1 Brain Connectomes Come of Age
Xiao-Jing Wang !, , Ulises Pereira ! , Marcello G. P. Rosa , Henry Kennedy $ Center for Neural Science, New York University, 4 Washington Place, New York, NY 10003, USA Biomedicine Discovery Institute and Australian Research Council Centre of Excellence for Integrative Brain Function, Monash University, Clayton, VIC 3800, Australia Stem Cell and Brain Research Institute, INSERM U846, 69500 Bron, France Correspondence: [email protected] Submitted to
Current Opinion in Neurobiology
Abstract.
Databases of directed- and weighted- connectivity for mouse, macaque and marmoset monkeys, have recently become available and begun to be used to build dynamical models. A hierarchical organization can be defined based on laminar patterns of cortical connections, possibly improved by thalamocortical projections. A large-scale model of the macaque cortex endowed with a laminar structure accounts for layer-dependent and frequency-modulated interplays between bottom-up and top-down processes. Signal propagation in a version of the model with spiking neurons displays a threshold of stimulus amplitude for the activity to gain access to the prefrontal cortex, reminiscent of the ignition phenomenon associated with conscious perception. These two examples illustrate how connectomics may inform theory leading to discoveries. Computational modeling raises open questions for future empirical research, in a back-and-forth collaboration of experimentalists and theorists. • Directed- and weighted inter-areal cortical connectivity matrices of macaque, marmoset and mouse exhibit similarities as well as marked differences. • The new connectomic data provide a structural basis for dynamical modeling multi-regional cortical circuit and understanding the global cortex. • Quantification of cortical hierarchy guides investigations of interplay between bottom-up and top-down information processes. Introduction
In 1991 Felleman and van Essen published a landmark work of neuroanatomy that combined data from the existing literature to establish a hierarchy of the macaque monkey cortex [1]. This paper provided an impetus for efforts that, 10 years later, led to an inter-areal cortical connectivity matrix, the Collation of Connectivity data on the Macaque brain (CoCoMac) [2]. The CoCoMac matrix was fairly rough, with connections between area pairs assigned as absent, weak or strong (hence not quantitatively weighted); many connections were missing because of lack of information. Nevertheless, it represented a pioneering event in the field now referred to as brain connectomics. The past two decades have seen significant advances [3–9]. Novel technologies have made it possible to determine wiring of neural circuits in the brain on microscopic, mesoscopic and macroscopic spatial scales [10–13]. Importantly, while it may be true that a picture is worth a thousand words, systematic measurements translated into precise numbers are essential for discovering general principles of large-scale cortical organization. This short review covers recent advances in our description of the cortico-cortical connections, and computational modeling based on the new quantitative databases. We shall summarize recent approaches and findings, as well as challenges that need to be addressed in order for the field to move ahead. The word “connectome” is currently used to refer to collations of data obtained with different methods, with multiple resolutions. The present review focuses on the connectome defined with cellular-resolution tracers, which at present can only be used in nonhuman animals.
From anatomy to multi-regional cortical dynamics
In recent years new databases of inter-areal connectivity have become available for both macaque monkey [14-15, 16••, 17], mice [18–19, 21••], and marmoset [22, 23••]. Using a systematic analysis of retrograde tracing, the weight of cortico-cortical connection is indexed between 0 and 1 (FLNs, or fraction of labeled neurons), which measures the weight of connection from a source area relative to all source areas for a target area [15]. Therefore, connections are weighted parametrically, which is much more informative than a binary matrix. It is also directed, unlike diffusion tractography which, although non-invasive, cannot differentiate fibers from area A to area B and those in the reversed direction. Whereas fiber amounts can be inferred from tractography, measurements from tract tracing are direct and thus constitute a “ground truth”. Indeed, the correlation is modest between log-transformed tractography and tracer connection weights in the macaque ( 𝑟 ≃ 0.59 [24]). Three results are noteworthy. First, the Felleman-van Essen cortical hierarchy is significantly improved by quantification, so that a directed connection from area A to area B is assigned an FLN value, and areas are arranged along a one-dimensional hierarchy numerically. Second, the weight of connection (if present) between two 3 areas decays exponentially with their distance (the exponential distance rule) [17]. Third, the weights of inter-areal connections are highly heterogeneous, spanning five orders of magnitude [15]. These three salient characteristics have also been shown in marmoset, another monkey species of growing interest in Neuroscience [22–23]. Therefore, a graph-theoretical view of cortical networks is inadequate unless spatial relationships between areas are taken into consideration [25]. This finding inspired a new class of generative models for the cortical networks that are explicitly spatially embedded [26], [14]. The new macaque connectivity matrix provides a structural scaffolding for the development of a dynamical model of multi-regional macaque cortex. As a sound practice in computational neuroscience, one must judiciously choose the level of complexity of a model that is proper to investigate specific scientific questions. In this case, the main question was: what is the biological mechanism that endows each area with appropriate temporal dynamics for its specialized function, such as rapid responses in the primary sensory areas and slow ramping activity underlying time integration in association areas? In the model, each area was mathematically modeled by a generic excitatory-inhibitory network, in accordance with the commonly accepted notion of a canonical circuit in the cortex [27–28]. The quantitative connection strengths, however, vary from one area to another. These variations were not random, but systematically change along low-dimensional axes across the cortical mantle. Chaudhuri et al. [29] considered the number of spines (loci of excitatory synapses) in the basal dendritic tree of pyramidal neurons, as a proxy of the strength of synaptic excitation per neuron, which displays an increasing gradient along the cortical hierarchy [30]. Interestingly, in this model, temporal dynamics of each area is dominated by a time constant that ranges from tens of milliseconds for early sensory areas to more than a second for prefrontal areas at the top of the cortical hierarchy, exactly what is desirable for functional differentiation. Importantly, the prevalent time constant of an area is not a monotonic function of its hierarchical position. For instance, the frontal eye field is at a relatively low position in the hierarchy [16], but it shows a long time constant by virtue of being part of the frontal lobe in close interactions with other frontal areas that display slow dynamics. The timescale spectrum in the cortex is constrained by both the macroscopic gradient of synaptic connection strength and the weighted inter-areal cortical network. The new concept of macroscopic gradients [31••] applies to both synaptic excitation and inhibition. For instance, counts of diverse inhibitory cell-types across the mouse cortex revealed that the density of GABAergic cells expressing calcium-binding protein parvalbumin (PV), which control spiking outputs of excitatory pyramidal neurons, is the highest in the primary visual cortex and much lower in association areas [32–33]. Assessment of such macroscopic gradients can be carried out using a variety of data, including levels of gene expression that encode receptors for synaptic excitation and inhibition [34••, 35•] and neuronal density [36]. This approach allows identification of the biological fingerprint of different cortical areas; these data can then be incorporated into dynamical computational modeling. 4 They also are valuable for comparison across species. In particular, we will discuss below the definition of cortical hierarchy in primates versus rodents.
Figure 1 : A multi-regional model of the macaque monkey cortex endowed with a laminar structure. (a) The scheme shows the four levels considered: a within-layer local microcircuit consisting of an excitatory(red) and an inhibitory (blue) population (upper left), a laminar circuit with two laminar modules (corresponding to supra- and infra-granular layers, lower left), an inter-areal circuit with laminar-specific projections (lower right), and a large-scale network of 30 cortical areas based on macaque anatomical connectivity (upper right). Each level is anatomically constrained. Only the connections at each level not shown at a lower level are plotted, for clarity. (b) Left panel: the superficial layer and deep layer display gamma (upper panel) and alpha (lower panel) oscillations. Right panel: the periodogram of the superficial layer shows gamma modulated by alpha wave (top), whereas the deep layer is dominated by alpha rhythmicity (bottom). (c) Granger causality as a function of frequency for feedforward signaling from V1 to V4 (green) and feedback (orange). (d) Cortical hierarchy deduced from the frequency-dependent Granger causality measure in the model (left panel) and in a monkey experiment (right panel). Reproduced with permission from [37] with experimental data from [76]. a b
Fig 1:Fig 2:
Fig 1:Fig 2: c Fig 5:Fig 6: d of increasing the input to the excitatory population on the peak power(left) and frequency (right) of the oscillations, for the case of both anisolated layer 2/3 microcircuit (gamma rhythm, top) and an isolatedlayer 5 microcircuit (alpha rhythm, bottom). Note that the power andfrequency curves saturate at high inputs, in line with evidence of non-linear effects on V1 gamma power for strong contrast ( , ).The local circuit presented here displays, therefore, the same in-creases of power and frequency of neural oscillations with increasinginput that are observed in recent electrophysiological studies of early visual areas. This makes the circuit a good starting point to understandrhythmic interactions in larger neural systems. Interlaminar level
Having characterized the neural dynamics of an isolated layer, we built alaminar circuit by considering several layers and adding interlaminarprojections between them. To investigate the interplay between gammaandalpha/low-betarhythms,weconsidertwodistinctlaminarmodules,onefor the generation of gamma and one for the generation of alpha rhythms.
Fig. 2. Local circuit model at the intralaminar level. ( A ) Scheme of the local circuit (top), with the excitatory and inhibitory population in red and blue, respectively,and examples of the oscillatory activity for an excitatory-inhibitory circuit in layer 2/3 (middle, in green) and layer 5/6 (bottom, in orange). ( B ) Power spectrum of thefiring rate of an isolated layer 2/3 as a function of input strength to the excitatory population. The spectrum of the spontaneous state (with zero input) has beensubtracted in each case to highlight changes induced by the input (see main text). As the input increases (which resembles the effect of increasing the contrast of avisual stimulus), the power of gamma rhythms becomes stronger, as in observations by Henrie and Shapley ( ). ( C ) Effect of the input to the excitatory population onthe power spectrum peak (left) and frequency (right) of the oscillations, for an isolated layer 2/3 (top) and an isolated layer 5/6 (bottom). Fig. 3. Cortical area model at the interlaminar level. ( A ) Scheme of the interlaminar circuit (left panel); self-connections within a given population are omitted in thefigure for clarity. Interlaminar connections considered in the model correspond to the strongest projections between layer 2/3 and layer 5/6 as found in experimentalstudies. Right: Power spectrum of layer 2/3 (top) and layer 5/6 (bottom) in the case of uncoupled, isolated layers (in black, for comparison) and interconnected network(green and orange, respectively). A background input of I = 8 was fed into the excitatory population of both layers. ( B ) Bottom: A set of 30 traces of activity in layer 5/6(in gray) and their average (in blue). The central peak of each trace was aligned at zero before averaging. Top: A periodogram of layer 2/3 showing the averaged powerfor a range of frequencies for the same temporal periods as the layer 5/6 traces. We can see the existence of a strong entrainment of gamma power to alpha phase, asin the experimental findings by Spaak et al . ( ). Input was I = 6 for supragranular and I = 8 for infragranular excitatory populations. ( C ) Effect of injecting externalcurrent to the excitatory population of layer 5/6 on the layer 2/3 gamma power and (dimensionless) firing rate (top left and right, respectively) and on layer 5/6 alphapower (bottom left). An inverse relationship between supragranular firing rate and alpha power is observed (bottom right), which highlights a possible link of enhancedalpha rhythms with activity suppression. S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E
Mejias et al . Sci. Adv. :e1601335 16 November 2016 on N o v e m be r , tt p :// ad v an c e s . sc i en c e m ag . o r g / D o w n l oaded f r o m coefficient (Fig. 6F) between anatomical projections and functional in-teractions, in accordance with experimental findings ( ).The laminar pattern of anatomical FF and FB projections is thedefining feature of the global anatomical hierarchy of visual areas( , ). Because our model displays a strong correlation between ana-tomical projections and functional interactions, it is interesting to testwhether the model predicts the emergence of a similar hierarchy at thefunctional level, as recently observed in vivo (
5, 6 ). We follow the sameprocedure as in the study by Bastos et al . ( ) to define the hierarchicaldistance between area pairs from mDAI values (see SupplementaryMethods), and after simulating the full large-scale model and comput-ing its mDAI values, we observe the emergence of a clear functionalhierarchy among visual areas, as shown in Fig. 6G. As in the experi-mental functional hierarchy ( ), early visual areas lie at the bottom ofthe hierarchy, followed by areas of the FEF (8l and 8m) and withextrastriate visual areas of the ventral and dorsal functional streams at the top. Areas within the same functional streams are clustered aroundsimilar hierarchical values (Fig. 6H), in agreement with the study byBastos et al . ( ). These results show that the spectral functional inter-actions, as well as the formation of a functional hierarchy observedexperimentally, can be explained within a computational frameworkof locally generated rhythms propagated through a multilaminarnetwork structure.As an interesting example of the predictive power of our large-scalemodel, we use it to investigate a complex phenomenon observed duringvisual attention tasks. It has been reported that, contrary to the anatom-ical hierarchy, the functional hierarchy is not fixed. More precisely,Bastos et al . ( ) found that the positions of visual areas in the functionalhierarchy are highly dynamic and switch locations in a context-dependentfashion. The ranking of areas in the functionally defined hierarchy inthe pre-cue period of the task differs significantly from that obtainedin the post-cue period, when top-down modulations from higher areas Fig. 6. Large-scale cortical network and functional hierarchy. ( A ) Illustration of the anatomical tract-tracing technique used to obtain the anatomical large-scalenetwork and, in particular, the fraction of supragranular labeled neurons (or SLNs, see main text for details). A high (low) value of SLN for a given projection indicatesthat the source area is lower (higher) than the target area (the injected area) in the anatomical hierarchy. ( B ) 3D plot of the macaque anatomical network obtained (onlyprojections with FLN of >0.005 are plotted, for clarity), with all 30 areas in their spatial positions. Connection strength is indicated by line width. ( C and D ) alpha (C) andgamma (D) power for eight selected cortical areas of interest (V1, V2, V4, DP, 8m, 8l, TEO, and 7A). ( E ) Correlation between SLN and DAI, as a function of frequency. Thecorrelation is positive in the gamma range and negative in the alpha range, indicating a prevalent involvement of these rhythms in FF and FB interactions, respectively.( F ) Correlation between SLN and the combined DAI across gamma (30 to 70 Hz) and alpha/low-beta (6 to 18 Hz) frequency ranges (named mDAI, see text for details).( G ) Functional hierarchy emerging from the frequency-specific interactions in the network and computed using the mDAI values as in Bastos et al . ( ). ( H ) Areasbelonging to the same type (early visual, ventral, dorsal, or frontal; indicated by color box) tend to be clustered in the same way as in the experimental observations.For all panels, visual input was simulated with an input current I = 8 to the supragranular excitatory population of V1, and in addition to this input, a background currentof I = 6 to all excitatory populations in the network was also considered. S C I E N C E A D V A N C E S | R E S E A R C H A R T I C L E
Mejias et al . Sci. Adv. :e1601335 16 November 2016 on N o v e m be r , tt p :// ad v an c e s . sc i en c e m ag . o r g / D o w n l oaded f r o m anatomical hierarchy (Figure 6A) (Markov et al., 2014b) ratherthan the functional hierarchy, thereby avoiding circularity. Wedefined each area in turn as the target area, and averaged itsGC influences to all other areas, separately for the bottom-upand top-down directions (Figure 6B). Theta-band influenceswere more bottom-up directed for seven of eight target areas(and not significantly different for the remaining area), beta-band influences were more top-down directed for all targetareas, and gamma-band influences were more bottom-updirected for all target areas. In the grand average across all 28pairs of areas and both animals, this pattern was highly signifi-cant (Figure 6C, p = 0 for each of the three frequency bands).The same held also for each monkey individually without align-ment of frequency bands between animals (Figure 7, p = 0 foreach of the three frequency bands and each animal).Additional analyses showed that this pattern was not dueto observation noise (Nalatore et al., 2007) (Figures S6A andS6B) or the bipolar derivation scheme (Figures S6C and S6D).Regarding the theta-band, we note that the visual corticaltheta rhythm is partly locked to microsaccades (Bosman et al.,2009). Therefore, theta-rhythmic microsaccades with corre-sponding retinal image motion and subsequent visual responsesmight contribute to the feedforward GC influences in the theta-band. For the gamma-band, an analysis that excluded micro-saccade effects left the pattern of GC influences unchanged(Figures S6E and S6F). We also performed a conditional GCinfluence analysis (Wen et al., 2013), which aimed at estimatingthe GC influences that two areas exert directly onto each other,while excluding influences mediated by any one of the remainingvisual areas. This analysis left the pattern of results unchangedfor gamma and beta, and suggested the involvement of largernetworks for theta (Figure S7). ABCDE
Figure 4. Granger-Causal Influences Correlate Directly withAnatomy and Establish a Functional Hierarchy (A)SchematicofretrogradeanatomicaltracingmethodandcalculationofSLNvalues. Retrograde tracer is injected into a target area and labels neurons inseveral source areas projecting to the target area. Source areas hierarchicallylower (higher) than the target area have a progressively higher (lower) pro-portion of labeled neurons in the supragranular layers, i.e., the lower (higher)the source area relative to the target area, the higher (lower) the SLN value ofthe source-to-target projection.(B)Spearmanrankcorrelationacrossareapairs,betweenDAIvaluesfromtwomonkeyswithECoGrecordingsandSLNvaluesfromanindependentsetof25monkeys. This DAI-SLN correlation was calculated per frequency bin of theDAI, resulting in the spectrum. The gray-shaded region shows the 99.9%confidence interval, corresponding to a 95% confidence interval after cor-recting for the multiple comparisons across frequencies. Theta and gammainfluences were related to anatomical feedforward projections, and beta in-fluences to feedback projections. To assess the theta peak with 1 Hz spectralresolution, the analysis used 1 s epochs and Hann tapering. Only SLN valuesbased on at least ten labeled neurons were included.(C) Correlation between SLN and the DAI combined across theta-, beta-, andgamma-bands as specified on the y axis.(D) Black dots indicate hierarchical levels for all areas, derived by taking eacharea in turn as target and assigning the hierarchical level to the other areasbased on their GC influences to the target. Error bars show the SEM acrosstarget areas. Red dotsindicate hierarchical levels after removing V1, revealingimmunity to this manipulation.(E) Red dots indicate hierarchical levels of the full model versus one with V1removed. Other colors indicate corresponding analyses after removing moreareas from the lower or upper end of the hierarchy.
Neuron , 390–401, January 21, 2015 ª Hierarchical information processing in the macaque cortex
Quantification of a cortical hierarchy is based on the observation that, in general, source neurons for a feedforward projection (e.g., from V1 to V2) reside in the superficial layers (above layer 4), whereas a feedback projection (V2 to V1) originates from neurons in the deep layers (below layer 4). Certain areas lack a prominent layer 4; some connections such as between V4 and FEF have similar proportions of source neurons in the superficial layers. Notwithstanding exceptions, it is clear that in order to investigate how bottom-up and top-down processes interact, a computational model should incorporate a laminar cortical structure. Mejias et al. [37] built such a model in which a local area has a superficial layer and a deep layer; each with an excitatory-inhibitory microcircuit (Fig. 1a). The superficial layer exhibits noisy synchronous oscillations in the gamma ( ≃ ≃ ≃ Figure 2 : Signal propagation and the ignition phenomenon in the cortex. (a) Top and middle: In a macaque cortex model of spiking neurons, as the amplitude of an input to V1 is gradually increased, the peak response in areas of the occipital lobe (black) grows gradually. By contrast, activity is absent in the prefrontal cortex unless the stimulus intensity exceeds a threshold level (red). The activity map is confined to the posterior part of the cortex when the input is weak; if the input is above the threshold, access to the PFC leads to enhanced activity throughout the cortex. Note that this model included only a subset of cortical areas for which connectivity data are available, therefore the activity map is restricted only to those areas in the model. Bottom: The thresholding effect disappears when top-down connections in the model are deleted, demonstrating an important role of long-range feedback loops. (b) The model behavior is akin to the all-or-none ignition phenomenon associated with consciousness, that was observed experimentally with humans. Panel (a) is reproduced from [46], (b) from [49].
Cortical hierarchy in mouse and marmoset
Cortico-cortical connectivity in mouse also displays a wide range of connection weights and the exponential distance rule [20, 53]. However, whether the mouse cortex displays a well-defined hierarchy remains unsettled. Previous studies note various biological entities are high in V1 and low in association areas, such as PV neuron density [32] and the T1w:T2w ratio from structural magnetic resonance imaging, which correlates with the level of myelin content in the grey matter [35, 54]. Such measures thus roughly change across the cortex in a way reminiscent of a
Without feedback
Above 2 figs - no fb FLN - vent (areas 0 - 4, 8, 18) vs ALL
CBA A c t i v i t y A c t i v i t y LowHigh Without feedbackShuffling FLNs
The relations between stimulus strength, attention,and conscious perception are complex because attentionmechanisms can also be activated automatically in abottom-up manner. When the stimuli have strong energy,sharp onsets or strong emotional content, they mighttrigger an activation of frontal eye fields or amygdalapathways, thus causing an amplification that can lowertheir threshold for conscious perception [35]. Thus, bothbottom-up stimulus strength and top-down attentionalamplification (whether triggered voluntarily or by auto-matic attraction) are jointly needed for conscious percep-tion, but they might not always be sufficient for astimulus to cross the threshold for conscious perception.Conscious perception must therefore be evaluated bysubjective report, preferably on a trial-by-trial basis.Verifying that the stimuli can be consciously perceived ina separate experimental block where they are attended,as done by Tse et al. [18], does not suffice to guaranteeconscious perception in a different block where attentionis diverted. One cannot simply assume that, by unmasking stimuli, one is studying the neural correlatesof conscious processing.
Distinguishing accessibility from access
The above distinctions lead us to proposal a formaldefinition of two types of non-conscious processes(Figure 1):(1)
Subliminal processing . We define subliminal proces-sing (etymologically ‘below the threshold’) as acondition of information inaccessibility where bot-tom-up activationisinsufficienttotrigger alarge-scalereverberating state in a global network of neuronswith long range axons. Simulations of a minimalthalamo-corticalnetwork[4]indicates thatsuchanon-linear self-amplifying system possesses a well-defineddynamical threshold. A processing stream thatexceeds a minimal activation level quickly growsuntil a full-scale ignition is seen, while a slightly
TRENDS in Cognitive Sciences
Top-down attention
Absent Present
Bottom-upstimulusstrength
WeakorinterruptedSufficientlystrong
Subliminal (unattended)
Subliminal (attended)
Preconscious Conscious • Very little activation• Activation is already weak in early extrastriate areas• Little or no priming• No reportability • Strong feedforward activation• Activation decreases with depth• Depth of processing depends on attention and task set• Activation can reach semantic level• Short-lived priming• No durable fronto- parietal activity• No reportability• Orientation of top-down attention• Amplification of sensori-motor activity• Intense activation spreading to parieto- frontal network• Long-distance loops and global synchrony• Durable activation, maintained at will• Conscious reportability• Intense activation, yet confined to sensori-motor processors• Occipito-temporal loops and local synchrony• Priming at multiple levels• No reportability while attention is occupied elsewhere
Figure1.
Proposeddistinctionbetweensubliminal,preconscious,andconsciousprocessing.Threetypesofbrainstatesareschematicallyshown,jointlydefinedbybottom-upstimulusstrength(ontheverticalaxisatleft)andtop-downattention(onthehorizontalaxis).Shadesofcolorillustratetheamountofactivationinlocalareas,andsmallarrowstheinteractionsamongthem.Largearrowsschematicallyillustratetheorientationoftop-downattentiontothestimulus,orawayfromit(‘task-unrelatedattention’).Dashedcurvesindicateacontinuumofstates,andthicklineswithseparatorsindicateasharptransitionbetweenstates.During subliminalprocessing ,activationpropagatesbut remains weak and quickly dissipating (decaying to zero after 1–2 seconds). A continuum of subliminal states can exist, depending on masking strength, top-downattention,andinstructions(seeBox1).During preconsciousprocessing ,activationcanbestrong,durable,andcanspreadtomultiplespecializedsensori-motorareas(e.g.frontaleyefields).However,whenattentionis orientedawayfromthestimulus(largeblackarrows),activationisblockedfromaccessinghigherparieto-frontalareasandestablishing long-distance synchrony. During conscious processing , activation invades a parieto-frontal system, can be maintained ad libidum in working memory, andbecomescapableofguidingintentionalactionsincludingverbalreports.Thetransitionbetweenpreconsciousandconsciousissharp,asexpectedfromthedynamicsofaself-amplifiednon-linearsystem[4].
Opinion
TRENDS in Cognitive Sciences
Vol.10 No.5 May 2006206
The relations between stimulus strength, attention,and conscious perception are complex because attentionmechanisms can also be activated automatically in abottom-up manner. When the stimuli have strong energy,sharp onsets or strong emotional content, they mighttrigger an activation of frontal eye fields or amygdalapathways, thus causing an amplification that can lowertheir threshold for conscious perception [35]. Thus, bothbottom-up stimulus strength and top-down attentionalamplification (whether triggered voluntarily or by auto-matic attraction) are jointly needed for conscious percep-tion, but they might not always be sufficient for astimulus to cross the threshold for conscious perception.Conscious perception must therefore be evaluated bysubjective report, preferably on a trial-by-trial basis.Verifying that the stimuli can be consciously perceived ina separate experimental block where they are attended,as done by Tse et al. [18], does not suffice to guaranteeconscious perception in a different block where attentionis diverted. One cannot simply assume that, by unmasking stimuli, one is studying the neural correlatesof conscious processing.
Distinguishing accessibility from access
The above distinctions lead us to proposal a formaldefinition of two types of non-conscious processes(Figure 1):(1)
Subliminal processing . We define subliminal proces-sing (etymologically ‘below the threshold’) as acondition of information inaccessibility where bot-tom-up activationis insufficient to trigger alarge-scalereverberating state in a global network of neuronswith long range axons. Simulations of a minimalthalamo-corticalnetwork[4]indicates thatsuchanon-linear self-amplifying system possesses a well-defineddynamical threshold. A processing stream thatexceeds a minimal activation level quickly growsuntil a full-scale ignition is seen, while a slightly
TRENDS in Cognitive Sciences
Top-down attention
Absent Present
Bottom-upstimulusstrength
WeakorinterruptedSufficientlystrong
Subliminal (unattended)
Subliminal (attended)
Preconscious Conscious • Very little activation• Activation is already weak in early extrastriate areas• Little or no priming• No reportability • Strong feedforward activation• Activation decreases with depth• Depth of processing depends on attention and task set• Activation can reach semantic level• Short-lived priming• No durable fronto- parietal activity• No reportability• Orientation of top-down attention• Amplification of sensori-motor activity• Intense activation spreading to parieto- frontal network• Long-distance loops and global synchrony• Durable activation, maintained at will• Conscious reportability• Intense activation, yet confined to sensori-motor processors• Occipito-temporal loops and local synchrony• Priming at multiple levels• No reportability while attention is occupied elsewhere
Figure1.
Proposeddistinctionbetweensubliminal,preconscious,andconsciousprocessing.Threetypesofbrainstatesareschematicallyshown,jointlydefinedbybottom-upstimulusstrength(ontheverticalaxisatleft)andtop-downattention(onthehorizontalaxis).Shadesofcolorillustratetheamountofactivationinlocalareas,andsmallarrowstheinteractionsamongthem.Largearrowsschematicallyillustratetheorientationoftop-downattentiontothestimulus,orawayfromit(‘task-unrelatedattention’).Dashedcurvesindicateacontinuumofstates,andthicklineswithseparatorsindicateasharptransitionbetweenstates.During subliminalprocessing ,activationpropagatesbut remains weak and quickly dissipating (decaying to zero after 1–2 seconds). A continuum of subliminal states can exist, depending on masking strength, top-downattention,andinstructions(seeBox1).During preconsciousprocessing ,activationcanbestrong,durable,andcanspreadtomultiplespecializedsensori-motorareas(e.g.frontal eyefields).However, whenattentionis orientedawayfrom thestimulus(largeblackarrows), activationisblockedfromaccessinghigherparieto-frontalareasandestablishing long-distance synchrony. During conscious processing , activation invades a parieto-frontal system, can be maintained ad libidum in working memory, andbecomescapableofguidingintentionalactionsincludingverbalreports.Thetransitionbetweenpreconsciousandconsciousissharp,asexpectedfromthedynamicsofaself-amplifiednon-linearsystem[4].
Opinion
TRENDS in Cognitive Sciences
Vol.10 No.5 May 2006206
Conscious processingSubliminal processing “ignition” ab diffusion distance between areas based on the transition probabilities of a hypothetical diffusion process. This distance produces a diffusion space where closer areas in this space share a larger number of paths connecting them, while areas far apart are less connected. In general, the diffusion distance depends on a low number of ‘principal directions’ or ‘principal gradient’ in diffusion space, providing the method a low dimensional embedding of the connectivity. Applying this approach to the whole mouse brain data of [59] and by choosing V1 as the origin in the diffusion space, a hierarchy among areas can be built by sorting areas by their diffusion distance to the origin. Figure 3 shows the pairwise correlations between the Harris hierarchy, the hierarchy deduced from the diffusion map, PV density and T1w:T2w ratio in the mouse brain. Intriguingly, Spearman correlation coefficient values are in the range of 0.35 to 0.5. The explanation of substantial but far from perfect correlations is presently unclear, indicating that future research is warranted to achieve a consensus on the definition of cortical hierarchy in the mouse. On the other hand, it is possible that a cortical hierarchy is flatter or less developed in rodents than primates [20]. This difference in organization could emerge from simple scaling laws [60-61], which predict that brain size is inversely correlated with “percent connectedness” (the fraction of brain cells with which any cell communicates directly). This, in turn, could have the effect of increasing the variety of inputs to any given cortical area, hence reducing the dominance of any single source, and “blurring” the definition of hierarchical levels. 9 Recent analyses based on a dataset of directed and weighted connections in the marmoset cortex shed light on this issue [22-23]. Marmosets, like macaques, are simian primates, but are much smaller (on average, the mass of the marmoset brain is 12 times smaller than that of M. fascicularis ). In line with the scaling hypothesis, previous studies have indicated that the sources of afferents to both sensory [60] and association [61] areas are more widely distributed spatially across the cortex in marmosets than in macaques. However, a recent comprehensive study of the cortical connectome using statistical techniques applied to retrograde tracer data also revealed that this is accomplished without loss of specificity: the cortical connectivity matrix is very similar to that in the macaque in terms of overall density (approximately 2/3 of the possible connections that could exist are observed experimentally in both species), but they both differ from the mouse (where 97% of the possible connections exist). The similarity between macaque and marmoset extends to more detailed properties of the connectome, such as occurrence of reciprocal versus unidirectional connections. Other properties of the marmoset connectome, such as the presence of a well-defined core-periphery arrangement and the log-normal distribution of connection weights, also bring the two primates in close alignment. Importantly for the present argument, the marmoset cortex is also characterized by a well-defined hierarchy, where areas belonging to the different sensory domains occupy defined levels, from primary visual, auditory and somatosensory areas, though several higher-order association areas, to sensory association and polysensory areas. These multiple hierarchies converge to a core of frontal, posterior parietal, rostral temporal areas, which occupy the highest hierarchical levels, and include the regions of the cortex that expanded most clearly during primate evolution [63]. Furthermore, the hierarchical levels defined by connectivity are highly correlated with structural measures such as neuronal density and number of spines in the basal dendritic trees of pyramidal cells [64], [36]. Thus, the current evidence suggests that brain mass (and hence the number of neurons) does not in isolation fully predict the characteristics of the hierarchical organization of the cortex across mammals, and point to specific differences between primates and rodents, which are likely to have emerged due to specific evolutionary pressures. Further studies in marmosets, including the integration of cellular connectivity data with high-resolution tractography and functional connectivity measures using neuroinformatic platforms [65-66] offer the promise of greater insight onto the correlation with non-invasive measurements in the human brain, which promise to increase our ability to investigate the bases of neuropsychiatric conditions [67-69]. 10
Figure 3 : Cortical hierarchy in the mouse can be defined by four different measures: T1w:T2w ratio and PV neuron density decrease, whereas the diffusion map measure and the Harris hierarchy based on layer-dependent connectivity generally increase with the hierarchy. The pairwise correlation between these four measures, however, typically have a correlation of about 0.3-0.5.
Looking into Future
In summary, directed- and weighted- inter-areal cortical connection matrices now exist for macaque, marmoset and mouse. These data are of a different kind from connectomics on µ m spatial scale, achieved using electron microscopy for much smaller animals such as Drosophila fly [70]. Combined with genetic tools, research in this direction blurs the boundary between macroscopic and mesoscopic connectomes towards cell-type specific connectivity. For monkeys, existent datasets are incomplete as they only include a subset of cortical areas compared to tractography, which provides a complete cortico-cortical connectivity matrix. This limitation makes it difficult for dynamical models to simulate functional connectivity, defined by the covariance of activities between cortical areas. A subnetwork does not encompass all areas and their feedback loops, and this could impact on global brain dynamics. While modeling has been attempted in this direction [71], it was done using tractography data, which has a poor signal-to-noise ratio (that is, it includes numerous false positives and false negatives) and is devoid of directionality information. Therefore, ongoing efforts to complete the full graph of monkey cortico-cortical connectivity should be a priority for the field.
Visual Auditory Medial Somatomotor Lateral Prefrontal a b cd e f Acknowledgments . This work was partly supported by the ONR Grant N00014-17-1-2041, US National Institutes of Health (NIH) grant 062349, and the Simons Collaboration on the Global Brain program grant 543057SPI to XJW; by research grants from the Australian Research Council (DP140101968, CE140100007) to MGPR. UP was supported by The Swartz Foundation. HK was supported by LABEX CORTEX (ANR-11-LABX-0042) of Université de Lyon (ANR-11-IDEX-0007) operated by the French National Research Agency (ANR) (HK.), ANR-11-BSV4-501, CORE-NETS (HK.), ANR-14-CE13-0033, ARCHI-CORE (HK.), ANR-15-CE32-0016, CORNET (HK.), Chinese Academy of Sciences President's International Fellowship Initiative. Grant No. 2018VBA0011 (HK).
References [1] D. J. Felleman and D. C. Van Essen, “Distributed hierarchical processing in the primate cerebral cortex,”
Cereb Cortex , vol. 1, pp. 1–47, 1991. [2] K. E. Stephan, L. Kamper, A. Bozkurt, G. A. Burns, M. P. Young, and R. Kötter, “Advanced database methodology for the collation of connectivity data on the macaque brain (cocomac),”
Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences , vol. 356, no. 1412, pp. 1159–1186, 2001. [3] M. Bota, H.-W. Dong, and L. W. Swanson, “From gene networks to brain networks,”
Nature Neurosci. , vol. 6, pp. 795–799, 2003. [4] O. Sporns, G. Tononi, and R. Kötter, “The human connectome: A structural description of the human brain,”
PLoS Comput. Biol. , vol. 1, p. e42, 2005. [5] J. W. Lichtman, J. Livet, and J. R. Sanes, “A technicolour approach to the connectome,”
Nature Reviews Neuroscience , vol. 9, pp. 417–422, 2008. [6] H. S. Seung,
Connectome: How the brain’s wiring makes who we are . Boston, MA: Houghton Mifflin Harcourt, 2012. [7] O. Sporns, “Contributions and challenges for network models in cognitive neuroscience,”
Nat. Neurosci. , vol. 17, pp. 652–660, 2014. [8] H. Barbas, “General cortical and special prefrontal connections: principles from structure to function,”
Annu. Rev. Neurosci. , vol. 38, pp. 269–289, 2015. [9] L. W. Swanson and J. W. Lichtman, “From Cajal to connectome and beyond,”
Annual Review of Neuroscience , vol. 39, pp. 197–216, 2016. [10] M. Helmstaedter, K. L. Briggman, and W. Denk, “3D structural imaging of the brain with photons and electrons,”
Curr. Opin. Neurobiol. , vol. 18, pp. 633–641, 2008. 13 [11] J. W. Bohland et al. , “A proposal for a coordinated effort for the determination of brainwide neuroanatomical connectivity in model organisms at a mesoscopic scale,”
PLoS Comput. Biol. , vol. 5, p. e1000334, 2009. [12] X.-J. Wang and H. Kennedy, “Brain structure and dynamics across scales: in search of rules,”
Curr. Opin. Neurobiol. , vol. 37, pp. 92–98, 2016. [13] C. C. Hilgetag, S. F. Beul, S. J. van Albada, and A. Goulas, “An architectonic type principle integrates macroscopic cortico-cortical connections with intrinsic cortical circuits of the primate brain,”
Network Neurosci. , p. doi.org/10.1162/netn_a_ 00100, 2019. [14] N. T. Markov, M. Ercsey-Ravasz, D. C. Van Essen, K. Knoblauch, Z. Toroczkai, and H. Kennedy, “Cortical high-density counterstream architectures,”
Science , vol. 342, p. 1238406, 2013. [15] N. T. Markov et al. , “A weighted and directed interareal connectivity matrix for macaque cerebral cortex,”
Cereb. Cortex , vol. 24, pp. 17–36, 2014. [16] N. T. Markov et al. , “Anatomy of hierarchy: feedforward and feedback pathways in macaque visual cortex,”
J. Comp. Neurol. , vol. 522, pp. 225–259, 2014. •• A modern analysis quantifies macaque cortical hierarchy and uncovers a counterstream architecture. [17] M. Ercsey-Ravasz et al. , “A predictive network model of cerebral cortical connectivity based on a distance rule,”
Neuron , vol. 80, pp. 184–197, 2013. [18] S. W. Oh et al. , “A mesoscale connectome of the mouse brain,”
Nature , vol. 508, pp. 207–214, 2014. [19] B. Zingg et al. , “Neural networks of the mouse neocortex,”
Cell , vol. 156, pp. 1096–1111, 2014. [20] R. Gămănuţ et al. , “The mouse cortical connectome, characterized by an ultra-dense cortical graph, maintains specificity by distinct connectivity profiles,”
Neuron , vol. 97, pp. 698–715, 2018. [21] J. A. Harris et al. , “Hierarchical organization of cortical and thalamic connectivity,”
Nature , vol. 575, pp. 195–202, 2019. •• A quantification of hierarchy in mouse cortex based on cell-type dependent connectivity and thalamo-cortical projection patterns. [22] P. Majka et al. , “Open access resource for cellular-resolution analyses of corticocortical connectivity in the marmoset monkey,”
Nature Communications , vol. 11, no. 1, pp. 1–14, 2020. 14 [23] P. Theodoni, P. Majka, D. H. Reser, D. K. Wójcik, M. G. Rosa, and X.-J. Wang, “Structural attributes and principles of the neocortical connectome in the marmoset monkey,”
BioRxiv , 2020. •• An analysis of marmoset cortico-cortical connectivity statistics. The authors found that marmoset cortex displays the exponential distance rule as in macaque [17], and report on a power law of the characteristic spatial length as a function of the brain’s gray matter volume across rodent and primate species. [24] C. J. Donahue et al. , “Using diffusion tractography to predict cortical connection strength and distance: a quantitative comparison with tracers in the monkey,”
J. Neurosci. , vol. 36, pp. 6758–6770, 2016. [25] Knoblauch K, Ercsey-Ravasz M, Kennedy H, Toroczkai Z (2016) The brain in space. In: Micro-, meso- and macro- connectomics of the brain, vol. 22 (Kennedy, H. et al., eds), pp 45-74: Springer, Heidelberg. [26] H. F. Song, H. Kennedy, and X.-J. Wang, “Spatial embedding of similarity structure in the cerebral cortex,”
Proc. Natl. Acad. Sci. (USA). , vol. 111, pp. 16580–16585, 2014. [27] R. J. Douglas and K. A. C. Martin, “Neuronal circuits of the neocortex,”
Annu Rev Neurosci , vol. 27, pp. 419–451, 2004. [28] R. J. Douglas and K. A. Martin, “Behavioral architecture of the cortical sheet,”
Curr. Biol. , vol. 22, pp. R1033–1038, 2012. [29] R. Chaudhuri, K. Knoblauch, M. A. Gariel, H. Kennedy, and X.-J. Wang, “A large-scale circuit mechanism for hierarchical dynamical processing in the primate cortex,”
Neuron , vol. 88, pp. 419–431, 2015. [30] G. N. Elston GN, and M. G. P. Rosa, “Morphological variation of layer III pyramidal neurones in the occipitotemporal pathway of the macaque monkey visual cortex”, Cereb. Cortex, vol. 8, pp. 278-294, 1998. [31] X.-J. Wang, “Macroscopic gradients of synaptic excitation and inhibition in the neocortex,”
Nature Reviews Neuroscience , vol. 21, pp. 169–178, 2020. •• This perspective marshals recent experimental and computational evidence in support of the new concept of macroscopic gradients of biological properties as a general principle of large-scale cortical organization. [32] Y. Kim et al. , “Brain-wide maps reveal stereotyped cell-type-based cortical architecture and subcortical sexual dimorphism,”
Cell , vol. 171, pp. 456–469, 2017. [33] X.-J. Wang and G. R. Yang, “A disinhibitory circuit motif and flexible information routing in the brain,”
Curr. Opin. Neurobiol. , vol. 49, pp. 75–83, 2018. 15 [34] J. B. Burt et al. , “Hierarchy of transcriptomic specialization across human cortex captured by structural neuroimaging topography,”
Nat. Neurosci. , vol. 21, pp. 1251–1259, 2018. •• Evidence that genes encoding receptors important for synaptic excitation and inhibition display macroscopic gradients in the human cortex. [35] B. D. Fulcher, J. D. Murray, V. Zerbi, and X. J. Wang, “Multimodal gradients across mouse cortex,”
Proc. Natl. Acad. Sci. U.S.A. , vol. 116, pp. 4689–4695, 2019. • Evidence that genes encoding receptors important for synaptic excitation and inhibition display macroscopic gradients in the mouse cortex. [36] N. Atapour, P. Majka, I. H. Wolkowicz, D. Malamanova, K. H. Worthy, and M. G. Rosa, “Neuronal distribution across the cerebral cortex of the marmoset monkey (
Callithrix jacchus ),”
Cerebral Cortex , vol. 29, no. 9, pp. 3836–3863, 2019. [37] J. F. Mejias, J. D. Murray, H. Kennedy, and X. J. Wang, “Feedforward and feedback frequency-dependent interactions in a large-scale laminar network of the primate cortex,”
Sci Adv , vol. 2, p. e1601335, 2016. [38] X.-J. Wang, “Neurophysiological and computational principles of cortical rhythms in cognition,”
Physiol. Rev. , vol. 90, pp. 1195–1268, 2010. [39] G. Buzsáki and X.-J. Wang, “Mechanisms of gamma oscillations,”
Annu. Rev. Neurosci. , vol. 35, pp. 203–225, 2012. [40] O. Jensen, P. Goel, N. Kopell, M. Pohja, R. Hari, and B. Ermentrout, “On the human sensorimotor-cortex beta rhythm: sources and modeling,”
Neuroimage , vol. 26, pp. 347–355, 2005. [41] J. H. Lee, M. A. Whittington, and N. J. Kopell, “Top-down beta rhythms support selective attention via interlaminar interaction: a model,”
PLoS Comput. Biol. , vol. 9, p. e1003164, 2013. [42] E. Spaak, M. Bonnefond, A. Maier, D. A. Leopold, and O. Jensen, “Layer-specific entrainment of gamma-band neural activity by the alpha rhythm in monkey visual cortex,”
Current Biology , vol. 22, no. 24, pp. 2313–2318, 2012. [43] T. van Kerkoerle et al. , “Alpha and gamma oscillations characterize feedback and feedforward processing in monkey visual cortex,”
Proc. Natl. Acad. Sci. U.S.A. , vol. 111, pp. 14332–14341, 2014. [44] A. M. Bastos et al. , “Visual areas exert feedforward and feedback influences through distinct frequency channels,”
Neuron , vol. 85, pp. 390–401, 2015. [45] S. Song, P. Sjöström, M. Reigl, S. Nelson, and D. Chklovskii, “Highly nonrandom features of synaptic connectivity in local cortical circuits,”
PLoS Biol. , vol. 3, p. e68, 2005. 16 [46] M. R. Joglekar, J. F. Mejias, G. R. Yang, and X.-J. Wang, “Inter-areal balanced amplification enhances signal propagation in a large-scale circuit model of the primate cortex,”
Neuron , vol. 98, pp. 222–234, 2018. • The ignition phenomenon associated with consciousness naturally emerges in a spiking network model of large-scale primate cortex. [47] M. Diesmann, M. Gewaltig, and A. Aertsen, “Stable propagation of synchronous spiking in cortical neural networks,”
Nature , vol. 402, pp. 529–533, 1999. [48] S. Dehaene, M. Kerszberg, and J. P. Changeux, “A neuronal model of a global workspace in effortful cognitive tasks,”
Proc. Natl. Acad. Sci. U.S.A. , vol. 95, pp. 14529–14534, 1998. [49] S. Dehaene, J. P. Changeux, L. Naccache, J. Sackur, and C. Sergent, “Conscious, preconscious, and subliminal processing: a testable taxonomy,”
Trends Cogn. Sci. (Regul. Ed.) , vol. 10, pp. 204–211, 2006. [50] G. A. Mashour, P. Roelfsema, J. P. Changeux, and S. Dehaene, “Conscious Processing and the Global Neuronal Workspace Hypothesis,”
Neuron , vol. 105, pp. 776–798, 2020. [51] B. van Vugt et al. , “The threshold for conscious report: Signal loss and response bias in visual and frontal cortex,”
Science , vol. 360, pp. 537–542, 2018. [52] G. A. Mashour, “The controversial correlates of consciousness,”
Science , vol. 360, pp. 493–494, 2018. [53] Horvat S, Gamanut R, Ercsey-Ravasz M, Magrou L, Gamanut B, Van Essen DC, Burkhalter A, Knoblauch K, Toroczkai Z, Kennedy H (2016) Spatial Embedding and Wiring Cost Constrain the Functional Layout of the Cortical Network of Rodents and Primates. PLoS biology 14:e1002512. [54] M. F. Glasser and D. C. Van Essen, “Mapping human cortical areas in vivo based on myelin content as revealed by T1- and T2-weighted MRI,”
J. Neurosci. , vol. 31, pp. 11597–11616, 2011. [55] J.H. Siegle et al., A survey of spiking activity reveals a functional hierarchy of mouse corticothalamic visual areas.
BiorRxiv, doi.org/10.1101/805010, 2019. [56] R.D. D'Souza et al. Canonical and noncanonical features of the mouse visual cortical hierarchy.
BiorRxiv. doi.org/10.1101/2020.03.30.016303, 2019. [57] D. S. Margulies et al. , “Situating the default-mode network along a principal gradient of macroscale cortical organization,”
Proc. Natl. Acad. Sci. U.S.A. , vol. 113, pp. 12574–12579, 2016. • Application of the mathematical diffusion map to brain connectivity data provides an independent approach for defining hierarchy and macroscopic gradients. 17 [58] R. R. Coifman and S. Lafon, “Diffusion maps,”
Applied and computational harmonic analysis , vol. 21, no. 1, pp. 5–30, 2006. [59] J. E. Knox et al. , “High-resolution data-driven model of the mouse connectome,”
Network Neuroscience , vol. 3, no. 1, pp. 217–236, 2018. [60] J. L. Ringo, “Neuronal interconnection as a function of brain size,”
Brain, Behavior and Evolution , vol. 38, no. 1, pp. 1–6, 1991. [61] M. G. Rosa et al. , “Cortical afferents of area 10 in cebus monkeys: Implications for the evolution of the frontal pole,”
Cerebral Cortex , 29, 1473–1495, 2019. [62] S. M. Palmer and M. G. Rosa, “Quantitative analysis of the corticocortical projections to the middle temporal area in the marmoset monkey: Evolutionary and functional implications,”
Cerebral Cortex , vol. 16, no. 9, pp. 1361–1375, 2006. [63] T. A. Chaplin, H.-H. Yu, J. G. Soares, R. Gattass, and M. G. Rosa, “A conserved pattern of differential expansion of cortical areas in simian primates,”
Journal of Neuroscience , vol. 33, no. 38, pp. 15120–15125, 2013. [64] G. N. Elston and M. G. Rosa, “Complex dendritic fields of pyramidal cells in the frontal eye field of the macaque monkey: Comparison with parietal areas 7a and lip,”
Neuroreport , vol. 9, no. 1, pp. 127–131, 1998. [65] Y. Hori et al. , “Comparison of resting-state functional connectivity in marmosets with tracer-based cellular connectivity,”
Neuroimage , vol. 204, p. 116241, 2020. [66] C. Liu et al. , “A resource for the detailed 3D mapping of white matter pathways in the marmoset brain,”
Nature Neuroscience , 23, 271-280, 2020. [67] X.-J. Wang, J. H. Krystal, Computational psychiatry.
Neuron , 84, 3:638-54, 2014. [68] C. Liu, C. C.-C. Yen, D. Szczupak, Q. Y. Frank, D. A. Leopold, and A. C. Silva, “Anatomical and functional investigation of the marmoset default mode network,”
Nature communications , vol. 10, no. 1, pp. 1–8, 2019. [69] R. L. Buckner and D. S. Margulies, “Macroscale cortical organization and a default-like apex transmodal network in the marmoset monkey,”
Nature communications , vol. 10, no. 1, pp. 1–12, 2019. [70] C. Xu, et al. A connectome of the adult Drosophila central brain.
BioRxiv , doi: https://doi.org/10.1101/2020.01.21.911859. [71] G. Deco, V. K. Jirsa, and A. R. McIntosh, “Emerging concepts for the dynamical organization of resting-state activity in the brain,”
Nat. Rev. Neurosci. , vol. 12, pp. 43–56, 2011. [72] J. J. Jun et al. , “Fully integrated silicon probes for high-density recording of neural activity,”
Nature , vol. 551, pp. 232–236, 2017. 18 [73] N. A. Steinmetz, P. Zatka-Haas, M. Carandini, and K. D. Harris, “Distributed coding of choice, action and engagement across the mouse brain,”
Nature , vol. 576, pp. 266–273, 2019. • Neuropixel recording from 30 thousands neurons reveals different brain regions are engaged in a simple discrimination task. [74] C. Stringer, M. Pachitariu, N. Steinmetz, C. B. Reddy, M. Carandini, and K. D. Harris, “Spontaneous behaviors drive multidimensional, brainwide activity,”
Science , 364, p. 255, 2019. •• Brainwide activity patterns during spontaneous behaviors inspire novel data analysis and raises new questions about distributed nature of brain dynamics. [75] C. Stringer, M. Pachitariu, N. Steinmetz, M. Carandini, and K. D. Harris, “High-dimensional geometry of population responses in visual cortex,”
Nature , vol. 571, pp. 361–365, 2019. [76] A. M. Bastos, W. M. Usrey, R. A. Adams, G. R. Mangun, P. Fries, and K. J. Friston, “Canonical microcircuits for predictive coding,”