward, then the Ato-C projection is typically strongly feedforward. Note there is no a-priori reason that prevents the Ato-C projection from being feedback, because neurons in A projecting to B and C are typically almost
completely non-overlapping (Salin and Bullier, 1995), and neurons are exclusively either feedback or
feedforward (Markov et al., 2014). The fact that inter-areal projections tend to confirm to a global
hierarchy is an empirical finding in the large set of available SLN data (Markov et al., 2014).
Given that the macaque SLN is correlated to the human DAI, we asked whether the human DAI also
establishes a hierarchy of human visual areas, similar to the DAI-based hierarchy of macaque visual
areas established previously (Bastos et al., 2015). Within each subject, we considered each area in
turn as a seed area. The DAI values of this seed area relative to all areas were ranked (with the DAI of
the seed to itself being zero, according to the DAI definition). Average ranking across all seven seed
areas was established for each subject. Figure 7 shows average rankings and their SEMs across the 43
subjects. Figure 7A shows the rankings derived from the DAI in the gamma-frequency band. The
human GC-based ranking is plotted against the macaque anatomy-based ranking (Markov et al.,
6
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
2014) and the two hierarchies are significantly correlated (left hemisphere: R=0.962, p=0.0005; right
hemisphere: R=0.945, p=0.0013; Pearson correlations in this and the following tests). Note, an
essentially perfect correlation existed for all areas, except TEO, which ranked slightly higher in the
human GC-based than the macaque anatomy-based hierarchy. Note also that hierarchical levels span
almost the entire possible range of 7 levels, suggesting slightly less than 7 levels as a first estimate of
a lower bound for the number of hierarchical levels present in the data (see below for a higher
estimate, when more areas are included).
The same procedure was repeated with the DAI in the alpha-beta frequency band (after inverting its
sign because of its negative correlation to SLN). The resulting relation to anatomy was less
pronounced, because early visual areas V1 and V2 were placed higher and the ventral-stream areas
V4 and TEO were placed lower than in the anatomical hierarchy (Figure 7B) (left hemisphere: R=0.75,
p=0.032; right hemisphere: R=0.812, p=0.0267). A hierarchy based on average DAI values from both
the gamma and the alpha-beta band (after inverting the latter) shows excellent correspondence
between functional and anatomical hierarchies (Figure 7C) (left hemisphere: R=0.901, p=0.0057; right
hemisphere: R=0.923, p=0.003). In comparison to the pure gamma-based hierarchy, the ventralstream areas V4 and TEO are moved lower. This might be due to strong feedback influences exerted
in the alpha-beta band (see Discussion).
Functional hierarchies built separately from the GC values in the two hemispheres were qualitatively
identical (compare the two columns of Figure 7). Further, the results remained qualitatively
unchanged when the analysis was repeated without alignment to the subject-wise peak frequencies
in the alpha-beta and gamma band (Figure S3). When the analysis was repeated for areas randomly
defined as contiguous patches of dipoles (with sizes similar to the actual areas), this produced
correlations distributed around zero. All observed correlations exceeded the 97.5th percentile of
these randomization distributions, i.e. they were significantly different in a two-sided test.
As an independent test for the presence of functional hierarchies, we computed repeated measures
one-way ANOVAs across subjects with the factor AREA and the dependent variable HIERARCHICAL
LEVEL. The observed F-values were compared to the distribution of F-values from 1000 trial shuffles.
All hierarchies were highly significant (p-values of 0 to 0.003).
Granger Causality establishes a functional hierarchy across 26 human visual areas
The analysis so far was limited to 7 areas, for which the homology between human and macaque is
well pronounced and for which quantitative retrograde tracing data were available. Given that the
anatomical hierarchy was strongly correlated with a GC-based hierarchy, we next investigated
whether a hierarchy can also be built for a larger set of human visual areas: V1, V2, V3, V3A, V3B,
V3C, V3D, V4, V4t, LO1, LO2, PITd, PITv, MT, VO1, MST, FST, ER, V7, IPS1, IPS2, IPS3, IPS4, 7A, FEF, 46
(see Experimental procedures for description). For these 26 areas, the functional hierarchical level
was derived based on the DAI metric, in the same fashion as for the 7 areas. Figure 8A shows the 26
areas according to their functional hierarchical level. If the hierarchical ranking had not been
consistent across seed areas within subjects, or across subjects, the different areas would rank
roughly at the same hierarchical level. Similarly, if ranking were not consistent across subjects, the
error bars would be large relative to the rank range. Thus, the fact that in Figure 8, the 26 areas span
almost the entire dynamic range of 26 hierarchical levels, with relatively small error bars, suggests
slightly less than 26 levels as a lower bound for the number of hierarchical levels present in the data.
It also suggests that the organization of most area-pair-wise GC influences corresponds to a global
hierarchy. In addition, the resulting hierarchy is in good agreement with the mosaic of evidence from
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
7
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
different sources including quantitative anatomy (Barone et al., 2000; Markov et al., 2014), general
anatomical connectivity and position (Felleman and Essen, 1991) and cognitive architectures
(Dehaene, 2005). The location of the frontal eye field (FEF) low in the hierarchy is in agreement with
recent quantitative analysis of retrograde tracing (Barone et al., 2000; Markov et al., 2014). In terms
of hierarchical discriminability of areas, it is clear that areas are more separable at the upper and
lower ends of the hierarchy, than they are in the middle of the hierarchy.
The same procedure repeated in the alpha-beta band led to several areas changing their hierarchical
level. To illustrate this, Figure 8B retains the x-axis ordering established in Figure 8A on the basis of
the gamma-band DAI, but reports on the y-axis the hierarchical level based on the alpha-beta band
DAI (a copy of each area’s gamma-DAI based hierarchical level is shown as a gray dot). In general,
ventral-stream areas (labeled green) V4, PITv, PITd and VO1 are moved to lower levels, whereas
many dorsal-stream areas (labeled red) including V3C, V3D, MST, FST, IPS3, IPS4 and 7A are shifted to
higher levels. Similarly V1 and V2 are moved to higher levels. These shifts could be due to relatively
strong alpha-beta-band influences exerted by the dorsal-stream areas onto the ventral stream areas,
and exempting the earliest areas V1 and V2 (see Discussion). Likewise, FEF is shifted to a higher level,
suggesting that it exerts relatively strong alpha-beta feedback.
The split between dorsal- and ventral-stream areas becomes even more apparent, when the
functional hierarchy is based on the DAIs from the gamma and alpha-beta bands combined
(Figure 8C). Compared to a pure gamma-based hierarchy (gray dots), it shows a consistent upward
shift for most dorsal-stream areas (V3C, V3D, MST, FST, V7, IPS1 to 4) and a consistent downward
shift for ventral-stream areas (V4, PITv, PITd, VO1).
We repeated the ANOVA analysis for the 26 areas, and again all hierarchies were significant (p-values
ranging from 0 to 0.016).
Discussion
We used MEG to study the spectral specificity of causal influences along inferred anatomical
feedforward and feedback projections in the human visual cortex. Projections were defined as
feedforward or feedback based on retrograde tracing data in homologous macaque visual areas.
Frequency-resolved Granger causality between human visual cortical areas showed that causal
influences along feedforward projections were predominant in the gamma band, while causal
interactions along feedback projections were predominant in the alpha-beta band. Based on these
results, we used all inter-areal GC values to build a functional hierarchy in human visual cortex, which
was strongly correlated to the hierarchy derived from macaque retrograde tracing data. Finally, we
extrapolated this to 26 human visual areas.
The present analyses are based on MEG data from 43 human subjects, which allowed us to use
statistics that are non-parametric counterparts of random-effect statistics. The previous analysis
using ECoG in macaques was restricted to two animals, as is typical for non-human primate studies,
and thereby to a fixed-effect analysis. Whereas a fixed-effect analysis allows inference only on the
tested sample, our present analysis allows generalization to the population of human subjects.
Fast versus slow oscillations for feedforward and feedback signaling
Several studies have reported that feedforward and feedback communication use respectively higher
and lower frequency ranges. Some studies compared task conditions and found that conditions
expected to emphasize feedforward signaling lead to stronger synchronization in relatively higher
8
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
frequency bands, whereas conditions expected to emphasize feedback signaling lead to stronger
synchronization in relatively lower frequency bands (Arnal et al., 2011; Buschman and Miller, 2007;
von Stein et al., 2000). Other studies compared anatomical structures and found that pathways or
layers involved in feedforward signaling show stronger synchronization at higher frequencies, and
pathways or layers involved in feedback signaling show stronger synchronization in lower
frequencies. For example, feedforward communication from medial entorhinal cortex to
hippocampal area CA1 uses higher gamma, whereas feedback communication from CA3 to CA1 uses
lower gamma (Colgin et al., 2009; Schomburg et al., 2014). In visual cortex, superficial layers, sending
feedforward projections, show predominantly gamma-band synchronization, whereas infragranular
layers, sending feedback projections, show predominantly alpha-beta-band synchronization (Buffalo
et al., 2011; Xing et al., 2012). Across cortical sublayers, gamma shows phase shifts consistent with
feedforward signaling (Livingstone, 1996; van Kerkoerle et al., 2014), and alpha shows phase shifts or
GC consistent with feedback signaling (Bollimunta et al., 2011; van Kerkoerle et al., 2014). Direct
quantification of directed influences by means of GC showed that inter-areal feedforward influences
are carried by gamma and feedback influences by alpha or beta rhythms. Between human auditory
area A1 and auditory association cortex, feedforward signaling uses gamma, and feedback signaling
delta/beta rhythms (Fontolan et al., 2014). Between macaque visual areas V1 and V4, feedforward
signaling uses gamma and feedback signaling alpha or beta (Bosman et al., 2012; van Kerkoerle et al.,
2014).
Finally, a recent study investigated eight visual areas in macaque and found that across the
numerous inter-areal projections, feedforward connectivity was systematically related to GC in the
theta and gamma bands, and feedback connectivity systematically to GC in the beta band (Bastos et
al., 2015). These earlier results from macaques agree fully with the present results for humans in the
gamma band, but they agree only partly in the other frequency bands. Our present analysis in
humans suggests that top-down influences are exerted by rhythms spanning alpha and beta
frequency ranges, even though previous studies clearly suggest separate alpha and beta rhythms
(Gregoriou et al., 2015; Wang, 2010; Womelsdorf et al., 2014). It is important to note that our
analysis merely reveals whether directed influences in a frequency band relate to feedforward or
feedback projections. Our results suggest that both alpha and beta subserve top-down signaling. Yet,
this does not contradict differential mechanisms and sub-functions of top-down influences (Fries,
2015). Alpha might mediate top-down influences suppressing irrelevant background (van Kerkoerle
et al., 2014), whereas beta might mediate top-down influences, that facilitate the bottom-up
communication of attended stimuli (Bastos et al., 2015; Lee et al., 2013). Our present analysis did not
reveal a convincing peak of GC influences in the theta-frequency range, even though theta-band
influences have been found in the earlier study with macaques (Bastos et al., 2015). The reason for
this difference is currently unclear, and we can only speculate that it might be due to differences
between ECoG and MEG, or between the stimulus and/or task conditions.
Compared to infragranular connectivity, supragranular connectivity shows fine-grained spatial
structure and interconnects some brain regions, like e.g. frontal cortex and ventral visual areas
(Markov et al., 2014), possibly allowing frontal areas to change their functional hierarchical level
dynamically (Bastos et al., 2015). In agreement with this, gamma-band GC influences between
macaque FEF and V4 predominate in the FEF-to-V4 direction after an attentional cue, but
subsequently predominate in the V4-to-FEF direction (Gregoriou et al., 2009). Between human
frontal cortex and high-level ventral visual cortex, gamma-band coherence shows a 20 ms phase lead
of frontal cortex (Baldauf and Desimone, 2014).
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
9
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
The present results, indicating a functional hierarchy of the human visual cortex, are relevant to
future investigation of predictive coding in human cortex. Feedforward and feedback influences have
been hypothesized to subserve the signaling of prediction errors and predictions, respectively. These
concepts are central to the theory of predictive coding, stating in essence that higher areas
constantly derive predictions based on incoming evidence and prior experience, and feed those
predictions back to lower areas, where they are subtracted from sensory evidence, such that only
prediction errors need to be forwarded to higher areas (Bastos et al., 2012). Crucially, in this
framework, the updating of predictions entails the accumulation of prediction errors over time, so
that predictions change more slowly than prediction errors. Therefore, if predictions and prediction
errors are mediated through inter-areal rhythmic synchronization, the communication of prediction
errors requires higher frequencies than the communication of predictions. This is in agreement with
several recent studies (Bastos et al., 2015; Fontolan et al., 2014; van Kerkoerle et al., 2014) and our
finding that feedforward signaling is dominated by gamma and feedback signaling is dominated by
alpha-beta rhythms.
Ventral stream areas
While the functional human hierarchy derived from the DAI was largely in agreement with the
macaque anatomical hierarchy, some important deviations were observed, especially when the
hierarchy was computed from the feedback causal asymmetries in the alpha-beta band. In this case,
the ventral stream areas V4 and TEO were much lower in the functional hierarchy than in the
anatomical one. Feedback causal influences were strongest onto area V4 and progressively
decreased towards the primary visual cortex. A similarly decreasing pattern has been observed for
the effect of attention on firing rates in areas V1, V2 and V4 (Buffalo et al., 2010). In this work, a
reverse order of attentional effects was shown, with a stronger and earlier enhancement in firing
rates in V4, and progressively smaller and later enhancement in V2 and in V1. This strong feedback
influence on V4 was also prominent in the hierarchy derived from the combined feedforward and
feedback causal asymmetries as manifested in the unexpectedly low hierarchical position of V4 (at a
similar level to V2).
These results suggest that feedback influences from higher ventral and dorsal areas primarily address
mid-level ventral stream areas such as V4 and TEO, and to a lesser degree affect the earlier visual
areas V2 and V1. It has already been suggested that early ventral stream areas and especially area V4
is well positioned for integrating top-down influences with bottom-up information (Roe et al., 2012),
mainly due to strong corticocortical (Cloutman, 2013; Markov et al., 2014) and
corticothalamocortical (Van Essen, 2005) anatomical connections with temporal, prefrontal and
parietal regions.
Also, details of the task may differentially impact the patterns of alpha-beta band top-down
influences. Our task required subjects to detect a speed change in a drifting grating and may
therefore predominantly engage the dorsal stream. It is conceivable that this contributed to strong
alpha-beta band influences from dorsal to ventral stream areas, which in turn places dorsal stream
areas higher and ventral stream areas lower in the functional hierarchy. Future studies might
systematically investigate changes in the functional hierarchy as a function of task requirements.
Functional hierarchy of 26 visual areas
Using the asymmetry in causal interactions, we estimated the functional hierarchy for 26 areas of the
human visual system. For several of the 26 areas, it is not possible to establish a clear homology to
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
10
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
areas in the non-human primate brain. Yet, for the areas, for which a homology can be hypothesized,
there was a good agreement between functional and anatomical hierarchical level (Felleman and
Essen, 1991; Markov et al., 2014). Furthermore, for areas defined only in humans, the functional
hierarchical level agreed with expectations derived from overall position in the brain and functional
characteristics. The strong effect of feedback causal influences on mid-level ventral stream areas was
observed also in this hierarchy.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
11
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
Experimental procedures
Experimental paradigm
The experimental paradigm (Figure 1A), was adopted from (van Pelt et al., 2012) and (Hoogenboom
et al., 2006) and is known to induce strong gamma band oscillations in occipital areas. Each trial was
initiated with a fixation point (Gaussian of diameter 0.5 deg) at the center of the screen for 1.5 s. The
fixation period was followed by the presentation of a circular sine-wave grating, centered at the
fixation point and contracting towards its center (spatial frequency: 3 cycles/deg, velocity:
0.66 deg/s, contrast: 100%). The screen background was black. At a random moment between 0.75
and 3 s after stimulus onset, velocity increased (step to 1 deg/s), which was reported by the subjects
with their right index finger. After the response followed a rest period of 1 s during which feedback
was given. A total of 180 trials were presented to each subject. Visual presentation was performed
on a translucent screen, situated 0.9 m in front of subjects’ eyes by using an LCD projector with a
60 Hz update rate. We quantified the MEG signal components locked to the LCD update for sensors
over visual cortex (Figure 1C), and found them to account for a very small part of the total power at
60 Hz (<1.94% of 60 Hz power during baseline, <1.42% of 60 Hz power during stimulation). Note that
data were acquired in Europe, i.e. the power line frequency was 50 Hz.
Data Acquisition
MEG: Recordings were performed with a whole-head MEG system (CTF Systems Inc.) comprising 275
axial gradiometers. MRI: Structural MR images were acquired using an Avanto 1.5 T whole body MRI
scanner (Siemens, Germany). Further details can be found in the corresponding section of
Supplemental Experimental Procedures.
Subjects
Datasets of 43 subjects were selected from a pool of 160 subjects, which had been recorded for
genotyping of parameters derived from visually induced gamma-band activity. Further details on
selection criteria can be found in the corresponding section of Supplemental Experimental
Procedures. All subjects gave written informed consent according to guidelines of the local ethics
committee (Commissie Mensgebonden Onderzoek Regio Arnhem-Nijmegen, The Netherlands).
Data preprocessing and spectral estimation
All MEG data analysis was performed in MATLAB, partly using the FieldTrip toolbox (Oostenveld et
al., 2011). Data segments which contained artifacts originating from jumps in the SQUIDs (superconducting quantum interference devices), muscle activity and eyes-movements were identified and
removed using semi-automated FieldTrip routines. Power line artifacts were estimated and removed
using a discrete Fourier Transform at 50, 100 and 150 Hz with a bin width of ±0.25 Hz. We used data
from 0.365 s after stimulus onset until the visual stimulus changed speed, thereby excluding
stimulus-onset or stimulus–change related non-stationarities. These data were divided into nonoverlapping 0.365 s epochs. This fixed epoch length provided a good compromise between
optimizing the use of the variable-length trials and maximizing the spectral resolution for the nonparametric spectral factorization employed in the calculation of Granger causality. Epochs taken from
early or late in the trial (median split) did not differ in their GC and were therefore pooled. Spectral
estimation was first performed at the sensor level. Per sensor, each data epoch was demeaned,
Hann-tapered, zero-padded to 1 s length and Fourier transformed to give the spectrum between 1
and 140 Hz (Percival and Walden, 2000).
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
12
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
Human Brain Model
For source analysis, structural MR images were acquired for each individual subject, and the cortical
mantle was extracted with the FreeSurfer longitudinal stream pipeline (Reuter et al., 2012). Cortical
parcellation and area definition used a recent visuotopic atlas provided by (Abdollahi et al., 2014),
the FreeSurfer software suite (http://FreeSurfer.net/) (Desikan et al., 2006; Destrieux et al., 2010;
Fischl et al., 2004) and the Caret software suite (http://brainvis.wustl.edu/) (Glasser and Van Essen,
2011; Van Essen et al., 2012).
Further details on the construction of the source space with the investigated visual areas can be
found in the corresponding section of Supplemental Experimental Procedures.
Selection of homologous brain areas between human and macaque
We aimed at relating functional connectivity derived from human MEG recordings to anatomical
connectivity derived from macaque retrograde tracing experiments. Therefore, the availability of
retrograde tract tracing data dictated the selection of visual areas. From the set of areas in the
occipital, parietal and temporal cortical regions with such tracing data, a subset was selected for
which evidence of homology between human and macaque brains exists in the literature. These
areas were V1, V2, MT, V4, DP, TEO and 7A and are shown in Figure 2B. The numbers of dipoles per
area are listed in Supplemental Table 1.
More details about the selected areas and their homology between species can be found in the
corresponding section of Supplemental Experimental Procedures.
Parcels representing areas V1 and V2 are quite extensive in comparison to the other parcels and
contain a large number of voxels that are far from the main locus of activation, as defined by visually
induced gamma power enhancement (Figure 2). In order to minimize the effect of the resulting
extensive smoothing, a more spatially confined representation for these areas has been employed. In
this representation, V1 and V2 parcels consist of the vertices within 2 cm from the local gamma
power maxima within the original representation of these areas.
Inverse solution
Subject-wise structural MRI scans were used to derive individual single-shell volume conductor
models (Nolte, 2003). The inverse solution was then performed using Linearly Constrained Minimum
Variance (LCMV) adaptive spatial filtering, often referred to as “Beamforming”. When determining
the LCMV solution, we used a regularization parameter of 20% of the covariance matrix.
For further details on the inverse solution please see the corresponding section in Supplemental
Experimental Procedures.
Granger Causality (GC)
In order to study the strength and directionality of causal influences between the investigated brain
areas, we employed GC (Granger, 1969), a statistical measure that quantifies the extent to which one
time series can predict another. The principle idea behind GC is that if the addition of the history of
signal A improves the prediction of signal B, as compared to the prediction of signal B based on its
own history alone, then signal A is said to ‘”Granger-cause” signal B.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
13
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
This measure of predictive causality can be estimated either in the time domain or in the frequency
domain (Geweke, 1982). In the latter case GC provides a spectrum of causal interactions between the
two signals as a function of frequency. This is of great utility for the analysis of MEG data, because
oscillations at distinct frequencies can be simultaneously realized in different brain areas and likely
subserve inter-areal communication (Fries, 2005; Fries, 2015). GC can be calculated using parametric
methods, based on multivariate autoregressive models. This approach requires the selection of
model parameters and can fail to capture complex spectral features (Mitra and Pesaran, 1999).
Therefore, we used a non-parametric approach, which estimates GC directly from a factorization of
the cross-spectral density (CSD) matrix (Dhamala et al., 2008). For the present study, the CSD of two
brain sources is simply the scalar product between their complex and complex-conjugate spectral
coefficient sequences. These coefficients are computed by multiplying the sensor-level spectral
decomposition with the spatial filter for each location.
The visuotopic source space used in this work consists of the set of points distributed within the
brain areas of interest, referred to as “parcels”. To calculate GC between two parcels, we first
averaged the CSD between all inter-parcel dipole pairs. Within parcels, the power spectral density
was computed by averaging all the within-parcel power-spectral density values.
Retrograde tracing database
Description of the anatomical dataset acquisition and analysis has been reported in (Markov et al.,
2014). Please see also Supplemental Experimental Procedures for additional information.
Selection of 26 visual areas from the human brain
For this analysis, we extended the initial set of 7 visual areas, for which anatomical connectivity
information was available, to an extended set of 26 areas involved in visual processing. The
additional areas were V3, V3A, V3B, V3C, V3D, V4t, LO1, LO2, VO1, MST, FST from the atlas of
(Abdollahi et al., 2014), and areas IPS1, IPS2, IPS3, IPS4, 46, ER from the visuotopic atlas of Caret (Van
Essen et al., 2012). Area PIT from the 7-area analysis was split into its two constituent areas PITv,
PITd from the atlas of (Abdollahi et al., 2014). Frontal eye fields (FEF) were manually selected. For
further details on the 26 selected visual areas, see Supplemental Experimental Procedures.
Statistical testing
Figures 4 and 5 use randomization-based non-parametric statistical tests with cluster-based multiple
comparison correction (Maris and Oostenveld, 2007). Each subject provided one feedforward and
one feedback GC spectrum, which were compared across subjects via t-tests. The t-values of
spectrally adjacent, significant t-tests were summed to generate the observed (first-level, clusterbased) test metric (not used for inference). We then performed 1000 randomizations. In each
randomization, feedforward and feedback labels were randomly exchanged per subject, and the test
metric was computed as for the non-randomized data. Per randomization, the largest test metric
entered into the randomization distribution. The observed test metrics were compared against the
randomization distribution. Comparisons between feedforward or feedback GC versus the bias
estimates proceeded accordingly. Similarly, the correlation coefficients between DAI and SLN values
(computed as explained in the Results section) were tested against zero (Figure 6) and against bias
estimates (Figure S2), i.e. each subject provided one correlation spectrum and either one null
spectrum or one bias-estimate spectrum.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
14
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
Figure legends
Figure 1. Experimental paradigm and visually induced changes in MEG signal power in frequency,
space and time
(A) Experimental paradigm. The fixation period was followed by the presentation of a circular sinewave grating, centered at the fixation point and contracting towards its center. At a random moment
between 0.75 and 3 s after stimulus onset, velocity increased, which was reported by the subjects
with their right index finger. Thicker red arrows indicate the speed change. (B) Spectrum of power
change during stimulation versus baseline (-0.5 to -0.2 sec). Each line represents the average across
subjects for one MEG sensor. (C) Sensor-level topography of power change from (B) in the frequency
range of 40 to 75 Hz. (D) Average power change as a function of time and frequency for selected
sensors over occipito-parietal areas, shown with * in (C). Color bar applies to (C) and (D).
Figure 2. Gamma power distribution on the cortical sheet and human-macaque homology
definition
(A) Gamma-power change distribution on the cortical sheet as derived from inverse solution. (B)
Selected human-macaque homologous visual areas. Displayed on flat cortical maps. The small black
spheres on the macaque flat cortex indicate the injection sites of retrograde tracer published in
(Markov et al., 2014). (C) Vertices representing the human homologous areas from (B) in the MEG
inverse solution on the very inflated Conte69 template brain.
Figure 3. Granger Causality limits the effect of field spread
For this investigation, each hemisphere was randomly segmented into 400 parcels according to the
Fast Marching algorithm (Sethian, 1999). There were on average 10 vertices per parcel. The parcel
closest to the peak of the average gamma power over all subjects was identified and was selected as
the seed parcel, marked here in black. (A, B) Coherence was computed for each subject between this
seed parcel and the remaining 399 parcels. These coherence values were subsequently averaged
across all subjects for each pair and each frequency. Finally for each pair, the coherence values
within the gamma band (40 to 75 Hz) were averaged. (C, D) GC from the seed to the rest of the
parcels for the same frequency range.
Figure 4. Granger Causality in feedforward and feedback directions
(A) GC spectrum averaged across all pairs of areas. GC values were averaged across all area pairs and
subjects, separately per hemisphere. Significance relative to the surrogate (random) data was
computed by comparing for all subjects the GC between the original and the surrogate conditions
using a permutation-based non-parametric statistical test. (B, C) GC spectra averaged separately for
inter-areal influences corresponding to feedforward (green) or feedback (black) projections. Two
non-parametric statistical tests were performed. One between the actual and surrogate average GC
spectra and one between the feedforward and feedback average GC spectra. This analysis was
performed separately for the left (B) and right (C) hemisphere. (D) GC spectra, averaged over
subjects, separately per area pair and hemisphere. Again two non-parametric statistical tests were
performed for each area pair: One between the actual and surrogate data and one between the
feedforward and feedback direction. Tests were corrected for multiple comparisons across
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
15
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
frequencies and across area pairs within each hemisphere. Left (right) hemisphere data are shown in
lower (upper) triangle. The two arrows above each subplot signify the feedforward (green arrow) and
feedback (black arrow) characteristic of the anatomical projections, as defined by the anatomical SLN
values (listed above and below the arrows). GC spectra in the feedforward direction are shown in
green, GC spectra in the feedback direction in black. See information at the bottom of the plot for
details on each subplot.
Figure 5. Granger Causality in feedforward and feedback directions after peak alignment
In order to account for inter-subject variability, two new frequency axes were devised, aligned to the
individual alpha-beta and gamma peaks. The alpha-beta frequency axis spanned from -5 to +5 Hz
around the alpha-beta peak. The gamma frequency axis spanned from -25 to +70 around the gamma
peak. There was no overlap between these two frequency ranges for any subject. Otherwise,
analyses and figure format is as in Figure 4.
Figure 6. DAI-SLN correlation spectra reveal systematic relation between GC and anatomy
Spearman correlation between DAI values (from human MEG) and SLN values (from macaque
retrograde tracing) across pairs of visual areas. The DAI-SLN correlation was determined per
frequency of the GC spectrum. Significance relative to zero was computed using a permutation-based
non-parametric statistical test against the null hypothesis of zero correlation, with cluster-based
multiple comparisons correction. This analysis was performed separately for the left (A) and right (B)
hemisphere. In both cases, a frequency cluster of negative correlation was identified in the alphabeta range and a cluster of positive correlation was identified in the gamma range.
Figure 7. Granger-Causality-based hierarchy correlates with anatomical hierarchy
The relation between the functional hierarchy of human visual areas and the anatomical hierarchy of
homologous macaque areas was determined separately for left and right hemispheres. The
functional hierarchy of the human visual areas was computed from the DAI values (see Experimental
procedures). Three cases were investigated, namely with the functional hierarchy computed from (A)
DAI in the gamma band, (B) DAI in the alpha-beta band and (C) the combined DAI in alpha-beta and
gamma bands. Early visual areas are depicted in black color, ventral-stream areas in green and
dorsal-stream areas in red. Inset brackets on bottom right of each subplot report Pearson correlation
between human functional and macaque anatomical hierarchical levels. Notice the push towards
lower hierarchical levels of midlevel ventral stream areas by the inclusion of feedback causal
influences in the alpha-beta band (compare (C) to (A)).
Figure 8. Functional hierarchy of 26 human visual areas
The functional hierarchy of 26 human visual areas was computed based on (A) DAI in the gamma
band, (B) DAI in the alpha-beta band and (C) the combined DAI in alpha-beta and gamma bands. The
functional hierarchy in each case is the average of the functional hierarchies in the left and right
hemisphere. Ventral-stream areas are depicted in green, dorsal-stream areas in red and all other
areas in black. In (A) the 26 areas have been ranked along the x-axis based on their average
hierarchical level depicted on the y-axis. In (B) and (C), the ranking from (A) has been preserved on
the x-axis in order to illustrate changes in the hierarchy for DAIs of different frequency bands.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
16
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
Author contributions
Conceptualization, G.M., J.V., and P.F.; Methodology, G.M., J.V. and J.S.; Formal Analysis, G.M., and
J.V.; Investigation, S.P., and H.K.; Resources, P.F.; Writing – Original Draft, G.M., and P.F.; Writing –
Review & Editing, G.M., J.V., S.P., J.S., H.K., and P.F.; Visualization, G.M.; Supervision, P.F.; Project
Administration, P.F.; Funding Acquisition, P.F.
Acknowledgements
We thank Marieke van de Nieuwenhuijzen for assisting in the recording of MEG datasets and Eric
Maris for help with non-parametric statistics. This work was supported by the Human Connectome
Project (WU-Minn Consortium, NIH grant 1U54MH091657 to P.F.), a European Young Investigator
Award (P.F.), the European Union (HEALTH-F2-2008-200728 to P.F.), the LOEWE program
(“Neuronale Koordination Forschungsschwerpunkt Frankfurt” to P.F.), NWO (VIDI grant to J.S.), the
Smart Mix Program of the Netherlands Ministry of Economic Affairs and the Netherlands Ministry of
Education, Culture and Science (BrainGain to P.F. and J.S.), ANR-11-BSV4-501 (H.K.), and LabEx
CORTEX (ANR-11-LABX-0042 to H.K.).
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
17
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
References
Abdollahi, R.O., Kolster, H., Glasser, M.F., Robinson, E.C., Coalson, T.S., Dierker, D., Jenkinson, M.,
Van Essen, D.C., and Orban, G.A. (2014). Correspondences between retinotopic areas and myelin
maps in human visual cortex. Neuroimage 99, 509-524.
Arnal, L.H., Wyart, V., and Giraud, A.-L. (2011). Transitions in neural oscillations reflect prediction
errors generated in audiovisual speech. Nat Neurosci 14, 797-801.
Baldauf, D., and Desimone, R. (2014). Neural mechanisms of object-based attention. Science 344,
424-427.
Barone, P., Batardiere, A., Knoblauch, K., and Kennedy, H. (2000). Laminar distribution of neurons in
extrastriate areas projecting to visual areas V1 and V4 correlates with the hierarchical rank and
indicates the operation of a distance rule. The Journal of neuroscience : the official journal of the
Society for Neuroscience 20, 3263-3281.
Bastos, A.M., Usrey, W.M., Adams, R.A., Mangun, G.R., Fries, P., and Friston, K.J. (2012). Canonical
microcircuits for predictive coding. Neuron 76, 695-711.
Bastos, A.M., Vezoli, J., Bosman, C.A., Schoffelen, J.-M., Oostenveld, R., Dowdall, J.R., De Weerd, P.,
Kennedy, H., and Fries, P. (2015). Visual areas exert feedforward and feedback influences through
distinct frequency channels. Neuron 85, 390-401.
Bollimunta, A., Chen, Y., Schroeder, C.E., and 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., and Ding, M. (2011). Neuronal mechanisms and attentional
modulation of corticothalamic alpha oscillations. The Journal of Neuroscience 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., and Fries, P. (2012). Attentional stimulus selection through selective
synchronization between monkey visual areas. Neuron 75, 875-888.
Bressler, S.L., and Richter, C.G. (2014). Interareal oscillatory synchronization in top-down neocortical
processing. Curr Opin Neurobiol 31C, 62-66.
Buffalo, E.A., Fries, P., Landman, R., Buschman, T.J., and Desimone, R. (2011). Laminar differences in
gamma and alpha coherence in the ventral stream. Proceedings of the National Academy of Sciences
108, 11262-11267.
Buffalo, E.A., Fries, P., Landman, R., Liang, H., and Desimone, R. (2010). A backward progression of
attentional effects in the ventral stream. Proc Natl Acad Sci U S A 107, 361-365.
Buschman, T.J., and Miller, E.K. (2007). Top-down versus bottom-up control of attention in the
prefrontal and posterior parietal cortices. Science 315, 1860-1862.
Cloutman, L.L. (2013). Interaction between dorsal and ventral processing streams: where, when and
how? Brain Lang 127, 251-263.
Colgin, L.L., Denninger, T., Fyhn, M., Hafting, T., Bonnevie, T., Jensen, O., Moser, M.B., and Moser, E.I.
(2009). Frequency of gamma oscillations routes flow of information in the hippocampus. Nature 462,
353-357.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
18
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
Dehaene, S., Duhamel, J.R., Hauser, M.D., and Rizzolatti, G. (2005). From monkey brain to human
brain: A Fyssen foundation symposium (Cambridge, MA: MIT Press).
Desikan, R.S., Ségonne, F., Fischl, B., Quinn, B.T., Dickerson, B.C., Blacker, D., Buckner, R.L., Dale,
A.M., Maguire, R.P., Hyman, B.T., et al. (2006). An automated labeling system for subdividing the
human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage 31, 968-980.
Destrieux, C., Fischl, B., Dale, A., and Halgren, E. (2010). Automatic parcellation of human cortical gyri
and sulci using standard anatomical nomenclature. Neuroimage 53, 1-15.
Dhamala, M., Rangarajan, G., and Ding, M. (2008). Estimating Granger causality from Fourier and
wavelet transforms of time series data. Physical review letters 100, 018701-018701.
Felleman, D.J., and Essen, D.C.V. (1991). Distributed hierarchical processing in the primate cerebral
cortex. Cereb Cortex 1, 1-47.
Fischl, B., van der Kouwe, A., Destrieux, C., Halgren, E., Ségonne, F., Salat, D.H., Busa, E., Seidman,
L.J., Goldstein, J., Kennedy, D., et al. (2004). Automatically parcellating the human cerebral cortex.
Cereb Cortex 14, 11-22.
Fontolan, L., Morillon, B., Liegeois-Chauvel, C., and Giraud, A.-L. (2014). The contribution of
frequency-specific activity to hierarchical information processing in the human auditory cortex. Nat
Commun 5, 4694-4694.
Fries, P. (2005). A mechanism for cognitive dynamics: neuronal communication through neuronal
coherence. Trends Cogn Sci 9, 474-480.
Fries, P. (2015). Rhythms for Cognition: Communication through Coherence. Neuron 88, 220-235.
Geweke, J. (1982). Measurement of linear dependence and feedback between multiple time series.
Journal of the American Statistical Association 77, 304-313.
Glasser, M.F., and Van Essen, D.C. (2011). Mapping human cortical areas in vivo based on myelin
content as revealed by T1- and T2-weighted MRI. J Neurosci 31, 11597-11616.
Granger, C.W.J. (1969). Investigating causal relations by econometric models and cross-spectral
methods. Econometrica: Journal of the Econometric Society, 424-438.
Gregoriou, G.G., Gotts, S.J., Zhou, H., and Desimone, R. (2009). High-frequency, long-range coupling
between prefrontal and visual cortex during attention. Science 324, 1207-1210.
Gregoriou, G.G., Paneri, S., and Sapountzis, P. (2015). Oscillatory synchrony as a mechanism of
attentional processing. Brain Res.
Haegens, S., Cousijn, H., Wallis, G., Harrison, P.J., and Nobre, A.C. (2014). Inter- and intra-individual
variability in alpha peak frequency. Neuroimage 92, 46-55.
Hoogenboom, N., Schoffelen, J.-M., Oostenveld, R., Parkes, L.M., and Fries, P. (2006). Localizing
human visual gamma-band activity in frequency, time and space. Neuroimage 29, 764-773.
Jutai, J.W., Gruzelier, J.H., and Connolly, J.F. (1984). Spectral analysis of the visual evoked potential
(VEP): effects of stimulus luminance. Psychophysiology 21, 665-672.
Lee, J.H., Whittington, M.A., and Kopell, N.J. (2013). Top-down beta rhythms support selective
attention via interlaminar interaction: a model. PLoS Comput Biol 9, e1003164-e1003164.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
19
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
Livingstone, M.S. (1996). Oscillatory firing and interneuronal correlations in squirrel monkey striate
cortex. Journal of neurophysiology 75, 2467-2485.
Maris, E., and Oostenveld, R. (2007). Nonparametric statistical testing of EEG-and MEG-data. Journal
of neuroscience methods 164, 177-190.
Markov, N.T., Vezoli, J., Chameau, P., Falchier, A., Quilodran, R., Huissoud, C., Lamy, C., Misery, P.,
Giroud, P., Ullman, S., et al. (2014). Anatomy of hierarchy: feedforward and feedback pathways in
macaque visual cortex. J Comp Neurol 522, 225-259.
Mitra, P.P., and Pesaran, B. (1999). Analysis of dynamic brain imaging data. Biophysical journal 76,
691-708.
Muckli, L., De Martino, F., Vizioli, L., Petro, L.S., Smith, F.W., Ugurbil, K., Goebel, R., and Yacoub, E.
(2015). Contextual Feedback to Superficial Layers of V1. Curr Biol.
Nolte, G. (2003). The magnetic lead field theorem in the quasi-static approximation and its use for
magnetoencephalography forward calculation in realistic volume conductors. Physics in medicine
and biology 48, 3637-3637.
Olman, C.A., Harel, N., Feinberg, D.A., He, S., Zhang, P., Ugurbil, K., and Yacoub, E. (2012). Layerspecific fMRI reflects different neuronal computations at different depths in human V1. PloS one 7,
e32536.
Oostenveld, R., Fries, P., Maris, E., and Schoffelen, J.-M. (2011). FieldTrip: Open source software for
advanced analysis of MEG, EEG, and invasive electrophysiological data. Comput Intell Neurosci 2011,
156869-156869.
Percival, D.B., and Walden, A.T. (2000). Wavelet Methods for Time Series Analysis (Cambridge Series
in Statistical and Probabilistic Mathematics) (Cambridge University Press).
Plomp, G., Quairiaux, C., Kiss, J.Z., Astolfi, L., and Michel, C.M. (2014). Dynamic connectivity among
cortical layers in local and large-scale sensory processing. Eur J Neurosci 40, 3215-3223.
Reuter, M., Schmansky, N.J., Rosas, H.D., and Fischl, B. (2012). Within-subject template estimation
for unbiased longitudinal image analysis. Neuroimage 61, 1402-1418.
Roe, A.W., Chelazzi, L., Connor, C.E., Conway, B.R., Fujita, I., Gallant, J.L., Lu, H., and Vanduffel, W.
(2012). Toward a unified theory of visual area V4. Neuron 74, 12-29.
Salin, P.A., and Bullier, J. (1995). Corticocortical connections in the visual system: structure and
function. Physiol Rev 75, 107-154.
Schoffelen, J.-M., and Gross, J. (2009). Source connectivity analysis with MEG and EEG. Hum Brain
Mapp 30, 1857-1865.
Schomburg, E.W., Fernández-Ruiz, A., Mizuseki, K., Berényi, A., Anastassiou, C.A., Koch, C., and
Buzsáki, G. (2014). Theta phase segregation of input-specific gamma patterns in entorhinalhippocampal networks. Neuron 84, 470-485.
Sethian, J.A. (1999). Level set methods and fast marching methods: evolving interfaces in
computational geometry, fluid mechanics, computer vision, and materials science, Vol 3 (Cambridge
university press).
Van Essen, D.C. (2005). Corticocortical and thalamocortical information flow in the primate visual
system. Prog Brain Res 149, 173-185.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
20
This is a post-print version of the following article: Alpha-beta and Gamma Rhythms Subserve Feedback and
Feedforward Influences among Human Visual Cortical Areas. Michalareas, G., Vezoli, J., van Pelt, S., Schoffelen,
J.M., Kennedy, H. and Fries, P. Neuron 89, January 20, 2016. http://dx.doi.org/10.1016/j.neuron.2015.12.018.
Van Essen, D.C., Glasser, M.F., Dierker, D.L., Harwell, J., and Coalson, T. (2012). Parcellations and
hemispheric asymmetries of human cerebral cortex analyzed on surface-based atlases. Cereb Cortex
22, 2241-2262.
van Kerkoerle, T., Self, M.W., Dagnino, B., Gariel-Mathis, M.-A., Poort, J., van der Togt, C., and
Roelfsema, P.R. (2014). Alpha and gamma oscillations characterize feedback and feedforward
processing in monkey visual cortex. Proc Natl Acad Sci U S A 111, 14332-14341.
van Pelt, S., Boomsma, D.I., and Fries, P. (2012). Magnetoencephalography in twins reveals a strong
genetic determination of the peak frequency of visually induced gamma-band synchronization. J
Neurosci 32, 3388-3392.
von Stein, A., Chiang, C., and König, P. (2000). Top-down processing mediated by interareal
synchronization. Proceedings of the National Academy of Sciences of the United States of America
97, 14748-14753.
Wang, X.-J. (2010). Neurophysiological and computational principles of cortical rhythms in cognition.
Physiol Rev 90, 1195-1268.
Wang, X., Chen, Y., and Ding, M. (2008). Estimating Granger causality after stimulus onset: a
cautionary note. Neuroimage 41, 767-776.
Womelsdorf, T., Valiante, T.A., Sahin, N.T., Miller, K.J., and Tiesinga, P. (2014). Dynamic circuit motifs
underlying rhythmic gain control, gating and integration. Nat Neurosci 17, 1031-1039.
Xing, D., Yeh, C.-I., Burns, S., and Shapley, R.M. (2012). Laminar analysis of visually evoked activity in
the primary visual cortex. Proc Natl Acad Sci U S A 109, 13871-13876.
Michalareas et al., Feedforward and Feedback Influences Among Human Visual Areas.
21
A
Experimental paradigm
OK
Baseline
fixation
1.5 sec
B
Feedback
0.5 sec
time
Power change spectrum
140
Power change (%)
Change,
Visual
Stimulation Response
0.75-3 sec in < 0.5 sec
0
-40
C
10
30
50
70 90
Frequency(Hz)
110
130
Gamma-power change topography
80
0
-40
Power change (%)
40
-80
D
Time-Frequency analysis
Frequency(Hz)
90
70
50
30
10
-0.2
0
0.2
stimulus
onset
0.4
0.6
0.8
1
Time(sec)
Figure 1
A
Gamma-power change topography
80
0
-40
Power change (%)
40
-80
B
Selected homologous areas
Macaque
Human
injection sites
Atlas Labels
V1
V1
V2
V2
hV4
V4
MT
MT
V7
DP
phPIT(ventral&dorsal)
G_pariet_inf-Angular
C
TEO
7A
Vertices within the selected areas
Figure 2
B
0.6
0.5
0.4
0.3
0.2
Coherence
A
0.1
0
C
D
0.024
0.020
0.012
GC
0.016
0.008
0.004
0
seed parcel
Figure 3
A
B
C
1
20
40
D
120
0.772
0.433
0.772
V2
0.433
0
1
0.321
V4
3.40
0
50
100
freq(Hz)
0.890
0.159
MT
3.14
NA
0.104
DP
4.37
V2
0.942
0.306
MT
V4
NA
0.119
TEO
3.78
V2
0.135
DP
V4
0.291
1
NA
0.000
7A
4.73
50
100
freq(Hz)
0.306
MT
0.946
0.054
TEO
0.364
DP
DP
0.511
MT
0.024
0.662
0.439
TEO
0
0.135
0.802
DP
0.364
0.177
0.333
TEO
0
0.439
TEO
V4
MT
0.486
50
100
freq(Hz)
0.341
MT
0.253
1.000
0.661
7A
50
100
freq(Hz)
0.333
TEO
MT
0
0.802
0.177
7A
1.000
0.661
7A
0
0.341
0.253
DP
7A
0.337
0.232
DP
1.68
1
50
100
freq(Hz)
0
TEO
0
7A
0
7A
0.706
0.000
7A
1.62
0.337
0.232
DP
1.85
1
0.024
1.90
TEO
DP
NA
0
0.424
1.56
1
7A
3.11
0
MT
50
100
freq(Hz)
TEO
V2
0
DP
0.000
0
0.662
2.48
1.47
1
0.054
TEO
2.62
0.838
NA
3.63
V4
0
0.424
V1
0
0.946
2.68
MT
7A
TEO
0
0.498
0
MT
0
V4
50
100
freq(Hz)
0.486
2.15
3.35
1
0.838
0.119
140
7A
3.04
V4
50
100
freq(Hz)
NA
120
5.05
V2
0
V4
7A
V1
DP
3.33
1
60
80
100
Frequency (Hz)
0
0.909
1.61
0
NA
DP
0.104
0
0.615
40
3.89
V2
0
0.498
2.56
V2
0
0.511
0
NA
3.87
V4
0
V2
3.86
1
0.942
0
MT
20
TEO
1.78
V4
0
V1
0.615
1
0
V2
50
100
freq(Hz)
0
140
4.86
2.90
3.27
3.11
0
V1
0
0.909
0
V1
MT
2.46
3.73
0
0.159
2.40
0
V1
0
0.291
0
V4
2.52
0
7A
0.936
0.890
0
0.936
V2
0
V1
TEO
V2
120
DP
3.75
2.40
0
DP
V1
2.84
0.981
60
80
100
Frequency (Hz)
V4
0.321
3.99
2.94
V1
MT
40
MT
0.981
V1
3.37
V2
20
V4
0
V4
1
Granger Causality
Feedforward(FF)
Feedback(FB)
Random FF
Random FB
Significant Frequencies
FF vs Random FF
FB vs Random FB
FF vs FB
RIGHT HEMISPHERE
V1
GC axes :
x10-²
V1
0
140
V2
V1
V1
V2
60
80
100
Frequency (Hz)
3.32
x10-²
Granger Causality
Feedforward(FF)
Feedback(FB)
Random FF
Random FB
Significant Frequencies
FF vs Random FF
FB vs Random FB
FF vs FB
Granger Causality
Granger Causality
0
3.08
x10-²
Granger Causality
Actual : Left
Actual : Right
Random : Left
Random : Right
Significant Frequencies
Actual vs Random
Left
Right
Granger Causality
2.33
x10-²
GC over FF and FB connections - Right Hemisphere
GC over FF and FB connections - Left Hemisphere
GC over all connections
TEO
0.706
0.000
7A
1
50
100
freq(Hz)
1.82
1
50
100
freq(Hz)
0
1
50
100
freq(Hz)
LEFT HEMISPHERE
Lower Area A
(accord. to SLN)
Feed-forward SLN
Feed-back SLN
Higher Area B
(accord. to SLN)
spectrum of GC from area A to area B
spectrum of GC from area B to area A
spectrum of RANDOM GC from area A to area B
spectrum of RANDOM GC from area B to area A
frequencies of significant GC from A to B
w.r.t corresponding RANDOM GC.
frequencies of significant GC from B to A
w.r.t corresponding RANDOM GC.
frequencies of significant difference between GC from A to B and
GC from B to A
Figure 4
A
B
C
GC over all connections
α/β +5 −20
peak
D
+20
γ
peak
0.772
0.433
0.772
0.981
V2
V1
3.22
0
V2
0.321
V4
3.51
0.890
0.159
MT
3.86
NA
V2
0.104
DP
5.14
0.942
0.306
MT
V4
NA
0.119
TEO
3.90
0.909
0.135
DP
V4
0.946
0.054
TEO
NA
0.000
7A
5.43
−5 αβ +5
0
0.364
DP
0.511
MT
NA
0.024
−5 αβ +5
DP
0.662
0.439
TEO
0.486
0
MT
0.802
0.177
7A
−5 αβ +5
MT
0.486
0.333
TEO
TEO
0.253
0
DP
1.000
0.661
7A
−5 αβ +5
7A
0
−5 αβ +5
7A
V4
0.802
0.177
7A
0.333
TEO
MT
1.000
0.661
7A
1.96
0
0.341
0.253
DP
7A
0.337
0.232
DP
1.61
−5 αβ +5
0
γ +20 +60
freq(Hz)
TEO
0.706
0.000
7A
1.84
0.337
0.232
DP
TEO
1.86
γ +20 +60
freq(Hz)
0.024
0
0.424
MT
0
γ +20 +60
freq(Hz)
TEO
0.341
NA
3.55
2.29
1.38
0
0.439
TEO
1.69
−5 αβ +5
7A
0
0.662
0
0.838
0.000
4.01
2.52
0.424
V2
0
MT
γ +20 +60
freq(Hz)
0.054
TEO
2.72
DP
MT
2.29
4.14
0
0.838
0
V4
γ +20 +60
freq(Hz)
0.364
NA
0
0.946
V4
0
V4
7A
DP
0
γ +20 +60
freq(Hz)
V1
0
0.498
V4
3.50
−5 αβ +5
TEO
3.42
1.59
2.90
4.45
γ +20 +60
freq(Hz)
0.498
0
V2
0.135
+60
5.38
V2
1.74
0
V2
0
V1
0.511
0
MT
3.68
3.27
0
0.615
DP
0
0.615
V4
0.119
+40
Frequency (Hz)
0
0.909
4.05
0
γ +20 +60
freq(Hz)
+20
7A
4.17
V2
0
V2
0
V1
0.306
NA
V1
0
MT
2.97
4.30
0
0
−5 αβ +5
DP
0.104
4.89
0.942
γ
peak
TEO
NA
V1
2.47
0
V1
7A
0.291
0
V4
2.97
0
TEO
0.936
MT
3.12
0
V1
DP
V2
0.159
V2
2.52
0
MT
0.291
V4
α/β +5 −20
peak
DP
0
0.936
0
−5
+60
Frequency (Hz)
0.890
V1
3.21
0.981
+40
3.84
V2
0
V4
V4
0
γ +20 +60
freq(Hz)
3.14
V1
0.321
4.39
−5 αβ +5
+20
MT
V4
0.433
γ
peak
Granger Causality
Feedforward(FF)
Feedback(FB)
Random FF
Random FB
Significant Frequencies
FF vs Random FF
FB vs Random FB
FF vs FB
x
RIGHT HEMISPHERE
V1
α/β +5 −20
peak
V2
V1
GC axes :
x10-²
0
−5
+60
Frequency (Hz)
V1
V1
+40
3.60
10-²
Granger Causality
Feedforward(FF)
Feedback(FB)
Random FF
Random FB
Significant Frequencies
FF vs Random FF
FB vs Random FB
FF vs FB
Granger Causality
Granger Causality
0
−5
V2
3.46
x10-²
Granger Causality
Actual : Left
Actual : Right
Random : Left
Random : Right
Significant Frequencies
Actual vs Random
Left
Right
GC over FF and FB connections - Right hemisphere
Granger Causality
2.50
x10-²
GC over FF and FB connections - Left hemisphere
0.706
0.000
0
7A
−5 αβ +5
γ +20 +60
freq(Hz)
1.86
γ +20 +60
freq(Hz)
0
−5 αβ +5
γ +20 +60
freq(Hz)
LEFT HEMISPHERE
Lower Area A
(accord. to SLN)
Feed-forward SLN
Feed-back SLN
Higher Area B
(accord. to SLN)
frequencies of significant GC from A to B
w.r.t corresponding RANDOM GC.
frequencies of significant GC from B to A
w.r.t corresponding RANDOM GC.
spectrum of GC from area A to area B
spectrum of GC from area B to area A
spectrum of RANDOM GC from area A to area B
spectrum of RANDOM GC from area B to area A
frequencies of significant difference between GC from A to B and
GC from B to A
αβ
γ
peak frequency in the alpha-beta range
peak frequency in the gamma range
Figure 5
A
B
DAI-SLN correlation - Left Hemisphere
0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
DAI-SLN correlation - Right Hemisphere
0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.226
−5 α/β +5−20
peak
γ
peak
+20
+40
+60
Frequency (Hz)
pos.cluster
neg.cluster
0.467
0.5
Correlation Coefficient
Correlation Coefficient
pos.cluster
neg.cluster
0.465
0.5
-0.178
−5 α/β +5−20
peak
γ
peak
+20
+40
+60
Frequency (Hz)
Figure 6
A
DAI from Gamma band
Functional hierarchy
level
Left hemisphere
Right hemisphere
7
6
5
4
3
V2
2
1
TEO
MT
V4
V1
1
2
Ventral
Dorsal
7A
DP
4
5
7
8
9
Anatomical hierarchy level
B
V4
V2
6
DP
MT
V1
[R=0.962, P=0.0005]
3
7A
TEO
1
[R=0.945, P=0.0013]
2
3
4
5
6
7
8
9
Anatomical hierarchy level
Functional hierarchy
level
DAI from Alpha-Beta band
Left hemisphere
7
6
TEO
3 V1 V2
V1
V4
2
TEO
V2
V4
[R=0.750, P=0.0320]
1
2
3
4
5
6
7
8
9
Anatomical hierarchy level
7A
DP
MT
MT
4
C
7A
DP
5
1
Right hemisphere
Ventral
Dorsal
[R=0.812, P=0.0267]
1
2
3
4
5
6
7
8
9
Anatomical hierarchy level
Functional hierarchy
level
DAI from combined Gamma and Alpha-Beta bands
Left hemisphere
7
6
5
V2
3
1
TEO
V4
V1
2
7A
DP
MT
TEO
4
Right hemisphere
Ventral
Dorsal
V1
[R=0.901, P=0.0057]
1
2
3
4
5
6
7
8
9
Anatomical hierarchy level
1
2
V2
7A
DP
MT
V4
[R=0.923, P=0.0030]
3
4
5
6
7
8
9
Anatomical hierarchy level
Figure 7
A
Hierarchy from DAI in Gamma band
Average Rank in Hierarchy
26
Ventral stream
Dorsal stream
ar.46
7A
21
IPS4
ER
IPS2
FST
16
V3C
11
6
V2
LO1
V3A
V3
V7
FEF
PITd
MT
V4t
IPS1
IPS3
MST
LO2
V4
V3B
VO1
PITv
V3D
V1
1
Rank
B
Hierarchy from DAI in Beta band
Average Rank in Hierarchy
26
Ventral stream
Dorsal stream
Hierarchical level in (A)
21
IPS4
MST
FEF
IPS2
ER
ar.46
V7
IPS3
V3D
V3C
16
IPS1
FST
LO2
11
V1
LO1
V2
6
V3
PITv
MT
V4t
V3B
V3A
VO1
PITd
V4
1
Rank based on DAI in gamma band (A)
C
Hierarchy from combined DAI in Gamma and Alpha/Beta bands
26
Average Rank in Hierarchy
7A
Ventral stream
Dorsal stream
Hierarchical level in (A)
21
ar.46
MST
FEF
16
LO2
11
V3B
6
V1
1
V2
V3
LO1
V4t
MT
V7
IPS4
V3D
V3C
IPS3
IPS1
IPS2
ER
7A
FST
PITv
VO1
PITd
V3A V4
Rank based on DAI in gamma band (A)
Figure 8
1
Supplemental Data
Figure S1. Distributions of individual peak frequencies in Gamma and Alpha-Beta bands
For each subject, Gaussian functions were fit to the GC spectrum averaged over all area pairs of both
hemispheres, separately for the frequency ranges 4-20 Hz (A, C) and 30-100 Hz (B, D). (A) Histogram
of peak frequencies across the 43 subjects in the 4-20 Hz range. (B) Histogram of goodness-of-fit
measure (r-square) of the Gaussian function in the 4-20 Hz range across the 43 subjects. (C) Same as
(A), but for 30-100 Hz range. (D) Same as (B), but for 30-100 Hz range. Notice that the goodness-offit distribution is heavily skewed towards 1, indicating good fitting results.
Figure S2. DAI-SLN correlation spectra versus randomized surrogate data and without
peak-frequency alignment.
In order to test whether there was any bias in the DAI-SLN correlation presented in Fig. 6, epochs
were shuffled prior to computing GC and the subsequent DAI-SLN correlation. In (A) and (B), for left
and right hemispheres respectively, this randomized surrogate correlation is depicted by the yellow
line. Significance was again assessed with the same non-parametric permutation based tests. Very
similar positive and negative frequency clusters were found in the gamma and alpha-beta ranges
respectively. (C,D) DAI-SLN correlation spectra without peak-frequency alignment. For these
analyses, the spectrum of each subject has not been aligned relative to the individual peaks in the
gamma and alpha-beta bands, as was done in Figure 6. Rather the DAI-SLN correlation spectra are
averaged over subjects without alignment. Significance is assessed relative to zero correlation as in
Fig. 6. (E,F) DAI-SLN correlation without peak-frequency alignment as in (C,D), and testing against
random correlation as in (A, B). Note that without peak-frequency alignment, peak correlation
values are slightly lower.
Figure S3. Granger-Causality-based hierarchy versus anatomical hierarchy without peakfrequency alignment.
This figure presents the relation of the human functional hierarchy derived for DAI values versus the
macaque anatomical hierarchy from SLN values, similar to Figure 7, but without alignment of
individual peak frequencies. In the same fashion as in Figure 7, three cases were investigated,
namely with the functional hierarchy computed from (A) DAI in the gamma band, (B) DAI in the
alpha-beta band and (C) the combined DAI in alpha-beta and gamma bands. Early visual areas are
depicted in black, ventral-stream areas in green, and dorsal-stream areas in red. Inset brackets on
bottom right of each subplot report Pearson correlation between human functional and macaque
anatomical hierarchical levels. Results were qualitatively the same as in Fig. 7.
Table S1. Number of dipoles within each of the 26 visual areas in the left and right
hemispheres of the MEG source model.
A
B
Hist. of GC Peak frequency in gamma range
10
15
N subjects
N subjects
8
6
4
0
52
C
54
56
58
60
62
64
Peak frequency(Hz)
66
68
0
70
0
D
Hist. of GC Peak frequency in alpha/beta range
20
0.2
0.4
R−square
0.6
0.8
1
Hist. of Goodness-of-fit of Gaussian in alpha/beta range
25
20
N subjects
15
N subjects
10
5
2
10
5
0
Hist. of Goodness-of-fit of Gaussian in gamma range
20
15
10
5
6
8
10
12
14
Peak frequency(Hz)
16
18
20
0
0
0.2
0.4
R−square
0.6
0.8
1
Supplemental Figure 1
Correlation between DAI and SLN
Left Hemisphere
B
Aligned Spectrum - Test vs. Random
actual
random
pos.cluster
neg.cluster
0.465
0.5
0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.226
-0.3
−5 α/β +5−20
peak
γ
peak
+20
+40
Correlation Coefficient
Correlation Coefficient
A
Right Hemisphere
0.3
0.2
0.1
0
-0.1
-0.2
-0.178
−5 α/β +5−20
peak
Frequency (Hz)
0.5
+20
+40
+60
Frequency (Hz)
0.410
0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
Non-Aligned Spectrum - Test vs. Zero
0.5
pos.cluster
neg.cluster
Correlatino Coefficient
Correlation Coefficient
γ
peak
D
Non-Aligned Spectrum - Test vs. Zero
20
pos.cluster
neg.cluster
0.421
0.4
0.3
0.2
0.1
0
-0.1
-0.2 -0.166
-0.193
1
40
60
80
100 120 140
Frequency (Hz)
E
1
20
40
60
80
100 120 140
Frequency (Hz)
F
Non-Aligned Spectrum - Test vs. Random
actual
random
pos.cluster
neg.cluster
0.410
0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
20
actual
random
pos.cluster
neg.cluster
0.421
0.4
0.3
0.2
0.1
0
-0.1
-0.2 -0.166
-0.193
1
Non-Aligned Spectrum - Test vs. Random
0.5
Correlation Coefficient
0.5
Correlation Coefficient
actual
random
pos.cluster
neg.cluster
0.467
0.4
+60
C
Aligned Spectrum - Test vs. Random
0.5
40
60
80
100 120 140
Frequency (Hz)
1
20
40
60
80
100 120 140
Frequency (Hz)
Supplemental Figure 2
A
DAI from Gamma band
Left hemisphere
Functional hierarchy
level
7
6
4
DP
2
1
V2
V1
[R=0.975, P=0.0002]
2
DP
MT
V4
V2
V1
7A
TEO
V4
3
1
Ventral
Dorsal
7A
TEO
MT
5
Right hemisphere
3
4
5
6
7
8
9
Anatomical hierarchy level
B
1
[R=0.952, P=0.0010]
2
3
4
5
6
7
8
9
Anatomical hierarchy level
DAI from Alpha-Beta band
Left hemisphere
Functional hierarchy
level
7
6
7A
DP
5
4
MT
V1
3
V2
V1 V2
TEO
V4
[R=0.693, P=0.0414]
1
C
2
3
4
5
6
7
8
9
Anatomical hierarchy level
7A
DP
MT
V4
2
1
Right hemisphere
Ventral
Dorsal
TEO
[R=0.770, P=0.0328]
1
2
3
4
5
6
7
8
9
Anatomical hierarchy level
Functional hierarchy
level
DAI from combined Gamma and Alpha-Beta bands
Left hemisphere
7
6
7A
DP
5
MT
4
3
2 V1
1
Right hemisphere
V2
Ventral
Dorsal
MT
TEO
V4
V1
V2
2
3
4
5
6
7
8
9
Anatomical hierarchy level
TEO
V4
[R=0.933, P=0.0020]
1
7A
DP
[R=0.930, P=0.0027]
1
2
3
4
5
6
7
8
9
Anatomical hierarchy level
Supplemental Figure 3
Area Name
V1
V2
V3
V3A
V3B
V3C
V3D
V4
V4t
LO1
LO2
PITd
PITv
MT
VO1
MST
FST
ER
V7
IPS1
IPS2
IPS3
IPS4
7A
FEF
N dipoles in Le Hemisphere
31
40
49
9
12
11
8
9
2
14
11
3
3
3
11
2
2
22
5
5
6
15
12
56
24
N dipoles in Right Hemisphere
28
38
44
10
8
11
8
10
3
15
9
2
2
8
7
2
2
22
6
5
6
15
12
58
20
Supplemental Table 1
2
Supplemental Experimental Procedures
Data Acquisition
MEG: Recordings were performed with a whole-head MEG system (CTF Systems Inc.) comprising 275
axial gradiometers, located at the Donders Institute of Radboud University Nijmegen (Nijmegen,
Netherlands). In addition, horizontal and vertical Electrooculograms (EOG) were recorded using
bipolar electrodes for off-line eye-movement artifact rejection. All signals were low-pass filtered at
300 Hz, sampled at 1200 Hz, and stored. Head position relative to the gradiometer array was
determined prior and after the MEG recording using 3 coils attached to subject’s nasion, left and
right ear canals.
MRI: Structural MR images were acquired using an Avanto 1.5 T whole body MRI scanner (Siemens,
Germany). A volume head coil was used for RF transmission and signal reception. The scan was
performed with a 3D MPRAGE sequence with 1 mm isotropic resolution.
Subjects
Datasets of 43 subjects were selected from a pool of 160 subjects, which had been recorded for
genotyping of parameters derived from visually induced gamma-band activity. The genotyping study
required a large number of subjects and was fully focused on gamma, and it therefore accepted
subjects, in which sub-gamma activity was confounded by artifacts. For the present study, subjects
were excluded if 1) their body carried ferromagnetic elements causing low-frequency MEG artifacts
(e.g. metallic orthodontic retainers), 2) they did not have normal vision, 3) they used medication,
4) they exhibited head movements larger than 10 mm during the experiment. This left 43 subjects
(36 right-handed) with an average age of 23.1 (range: 18 to 34). All subjects gave written informed
consent according to guidelines of the local ethics committee (Commissie Mensgebonden
Onderzoek Regio Arnhem-Nijmegen, The Netherlands).
Human Brain Model
For source analysis, structural MR images were acquired for each individual subject, and the cortical
mantle was extracted with the FreeSurfer longitudinal stream pipeline (Reuter et al., 2012). Cortical
parcellation and area definition used a recent visuotopic atlas provided by (Abdollahi et al., 2014),
the FreeSurfer software suite (http://FreeSurfer.net/) (Fischl et al., 2004; Desikan et al., 2006;
Destrieux et al., 2010) and the Caret software suite (http://brainvis.wustl.edu/) (Glasser and Van
Essen, 2011; Van Essen et al., 2012).
The cortical sheet extracted from FreeSurfer was transformed and registered into the Caret
representation through the Caret pipeline “freesurfer_to_fs_LR”. In Caret, the Conte69 atlas is
defined and contains a composite parcellation (Van Essen et al., 2012) which includes multiple
visuotopic areas in the occipital, parietal and temporal lobe, which have been mapped based on a
number of studies. The recent visuotopic atlas provided by (Abdollahi et al., 2014) is based on
Maximum Probability Maps (MPMs) for 18 human visual areas, that have been built based on recent
Multimodal Surface Matching (MSM) algorithms for inter-subject registration, which utilize as
registration features not only structural cortical characteristics but also retinotopic activation maps.
The resulting parcellation provides a more refined cortical representation of these 18 cortical areas
as compared to previous studies. This parcellation atlas is represented on the fs_LR cortical surface
3
template, similarly to the Conte69 visuotopic atlas. Hereafter, for convenience, this parcellation will
be referred to in this text as the “Abdollahi2014” parcellation. Finally, one of the parcellation
schemes of FreeSurfer, “aparc.a2009s” (Destrieux et al., 2010), provides parcels segregating all the
main gyri and sulci of the human cortex and was used in addition to the visuotopic Conte69 and the
parcellations from Caret, as explained in more detail below.
Both, the FreeSurfer and Caret cortical sheet representations contain more than 105 vertices per
hemisphere and were subsampled for MEG source analysis with the MNE software suite (Gramfort
et al., 2014). Through recursive icosahedron subdivision, this software subsamples the dense
FreeSurfer cortical representation by a fixed number of 4098 vertices per hemisphere. Those
vertices were mapped to their nearest neighbors in the Caret common template surface
representation and thereby to the Conte69 atlas. The vertex positions were then transformed into
the head coordinate system, where the MEG measurements were acquired and the inverse solution
was computed.
Selection of 7 homologous brain areas between human and macaque
V1, V2: In both macaque and human brains, V1 and V2 are the largest cortical areas and have the
clearest homology between these species (Van Essen, 2005). They are both available in the
Abdollahi2014 atlas labelled as V1 and V2 respectively.
MT: In the macaque, area MT is located on the posterior bank of the superior temporal sulcus, is
characterized by its distinct heavy myelination, and contains a complete topographic representation
of the contralateral visual field (Van Essen et al., 1981). In the human brain, area MT has been
consistently mapped at a corresponding location, has a similar visuotopic organization and is
characterized by its motion-selective activation (Huk et al., 2002; Fischl et al., 2008; Kolster et al.,
2010). The area used here is area MT/V5 in the so called MT/V5+ complex identified in the human
brain (Kolster et al., 2010). This area is available in the Abdollahi2014 atlas, labelled as MT.
V4 (hV4): In the macaque, area V4 has been mapped both physiologically and with functional MRI. It
consists of 2 parts, a dorsal lower field representation and a ventral upper-field representation. The
horizontal meridian is represented anterior to both ventral and dorsal V4 in the macaque brain (Fize
et al., 2003; Janssens et al., 2014; Kolster et al., 2014). In the human brain the strongest evidence of
homology comes mainly from the ventral part of human V4, which has been mapped by various
studies in a corresponding location and with similar visuotopic organization (Brewer et al., 2005; Van
Essen, 2005; Swisher et al., 2007; Kolster et al., 2010). This area is available in the Abdollahi2014
atlas, labelled as hV4.
DP (V7): In the macaque brain, visual area DP occupies the dorsal aspect of the prelunate gyrus
immediately anterior to area V3A. In the human brain, an area located also just anteriorly to area
V3A (as defined in (Tootell et al., 1998) or V3D as defined in (Georgieva et al., 2009) and (Abdollahi
et al., 2014)) with a putative homology in its retinotopic representation has been identified by
(Tootell et al., 1998) and given the name V7. This area is available in the Abdollahi2014 atlas,
labelled as V7.
TEO (PIT): In the macaque brain, according to (Felleman and Essen, 1991), the posterior part of the
inferotemporal cortex is divided into two regions labelled as PITd (dorsal) and PITv (ventral). These
areas coincide largely with the architectonic Temporal Occipital area (TEO). In the human brain, the
putative homologous areas have been mapped using functional MRI, have been shown to be
4
involved in processing two-dimensional shape, and they have been given the same names as in the
macaque (Kolster et al., 2010). These areas are available in the Abdollahi2014 atlas labelled as
phPITd and phPITv. Here, these ventral and dorsal parcels have been collapsed into a single parcel
called PIT.
7A (Gyrus parietalis inferior – Angular gyrus): In the macaque brain, area 7A occupies the caudal part
of the Inferior Parietal Lobule (IPL) and extends only a short distance into the intraparietal sulcus
(Felleman and Essen, 1991). In an alternative, widely used parcellation scheme of the macaque IPL
by (Pandya and Seltzer, 1982) this area was attributed to 2 areas of the caudal IPL, namely the areas
labelled as PG and Opt. In the human brain, (Caspers et al., 2011) identified 2 homologous areas
through diffusion-weighted MRI, spanning the caudal part of human IPL and showing strong
resemblance in their connectivity patterns with that of the macaque. In the human visuotopic
parcellation scheme available in the Conte69 atlas, no parcel covers the caudal part of IPL, where
this homology has been identified. In FreeSurfer’s “aparc.a2009s” parcellation scheme (Destrieux et
al., 2010) the caudal IPL is covered by the parcel labelled as “G_pariet_inf-Angular”, which primarily
encapsulates the angular gyrus and is segregated from rostral IPL by the sulcus intermedius premus.
Dorsally it extends into the intrapariatal sulcus (IPS). Based on the above anatomical characteristics,
this parcel was selected as homologous to macaque area 7A.
As mentioned above, the dense cortical surface representation of the common template “fs_LR”
surface was down-sampled for the MEG source analysis. Accordingly, the representation of the
selected areas was defined in this subsampled space.
Details on the Selection of 26 visual areas from the human brain
V1, V2, V4, MT, V7, 7A: These areas are the same that were used in the 7-area analysis (see above
and Experimental Procedures).FEF: The Frontal Eye Field is a cortical area located in humans around
the intersection of the pre-central sulcus with the dorsal portion of the superior frontal sulcus. It is
considered a “hub”, having anatomical connections with visual, parietal, temporal and prefrontal
areas. Functionally it is involved in oculomotor behavior, visuo-spatial attention, visual awareness
and perceptual modulation (Vernet et al., 2014). In the visuotopic parcellation of the Conte69 atlas,
an FEF parcel is not available. The most relevant parcel that is available is Brodmann area 8, in which
FEF occupies only a relatively small portion. In order to represent FEF, only the ventral branch of the
superior precentral sulcus and its intersection with the superior frontal sulcus were manually
selected as the parcel that represents FEF (Amiez and Petrides, 2009).
The following areas are available in the atlas dataset of (Abdollahi et al., 2014) which in the following
text will be referred to as “Abdollahi2014”:
V3: Area V3, in both human and monkey, consists of a lower and an upper field representation,
neighboring area V2 dorsally and ventrally respectively. Although these dorsal and ventral parts have
been shown to have some asymmetries in their function, architecture and connectivity, they are
considered in both human and monkey as the constituents of a single retinotopic area (Van Essen,
2005).
V3A, V3B, V3C, V3D: These areas that span the cortical surface adjoined dorsally to area V3 were
identified by (Georgieva et al., 2009). Therein, they are jointly referred to as the “V3A complex” and
they represent the same cortical areas previously referred to as V3A and V3B (Wandell et al., 2007).
The dorsal areas of the complex, V3C and V3D, adjoin ventrally area V7. These areas contain neurons
5
with larger retinotopic fields than areas V1, V2 and V3. These areas have been implicated in motion
processing, with areas V3A and V3B (corresponding to area V3B in (Fischer, 2011)) having a higher
preference for externally induced retinal motion, and areas V3C and V3D (corresponding to area V3A
in (Fischer, 2011)) having a higher preference for real motion.
LO1, LO2: Lateral Occipital Areas 1 and 2. These areas are located in the fundus of the lateral
occipital sulcus between dorsal V3 and MT, with LO2 anterior to LO1. The retinotopy of these areas
is less precise and more variable than early visual areas. It has been suggested that these two areas
represent shape information, with LO1 extracting boundary information and LO2 extracting regions
and representing shape (Larsson and Heeger, 2006; Wandell et al., 2007).
PITv, PITd: These two areas are the constituents of area PIT used in the 7-area analysis. In the 26area analysis PIT is replaced by these two areas.
MT, pMSTv, pFST, pV4t: These four areas comprise the human middle temporal MT/V5+ complex,
homologous to the MT/V5 complex in the monkey brain. This complex consists of the middle
temporal area (MT/V5), the putative ventral part of the medial superior temporal area (pMSTv), the
putative fundus of the superior temporal area (pFST) and the putative V4 transitional zone. Area V4t
is adjoined ventrally to MT/V5 and dorsally to phPITd. Area pMSTv is located rostrally to area
MT/V5. Area pFST is located anteriorly to area V4t and ventrally to area pMSTv. All four areas
respond strongly to motion, have much larger receptive fields than V1 and are considered parts of
the dorsal stream. Within the MT/V5+ complex, area pMSTv has the largest receptive fields. The
more ventral areas of the complex, V4t and pFST, respond stronger to shapes, while the more dorsal
areas, MT/V5 and pMSTv respond stronger to 3D structure from motion (Kolster et al., 2010). Area
MT here is the same as the one used in the 7-area analysis.
VO1: Area VO1 is part of Ventral Occipital areas(VO) located on the fusiform gyrus anteriorly of area
hV4 and is considered part of the ventral stream. It has a preference for foveal and parafoveal
eccentricities. Due to the fact that VO areas show little object selectivity, it has been hypothesized
that they play an intermediate visual processing role between early visual cortex and higher order
object-selective cortex (Brewer et al., 2005; Arcaro et al., 2009; Kolster et al., 2010).
The following areas are available in the visuotopic parcellation of the Conte69 atlas:
IPS1, IPS2, IPS3, IPS4: These four parcels are subdivisions of the posterior medial bank of the intrapariatel sulcus(IPS) with distinct retinotopic maps, numbered from posterior to anterior locations.
Numerous functions related to visual processing such as visual attention, saccadic preparation and
multisensory integration have been shown to activate these areas (Swisher et al., 2007).
ER: The entorhinal cortex, located in the medial temporal lobe, comprises of a set of cortical areas
that are known to be a major gateway to and from the hippocampus. Based on its anatomical
connectivity and functional properties, it has been categorized as multimodal association cortex.
Although it is not retinotopically organized, it has strong connections with some visual areas
(Felleman and Essen, 1991) and it has recently been shown to play an important role in spatial
navigation (Moser et al., 2008).
Area 46: Brodmann area 46 corresponds with the dorsolateral prefrontal cortex in and around the
middle third of the principal sulcus. It has strong connections mainly with parietal and temporal
cortex (Petrides and Pandya, 1999). There has been strong evidence that functionally it is involved in
maintaining and manipulating information in working memory (Krawczyk, 2002).
6
Retrograde tracing database
Description of the anatomical dataset acquisition and analysis has been reported in (Markov et al.,
2014). The values that we used correspond to multiple injections each into V1, V2, V4 and single
injections into areas MT, DP, TEO and 7A. Only SLN values based on at least 10 labeled neurons were
included. Three projections had fewer than 10 neurons labeled, and their SLN was therefore
considered unreliably quantified. These projections were V1-to-DP, V1-to-7A and V2-to-7A. The V1to-TEO projection was reported as missing i.e. no labeled neurons found in V1 after TEO injection.
For these projections, there is strong evidence in the literature that they are present and all have a
feedforward characteristic, and this was used in place of SLN evidence. Updates, atlases and
additional information concerning the anatomical dataset that was used for this work is available at
www.core-nets.org.
Inverse solution
Subject-wise structural MRI scans were used to derive individual single-shell volume conductor
models (Nolte, 2003). The inverse solution was then performed using Linearly Constrained Minimum
Variance (LCMV) adaptive spatial filtering, often referred to as “Beamforming” (Van Veen et al.,
1997). The beamformer gives, for each of a set of predefined locations, a coefficient matrix such that
the multiplication of the matrix with the sensor-level signals provides an estimate of the dipole
moment at that location, while minimizing influences of other, linearly uncorrelated sources. We
used this approach to scan a set of dipole locations on the cortical mesh. As the beamformer output
is simply a linear projection of sensor-level data, it does not only apply to the raw signal of the
different MEG sensors but also to any linear decomposition, like the Fourier transform of the raw
signal.
There are beamformer methods that derive spatial filters for each frequency separately (Gross et al.,
2001). They are mostly suited to provide source estimates for selected frequency bands. The current
work investigates the entire spectrum, and focuses on GC, whose estimation entails multiple
neighboring frequencies. Therefore, we opted to estimate the source spectrum by means of the
LCMV broadband inverse solution. The LCMV solution is based on the broadband covariance matrix
of the non-averaged MEG sensor data. When determining the LCMV solution, we used a
regularization parameter of 20% of the covariance matrix.
The set of brain locations for which the inverse solution was computed comprised of all the points
distributed within the human visual cortical areas of interest in each hemisphere.
Supplemental References
Abdollahi RO, Kolster H, Glasser MF, Robinson EC, Coalson TS, Dierker D, Jenkinson M, Van Essen DC,
Orban GA (2014) Correspondences between retinotopic areas and myelin maps in human
visual cortex. Neuroimage 99:509-524.
Amiez C, Petrides M (2009) Anatomical organization of the eye fields in the human and non-human
primate frontal cortex. Prog Neurobiol 89:220-230.
Arcaro MJ, McMains SA, Singer BD, Kastner S (2009) Retinotopic organization of human ventral
visual cortex. J Neurosci 29:10638-10652.
Brewer AA, Liu J, Wade AR, Wandell BA (2005) Visual field maps and stimulus selectivity in human
ventral occipital cortex. Nat Neurosci 8:1102-1109.
7
Caspers S, Eickhoff SB, Rick T, von Kapri A, Kuhlen T, Huang R, Shah NJ, Zilles K (2011) Probabilistic
fibre tract analysis of cytoarchitectonically defined human inferior parietal lobule areas
reveals similarities to macaques. Neuroimage 58:362-380.
Desikan RS, Ségonne F, Fischl B, Quinn BT, Dickerson BC, Blacker D, Buckner RL, Dale AM, Maguire
RP, Hyman BT, Albert MS, Killiany RJ (2006) An automated labeling system for subdividing
the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage
31:968-980.
Destrieux C, Fischl B, Dale A, Halgren E (2010) Automatic parcellation of human cortical gyri and sulci
using standard anatomical nomenclature. Neuroimage 53:1-15.
Felleman DJ, Essen DCV (1991) Distributed hierarchical processing in the primate cerebral cortex.
Cereb Cortex 1:1-47.
Fischer E (2011) Visual motion and self-motion processing in the human brain, MPI Series in
Biological Cybernetics, Bd. 31: Logos Verlag Berlin GmbH.
Fischl B, Rajendran N, Busa E, Augustinack J, Hinds O, Yeo BTT, Mohlberg H, Amunts K, Zilles K (2008)
Cortical folding patterns and predicting cytoarchitecture. Cereb Cortex 18:1973-1980.
Fischl B, van der Kouwe A, Destrieux C, Halgren E, Ségonne F, Salat DH, Busa E, Seidman LJ, Goldstein
J, Kennedy D, Caviness V, Makris N, Rosen B, Dale AM (2004) Automatically parcellating the
human cerebral cortex. Cereb Cortex 14:11-22.
Fize D, Vanduffel W, Nelissen K, Denys K, Chef d'Hotel C, Faugeras O, Orban GA (2003) The
retinotopic organization of primate dorsal V4 and surrounding areas: A functional magnetic
resonance imaging study in awake monkeys. J Neurosci 23:7395-7406.
Georgieva S, Peeters R, Kolster H, Todd JT, Orban GA (2009) The processing of three-dimensional
shape from disparity in the human brain. J Neurosci 29:727-742.
Glasser MF, Van Essen DC (2011) Mapping human cortical areas in vivo based on myelin content as
revealed by T1- and T2-weighted MRI. J Neurosci 31:11597-11616.
Gramfort A, Luessi M, Larson E, Engemann DA, Strohmeier D, Brodbeck C, Parkkonen L, Hämäläinen
MS (2014) MNE software for processing MEG and EEG data. NeuroImage 86:446-460.
Gross J, Kujala J, Hämäläinen M, Timmermann L, Schnitzler A, Salmelin R (2001) Dynamic imaging of
coherent sources: studying neural interactions in the human brain. Proceedings of the
National Academy of Sciences 98:694-699.
Huk AC, Dougherty RF, Heeger DJ (2002) Retinotopy and functional subdivision of human areas MT
and MST. J Neurosci 22:7195-7205.
Janssens T, Zhu Q, Popivanov ID, Vanduffel W (2014) Probabilistic and single-subject retinotopic
maps reveal the topographic organization of face patches in the macaque cortex. J Neurosci
34:10156-10167.
Kolster H, Peeters R, Orban GA (2010) The retinotopic organization of the human middle temporal
area MT/V5 and its cortical neighbors. J Neurosci 30:9801-9820.
Kolster H, Janssens T, Orban GA, Vanduffel W (2014) The retinotopic organization of macaque
occipitotemporal cortex anterior to V4 and caudoventral to the middle temporal (MT)
cluster. J Neurosci 34:10168-10191.
Krawczyk DC (2002) Contributions of the prefrontal cortex to the neural basis of human decision
making. Neurosci Biobehav Rev 26:631-664.
Larsson J, Heeger DJ (2006) Two retinotopic visual areas in human lateral occipital cortex. J Neurosci
26:13128-13142.
Markov NT, Vezoli J, Chameau P, Falchier A, Quilodran R, Huissoud C, Lamy C, Misery P, Giroud P,
Ullman S, Barone P, Dehay C, Knoblauch K, Kennedy H (2014) Anatomy of hierarchy:
feedforward and feedback pathways in macaque visual cortex. J Comp Neurol 522:225-259.
Moser EI, Kropff E, Moser M-B (2008) Place cells, grid cells, and the brain's spatial representation
system. Annu Rev Neurosci 31:69-89.
8
Nolte G (2003) The magnetic lead field theorem in the quasi-static approximation and its use for
magnetoencephalography forward calculation in realistic volume conductors. Physics in
medicine and biology 48:3637-3637.
Pandya DN, Seltzer B (1982) Intrinsic connections and architectonics of posterior parietal cortex in
the rhesus monkey. J Comp Neurol 204:196-210.
Petrides M, Pandya DN (1999) Dorsolateral prefrontal cortex: comparative cytoarchitectonic analysis
in the human and the macaque brain and corticocortical connection patterns. Eur J Neurosci
11:1011-1036.
Reuter M, Schmansky NJ, Rosas HD, Fischl B (2012) Within-subject template estimation for unbiased
longitudinal image analysis. Neuroimage 61:1402-1418.
Swisher JD, Halko MA, Merabet LB, McMains SA, Somers DC (2007) Visual topography of human
intraparietal sulcus. J Neurosci 27:5326-5337.
Tootell RB, Hadjikhani N, Hall EK, Marrett S, Vanduffel W, Vaughan JT, Dale AM (1998) The
retinotopy of visual spatial attention. Neuron 21:1409-1422.
Van Essen DC (2005) Surface-Based Comparisons of Macaque and Human Cortical Organization. In:
From Monkey Brain to Human Brain: A Fyssen Foundation Symposium, pp 3-19: Bradford
Books.
Van Essen DC, Maunsell JH, Bixby JL (1981) The middle temporal visual area in the macaque:
myeloarchitecture, connections, functional properties and topographic organization. J Comp
Neurol 199:293-326.
Van Essen DC, Glasser MF, Dierker DL, Harwell J, Coalson T (2012) Parcellations and hemispheric
asymmetries of human cerebral cortex analyzed on surface-based atlases. Cereb Cortex
22:2241-2262.
Van Veen BD, van Drongelen W, Yuchtman M, Suzuki A (1997) Localization of brain electrical activity
via linearly constrained minimum variance spatial filtering. Biomedical Engineering, IEEE
Transactions on 44:867-880.
Vernet M, Quentin R, Chanes L, Mitsumasu A, Valero-Cabré A (2014) Frontal eye field, where art
thou? Anatomy, function, and non-invasive manipulation of frontal regions involved in eye
movements and associated cognitive operations. Front Integr Neurosci 8:66-66.
Wandell BA, Dumoulin SO, Brewer AA (2007) Visual field maps in human cortex. Neuron 56:366-383.