Neuron

Review

Rapid Neocortical Dynamics:Cellular and Network Mechanisms

Bilal Haider1 and David A. McCormick1,*1Department of Neurobiology, Kavli Institute for Neuroscience, Yale University School of Medicine, 333 Cedar Street, New Haven,CT 06510, USA*Correspondence: david.mccormick@yale.eduDOI 10.1016/j.neuron.2009.04.008

The highly interconnected local and large-scale networks of the neocortical sheet rapidly and dynamicallymodulate their functional connectivity according to behavioral demands. This basic operating principle ofthe neocortex is mediated by the continuously changing flow of excitatory and inhibitory synaptic barragesthat not only control participation of neurons in networks but also define the networks themselves. The rapidcontrol of neuronal responsiveness via synaptic bombardment is a fundamental property of corticaldynamics that may provide the basis of diverse behaviors, including sensory perception, motor integration,working memory, and attention.

IntroductionTypical diagrams of the neocortical sheet suggest a static,

compartmentalized structure in which signals travel in a well-

prescribed manner, flowing across and within layers in a

manner dictated predominantly by patterns of anatomical

connectivity. While these relatively hard-wired pathways

certainly establish routes for information transfer within cortical

networks, it is the dynamic, moment-to-moment fluctuations in

activity traversing these pathways that determine the functional

connectivity of cortical networks. Since an individual cortical

neuron receives input from tens of thousands of other cortical

neurons, and in turn projects its output activity patterns to

thousands of recipient neurons, the influence of one neuron

(or group of neurons) upon another depends critically upon

the activity state of all of the neurons that are interconnected

with the cells or networks under consideration. In other words,

although there is a strong anatomical bias to neuronal interac-

tions, exactly which neuronal subpopulations actively commu-

nicate at any particular moment in time (i.e., functional connec-

tivity) depends upon the state of activity in the network itself

and can change rapidly to meet behavioral demands. Although

this framework of functional cortical connectivity is by no

means new (Hebb, 1949; Lorente de No, 1938), it nonetheless

has motivated a considerable number of studies that collec-

tively have yielded insights into the subtle spatial and temporal

dynamics of local recurrent excitatory and inhibitory cortical

networks (e.g., see Douglas and Martin, 2007a; Gilbert and

Sigman, 2007; Thomson et al., 2002). Through such work, it

is becoming clear that a basic operation of the cerebral cortex

is to control the flow of neuronal communication by transiently

linking specific groups of neurons (subnetworks) together

by dynamic modulation of neuronal responsiveness. These

rapid changes in responsiveness (over milliseconds to

seconds) occur largely through alterations in ongoing synaptic

activity generated by local, as well as long-range, neuronal

interactions and may underlie changes in functional connec-

tivity that enable the flexibility of sensory-motor behaviors at

these same timescales.

The purpose of this review is to provide a brief overview of

rapid (milliseconds to seconds) network dynamics in the cortex

that may be mediated by changes in neuronal responsiveness

and interactions resultant from synaptic bombardment. Many

excellent reviews have been written on other aspects of cortical

dynamics, including the roles of synaptic plasticity, intrinsic

membrane properties, neuromodulators, and changes in

anatomical connectivity (Alvarez and Sabatini, 2007; Bean,

2007; Caporale and Dan, 2008; Kerchner and Nicoll, 2008; Lli-

nas, 1988; Luo et al., 2008; Magee and Johnston, 2005; Malenka

and Bear, 2004; McCormick, 1992; Sjostrom et al., 2008). Here,

we begin by considering the main cellular and network mecha-

nisms that may rapidly alter the responsiveness and functional

connectivity of cortical neurons and examine the predictions of

computational models regarding the impact of synaptic

barrages on neuronal input-output relations in vivo. We will

then review experimental work demonstrating that rapid alter-

ations in excitatory and inhibitory synaptic barrages generated

within specific subnetworks mediate fast changes in cortical

excitability and information flow. We will take these insights

and evaluate how such state changes via synaptic bombard-

ment affect the sensory response properties of cortical neurons

in vivo and attempt to link the described cellular and network

mechanisms to recent evidence for cortical activity states

observed in awake and behaving animals. We will conclude by

outlining some key experiments dissecting rapid changes in

functional connectivity that may be testable in the near future

with newly emerging techniques. Our overall synthesis suggests

that concerted changes in synaptic activity in local networks may

serve as the key mechanism for determining not only action

potential rate and timing in single neurons but also serve as

a context of past and present network activity, linking ensembles

of neurons together in a behaviorally relevant fashion. The

dynamic control of neuronal interactions through synaptic

barrages generated in the local network may be a unifying prin-

ciple underlying diverse phenomena such as attention,

sensory-motor integration, working memory, and the sleep-

wake cycle.

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 171

Neuron

Review

A B Figure 1. Rapid Cortical Network DynamicsAre Mediatedby Modulation ofNeuronalGain(A) Schematic of various forms of response modu-lation. Input gain (dashed gray) shifts the thresholdand saturation points without altering the ratio ofinput to output. Multiplicative gain increases (red)or divisively decreases (blue) the responsivenessat all levels of input by a similar percentage andresults in a change in the slope of the input-outputfunction.(B) Schematic diagram illustrating five sets ofneurons or neuronal groups that are anatomicallyinterconnected (double-headed arrows), with themiddle group (1) receiving an input. If neuronalgroup2 has high gain, then groups 1, 2, and 3 exhibitenhanced functional connectivity (dashed arrows),with activity flowing easily between groups 1 and3. Likewise, if group 5 has low gain, then the flowof information and interactions between groups 1,4, and 5 will be limited.Note that the responsiveness(gain) within and among neuronal groups is deter-mined by local recurrent excitation and inhibition.

Modulating Responsiveness Dynamically ReconfiguresFunctional ConnectivityNeurons in the neocortex are interconnected directly with each

other and indirectly through other interposed neurons. Conse-

quently, a major mechanism by which neocortical subnetworks

are rapidly formed and broken is through the modulation of

neuronal excitability (Figure 1). Of course, changes in nearly

any property of the cortical network will affect neuronal respon-

siveness and network interactions. Classic examples include

changes in ionic concentrations and currents (Bean, 2007),

synaptic plasticity (Dan and Poo, 2006), and the actions of neuro-

modulators (McCormick, 1992). Here, we will focus on rapid

changes in network dynamics (milliseconds to seconds) and

the subsequent changes in synaptic barrages in recipient

neurons as a possible mediator of response modulation, since

network reconfiguration at this timescale is likely to be of utmost

importance to active waking behavior.

Let us first consider the relationship of neuronal discharge to

the amplitude of incoming synaptic barrages. The relationship

between neuronal response magnitude as a function of input

amplitude, or input-output curve, generally exhibits a sigmoidal

shape such that increasing the amplitude of weak inputs causes

a gradual increase in action potential response, while increasing

the amplitude of intermediate-sized inputs results in a steep

increase in spike response. Larger inputs typically elicit response

saturation (Figure 1A). To alter neuronal responsiveness to an

input, this whole curve may move leftward or rightward along

the input axis, altering the threshold and response saturation

points, while preserving the shape of the input-output function

(Figure 1A, dashed lines). Alternatively, the input-output curve

may shift up or down along the output axis, changing the sensi-

tivity of the neuron and the maximal response rate. Finally, the

ratio of response magnitude relative to input amplitude may

change by a constant gain factor (i.e., stretch along the output

axis), producing a multiplicative (Figure 1A, red) or divisive

(blue) change, thus varying the slope of the input-output curve.

Although all of these alterations represent changes in neural

responsiveness, the last, a multiplicative change of the input-

output curve, is typically referred to as gain modulation (Cardin

172 Neuron 62, April 30, 2009 ª2009 Elsevier Inc.

et al., 2008; Chance et al., 2002; McAdams and Maunsell,

1999; Salinas and Abbott, 1996; Treue and Martinez Trujillo,

1999).

How might cortical neurons rapidly alter the responsiveness,

or gain, of other cortical neurons? The major mechanism by

which neurons influence one another is through the generation

of action potentials that are visible in target structures as post-

synaptic potentials. Bombardment of neurons by synaptic

potentials strongly influences neuronal excitability by controlling

membrane potential, membrane conductance, and membrane

potential variance. Together, these factors ultimately control

not only spike number but also spike timing (Chance et al.,

2002; Ho and Destexhe, 2000; Kuhn et al., 2004; Rudolph and

Destexhe, 2003; Shu et al., 2003a). In other words, patterns of

synaptic bombardment control, on a moment-to-moment basis,

the probability of spike generation in the recipient neurons.

These changes in spike probability are highly dynamic in time

and provide ‘‘windows of opportunity’’ for interactions of cells

with coactivated groups of cortical and subcortical neurons

(Figure 2). These variations in neuronal interactions through

synaptic bombardment occur at multiple timescales, from hours

in the case of the sleep-wake cycles to milliseconds in the case

of higher-frequency fluctuations in membrane potential (Buzsaki

and Draguhn, 2004). The changing form of ongoing synaptic

barrages continuously determines whether or not a neuron

responds to incoming synaptic inputs, particularly those of small

to medium amplitudes (Figure 1A), as typically found between

single cortical neurons. If the modulated neuron or group of

neurons is interposed between other cells, then its level of

responsiveness will in turn determine the degree to which the

other neurons interact (Figure 1B). When the window of opportu-

nity becomes relatively short (within the integration time constant

of a single neuron, e.g., window 3 in Figure 2), then the precise

timing of spikes may play an important role in the interactions

within and between cortical networks.

Models of Rapid Response ModulationModulation of the input/output function lies at the heart of rapid

changes in cortical functional connectivity. Could rapid changes

Neuron

Review

in membrane potential, conductance, and variance during

different behavioral states underlie the well-known changes in

neuronal responsiveness occurring in cortical neurons accord-

ing to context and behavior? This question has been explored

through several computational studies (Azouz, 2005; Chance

et al., 2002; Compte et al., 2003b; El Boustani et al., 2007; Ho

and Destexhe, 2000; Kuhn et al., 2004; Murphy and Miller,

2003; Prescott and De Koninck, 2003; Rudolph and Destexhe,

2003). These studies have revealed that there are several mech-

anisms for modulation of neuronal responsiveness, the mix

of which can cause a multiplicative, or nearly multiplicative,

increase in neuronal gain. In the absence of significant levels of

membrane potential variance (also called ‘‘noise’’), tonic depo-

larization increases neuronal responsiveness by shifting input-

output curves along the input axis (Figure 1A; change input

offset) (Cardin et al., 2008; Chance et al., 2002; Compte et al.,

2003b; Ho and Destexhe, 2000; Shu et al., 2003a). A similar

effect can be obtained by decreases in membrane conductance

in the absence of membrane potential noise (Chance et al., 2002;

Ho and Destexhe, 2000; Shu et al., 2003a). These lateral shifts in

the input-output curve result in a large enhancement of the

responsiveness to small inputs and a moderate or only small

enhancement of larger inputs (dashed gray lines in Figure 1A).

If the input range is restricted to only the initial portions of the

Figure 2. Synaptic Bombardment in Active Cortical Networks In VivoIllustrated are simultaneous recordings of the extracellular local field potential(LFP, top), the multiple unit activity (MU, middle), and intracellular membranepotential (bottom) from a cortical pyramidal cell during the generation of oneUp state (bordered by two Down states). Note that the Up state is associatedwith a marked increase in local network activity, depolarization of the neuron(�25 mV), a marked increase in membrane potential variance (SD of Up =2.5 mV), and the presence of higher-frequency (gamma, �40 Hz) oscillationsin the LFP. Periods of depolarization mediated by network activity provide‘‘windows of opportunity’’ based upon increased neuronal gain and may beof long (example 1), medium (example 2), or short (example 3) duration. Thus,various frequency components of synaptic activity interact to initiate actionpotentials. Inset expands window 3, where the depolarizing half (�12.5 ms)of one full LFP oscillation cycle provides a short temporal window for theintegration of both excitatory and inhibitory postsynaptic potentials (PSPs).

input-output curve, then depolarization or decreases in conduc-

tance (in the absence of membrane potential variance) can result

in what appears to be a change in gain (e.g., Figure 1A; portions

where dashed gray overlaps with solid red or blue). It is thus crit-

ical to determine the input-output relationship over a wide range

of subthreshold as well as saturating inputs. In real neurons

in vivo, the membrane potential of neurons exhibits substantial

rapid variations, during all behavioral states, owing to fluctua-

tions in synaptic bombardment (Figure 2) (Steriade et al.,

2001). The addition of membrane potential variance causes

neurons to discharge in a probabilistic fashion across a wide

range of synaptic barrage amplitudes (Ho and Destexhe, 2000;

Shu et al., 2003a). By providing a variable level of depolarization,

membrane variance can actually enhance the probability that

subthreshold inputs become suprathreshold, thus enhancing

the responsiveness to weak inputs (Chance et al., 2002; Ho

and Destexhe, 2000; Shu et al., 2003a). Various combinations

of changes in membrane potential, membrane conductance,

and membrane potential variance can be utilized to result in

several types of shifts in neuronal responsiveness, including

changes in the slope of the input-output function (Ayaz and

Chance, 2009; Chance et al., 2002; Ho and Destexhe, 2000;

Shu et al., 2003a). Indeed, it has been suggested that increases

in neuronal gain can be achieved through a simultaneous

decrease in membrane conductance and variance, such as

could occur from the concerted withdrawal of barrages of

synaptic potentials, the composition of which is perfectly

balanced so as to not depolarize or hyperpolarize the neuronal

membrane potential (Chance et al., 2002). Such a model predicts

that increases in neuronal gain are associated with decreases in

local network activity.

A consequence of membrane potential variance is that the

average frequency of firing in response to an input can follow

a power law function [output = (input)x; Figure 3A] (Miller and

Troyer, 2002). This power law relationship between membrane

potential and firing rate is not merely a theoretical possibility; it

is often observed in cortical neurons in vivo when the cells are

synaptically driven with sensory stimulation (Anderson et al.,

2000; Priebe and Ferster, 2008; Sanchez-Vives et al., 2000)

(Figure 3B). Such a relationship between membrane potential

and firing rate has broad and important implications for how

any tonic changes in membrane potential (such as from a neuro-

modulatory input) affect the gain of these cells (e.g., Disney et al.,

2007; Thurley et al., 2008). For instance, depolarization of

cortical neurons by a few millivolts, as occurs with barrages of

synaptic potentials, will result in only a small increase in activity

at the lower end of the membrane potential-firing rate relation-

ship but a much larger increase in the firing rate at the high

end of this relationship (Figures 3A and 3B). The resulting larger

increase in firing rate with higher levels of input (Figures 3A and

3B) can result in a percent enhancement that is quite similar

(but not identical) at each point of the input-output curve

(Figure 3C). In other words, over the range of voltages in which

natural synaptic activity activates action potentials in cortical

neurons, the presence of membrane potential variance induces

a nonlinear relationship between membrane potential and firing

rate. The presence of this power law behavior can cause simple

depolarization to nonetheless appear as a multiplicative-like

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 173

Neuron

Review

A B C

D E F

Figure 3. Modulation of Neuronal Gainthrough an Interaction of MembranePotential and Intrinsic Nonlinearities inInput-Output Transformation of CorticalNeurons(A) In the absence of neuronal variance, neuronsrespond to depolarization with an abrupt increasein firing rate once threshold has been reached(black). However, in the presence of membranepotential variance, the average relationshipbetween firing rate and membrane potential canexhibit a power law (red).(B) Intracellular recordings of cortical neuronalresponses to visual stimuli in vivo have revealed,on average, a power law relationship betweenmembrane potential and firing rate.(C) In the presence of a power law relationshipbetween membrane potential and firing rate,changes in either excitatory or inhibitory back-ground conductance that depolarize or hyperpo-larize neurons (±2–5 mV), respectively, can resultin multiplicative-like changes in the input-outputrelationship of cortical neurons.(D) Depolarization induced through the intracel-lular injection of current results in a multiplica-tive-like change in the contrast response functioncurve of visual cortical neurons.(E) Scaling the hyperpolarized curve from (D) witha constant gain factor reveals a multiplicative-likegain change. (F) Depolarization that occurs spon-taneously owing to synaptic bombardment in vivocan also result in multiplicative-like increases inneuronal responsiveness of the contrast responsefunction.(B) Adapted by permission from MacmillanPublishers Ltd: Nature Neuroscience (Priebe et al.,2004), copyright 2004; (C) modified from Murphyand Miller (2003); (D) modified from Sanchez-Viveset al. (2000); (E) modified from Haider et al. (2007).

increase in gain (Murphy and Miller, 2003). These results suggest

that multiplicative-like gain modulation to sensory stimuli could

be obtained simply by depolarizing neurons, through the release

of neuromodulators (McCormick, 1992), the alteration of intrinsic

ionic conductances (Sanchez-Vives et al., 2000), or by barrages

of synaptic activity (Haider et al., 2007; Ho and Destexhe, 2000;

Shu et al., 2003a).

In summary, tonic changes in membrane potential in the pres-

ence of membrane potential variance—resulting from the high

variability of synaptic barrages—can result in rapid, nonlinear

changes in neuronal responsiveness and gain. Membrane

potential variance causes neurons to respond in a probabilistic

manner as opposed to all-or-none threshold elements, and

increasing depolarization can have large effects on the average

input-output relations of cortical neurons. As we will see, exper-

imental interrogation of synaptic barrages in vivo during active

cortical states strongly suggests that such a multiplicative-like

enhancement of responsiveness can indeed occur by transient

epochs of synaptic depolarization.

Local Network Activity Dominates the MembranePotential of Cortical Neurons In VivoCortical neurons receive the great majority of their synaptic

inputs from neurons within the local network (i.e., <1 mm distant)

(Binzegger et al., 2004). Consequently, the activity of the local

network will have a dominating, but not exclusive, influence on

174 Neuron 62, April 30, 2009 ª2009 Elsevier Inc.

the membrane potential of nearby cells. This is important since

investigations of the effects of changes in membrane potential,

membrane conductance, and membrane potential variance

demonstrate that the primary determinant of the probability of

action potential generation is the level of membrane potential

depolarization (Shu et al., 2003a). The closer a neuron is to action

potential threshold (usually around�53 mV for cortical pyramidal

neurons) the more likely it is to fire upon stimulation, all else being

equal. In the absence of any synaptic input, the membrane

potential of a cortical neuron in vivo is markedly hyperpolarized

(e.g., between �80 and �65 mV, depending on layer, cell type,

and state of the animal) and dominated by K+ conductances

(Pare et al., 1998). Under these conditions, the typical cortical

neuron is �10–25 mV below action potential threshold. Depola-

rizing a typical cortical pyramidal cell that has an input resistance

of 50 MU by 10 mV with a balanced barrage of excitatory and

inhibitory synaptic potentials would result in approximately

a 30% increase in membrane conductance (Shu et al., 2003a).

Although this increase in conductance by itself would decrease

neuronal responsiveness, the increase in excitability provided

by the depolarization is six times greater and therefore swamps

the conductance effect (Shu et al., 2003a). Of course, it should

be kept in mind that increases in membrane conductance can

themselves change the spatiotemporal integrative properties of

cortical neurons (Bernander et al., 1991; Borg-Graham et al.,

1998; Holt and Koch, 1997). In general, however, balanced

Neuron

Review

barrages of excitatory and inhibitory synaptic potentials (dis-

cussed in detail later) have their effects on action potential

discharge largely through determining the membrane potential

level of the postsynaptic neuron.

How many neurons must discharge to cause a typical cortical

pyramidal cell to fire? The average monosynaptic excitatory

postsynaptic potential (EPSP) from one cortical neuron to

another is on the order of 0.5 mV, far too small to elicit an action

potential in a quiescent neuron (Markram et al., 1997; Thomson

et al., 2002). A back-of-the-envelope calculation suggests that

on average approximately 50–100 excitatory synaptic potentials

(along with balanced inhibitory potentials) would need to be

received every 10 ms to keep the typical cortical pyramidal cell

depolarized by the 20 mV between resting membrane potential

and firing threshold. Since cortical neurons receive inputs from

approximately 3,000–10,000 other neurons (Binzegger et al.,

2004; Larkman, 1991), this indicates that even a low average

level of action potential activity (e.g., 0.1–1 Hz) in the presynaptic

network could provide this maintained depolarization, depend-

ing also upon the level of activation of inhibitory networks. This

average firing rate is well within the range of cortical neurons

active in the waking state (Beloozerova et al., 2003; Shoham

et al., 2006).

It is important to keep in mind that synaptic potentials arrive

throughout the dendrites, soma, and initial segment of the

axon and interact in a spatially and temporally complex manner,

the details of which are beyond the scope of this review, although

this important issue has received considerable attention (Larkum

et al., 2004; London and Hausser, 2005; Magee, 2000; Major

et al., 2008; Milojkovic et al., 2005; Murayama et al., 2009;

Nevian et al., 2007; Rudolph and Destexhe, 2003; Williams,

2004). Since nearly all action potentials that are responsible for

synaptic communication between cortical neurons are initiated

in the axon initial segment (Kole et al., 2008; McCormick et al.,

2007; Shu et al., 2007; Stuart et al., 1997; Yu et al., 2008), the

fluctuations as they appear at the level of the soma and axon

initial segment are of particular importance. Luckily, owing to

its large size, the soma is by far the most common place to

record the intracellular membrane potential fluctuations of

cortical neurons. For our present purposes, we propose that

the precise and rapid control of somatic membrane potential,

conductance, and variance is achieved through control of

concerted fluctuations in local network activity that can be

engaged by local and distant cortical, as well as subcortical,

neurons. Importantly, long-range connections are overwhelm-

ingly excitatory, but synapse onto both local inhibitory and excit-

atory neurons (McGuire et al., 1991), which are themselves

densely interconnected with one another. Experimental investi-

gations into the influence of synaptic bombardment on neuronal

responsiveness in local cortical networks have largely focused

on ‘‘spontaneous’’ cortical activity, of which the slow oscillation

is perhaps the best-characterized example.

The Cortical Slow Oscillation: A Model Systemto Investigate Neuronal Responsiveness and Excitatory-Inhibitory Interactions during Active Cortical StatesHow can we begin to examine the synaptic mechanisms of rapid

changes in the state and sensory responsiveness of single

cortical neurons and their extended interactions within an active

cortical network? Studies at multiple levels and in many systems

indicate that one key consequence of dense, local connectivity is

that the cortex exhibits spontaneous, persistent activity in the

absence of any external stimulation (for review, see Destexhe

et al., 2003; Major and Tank, 2004). This spontaneous, or

ongoing, activity is present during active behavior, quiet resting,

sleep, anesthesia, and even in cortical slices in vitro. However,

the spatial and temporal patterns of spontaneous activity in the

cortex differ markedly across these different conditions.

Studies in anesthetized cat visual cortex indicate that both

extended and local cortical regions exhibit dynamic activity

correlations varying over tens to hundreds of milliseconds.

Importantly, local regions that exhibit the strongest correlation

of excitability correspond to regions of cortex that exhibit similar

stimulus preferences, e.g., for the orientation of a drifting bar of

light (Arieli et al., 1996; Kenet et al., 2003). If a network is stimu-

lated appropriately and enters into an excitable state, temporally

coincident responses may be transmitted throughout this tran-

siently interconnected group of neurons, facilitating their interac-

tion (Abeles et al., 1995; Fujisawa et al., 2008; Harris, 2005;

Rutishauser and Douglas, 2008; Tiesinga et al., 2008). These

studies indicate that local cortical networks have a strong

tendency to transiently enter into stable states of enhanced

excitability that dynamically interact with sensory stimulation.

There is reason to believe, as we will discuss, that network acti-

vation in the waking state may exhibit at least some characteris-

tics similar to those described during anesthesia (Abeles et al.,

1995; Fujisawa et al., 2008; Poulet and Petersen, 2008; Steriade

et al., 2001). What are the cellular mechanisms that allow local

cortical networks to rapidly switch into and out of these semista-

ble and excitable states?

In the cortex, perhaps the most well-studiedpatternof recurrent

network activity is that of the slow (<1 Hz) oscillation, which is

characterized by a rhythmic alternation between activated and

deactivated states (Figure 2). The cortical slow oscillation has

been used to investigate a diverse array of cortical network

dynamics, ranging from sensory processing (Haider et al., 2007;

Hasenstaub et al., 2007; Lampl et al., 1999; Petersen et al.,

2003; Sachdev et al., 2004), short-term synaptic plasticity

(Crochet et al., 2005; Reig et al., 2006; Shu et al., 2006), commu-

nication between brain areas (Hahnetal., 2007), and the dynamics

of local recurrent networks (Cossart et al., 2003; Destexhe et al.,

2003; Haider et al., 2006; Luczak et al., 2007; Sanchez-Vives

and McCormick, 2000). This experimental work has also inspired

a series of computational studies (Compte et al., 2003b; Destexhe

et al., 2007; El Boustani et al., 2007; Holcman and Tsodyks, 2006;

Parga and Abbott, 2007; Rudolph et al., 2005).

Examination of the cortical slow oscillation in anesthetized and

naturally sleeping animals has yielded considerable insight into

the basic mechanisms of rapid changes in the state of cortical

networks, stable propagation of these activity states, and the

synchronization of activity across cortical regions. We propose

that insights gained from these studies provide useful and test-

able predictions about the mechanisms underlying rapid

changes in functional connectivity of cortical networks during

behavior. The cortical slow oscillation was first described

in studies of anesthetized rat auditory (Metherate and Ashe,

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 175

Neuron

Review

1993) and cat association cortex in vivo (Steriade et al., 1993b),

where rhythmic oscillations of the local field potential (LFP),

approximately once every second or two, were seen mirrored

in the membrane potential of nearby neurons (e.g., Figure 2).

This rhythmic cycle of depolarization and hyperpolarization

survived thalamic lesions and transection of the corpus callosum

(Steriade et al., 1993a) and even occurs in slices of many cortical

regions (Cossart et al., 2003; Sanchez-Vives and McCormick,

2000; Shu et al., 2003a, 2003b), indicating that the core mecha-

nisms of its generation exist in local cortical circuits. Similar

epochs of depolarization and hyperpolarization were found in

striatal and corticostriatal neurons of anesthetized rodents,

where it results from synaptic bombardment originating in

cortical networks (Cowan and Wilson, 1994; Stern et al., 1997).

Importantly, this type of activity occurs not only in anesthesia

but also naturally during slow wave sleep (SWS) (Destexhe

et al., 1999; Steriade et al., 2001) and may even occur during

quiet waking (Petersen et al., 2003; Poulet and Petersen,

2008). The activation and withdrawal of persistent activity during

neocortical Up and Down states has been compared to pharma-

cologically induced network activity (Tahvildari et al., 2008),

spontaneous network activity in organotypic cortical cultures

(Blackwell et al., 2003; Johnson and Buonomano, 2007),

and subthreshold depolarizing events in cortical dendrites

(Major et al., 2008; Milojkovic et al., 2005). While the ‘‘Up/

Down’’ terminology may be convenient for comparison of

grossly similar cortical dynamics, the mechanisms of generation,

time course, and consequences for neural responsiveness are

dramatically different in these varying depolarizing-hyperpolariz-

ing phenomena (for review, see Major and Tank, 2004) and

different still from the pattern of activity observed in striatal

neurons (Wilson and Kawaguchi, 1996). Our perspective is that

the cortical Up state, or active state of the slow oscillation, as

originally identified, is a synaptically generated, rhythmic oscilla-

tion of cortical origin that is seen simultaneously in the somatic,

dendritic, and proximal axonal membrane potential of single

neurons and in the local cortical network, as typically reflected

in the local field potential (Haider et al., 2006; Kerr et al., 2005;

Lampl et al., 1999; Sanchez-Vives and McCormick, 2000; Shu

et al., 2006; Steriade et al., 1993b; Waters and Helmchen, 2006).

Examination of the slow oscillation reveals that the transitions

between semistable states are rapid, occurring in approximately

50–100 ms (Figure 2). During the activated period, the membrane

potential of an individual cortical neuron is characterized by

depolarization of 15–25 mV, a modest increase in membrane

conductance, and irregular action potential firing driven by irreg-

ularity in the depolarizing potentials that reach action potential

threshold (Compte et al., 2003a, 2003b; Softky and Koch,

1993; van Vreeswijk and Sompolinsky, 1996). This firing irregu-

larity is also an important hallmark of neuronal activity in awake,

behaving animals (Compte et al., 2003a; Shadlen and Newsome,

1998). Many studies have demonstrated that activity in the local

network generates the depolarized and moderately variable (SD

of 2–3 mV) membrane potential of the Up state (Anderson et al.,

2000; Destexhe et al., 2003; Haider et al., 2006; Petersen et al.,

2003; Sanchez-Vives and McCormick, 2000; Shu et al., 2003b)

through an interaction of inhibitory and excitatory synaptic

potentials (Haider et al., 2006; Hasenstaub et al., 2005; Miura

176 Neuron 62, April 30, 2009 ª2009 Elsevier Inc.

et al., 2007; Shu et al., 2003b; van Vreeswijk and Sompolinsky,

1996; Vogels and Abbott, 2009). Long-range connections play

a key role in the communication, propagation, and synchroniza-

tion of this locally generated activity (Timofeev et al., 2000;

Volgushev et al., 2006).

Response Modulation in Active Cortical Networks:Excitatory-Inhibitory Interactions and TimingThe flow of synaptic activity through the cortex, which is highly

nonlinear and only partially hierarchical, is dependent on several

factors, including the amplitude of the particular synaptic event

in question (i.e., the ‘‘signal’’) in relation to the membrane poten-

tial and spike threshold, as well as the timing of the synaptic

event in relation to other synaptic potentials (i.e., the ‘‘back-

ground’’). As we have discussed, the nonlinear transformation

of membrane potential depolarization to spike output will take

place in each of the neurons in an ensemble, partially deter-

mining the ensuing pattern of synaptic activity and shaping

how each neuron interacts with the evolving barrage of synaptic

potentials. Neurons that maintain enhanced excitability will

continue to interact and remain in the ensemble, while those

neurons that are refractory (or actively inhibited) will transiently

fall out of ensemble membership. Given these complex proper-

ties of network dynamics, it becomes critical to understand the

mechanisms by which the amplitude-time course of the

membrane potential of cortical neurons is controlled and how

this interacts with synaptic barrages that might be considered

‘‘signals.’’ This key question has recently been addressed using

the cortical slow oscillation as a model, since it is characterized

by alternating periods of neuronal quiescence and recurrent

network activity. This property of the slow oscillation allows the

investigator to examine the influence of background, or sponta-

neous, synaptic barrages on, for example, responses to sensory

evoked synaptic potentials constituting ‘‘signals.’’

Intracellular recordings of synaptic activity in vivo during Up

states of the slow oscillation, as well as during sensory-evoked

responses, reveal that cortical networks operate through

a remarkable balance of synaptic excitation and inhibition

(Haider et al., 2006; Higley and Contreras, 2006; Marino et al.,

2005; Shu et al., 2003b; Wehr and Zador, 2003). In general, as

synaptic excitation to a cortical region increases, so does the

level of inhibition, owing to the highly recurrent nature of intra-

cortical excitatory and inhibitory networks (Binzegger et al.,

2004; Douglas and Martin, 2007b; Haider et al., 2006; Kapfer

et al., 2007; Murphy and Miller, 2009; Sanchez-Vives and

McCormick, 2000; Shu et al., 2003b; Silberberg and Markram,

2007; van Vreeswijk and Sompolinsky, 1996; Vogels and Abbott,

2009). This general balancing of excitatory and inhibitory inputs

to cortical neurons and networks must be generated in large

part locally, since long-range inhibitory connections are rare in

the neocortex. Intracellular analysis of sustained activity in

cortical networks as seen during Up states reveals that while

the general level of membrane potential is set by the excit-

atory-inhibitory balance and the intensity of synaptic barrages,

the moment-to-moment variations of membrane potential are

determined by rapid fluctuations in the timing of excitatory,

and especially inhibitory, synaptic potentials. In other words,

spike generation on a short timescale is determined by rapid

Neuron

Review

departures from a precise excitatory-inhibitory balance, which

lies on top of a stable depolarization that is mediated by a more

prolonged excitatory-inhibitory balance (Gabernet et al., 2005;

Hasenstaub et al., 2005; Higley and Contreras, 2006; Kapfer

et al., 2007; Okun and Lampl, 2008; Pouille and Scanziani,

2001; Rudolph et al., 2007; Vogels and Abbott, 2009). This is

important because there are generally two synaptic components

underlying action potential timing in cortical neurons. The

broader, less temporally precise component is generated

through recurrent excitatory and inhibitory interactions that set

the basal—yet variable—membrane potential of the neuron.

These broad changes in excitability are visible preferentially at

lower frequencies (<10 Hz). In addition to this component, rapid,

higher-frequency (30–100 Hz) fluctuations in inhibitory and excit-

atory synaptic potentials determine the precise probability and

timing of action potential generation, even to the millisecond level

(Cardin et al., 2009; Hasenstaub et al., 2005; Higley and Contre-

ras, 2006; Mainen and Sejnowski, 1995; Nowak et al., 1997;

Pouille and Scanziani, 2001). Importantly, it is the higher-

frequency components of membrane potential fluctuations that

most strongly influence the precise timing of action potentials

(Mainen and Sejnowski, 1995; Nowak et al., 1997) and have

thus been proposed to play an important role in determining func-

tional connectivity among populations of cortical neurons (Sali-

nas and Sejnowski, 2001; Womelsdorf et al., 2007).

These results emphasize the interactions of different temporal

bandwidths and ‘‘windows of opportunity’’ in cortical processing

(Figure 2; see also Canolty et al., 2006; Sirota et al., 2008). Depo-

larizing a cortical neuron by 15–25 mV from true resting potential

to firing threshold may require the summed efforts of prolonged

(e.g., seconds or longer) depolarizations mediated by ongoing

recurrent network activation (e.g., window 1, Figure 2), in addi-

tion to shorter-duration (e.g., tens to hundreds of ms) events

that are brought about through temporary coactivation of

a subpopulation of cortical neurons resulting in rapid changes

in the excitatory/inhibitory balance (e.g., windows 2 and 3,

Figure 2). One advantage to such a framework of temporarily

increased recurrent excitation balanced with inhibition is that

the network can maintain a dynamic range that is close to spike

threshold, and thereby is capable of responding rapidly to small

and fast fluctuations in ensemble dynamics (Kapfer et al., 2007;

Murphy and Miller, 2009; Pouille and Scanziani, 2001; Rudolph

and Destexhe, 2003; Silberberg and Markram, 2007; Swadlow,

2002; Vogels and Abbott, 2009; Wilent and Contreras, 2005). A

concomitant function of this fast-responding, balanced regime

is an ongoing check on runaway excitation that is present in

pathological states, such as cortical seizure activity (McCormick

and Contreras, 2001) and perhaps even during disordered infor-

mation flow characteristic of higher cognitive dysfunctions

(Aghajanian, 2009; Vogels and Abbott, 2009).

The chief advantage of this relatively balanced activity is its

dynamic nature: momentary excesses of excitation, or rapid with-

drawal of inhibition, may transiently depolarize the membrane

potential and lead to enhanced spike generation. The time frame

of these imbalances of excitation or inhibition may be as little as

only a few milliseconds, as evidenced by regular-spiking (RS)

pyramidal neuron firing leading fast-spiking (FS) inhibitory neuron

activity by�4 ms during fast network activation; this temporal lag

has also been confirmed by direct measurement of EPSPs and

IPSPs in nearby pairs of pyramidal neurons (Hasenstaub et al.,

2005; Okun and Lampl, 2008). Moreover, studies in behaving

animals suggest that a time window equal to the period of the

gamma oscillation (�10–30 ms) may be a critical constraint for

the dynamic coupling of interacting neural populations (for

review, see Fries et al., 2007; Harris, 2005). In support of this

crucial temporal window, the membrane potential trajectory

preceding sensory evoked spikes is often composed of a rapid

hyperpolarization-depolarization sequence occurring within

�20 ms (around 50 Hz), presumably mediated by a dynamic inter-

action of inhibitory and excitatory neurons and synaptic poten-

tials (Hasenstaub et al., 2005; Nowak et al., 1997; Poulet and

Petersen, 2008). An important consequence of a hyperpolarizing

potential followed by a rapid depolarizing potential is a lower

spike threshold and a more rapid action potential response

(Azouz and Gray, 2003). Furthermore, this time window of

10–30 ms is also on the same order as the membrane integration

time for cortical pyramidal cells in vivo during spontaneous

activity, further constraining the summation of synaptic barrages

(Leger et al., 2005). Therefore, through multiple mechanisms, this

narrow temporal window dictates how effectively the network

responds to the timing of synaptic potentials (Azouz, 2005; Singer

and Gray, 1995). Accordingly, intracellular studies have revealed

that facilitation of spiking responses during increased levels of

network activity is accompanied by a heightened sensitivity to

rapid depolarizations (indicative of synchronized synaptic

barrages), due to a shorter membrane time constant and accu-

mulation of sodium channel inactivation (Azouz and Gray, 1999;

Henze and Buzsaki, 2001; Nowak et al., 1997). Together these

observations suggest that rapid, balanced interactions between

excitatory and inhibitory circuits in active local networks dictate

neuronal sensitivity to synchronized synaptic inputs.

The excitatory-inhibitory interactions leading to such precise

moments of enhanced excitability have also been examined

in vitro, where spike-triggered averages of randomly fluctuating

artificial excitatory and inhibitory conductance stimuli revealed

that the withdrawal of inhibition may be at least as important

as increased excitation for the production of action potentials

(Hasenstaub et al., 2005; Piwkowska et al., 2008; Tiesinga

et al., 2008). Analytical methods have also suggested that

withdrawal of inhibition may be critical for production of spikes

in vivo (Azouz and Gray, 2008; Rudolph et al., 2007). Further,

when synaptic conductances are pharmacologically separated

in vivo, it is observed that there is more variance in the inhibitory

rather than the excitatory conductances, mirroring the observed

increase in power at higher frequencies in IPSPs over EPSPs

during network activation (Haider et al., 2006; Hasenstaub

et al., 2005). These results strongly suggest that fast variations

in membrane potential that define windows of neuronal excit-

ability depend critically upon distinct subnetworks of inhibitory

neurons and their interactions with locally and distantly projec-

ting excitatory cells.

Of the wide variety of inhibitory interneurons in the cerebral

cortex (Markram et al., 2004), the activity of fast-spiking (FS)

interneurons, especially basket and chandelier cells, is of partic-

ular relevance to the control of rapid variations in membrane

potential at the soma and proximal axon, since FS interneurons

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 177

Neuron

Review

synapse predominantly onto the soma, axon initial segment, and

proximal dendrites of cortical pyramidal cells (Buzsaki and Chro-

bak, 1995; Csicsvari et al., 2003; Penttonen et al., 1998; Peters,

1984). Not only are the synaptic terminals of these cells well-

placed to control the spike output of cortical pyramidal neurons,

but many of the physiological aspects of FS interneurons allow

these cells to communicate activity across a broad range of

temporal frequencies. FS interneurons generate short-duration

action potentials and are capable of discharging at high frequen-

cies with little spike frequency adaptation (McCormick et al.,

1985; Nowak et al., 2003). The responses of their membranes

to the injection of broad-band fluctuations show less decrement

of higher frequencies in comparison to other cell types (Destexhe

et al., 2001; Hasenstaub et al., 2005), a fact that results in part

from the shorter membrane time constant of FS interneurons

(Cruikshank et al., 2007). FS interneurons are better able to main-

tain synaptic transmission at higher frequencies than RS pyra-

midal cells (Galarreta and Hestrin, 1998; Varela et al., 1999),

and they receive kinetically faster EPSPs from presynaptic excit-

atory cells (Thomson et al., 2002). FS interneurons also form

direct electrical synapses with one another, enhancing their

synchronization (Connors and Long, 2004). FS interneurons

in vivo discharge with a wide range of frequencies, including

synchronization to gamma (30–80 Hz) oscillations (Cardin

et al., 2007; Csicsvari et al., 1999; Hasenstaub et al., 2005; Pent-

tonen et al., 1998), and are characterized by a marked sensitivity

to the activation of afferent inputs (Azouz et al., 1997; Cruikshank

et al., 2007; Hirsch et al., 2003; Jones et al., 2000; Nowak et al.,

2007; Swadlow et al., 1998). Finally, direct and temporally

precise activation of FS neurons through genetically encoded

light-activated channels results in the specific enhancement of

higher-frequency oscillations in cortical networks (Cardin et al.,

2009). Together, these results indicate that the cortical FS inter-

neuron network is responsible for conveying broad-band

frequency fluctuations in postsynaptic targets, thereby control-

ling the probability and timing of action potential generation in

pyramidal neurons that contribute to both local network interac-

tions as well as more distant cortico-cortical communication.

All of these results together suggest that rapid modulation of

functional connectivity across cortical areas is critically regulated

by local inhibitory subnetworks responsible for precise spike

timing in nearby pyramidal neurons and thus proper control of

information flow. The cortex contains numerous subtypes of

interneurons, in addition to the FS subtype, that collectively

exhibit a rich repertoire of local network excitatory and inhibitory

interactions, the details of which are only beginning to be re-

vealed (Fuentealba et al., 2008; Kapfer et al., 2007; Kawaguchi,

1997; Klausberger et al., 2003; Markram et al., 2004; Silberberg

and Markram, 2007). Similarly, subclasses of cortical pyramidal

cell can be distinguished based upon their anatomical connec-

tions and morphologies (DeFelipe et al., 2002), intrinsic electro-

physiological properties (Connors and Gutnick, 1990; Douglas

and Martin, 2007a; McCormick et al., 1985; Nowak et al., 2003),

and responses to neurotransmitters (Gil et al., 1997; McCormick,

1992). As with interneurons, these properties are nonrandomly

distributed, yielding strong correlations between all of these

measures (Nelson et al., 2006). Of particular relevance to our

discussion is the finding that connections between cortical

178 Neuron 62, April 30, 2009 ª2009 Elsevier Inc.

neurons are also highly nonrandom, with preferred pathways

and subcircuits appearing to be the rule rather than the exception

(Petreanu et al., 2009; Sawatari and Callaway, 2000; Song et al.,

2005).

The distribution of single excitatory postsynaptic potential

amplitudes between cortical neurons is highly skewed, with

a small but significant fraction exhibiting large values (e.g.,

several mV; Feldmeyer et al., 1999; Markram et al., 1997). During

any moment in time, the pathways that exhibit large amplitudes

may be considered preferred, while those eliciting small-ampli-

tude responses (or failures) are nonpreferred, with a more

moderate range of synaptic strengths in between these two

extremes. The occurrence of bidirectional connectivity among

nearby pyramidal neurons is substantially higher than chance,

as is the probability that a third neuron forms a functionally con-

nected ‘‘triplet’’ with these two interconnected neurons (Song

et al., 2005). Moreover, the strength of synaptic connections

between neurons within the input layer 4 may be quite specific

and restricted to only those neurons that share the same feed-

forward inputs from the thalamus (Feldmeyer et al., 1999). The

feed-forward transfer of information from input layer 4 neurons

to superficial layer neurons is also highly specific; intercon-

nected neurons in layer 2/3 also share inputs from common

pools of layer 4 neurons. However, these pools of feed-forward

inputs are not shared if the neurons in layers 2/3 are not con-

nected to one another (Feldmeyer et al., 1999; Shepherd and

Svoboda, 2005; Yoshimura et al., 2005). Thus, nearby pyramidal

neurons form highly specific connections that establish unique,

but flexible, cortical subnetworks.

Of particular importance to our discussion is the finding that,

regardless of the connectivity relationships among pairs of

neurons, superficial layers receive equipotent global excitatory

synaptic inputs from deep layers and equipotent global inhibitory

input from within a layer (Yoshimura et al., 2005). This raises the

intriguing possibility that nearby and spatially overlapping

neuronal populations that are preferentially interconnected

may be modulated cohesively by more global excitatory and

inhibitory background activity. Moreover, the amplitude of post-

synaptic potentials between cortical pyramidal neurons may be

facilitated by depolarization of the somatic membrane potential

(Shu et al., 2006), adding an additional ‘‘analog’’ mechanism of

enhanced interactions among functionally coactive subnetworks

of pyramidal neurons. All of these studies together suggest that

the functional connectivity among neurons in preferred subnet-

works could be rapidly modulated by synaptic barrages resulting

from low levels of sustained activity across the larger neural pop-

ulation, as we have described. What is the evidence that ongoing

changes in global excitatory and inhibitory background activity

levels may rapidly modulate responsiveness to specific

‘‘signals,’’ such as sensory evoked synaptic potentials?

Rapid Modulation of Sensory Responsivenessin Recurrent Networks In VivoIt is essential to efficient behavior that the responsiveness of

cortical neurons and networks be rapidly modulated according

to changes in context or shifts in behavioral demands. In the

visual cortex, several studies have examined how ongoing

changes in spontaneous activity (as measured, for example, by

Neuron

Review

the EEG) affect visual response properties (e.g., Haider et al.,

2007; Livingstone and Hubel, 1981; Worgotter et al., 1998).

Perhaps the simplest andmost instructive example for our discus-

sion is a study in which the sensory response during the lack of

network activity (during the Down state of the slow oscillation)

was compared with sensory responses elicited when the cortical

network was strongly activated (during the Up state) (Haider et al.,

2007). These experiments clearly showed that the spiking

response to short duration (50 ms) visual stimulation of the recep-

tive field was enhanced, more than two-fold, when primary visual

cortical networks were active versus silent. Importantly, this

enhancement during the Up state was continuously modulated

by ongoing variations in network activity levels. Comparing visual

responses elicited during different levels of the activated (Up)

state revealed that the input-output curve underwent a multiplica-

tive-like increase in gain (of up to 60%) with depolarization of the

recorded neuron induced by increases in local network activity

(Figure 3F). Responses to the intracellular injection of synaptic

conductance-like stimuli in these same neurons also revealed

response enhancement, indicating that the increase in respon-

siveness was due to the neuronal membrane potential level and

not purely the result of alterations in presynaptic activity (Haider

et al., 2007; Shu et al., 2003a). In fact, the dominant factor deter-

mining the enhanced cortical response in this case was the level

of membrane potential depolarization, not the magnitude of

the sensory-evoked synaptic potential, since direct depolariza-

tion of visual cortical neurons also caused a multiplicative-like

increase in neuronal gain (Figure 3D) (Sanchez-Vives et al.,

2000). Could such simple mechanisms underlie response

enhancement as observed in behaving animals?

Perhaps the most well-known example of cortical response

enhancement due to cognitive state is found in tasks engaging

attention. Attentional signals have been proposed to arise from

higher cortical areas (‘‘top-down’’) that facilitate the interactions

between neurons in lower cortical areas that encode object

features that are to be detected (Hahnloser et al., 2002; Miller

and Cohen, 2001; Reynolds and Chelazzi, 2004). Attentional

modulation of neuronal responsiveness is a powerful example

of interareal gain control and has been particularly well studied

in the visual system (Desimone and Duncan, 1995; Fries et al.,

2001; Lee and Maunsell, 2009; Martinez-Trujillo and Treue,

2004; Maunsell and Treue, 2006; McAdams and Reid, 2005;

Reynolds and Heeger, 2009; Reynolds et al., 2000). Spatial

attention can rapidly increase neuronal response amplitude by

up to 40% or more, with results that are consistent with both

a shift along the input axis (e.g., contrast gain enhancement

[Reynolds et al., 2000]) as well as multiplicative response gain

(Martinez-Trujillo and Treue, 2004; Williford and Maunsell,

2006; see also Lee and Maunsell, 2009; Reynolds and Heeger,

2009). Although intracellular examination of the synaptic mecha-

nisms producing these alterations is not yet possible in behaving

animals, in vitro recording, computational models, and intracel-

lular recording studies in anesthetized animals in vivo all suggest

that response facilitation may occur through simple depolariza-

tion of membrane potential via balanced excitatory and inhibitory

synaptic bombardment.

Can small amounts of depolarization have a behaviorally rele-

vant effect? The firing rate of cortical neurons is rather sensitive

to membrane potential, increasing by an average of three to

seven spikes/s per mV of depolarization (depending on the posi-

tion along the input-output curve, Figure 1A) (Carandini and

Ferster, 2000; Haider et al., 2007; Sanchez-Vives et al., 2000).

These results suggest that the ongoing variations in membrane

potential that occur in the waking state, which cover a range of

�10 mV, remarkably similar to the range of the Up state (Steriade

et al., 2001), may have a very strong, multiplicative-like effect on

neuronal responsiveness. Indeed, studies of behaving primates

occasionally show increases in spontaneous activity of a few

spikes/s upon engagement of selective attention, even in the

absence of visual stimuli (Luck et al., 1997; Reynolds et al.,

2000; Sundberg et al., 2009). We suspect that attention may

result in a slight depolarization of cortical neurons in the attended

sensory dimension (which may be spatial location, form, color,

motion, etc.), a mechanism that should also be accompanied

by an increase in background activity levels. Since depolariza-

tion of %1 mV may cause a significant increase in neuronal

responsiveness, in both excitatory and inhibitory neurons (e.g.,

three to seven spikes/s; Carandini and Ferster, 2000; Contreras

and Palmer, 2003), we suspect that gain modulation in subnet-

works of excitatory and inhibitory neurons engaged during

behavior may be achieved by moderate average changes in

the membrane potential, ranging from submillivolt up to several

millivolts. These small, rapid changes in membrane potential

are distinct from those mediated by classical Up and Down

states, although they may use similar network mechanisms.

How is membrane potential rapidly modulated in individual

neurons along with changing background activity levels? All of

our observations thus far suggest a basic mechanism for

sensory response facilitation: a balanced mixture of excitatory

and inhibitory synaptic bombardment tonically depolarizes

target neurons. Importantly, as long as the reversal potential of

this mixture of excitatory and inhibitory barrages is more positive

than the resting membrane potential, the neuron will rapidly

depolarize. This depolarization will be counteracted by the acti-

vation of intrinsic K+ conductances and facilitated by the activa-

tion of the persistent Na+ current (Bean, 2007; Llinas, 1988;

Waters and Helmchen, 2006). This net synaptic depolarization

in the presence of a noisy baseline can yield a nonlinear, multipli-

cative-like increase in neuronal gain (Murphy and Miller, 2003).

This enhanced activity is subsequently disseminated within a

coactivated ensemble of neurons, both local and distant (Fig-

ure 1B). The increased responsiveness of these coactivated

neurons may result in facilitated interactions between the

members, resulting in a competitive advantage over other

groups of neurons. Thus, moderate amounts of depolarization

in subgroups of local cortical neurons may have a powerful,

network-wide effect upon responsiveness and provide the

main cellular mechanism for rapidly altering functional connec-

tivity in cortical networks. The facilitating interactions may then

allow the activity of coactivated neurons to temporarily increase

their discharge in a ‘‘pop-out’’ effect employing ‘‘winner-take-

all’’ or ‘‘attractor’’ dynamics. Such dynamics are likely to underlie

diverse behavioral phenomena such as binocular rivalry, contex-

tual modulation, binary decision making, and working memory

(Albright and Stoner, 2002; Ardid et al., 2007; Barraclough

et al., 2004; Gilbert and Sigman, 2007; Hahnloser et al., 2002;

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 179

Neuron

Review

Logothetis, 1998; Luna et al., 2005; Rutishauser and Douglas,

2008; Vogels et al., 2005; Wang, 2008; Wong et al., 2007). Since

the activity of neuronal populations is constantly changing and

influencing other populations, a continual and dynamic flow of

synaptic barrages provides the context upon which neuronal

signals interact. The moment-to-moment evolution of synaptic

activity not only links groups of neurons and cortical territories

together but also links the past to the present and future of

cortical network behavior (e.g., Harris, 2005).

Rapid Modulation of Membrane Potentialand Background Activity during WakingIn the preceding sections, we have detailed how specific proper-

ties of balanced excitatory and inhibitory synaptic barrages may

result in transient depolarizations that interact with nonlinear

spike transformation to modulate the gain and functional

connectivity of cortical networks. What is the evidence that the

mechanisms revealed in anesthetized animals and in slices

in vitro are present in awake animals? Intracellular recordings

in waking rodents, cats, and monkeys indicate that neuronal

activity in the awake cortex exhibits an activated baseline state,

similar to that of the Up state of slow wave sleep (Chen and Fetz,

2005; Crochet and Petersen, 2006; Lee et al., 2006; Steriade

et al., 2001). Indeed, the transition from slow wave sleep to

waking is associated with a loss of Down states, giving the

impression that the waking state is a prolonged Up-like state

that merely lacks rhythmic hyperpolarization or deactivation

(Steriade et al., 2001). These findings suggest that under-

standing the cellular mechanisms of the Up state of slow wave

sleep and anesthesia may yield clues about the basic mecha-

nisms of the baseline waking state (Destexhe et al., 2007).

Interwoven with this baseline activation, there are rapid modu-

lations of neuronal responsiveness and activity, in relation to

behavioral demands and cortical processing (Brecht et al.,

2004b; Ferezou et al., 2006; Fries et al., 2007; Pesaran et al.,

2002; Poulet and Petersen, 2008; Singer and Gray, 1995). The

baseline membrane potential of the waking animal exhibits

a steady depolarization such that it is typically within a few milli-

volts of firing threshold, but variations of up to several millivolts

rapidly occur around this baseline level (Chen and Fetz, 2005;

Crochet and Petersen, 2006; Lee et al., 2006; Steriade et al.,

2001). There are further important variations on this theme of

sustained activation or depolarization during waking. First, the

waking cortex exhibits a broad range of states that correlate

with the general level of arousal (e.g., drowsiness, quiet waking,

alert, attentive, active behavior, etc. [e.g., Bezdudnaya et al.,

2006]) as well as the specifics of the behavioral task at hand.

Second, the level of depolarization and synaptic activity is likely

to vary dramatically between different cell types and layers,

since extracellular recordings in awake animals reveal large

differences in spontaneous and evoked discharge rates between

excitatory and FS inhibitory cells, as well as in cells across

cortical lamina (Beloozerova et al., 2003; Brecht et al., 2004b;

de Kock et al., 2007; Houweling and Brecht, 2008; Mitchell

et al., 2007).

As mentioned, a key difference between slow wave sleep and

active waking is that the synchronized and prolonged hyperpola-

rizing epochs similar to Down states appear to be relatively rare,

180 Neuron 62, April 30, 2009 ª2009 Elsevier Inc.

especially during active behavior. In addition, while the transi-

tions between Up and Down states of slow wave sleep and anes-

thesia exhibit broad propagation and synchrony, the occurrence

and synchronization of faster oscillations of membrane potential

and action potential activity of the attentive and waking brain are

much more restricted to select populations of neurons (Destexhe

et al., 1999; Fujisawa et al., 2008; Poulet and Petersen, 2008;

Tsujimoto et al., 2008).

In the inattentive or quiet resting state, the situation may be

different. Recent whole-cell recordings and voltage-sensitive

dye imaging of somatosensory cortex in quietly resting rodents

indicate that large regions of cortex exhibit spatiotemporally

correlated waves of hyperpolarization and depolarization at

1–5 Hz, in some ways similar to Up and Down states (Crochet

and Petersen, 2006; Poulet and Petersen, 2008). These large,

slow oscillations of membrane potential are highly synchronized

between neighboring neurons and thus generate slow oscilla-

tions in the local field potential (Poulet and Petersen, 2008).

During active behavior (i.e., vibrissal whisking), the amplitude

of these waves is dramatically reduced and is replaced with

membrane potential dynamics that more closely resemble those

previously observed in awake primates and cats (Crochet and

Petersen, 2006; Ferezou et al., 2007; Poulet and Petersen,

2008). Interestingly, action potentials in individual cortical

neurons during active whisking of the vibrissa are each preceded

by a large (5–10 mV) fast depolarizing, presumably synaptic,

event. In contrast to the slow fluctuations of the quiet waking

state, this fast synaptic event during active behavior does not

occur throughout the local network (Poulet and Petersen,

2008). This is expected if the action potential activity of individual

cells is driven by unique combinations of relatively sparse, but

synchronized, discharge within the large numbers of presynaptic

cortical neurons that influence the recorded cell (DeWeese and

Zador, 2006; Houweling and Brecht, 2008). We hypothesize

that each action potential is initiated by the interaction of multiple

synaptic components: one which collectively sets the general

level of membrane potential properties (particularly the level of

depolarization) and another which occurs in a well-defined

subnetwork of cortical neurons, the synchronized activity of

which determines the precise timing of the action potential.

How likely is it that the maintained increase in neuronal firing

occurring during waking has the same basis as the maintained

increase in activity associated with the Up state of the slow oscil-

lation? These two periods of activity are similar in that they both

are associated with a highly variable pattern of action potential

generation (Compte et al., 2003a; Hasenstaub et al., 2005; Shad-

len and Newsome, 1998), which is indicative of underlying varia-

tion in excitatory and inhibitory synaptic interactions (Mainen and

Sejnowski, 1995; McCormick et al., 2007; Shadlen and News-

ome, 1998; Vogels and Abbott, 2009). In addition, both changes

in persistent activity in waking animals and transitions into and

out of Up states can be very rapid, within 50–100 ms (Figure 2).

Finally, cortical Up states are typically associated with high-

frequency LFP oscillations (Figure 2) (Buzsaki and Draguhn,

2004; Destexhe et al., 2007; Hasenstaub et al., 2005) as is often

seen during increases in cortical activity in behaving animals

(Csicsvari et al., 2003; Fries et al., 2001, 2007; Pesaran et al.,

2008; Singer and Gray, 1995). In contrast to persistent cortical

Neuron

Review

activity in the waking animal, however, cortical Up states occur

nearly simultaneously throughout all or nearly all neurons in the

local network (Contreras and Steriade, 1995; Haider et al.,

2006; Hasenstaub et al., 2005; Lampl et al., 1999; Volgushev

et al., 2006), while persistent activity in the waking state can be

much sparser, with neighboring neurons capable of behaving

relatively independently (Genovesio et al., 2005). In particular,

intracellular recordings in awake animals suggests that single

spikes in a given neuron are preceded by relatively large-ampli-

tude synaptic potentials that are not shared or visible in the

membrane potential of nearby neurons (Chen and Fetz, 2005;

Poulet and Petersen, 2008). Similarly, the simultaneously

recorded activity across large populations of neurons may be

highly independent and exhibit significant trial-to-trial variability

(Greenberg et al., 2008; Kerr et al., 2007; Ohki et al., 2005). If

nearby neurons receive relatively unique constellations of strong

connections amidst a sea of weaker ones, it becomes possible

that highly specific and sparse transmission of sensory

responses occurs even if background activity levels are shared

(Olshausen and Field, 2004). In this manner, small differences

(<1 to a few millivolts) in synaptic potentials, riding on top of a

depolarized baseline membrane potential, would produce

nonlinear increases in spike output, as we have already detailed,

and facilitate the temporary interaction of specific pathways. We

suggest that although similar cellular mechanisms may be

involved in both awake and anesthetized or sleeping cortical

dynamics, network interactions in awake behaving animals are

less synchronized as compared to sleeping or anesthetized

conditions, restricting interactions to select subnetworks of

neurons that vary with the behavioral task at hand. We presume

that this parcellation of cortical networks in behaving animals

depends upon the rapid interaction of cortico-cortical excitatory

and local inhibitory subnetworks, utilizing similar though not

identical mechanisms as those observed in anesthetized and

sleeping animals.

Cortical Network Dynamics during Behavior: NeuralVariability Reflects Changes in Functional ConnectivityThe mechanisms that enable local cortical networks to quickly

change their responsiveness through recurrent interactions

have strong implications for rapid cortical dynamics in awake,

behaving animals. As we have seen, recurrent networks in the

cerebral cortex may control the amplitude, variability, and timing

of action potential responses, so as to provide appropriate

control over the spatiotemporal flow of neuronal activity elicited

by the rapidly changing demands of behavior. These issues

have been addressed in vivo with a variety of preparations,

most notable of which are recent investigations of trial-to-trial

response variability and contextual modulation of response

amplitude and timing.

Repeated presentation of a sensory stimulus, or performance

of a mental process or task, usually involves significant trial-

to-trial variability (Shadlen and Newsome, 1998). Although

portions of this variability arise from the inability to precisely

control experimental variables (such as eye movements), a large

part of it derives from spontaneous fluctuations in the activity

and state of the cortical network (Arieli et al., 1996; Fox et al.,

2007; Gur et al., 1997). A case in point: spontaneous variations

in firing of motor cortical neurons during a delay period spanning

stimulus delivery and ensuing motor command (i.e., when the

movement was being planned) are highly correlated with the

subsequent variability in velocity of the instructed movement

(Churchland et al., 2006a). These firing rate variations were

moderate (tens of spikes per second) and therefore may be

reflective of small to moderate (<2–4 mV) changes in membrane

potential. Indeed, if depolarization of neurons was responsible

for increases in neuronal responsiveness, it may also explain

the decrease in reaction time that is associated with trials in

which the patterns of spikes (within the spike train) exhibited

less variability (Churchland et al., 2006b). Sustained depolariza-

tion should cause spikes to occur earlier and more frequently

(since the membrane potential is closer to threshold) and should

also decrease long interspike intervals. These effects of depolar-

ization should enhance neuronal interactions and decrease

behavioral reaction times. Such results support our conjecture

that trial-to-trial variability is indicative of alterations in functional

connectivity, as determined by the interplay of ongoing patterns

of synaptic activity with the anatomical connections of the

cortex, and illustrate that changes in background activity reflect

changes in behavioral performance.

While it is not yet routinely possible to directly examine

membrane potential dynamics underlying variations in functional

connectivity in awake, behaving animals, it is possible to examine

changes in functional connectivity by performing multiple single-

unit recordings. Alterations in functional connectivity that are

driven by behavioral context have been recently demonstrated

in the medial temporal area (MT) of the monkey visual system (Co-

hen and Newsome, 2008). Pairs of neurons in this motion-sensi-

tive visual area changed the degree of their correlated activity

during the display of noncoherent ‘‘noisy’’ stimuli depending on

whether the animal was attempting to detect motion in the direc-

tion shared by the receptive field properties of the two neurons, or

in an orthogonal, nonshared direction (Cohen and Newsome,

2008). This change in correlated activity that depended upon

behavioral context indicates that single cells were able to rapidly

and flexibly participate in different ensembles of interacting

neurons, even in the presence of identical visual stimuli. Modula-

tion of neuronal interactions by internal activity states has also

been demonstrated in simultaneous ensemble recordings from

>100 rat prefrontal cortical neurons while animals ran a maze

for a food reward. These experiments showed that putative

monosynaptic influences appeared in spike cross-correlation

histograms transiently and only during specific portions of the

task (Fujisawa et al., 2008) (Figures 4A and 4B). Subsequently,

diagrams of functionally interconnected networks of cells re-

vealed large, distributed functional neuronal ensembles in the

medial prefrontal cortex, in which the participating members

change according to the behavioral performance of the animal

and the spike history of the neuron and network (Figures 4C

and 4D). These recordings also support the idea that the cortex

operates through a relatively sparse code in which individual

neurons discharge relatively rarely (average firing rate of <1 Hz),

but with transient and marked increases in firing occurring in

response to activation of particular patterns of network activity

tied to unique behavioral states (Olshausen and Field, 2004; Sho-

ham et al., 2006). Such rapid alterations in prefrontal network

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 181

Neuron

Review

A

B

C

D

Figure 4. State-Dependent Modifications ofFunctional Connectivity during Behavior(A) Short-latency (putative monosynaptic) neuronalinteractions between simultaneously recordedputative pyramidal (red) and inhibitory (blue) cellsin rat prefrontal cortex during the performance ofa maze task, as judged from extracellular record-ings and cross-correlation spike histograms.(B) For this pair, the strength of interaction varied asa function of spatial position in the maze, with thestrongest interactions apparent near the decisionpoint of the maze (middle).(C) Local network diagram constructed fromensembles of simultaneously recorded pyramidaland putative inhibitory neurons reveal similarshort-latency enhancement of monosynapticinteractions that are spatially dependent andwhose strength varies significantly with behavioraldemands. Interactions in red (1, 2) are significantlyenhanced on rightward turns during maze running.(D) Same ensemble of neurons upon leftward turnsduring maze running. Some functional connectionsthat are spatially modulated are present in bothdirections of running (2), while others appear (3)or disappear (1) during leftward turns, indicatingrapid changes in functional connectivity duringongoing behavior. Adapted by permission fromMacmillan Publishers Ltd: Nature Neuroscience(Fujisawa et al., 2008), copyright 2008.

dynamics could readily communicate with specific motor cortical

ensembles to control the time of action (Narayanan and Laubach,

2006).

Although the cellular mechanisms underlying transient periods

of functional connectivity cannot be examined with extracellular

recordings, several possibilities exist, including synaptic facilita-

tion and depression, or network interactions involving the large

number of interposed neurons that are not recorded. For

example, the facilitation of a monosynaptic excitatory interaction

between two neurons may be achieved through the rapid

depolarization of both the pre- and postsynaptic cells by a third

neuron (or neurons), allowing the transient appearance of

this monosynaptic connection in the extracellularly recorded

spike-rate histograms during restricted portions of behavior.

Keep in mind that this depolarization could result from increases

in excitation, withdrawal of inhibition, or some balanced mixture

between the two. A mechanism relying upon depolarization of

interacting or interposed neuronal groups to rapidly facilitate

functional connections, in highly specific task-dependent

ways, implies that the patterns of depolarization in individual

cortical neurons selectively regulate how information flows

through the cortical network. This routing of information by

transient increases in net depolarizing synaptic bombardment,

or withdrawal of net hyperpolarizing synaptic activity, will occur

in nearby and often spatially overlapping neuronal populations.

Changes in synaptic bombardment could modulate overlapping

circuits differentially owing to differences in the timing of

synaptic potentials received by these networks, differences

in the amplitude of synaptic interconnections, or differences

in intrinsic properties of the constituent neurons, as we have

discussed. Of course, these possibilities are not mutually exclu-

sive and will require well-formulated hypotheses and experi-

ments to elucidate the underlying mechanisms in behaving

animals.

182 Neuron 62, April 30, 2009 ª2009 Elsevier Inc.

Future Directions for Examinationof Rapid Cortical DynamicsThe functional analysis of neuronal circuits has always depended

upon the application of some variation of three basic techniques:

record, stimulate, or lesion. Current investigations into the

dynamics of cortical networks are no exception, although recent

advances in all three of these approaches have greatly increased

the finesse of the experimenter. Ideally, to fully investigate

cortical network dynamics, an investigator would have detailed

information on the connectivity pattern and activity state of all

the cells involved, as well as the ability to rapidly manipulate

and monitor any of the constituent cells during natural behavior.

The densely interconnected and ever-changing activity of intact

cortical networks makes this ideal very far from practical reach

now, and maybe forever. However, the application of current

technologies, as well as those that are just being developed,

promises to yield information that, while not all encompassing,

is certainly toward this ideal.

Monitoring the activity of multiple interconnected cortical

neurons is now possible through either arrays of extracellular

recording electrodes (for review, see Buzsaki, 2004) or the bulk

application of Ca2+-sensitive dyes (Garaschuk et al., 2006; Kerr

et al., 2005; Ohki et al., 2005; Sato et al., 2007). Combining the

latter with mice in which particular subpopulations of neurons

are genetically labeled with fluorescent proteins may yield valu-

able information on the dynamic interactions of these identified

cell types with other subpopulations of cortical neurons or during

sensory stimulation (Sohya et al., 2007). This approach is being

developed to study the dynamics of neural activity in vivo, using

two-photon microscopy (Svoboda and Yasuda, 2006), and could

be readily applied to active cortical slices in vitro (Sanchez-Vives

and McCormick, 2000). One serious limitation to this technique

is the relatively poor time resolution, primarily owing to the

slow kinetics of neuronal Ca2+ levels (Helmchen et al., 1996).

Neuron

Review

Unfortunately, many of the most interesting and relevant

cortical dynamics occur at a rate that is more rapid than can

be currently resolved using internal Ca2+ levels as a reporter.

These limitations can be overcome with the use of voltage-

sensitive dyes, but when applied extracellularly these suffer

from nonspecific staining of all (including nonneuronal) cortical

tissue and therefore are unable to reveal single-cell dynamics

(Grinvald and Hildesheim, 2004; Homma et al., 2009). While

combined extracellular or intracellular recording and imaging

may yield precise temporal and/or membrane potential data,

this will necessarily be limited to a small number of cells and

requires a suitable preparation and significant effort and exper-

tise on the part of the experimenter. Nonetheless, the combina-

tion of imaging and traditional electrophysiological recording

methods will certainly be advantageous for the monitoring

of cortical network dynamics. Important questions to be

answered include: What is the network basis of large, fast

synaptic events that give rise to action potentials? Do action

potentials arise from the activation of a unique combination of

sparse yet reliable neurons in restricted subnetworks, or do

spikes mainly arise from distributed and heterogeneous patterns

of activity? If there are repeating patterns, do they occupy

succinct spatial and temporal bandwidths? How do these two

different components interact to give rise to sensory-evoked

action potentials?

On the side of stimulation and lesion techniques, it is now

possible, using optogenetic techniques, to selectively depolarize

or hyperpolarize the membrane potential of selected subgroups

of neurons with the application of light (Boyden et al., 2005;

Zemelman et al., 2002; Zhang et al., 2007). Channelrhodopsin

and halorhodopsin are two light-sensitive and genetically encod-

able proteins that depolarize or hyperpolarize the membrane

potential and can be delivered to subpopulations of neurons

through the use of cell-specific promoters, retroviral delivery, or

in utero electroporation (Cardin et al., 2009; Gradinaru et al.,

2007; Huber et al., 2008; Nagel et al., 2003; Petreanu et al.,

2007). This methodology has already been used to optically stim-

ulate neural populations that drive perception and behavior (Dou-

glass et al., 2008; Huber et al., 2008) and has also been used to

change global brain state (Adamantidis et al., 2007). It is likely

that the expression of these light-sensitive channels will be under

increasingly more precise genetic control (e.g., the Cre-lox

system), allowing investigators to make use of the ever increasing

variety of mice that express Cre in specific subpopulations of

neurons (Cardin et al., 2009; Luo et al., 2008). The resulting

expression of light-sensitive channels/transporters in subpopula-

tions of neurons will allow investigators to control the membrane

potential of these cells specifically and noninvasively. This tech-

nique could be used to examine the influence of changes in

membrane potential of a particular subclass of neurons (e.g.,

FS interneurons) on the operation of the network (Cardin et al.,

2009). The targeting of light-sensitive channels/transporters to

particular locations in neurons (e.g., axon initial segment, apical

and basal dendrites) would also be useful for examining the role

of each of these substructures in cortical network dynamics

(Gradinaru et al., 2007). One general question that could be ad-

dressed is: Does subthreshold manipulation of membrane poten-

tial, conductance, and variance in subnetworks and/or substruc-

tures of pyramidal neurons confer a competitive advantage such

that these cells exhibit enhanced network interactions, e.g.,

during bistable sensory perception or binary decision making?

As all techniques have their advantages and limitations, signifi-

cant difficulties to these optogenetic strategies remain, particu-

larly in the titration of expression level needed for subthreshold

or suprathreshold activation, the cell-type specificity of channel

expression, as well as the toxicity or compensatory changes

induced by specific levels of protein expression or ionic disrup-

tion. Careful researchers will always keep in mind that any lesion

or stimulation technique suffers from problems of interpretation

due to, for example, rapid compensatory mechanisms.

As always, constrained and testable hypotheses on the mech-

anisms of cortical dynamics are of utmost value to the experi-

menter. One key hypothesis put forth in our discussion is that

behaviorally relevant changes in neuronal excitability should be

associated with changes in membrane potential, conductance,

and variance. As we have emphasized throughout this review,

the interaction of known (e.g., sensory) ‘‘signals’’ with changes

in background activity (‘‘noise’’) are particularly important for

the rapid modulation of neuronal responsiveness. For example,

increases in neuronal excitability associated with shifts in atten-

tion may be mediated by either depolarization of cortical neurons

through changes in ongoing barrages of synaptic activity (Haider

et al., 2007; Murphy and Miller, 2003) or by decreases in

membrane potential variance and conductance owing to

decreases in ongoing synaptic barrages (Chance et al., 2002).

Both of these hypotheses are testable with intracellular and/or

whole-cell recordings in awake behaving animals, a technique

that is thankfully starting to move out of the shadows as a party

trick and into the limelight as a serious tool for the neurophysiol-

ogist (Brecht et al., 2004a; Chen and Fetz, 2005; Houweling and

Brecht, 2008; Lee et al., 2006; Petersen et al., 2003; Poulet and

Petersen, 2008; Timofeev et al., 2001).

Finally, our understanding of feed-forward (i.e., thalamic) acti-

vation of sensory cortical networks is relatively well-understood

compared to our understanding of the mechanisms by which

long-range and feed-back connections activate cortical circuits.

This is of crucial importance to our understanding of rapid

cortical dynamics, for it is the distributed activity traversing

long-range connections across multiple cortical areas that

underlies the unity and diversity of behavior. We suspect that

long-range synaptic connections have the same specificity and

‘‘preferred’’ functionality as observed in local cortical subnet-

works, and these properties may allow, for example, specific

subnetworks of neurons to perform local versus long-range

computations (Le Be et al., 2007; Petreanu et al., 2007). In

combination with the above-mentioned techniques, it may

become possible to selectively perturb and reroute cortical

activity during behavior and functionally manipulate higher-level

perception, e.g., altering motion discrimination by manipulating

neuronal signals in primate MT or altering the memory of specific

visual cues by perturbing persistent activity in prefrontal cortex.

Given the pace of technological development and experimental

application in just the last 30 years since the first single-channel

patch-clamp recordings (Neher and Sakmann, 1976; Sakmann,

2006), this short list of dream experiments will likely come to

reality in the not-so-distant future.

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 183

Neuron

Review

ConclusionA fundamental task of cortical network operation is the dynamic

and rapid modulation of neuronal responsiveness according to

the ever-changing demands of behavior. Anatomical constraints

dictate that the majority of a single neuron’s activity is deter-

mined by fluctuations in local network activity levels, which

necessarily engage the unique recurrent properties of excitatory

and inhibitory subnetworks. We have outlined how the challenge

of rapid flexibility faced by the cortex could be solved by a simple

and general mechanism: modulation of neuronal membrane

potential through concerted synaptic bombardment in specific

excitatory and inhibitory subnetworks of functionally connected

neurons. This bombardment causes a noisy and elevated level of

depolarization that can enhance responsiveness rapidly and in

a multiplicative manner to a variety of inputs. Precise control of

membrane potential fluctuations via local network activity may

serve as the main mechanism to dynamically link one cortical

area to another, one cortical neuron to another, and even

substructures of individual neurons (e.g., dendritic branches)

together. Therefore, synaptic barrages operate in a holistic

manner: functionally associating and dissociating information

across vast cortical territory, while simultaneously modulating

interactions both between and within individual cortical neurons.

This fundamental operating principle of the cortex utilizes the

interaction of a highly recurrent local cortical network architec-

ture with intrinsic nonlinearities to rapidly and dynamically alter

neuronal responsive to incoming inputs. These enhanced local

network interactions in turn facilitate communication with distant

cortical neurons through long-range connections. By selectively

and temporally modulating the ongoing patterns of activity in

distal cortical regions, shared windows of excitability— deter-

mined by the biophysical rules of recurrent networks—serve to

transiently link one cortical area to another, facilitating network

computation and synchronous arrival of information in down-

stream targets. Our understanding of the rapid control of

synaptic barrages in local cortical networks has gained consid-

erably from the study of activated cortical states in a diversity

of preparations, from anesthetized, sleeping, and behaving

animals, to cortical slices in vitro and models in silico. A

continued dialog between these and other rapidly emerging

physiological and computational approaches will be necessary

to unravel the full details of local and large-scale network interac-

tions of the cerebral cortex.

ACKNOWLEDGMENTS

We thank past and present members of the McCormick lab, James Mazer, JonTouryan, Mark Laubach, and the anonymous reviewers for critical and helpfulcomments. This work was supported by the NIH, NINDS, NEI, and the KavliInstitute for Neuroscience.

REFERENCES

Abeles, M., Bergman, H., Gat, I., Meilijson, I., Seidemann, E., Tishby, N., andVaadia, E. (1995). Cortical activity flips among quasi-stationary states. Proc.Natl. Acad. Sci. USA 92, 8616–8620.

Adamantidis, A.R., Zhang, F., Aravanis, A.M., Deisseroth, K., and de Lecea, L.(2007). Neural substrates of awakening probed with optogenetic control ofhypocretin neurons. Nature 450, 420–424.

184 Neuron 62, April 30, 2009 ª2009 Elsevier Inc.

Aghajanian, G.K. (2009). Modeling ‘‘psychosis’’ in vitro by inducing disorderedneuronal network activity in cortical brain slices. Psychopharmacology (Berl),in press. Published online February 25, 2009. 10.1007/s00213-009-1484-9.

Albright, T.D., and Stoner, G.R. (2002). Contextual influences on visualprocessing. Annu. Rev. Neurosci. 25, 339–379.

Alvarez, V.A., and Sabatini, B.L. (2007). Anatomical and physiological plasticityof dendritic spines. Annu. Rev. Neurosci. 30, 79–97.

Anderson, J.S., Lampl, I., Gillespie, D.C., and Ferster, D. (2000). The contribu-tion of noise to contrast invariance of orientation tuning in cat visual cortex.Science 290, 1968–1972.

Ardid, S., Wang, X.J., and Compte, A. (2007). An integrated microcircuit modelof attentional processing in the neocortex. J. Neurosci. 27, 8486–8495.

Arieli, A., Sterkin, A., Grinvald, A., and Aertsen, A. (1996). Dynamics of ongoingactivity: explanation of the large variability in evoked cortical responses.Science 273, 1868–1871.

Ayaz, A., and Chance, F.S. (2009). Gain modulation of neuronal responsesby subtractive and divisive mechanisms of inhibition. J. Neurophysiol. 101,958–968.

Azouz, R. (2005). Dynamic spatiotemporal synaptic integration in corticalneurons: neuronal gain, revisited. J. Neurophysiol. 94, 2785–2796.

Azouz, R., and Gray, C.M. (1999). Cellular mechanisms contributing toresponse variability of cortical neurons in vivo. J. Neurosci. 19, 2209–2223.

Azouz, R., and Gray, C.M. (2003). Adaptive coincidence detection anddynamic gain control in visual cortical neurons in vivo. Neuron 37, 513–523.

Azouz, R., and Gray, C.M. (2008). Stimulus-selective spiking is driven by therelative timing of synchronous excitation and disinhibition in cat striate neuronsin vivo. Eur. J. Neurosci. 28, 1286–1300.

Azouz, R., Gray, C.M., Nowak, L.G., and McCormick, D.A. (1997). Physiolog-ical properties of inhibitory interneurons in cat striate cortex. Cereb. Cortex 7,534–545.

Barraclough, D.J., Conroy, M.L., and Lee, D. (2004). Prefrontal cortex anddecision making in a mixed-strategy game. Nat. Neurosci. 7, 404–410.

Bean, B.P. (2007). The action potential in mammalian central neurons. Nat.Rev. Neurosci. 8, 451–465.

Beloozerova, I.N., Sirota, M.G., and Swadlow, H.A. (2003). Activity of differentclasses of neurons of the motor cortex during locomotion. J. Neurosci. 23,1087–1097.

Bernander, O., Douglas, R.J., Martin, K.A., and Koch, C. (1991). Synapticbackground activity influences spatiotemporal integration in single pyramidalcells. Proc. Natl. Acad. Sci. USA 88, 11569–11573.

Bezdudnaya, T., Cano, M., Bereshpolova, Y., Stoelzel, C.R., Alonso, J.M., andSwadlow, H.A. (2006). Thalamic burst mode and inattention in the awakeLGNd. Neuron 49, 421–432.

Binzegger, T., Douglas, R.J., and Martin, K.A. (2004). A quantitative map of thecircuit of cat primary visual cortex. J. Neurosci. 24, 8441–8453.

Blackwell, K.T., Czubayko, U., and Plenz, D. (2003). Quantitative estimate ofsynaptic inputs to striatal neurons during up and down states in vitro. J. Neuro-sci. 23, 9123–9132.

Borg-Graham, L.J., Monier, C., and Fregnac, Y. (1998). Visual input evokestransient and strong shunting inhibition in visual cortical neurons. Nature393, 369–373.

Boyden, E.S., Zhang, F., Bamberg, E., Nagel, G., and Deisseroth, K. (2005).Millisecond-timescale, genetically targeted optical control of neural activity.Nat. Neurosci. 8, 1263–1268.

Brecht, M., Fee, M.S., Garaschuk, O., Helmchen, F., Margrie, T.W., Svoboda,K., and Osten, P. (2004a). Novel approaches to monitor and manipulate singleneurons in vivo. J. Neurosci. 24, 9223–9227.

Brecht, M., Schneider, M., Sakmann, B., and Margrie, T.W. (2004b). Whiskermovements evoked by stimulation of single pyramidal cells in rat motor cortex.Nature 427, 704–710.

Neuron

Review

Buzsaki, G. (2004). Large-scale recording of neuronal ensembles. Nat. Neuro-sci. 7, 446–451.

Buzsaki, G., and Chrobak, J.J. (1995). Temporal structure in spatiallyorganized neuronal ensembles: a role for interneuronal networks. Curr. Opin.Neurobiol. 5, 504–510.

Buzsaki, G., and Draguhn, A. (2004). Neuronal oscillations in cortical networks.Science 304, 1926–1929.

Canolty, R.T., Edwards, E., Dalal, S.S., Soltani, M., Nagarajan, S.S., Kirsch,H.E., Berger, M.S., Barbaro, N.M., and Knight, R.T. (2006). High gamma poweris phase-locked to theta oscillations in human neocortex. Science 313, 1626–1628.

Caporale, N., and Dan, Y. (2008). Spike timing-dependent plasticity: a Hebbianlearning rule. Annu. Rev. Neurosci. 31, 25–46.

Carandini, M., and Ferster, D. (2000). Membrane potential and firing rate in catprimary visual cortex. J. Neurosci. 20, 470–484.

Cardin, J.A., Palmer, L.A., and Contreras, D. (2007). Stimulus feature selec-tivity in excitatory and inhibitory neurons in primary visual cortex. J. Neurosci.27, 10333–10344.

Cardin, J.A., Palmer, L.A., and Contreras, D. (2008). Cellular mechanismsunderlying stimulus-dependent gain modulation in primary visual cortexneurons in vivo. Neuron 59, 150–160.

Cardin, J.A., Carlen, M., Meletis, K., Knoblich, U., Zhang, F., Deisseroth, K.,Tsai, L.-H., and Moore, C.I. (2009). Activation of fast spiking interneuronsinduces gamma oscillations and shapes sensory transmission. Nature,in press.

Chance, F.S., Abbott, L.F., and Reyes, A.D. (2002). Gain modulation frombackground synaptic input. Neuron 35, 773–782.

Chen, D., and Fetz, E.E. (2005). Characteristic membrane potential trajectoriesin primate sensorimotor cortex neurons recorded in vivo. J. Neurophysiol. 94,2713–2725.

Churchland, M.M., Afshar, A., and Shenoy, K.V. (2006a). A central source ofmovement variability. Neuron 52, 1085–1096.

Churchland, M.M., Yu, B.M., Ryu, S.I., Santhanam, G., and Shenoy, K.V.(2006b). Neural variability in premotor cortex provides a signature of motorpreparation. J. Neurosci. 26, 3697–3712.

Cohen, M.R., and Newsome, W.T. (2008). Context-dependent changes infunctional circuitry in visual area MT. Neuron 60, 162–173.

Compte, A., Constantinidis, C., Tegner, J., Raghavachari, S., Chafee, M.V.,Goldman-Rakic, P.S., and Wang, X.J. (2003a). Temporally irregular mnemonicpersistent activity in prefrontal neurons of monkeys during a delayed responsetask. J. Neurophysiol. 90, 3441–3454.

Compte, A., Sanchez-Vives, M.V., McCormick, D.A., and Wang, X.J. (2003b).Cellular and network mechanisms of slow oscillatory activity (<1 Hz) and wavepropagations in a cortical network model. J. Neurophysiol. 89, 2707–2725.

Connors, B.W., and Gutnick, M.J. (1990). Intrinsic firing patterns of diverseneocortical neurons. Trends Neurosci. 13, 99–104.

Connors, B.W., and Long, M.A. (2004). Electrical synapses in the mammalianbrain. Annu. Rev. Neurosci. 27, 393–418.

Contreras, D., and Steriade, M. (1995). Cellular basis of EEG slow rhythms:a study of dynamic corticothalamic relationships. J. Neurosci. 15, 604–622.

Contreras, D., and Palmer, L. (2003). Response to contrast of electrophysio-logically defined cell classes in primary visual cortex. J. Neurosci. 23, 6936–6945.

Cossart, R., Aronov, D., and Yuste, R. (2003). Attractor dynamics of networkUP states in the neocortex. Nature 423, 283–288.

Cowan, R.L., and Wilson, C.J. (1994). Spontaneous firing patterns and axonalprojections of single corticostriatal neurons in the rat medial agranular cortex.J. Neurophysiol. 71, 17–32.

Crochet, S., and Petersen, C.C. (2006). Correlating whisker behavior withmembrane potential in barrel cortex of awake mice. Nat. Neurosci. 9, 608–610.

Crochet, S., Chauvette, S., Boucetta, S., and Timofeev, I. (2005). Modulation ofsynaptic transmission in neocortex by network activities. Eur. J. Neurosci. 21,1030–1044.

Cruikshank, S.J., Lewis, T.J., and Connors, B.W. (2007). Synaptic basis forintense thalamocortical activation of feedforward inhibitory cells in neocortex.Nat. Neurosci. 10, 462–468.

Csicsvari, J., Hirase, H., Czurko, A., Mamiya, A., and Buzsaki, G. (1999).Oscillatory coupling of hippocampal pyramidal cells and interneurons in thebehaving Rat. J. Neurosci. 19, 274–287.

Csicsvari, J., Jamieson, B., Wise, K.D., and Buzsaki, G. (2003). Mechanismsof gamma oscillations in the hippocampus of the behaving rat. Neuron 37,311–322.

Dan, Y., and Poo, M.M. (2006). Spike timing-dependent plasticity: fromsynapse to perception. Physiol. Rev. 86, 1033–1048.

de Kock, C.P., Bruno, R.M., Spors, H., and Sakmann, B. (2007). Layer- andcell-type-specific suprathreshold stimulus representation in rat primarysomatosensory cortex. J. Physiol. 581, 139–154.

DeFelipe, J., Alonso-Nanclares, L., and Arellano, J.I. (2002). Microstructure ofthe neocortex: comparative aspects. J. Neurocytol. 31, 299–316.

Desimone, R., and Duncan, J. (1995). Neural mechanisms of selective visualattention. Annu. Rev. Neurosci. 18, 193–222.

Destexhe, A., Contreras, D., and Steriade, M. (1999). Spatiotemporal analysisof local field potentials and unit discharges in cat cerebral cortex during naturalwake and sleep states. J. Neurosci. 19, 4595–4608.

Destexhe, A., Rudolph, M., Fellous, J.M., and Sejnowski, T.J. (2001). Fluctu-ating synaptic conductances recreate in vivo-like activity in neocorticalneurons. Neuroscience 107, 13–24.

Destexhe, A., Rudolph, M., and Pare, D. (2003). The high-conductance state ofneocortical neurons in vivo. Nat. Rev. Neurosci. 4, 739–751.

Destexhe, A., Hughes, S.W., Rudolph, M., and Crunelli, V. (2007). Are cortico-thalamic ‘up’ states fragments of wakefulness? Trends Neurosci. 30, 334–342.

DeWeese, M.R., and Zador, A.M. (2006). Non-Gaussian membrane potentialdynamics imply sparse, synchronous activity in auditory cortex. J. Neurosci.26, 12206–12218.

Disney, A.A., Aoki, C., and Hawken, M.J. (2007). Gain modulation by nicotine inmacaque v1. Neuron 56, 701–713.

Douglas, R.J., and Martin, K.A. (2007a). Mapping the matrix: the ways ofneocortex. Neuron 56, 226–238.

Douglas, R.J., and Martin, K.A. (2007b). Recurrent neuronal circuits in theneocortex. Curr. Biol. 17, R496–R500.

Douglass, A.D., Kraves, S., Deisseroth, K., Schier, A.F., and Engert, F. (2008).Escape behavior elicited by single, channelrhodopsin-2-evoked spikes inzebrafish somatosensory neurons. Curr. Biol. 18, 1133–1137.

El Boustani, S., Pospischil, M., Rudolph-Lilith, M., and Destexhe, A. (2007).Activated cortical states: experiments, analyses and models. J. Physiol. (Paris)101, 99–109.

Feldmeyer, D., Egger, V., Lubke, J., and Sakmann, B. (1999). Reliable synapticconnections between pairs of excitatory layer 4 neurones within a single‘barrel’ of developing rat somatosensory cortex. J. Physiol. 521, 169–190.

Ferezou, I., Bolea, S., and Petersen, C.C. (2006). Visualizing the cortical repre-sentation of whisker touch: voltage-sensitive dye imaging in freely movingmice. Neuron 50, 617–629.

Ferezou, I., Haiss, F., Gentet, L.J., Aronoff, R., Weber, B., and Petersen, C.C.(2007). Spatiotemporal dynamics of cortical sensorimotor integration inbehaving mice. Neuron 56, 907–923.

Fox, M.D., Snyder, A.Z., Vincent, J.L., and Raichle, M.E. (2007). Intrinsic fluc-tuations within cortical systems account for intertrial variability in humanbehavior. Neuron 56, 171–184.

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 185

Neuron

Review

Fries, P., Reynolds, J.H., Rorie, A.E., and Desimone, R. (2001). Modulation ofoscillatory neuronal synchronization by selective visual attention. Science 291,1560–1563.

Fries, P., Nikolic, D., and Singer, W. (2007). The gamma cycle. Trends Neuro-sci. 30, 309–316.

Fuentealba, P., Begum, R., Capogna, M., Jinno, S., Marton, L.F., Csicsvari, J.,Thomson, A., Somogyi, P., and Klausberger, T. (2008). Ivy cells: a population ofnitric-oxide-producing, slow-spiking GABAergic neurons and their involve-ment in hippocampal network activity. Neuron 57, 917–929.

Fujisawa, S., Amarasingham, A., Harrison, M.T., and Buzsaki, G. (2008).Behavior-dependent short-term assembly dynamics in the medial prefrontalcortex. Nat. Neurosci. 11, 823–833.

Gabernet, L., Jadhav, S.P., Feldman, D.E., Carandini, M., and Scanziani, M.(2005). Somatosensory integration controlled by dynamic thalamocorticalfeed-forward inhibition. Neuron 48, 315–327.

Galarreta, M., and Hestrin, S. (1998). Frequency-dependent synaptic depres-sion and the balance of excitation and inhibition in the neocortex. Nat. Neuro-sci. 1, 587–594.

Garaschuk, O., Milos, R.I., and Konnerth, A. (2006). Targeted bulk-loading offluorescent indicators for two-photon brain imaging in vivo. Nat. Protocols 1,380–386.

Genovesio, A., Brasted, P.J., Mitz, A.R., and Wise, S.P. (2005). Prefrontalcortex activity related to abstract response strategies. Neuron 47, 307–320.

Gil, Z., Connors, B.W., and Amitai, Y. (1997). Differential regulation of neocor-tical synapses by neuromodulators and activity. Neuron 19, 679–686.

Gilbert, C.D., and Sigman, M. (2007). Brain states: top-down influences insensory processing. Neuron 54, 677–696.

Gradinaru, V., Thompson, K.R., Zhang, F., Mogri, M., Kay, K., Schneider, M.B.,and Deisseroth, K. (2007). Targeting and readout strategies for fast opticalneural control in vitro and in vivo. J. Neurosci. 27, 14231–14238.

Greenberg, D.S., Houweling, A.R., and Kerr, J.N. (2008). Population imaging ofongoing neuronal activity in the visual cortex of awake rats. Nat. Neurosci. 11,749–751.

Grinvald, A., and Hildesheim, R. (2004). VSDI: a new era in functional imagingof cortical dynamics. Nat. Rev. Neurosci. 5, 874–885.

Gur, M., Beylin, A., and Snodderly, D.M. (1997). Response variability ofneurons in primary visual cortex (V1) of alert monkeys. J. Neurosci. 17,2914–2920.

Hahn, T.T., Sakmann, B., and Mehta, M.R. (2007). Differential responses ofhippocampal subfields to cortical up-down states. Proc. Natl. Acad. Sci.USA 104, 5169–5174.

Hahnloser, R.H., Douglas, R.J., and Hepp, K. (2002). Attentional recruitment ofinter-areal recurrent networks for selective gain control. Neural Comput. 14,1669–1689.

Haider, B., Duque, A., Hasenstaub, A.R., and McCormick, D.A. (2006).Neocortical network activity in vivo is generated through a dynamic balanceof excitation and inhibition. J. Neurosci. 26, 4535–4545.

Haider, B., Duque, A., Hasenstaub, A.R., Yu, Y., and McCormick, D.A. (2007).Enhancement of visual responsiveness by spontaneous local network activityin vivo. J. Neurophysiol. 97, 4186–4202.

Harris, K.D. (2005). Neural signatures of cell assembly organization. Nat. Rev.Neurosci. 6, 399–407.

Hasenstaub, A., Shu, Y., Haider, B., Kraushaar, U., Duque, A., and McCor-mick, D.A. (2005). Inhibitory postsynaptic potentials carry synchronizedfrequency information in active cortical networks. Neuron 47, 423–435.

Hasenstaub, A., Sachdev, R.N., and McCormick, D.A. (2007). State changesrapidly modulate cortical neuronal responsiveness. J. Neurosci. 27, 9607–9622.

Hebb, D.O. (1949). The Organization of Behavior: A NeuropsychologicalTheory (New York: Wiley).

186 Neuron 62, April 30, 2009 ª2009 Elsevier Inc.

Helmchen, F., Imoto, K., and Sakmann, B. (1996). Ca2+ buffering and actionpotential-evoked Ca2+ signaling in dendrites of pyramidal neurons. Biophys.J. 70, 1069–1081.

Henze, D.A., and Buzsaki, G. (2001). Action potential threshold of hippocampalpyramidal cells in vivo is increased by recent spiking activity. Neuroscience105, 121–130.

Higley, M.J., and Contreras, D. (2006). Balanced excitation and inhibitiondetermine spike timing during frequency adaptation. J. Neurosci. 26, 448–457.

Hirsch, J.A., Martinez, L.M., Pillai, C., Alonso, J.M., Wang, Q., and Sommer,F.T. (2003). Functionally distinct inhibitory neurons at the first stage of visualcortical processing. Nat. Neurosci. 6, 1300–1308.

Ho, N., and Destexhe, A. (2000). Synaptic background activity enhances theresponsiveness of neocortical pyramidal neurons. J. Neurophysiol. 84,1488–1496.

Holcman, D., and Tsodyks, M. (2006). The emergence of Up and Down statesin cortical networks. PLoS Comput. Biol. 2, e23.

Holt, G.R., and Koch, C. (1997). Shunting inhibition does not have a divisiveeffect on firing rates. Neural Comput. 9, 1001–1013.

Homma, R., Baker, B.J., Jin, L., Garaschuk, O., Konnerth, A., Cohen, L.B.,Bleau, C.X., Canepari, M., Djurisic, M., and Zecevic, D. (2009). Wide-fieldand two-photon imaging of brain activity with voltage- and calcium-sensitivedyes. Methods Mol. Biol. 489, 43–79.

Houweling, A.R., and Brecht, M. (2008). Behavioural report of single neuronstimulation in somatosensory cortex. Nature 451, 65–68.

Huber, D., Petreanu, L., Ghitani, N., Ranade, S., Hromadka, T., Mainen, Z., andSvoboda, K. (2008). Sparse optical microstimulation in barrel cortex driveslearned behaviour in freely moving mice. Nature 451, 61–64.

Johnson, H.A., and Buonomano, D.V. (2007). Development and plasticity ofspontaneous activity and Up states in cortical organotypic slices. J. Neurosci.27, 5915–5925.

Jones, M.S., MacDonald, K.D., Choi, B., Dudek, F.E., and Barth, D.S. (2000).Intracellular correlates of fast (>200 Hz) electrical oscillations in rat somatosen-sory cortex. J. Neurophysiol. 84, 1505–1518.

Kapfer, C., Glickfeld, L.L., Atallah, B.V., and Scanziani, M. (2007). Supralinearincrease of recurrent inhibition during sparse activity in the somatosensorycortex. Nat. Neurosci. 10, 743–753.

Kawaguchi, Y. (1997). Selective cholinergic modulation of cortical GABAergiccell subtypes. J. Neurophysiol. 78, 1743–1747.

Kenet, T., Bibitchkov, D., Tsodyks, M., Grinvald, A., and Arieli, A. (2003). Spon-taneously emerging cortical representations of visual attributes. Nature 425,954–956.

Kerchner, G.A., and Nicoll, R.A. (2008). Silent synapses and the emergence ofa postsynaptic mechanism for LTP. Nat. Rev. Neurosci. 9, 813–825.

Kerr, J.N., Greenberg, D., and Helmchen, F. (2005). Imaging input and outputof neocortical networks in vivo. Proc. Natl. Acad. Sci. USA 102, 14063–14068.

Kerr, J.N., de Kock, C.P., Greenberg, D.S., Bruno, R.M., Sakmann, B., andHelmchen, F. (2007). Spatial organization of neuronal population responsesin layer 2/3 of rat barrel cortex. J. Neurosci. 27, 13316–13328.

Klausberger, T., Magill, P.J., Marton, L.F., Roberts, J.D., Cobden, P.M., Buz-saki, G., and Somogyi, P. (2003). Brain-state- and cell-type-specific firing ofhippocampal interneurons in vivo. Nature 421, 844–848.

Kole, M.H., Ilschner, S.U., Kampa, B.M., Williams, S.R., Ruben, P.C., andStuart, G.J. (2008). Action potential generation requires a high sodium channeldensity in the axon initial segment. Nat. Neurosci. 11, 178–186.

Kuhn, A., Aertsen, A., and Rotter, S. (2004). Neuronal integration of synapticinput in the fluctuation-driven regime. J. Neurosci. 24, 2345–2356.

Lampl, I., Reichova, I., and Ferster, D. (1999). Synchronous membrane poten-tial fluctuations in neurons of the cat visual cortex. Neuron 22, 361–374.

Larkman, A.U. (1991). Dendritic morphology of pyramidal neurones of thevisual cortex of the rat: III. Spine distributions. J. Comp. Neurol. 306, 332–343.

Neuron

Review

Larkum, M.E., Senn, W., and Luscher, H.R. (2004). Top-down dendritic inputincreases the gain of layer 5 pyramidal neurons. Cereb. Cortex 14, 1059–1070.

Le Be, J.V., Silberberg, G., Wang, Y., and Markram, H. (2007). Morphological,electrophysiological, and synaptic properties of corticocallosal pyramidal cellsin the neonatal rat neocortex. Cereb. Cortex 17, 2204–2213.

Lee, J., and Maunsell, J.H. (2009). A normalization model of attentional modu-lation of single unit responses. PLoS ONE 4, e4651.

Lee, A.K., Manns, I.D., Sakmann, B., and Brecht, M. (2006). Whole-cell record-ings in freely moving rats. Neuron 51, 399–407.

Leger, J.F., Stern, E.A., Aertsen, A., and Heck, D. (2005). Synaptic integrationin rat frontal cortex shaped by network activity. J. Neurophysiol. 93, 281–293.

Livingstone, M.S., and Hubel, D.H. (1981). Effects of sleep and arousal on theprocessing of visual information in the cat. Nature 291, 554–561.

Llinas, R.R. (1988). The intrinsic electrophysiological properties of mammalianneurons: insights into central nervous system function. Science 242, 1654–1664.

Logothetis, N.K. (1998). Single units and conscious vision. Philos. Trans. R.Soc. Lond. B Biol. Sci. 353, 1801–1818.

London, M., and Hausser, M. (2005). Dendritic computation. Annu. Rev. Neu-rosci. 28, 503–532.

Lorente de No, R. (1938). Cerebral Cortex: Architecture, intracortical connec-tions, motor projections. In Physiology of the Nervous System, J.F. Fulton, ed.(New York: Oxford University Press).

Luck, S.J., Chelazzi, L., Hillyard, S.A., and Desimone, R. (1997). Neural mech-anisms of spatial selective attention in areas V1, V2, and V4 of macaque visualcortex. J. Neurophysiol. 77, 24–42.

Luczak, A., Bartho, P., Marguet, S.L., Buzsaki, G., and Harris, K.D. (2007).Sequential structure of neocortical spontaneous activity in vivo. Proc. Natl.Acad. Sci. USA 104, 347–352.

Luna, R., Hernandez, A., Brody, C.D., and Romo, R. (2005). Neural codes forperceptual discrimination in primary somatosensory cortex. Nat. Neurosci.8, 1210–1219.

Luo, L., Callaway, E.M., and Svoboda, K. (2008). Genetic dissection of neuralcircuits. Neuron 57, 634–660.

Magee, J.C. (2000). Dendritic integration of excitatory synaptic input. Nat. Rev.Neurosci. 1, 181–190.

Magee, J.C., and Johnston, D. (2005). Plasticity of dendritic function. Curr.Opin. Neurobiol. 15, 334–342.

Mainen, Z.F., and Sejnowski, T.J. (1995). Reliability of spike timing in neocor-tical neurons. Science 268, 1503–1506.

Major, G., and Tank, D. (2004). Persistent neural activity: prevalence andmechanisms. Curr. Opin. Neurobiol. 14, 675–684.

Major, G., Polsky, A., Denk, W., Schiller, J., and Tank, D.W. (2008). Spatio-temporally graded NMDA spike/plateau potentials in basal dendrites ofneocortical pyramidal neurons. J. Neurophysiol. 99, 2584–2601.

Malenka, R.C., and Bear, M.F. (2004). LTP and LTD: an embarrassment ofriches. Neuron 44, 5–21.

Marino, J., Schummers, J., Lyon, D.C., Schwabe, L., Beck, O., Wiesing, P.,Obermayer, K., and Sur, M. (2005). Invariant computations in local corticalnetworks with balanced excitation and inhibition. Nat. Neurosci. 8, 194–201.

Markram, H., Lubke, J., Frotscher, M., Roth, A., and Sakmann, B. (1997). Phys-iology and anatomy of synaptic connections between thick tufted pyramidalneurones in the developing rat neocortex. J. Physiol. 500, 409–440.

Markram, H., Toledo-Rodriguez, M., Wang, Y., Gupta, A., Silberberg, G., andWu, C. (2004). Interneurons of the neocortical inhibitory system. Nat. Rev. Neu-rosci. 5, 793–807.

Martinez-Trujillo, J.C., and Treue, S. (2004). Feature-based attention increasesthe selectivity of population responses in primate visual cortex. Curr. Biol. 14,744–751.

Maunsell, J.H., and Treue, S. (2006). Feature-based attention in visual cortex.Trends Neurosci. 29, 317–322.

McAdams, C.J., and Maunsell, J.H. (1999). Effects of attention on orientation-tuning functions of single neurons in macaque cortical area V4. J. Neurosci. 19,431–441.

McAdams, C.J., and Reid, R.C. (2005). Attention modulates the responses ofsimple cells in monkey primary visual cortex. J. Neurosci. 25, 11023–11033.

McCormick, D.A. (1992). Neurotransmitter actions in the thalamus and cere-bral cortex and their role in neuromodulation of thalamocortical activity.Prog. Neurobiol. 39, 337–388.

McCormick, D.A., and Contreras, D. (2001). On the cellular and network basesof epileptic seizures. Annu. Rev. Physiol. 63, 815–846.

McCormick, D.A., Connors, B.W., Lighthall, J.W., and Prince, D.A. (1985).Comparative electrophysiology of pyramidal and sparsely spiny stellateneurons of the neocortex. J. Neurophysiol. 54, 782–806.

McCormick, D.A., Shu, Y., and Yu, Y. (2007). Neurophysiology: Hodgkin andHuxley model–still standing? Nature 445, E1–E2.

McGuire, B.A., Gilbert, C.D., Rivlin, P.K., and Wiesel, T.N. (1991). Targets ofhorizontal connections in macaque primary visual cortex. J. Comp. Neurol.305, 370–392.

Metherate, R., and Ashe, J.H. (1993). Ionic flux contributions to neocorticalslow waves and nucleus basalis-mediated activation: whole-cell recordingsin vivo. J. Neurosci. 13, 5312–5323.

Miller, E.K., and Cohen, J.D. (2001). An integrative theory of prefrontal cortexfunction. Annu. Rev. Neurosci. 24, 167–202.

Miller, K.D., and Troyer, T.W. (2002). Neural noise can explain expansive,power-law nonlinearities in neural response functions. J. Neurophysiol. 87,653–659.

Milojkovic, B.A., Radojicic, M.S., and Antic, S.D. (2005). A strict correlationbetween dendritic and somatic plateau depolarizations in the rat prefrontalcortex pyramidal neurons. J. Neurosci. 25, 3940–3951.

Mitchell, J.F., Sundberg, K.A., and Reynolds, J.H. (2007). Differential attention-dependent response modulation across cell classes in macaque visual areaV4. Neuron 55, 131–141.

Miura, K., Tsubo, Y., Okada, M., and Fukai, T. (2007). Balanced excitatory andinhibitory inputs to cortical neurons decouple firing irregularity from rate modu-lations. J. Neurosci. 27, 13802–13812.

Murayama, M., Perez-Garci, E., Nevian, T., Bock, T., Senn, W., and Larkum,M.E. (2009). Dendritic encoding of sensory stimuli controlled by deep corticalinterneurons. Nature 457, 1137–1141.

Murphy, B.K., and Miller, K.D. (2003). Multiplicative gain changes are inducedby excitation or inhibition alone. J. Neurosci. 23, 10040–10051.

Murphy, B.K., and Miller, K.D. (2009). Balanced amplification: a new mecha-nism of selective amplification of neural activity patterns. Neuron 61, 635–648.

Nagel, G., Szellas, T., Huhn, W., Kateriya, S., Adeishvili, N., Berthold, P., Olliq,D., Hegemann, P., and Bamberg, E. (2003). Channelrhodopsin-2, a directlylight-gated cation-selective membrane channel. Proc. Natl. Acad. Sci. USA100, 13940–13945.

Narayanan, N.S., and Laubach, M. (2006). Top-down control of motor cortexensembles by dorsomedial prefrontal cortex. Neuron 52, 921–931.

Neher, E., and Sakmann, B. (1976). Single-channel currents recorded frommembrane of denervated frog muscle fibres. Nature 260, 799–802.

Nelson, S.B., Sugino, K., and Hempel, C.M. (2006). The problem of neuronalcell types: a physiological genomics approach. Trends Neurosci. 29, 339–345.

Nevian, T., Larkum, M.E., Polsky, A., and Schiller, J. (2007). Properties of basaldendrites of layer 5 pyramidal neurons: a direct patch-clamp recording study.Nat. Neurosci. 10, 206–214.

Nowak, L.G., Sanchez-Vives, M.V., and McCormick, D.A. (1997). Influence oflow and high frequency inputs on spike timing in visual cortical neurons. Cereb.Cortex 7, 487–501.

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 187

Neuron

Review

Nowak, L.G., Azouz, R., Sanchez-Vives, M.V., Gray, C.M., and McCormick,D.A. (2003). Electrophysiological classes of cat primary visual cortical neuronsin vivo as revealed by quantitative analyses. J. Neurophysiol. 89, 1541–1566.

Nowak, L.G., Sanchez-Vives, M.V., and McCormick, D.A. (2007). Lack oforientation and direction selectivity in a subgroup of fast-spiking inhibitoryinterneurons: Cellular and synaptic mechanisms and comparison with otherelectrophysiological cell types. Cereb. Cortex 18, 1058–1078.

Ohki, K., Chung, S., Ch’ng, Y.H., Kara, P., and Reid, R.C. (2005). Functionalimaging with cellular resolution reveals precise micro-architecture in visualcortex. Nature 433, 597–603.

Okun, M., and Lampl, I. (2008). Instantaneous correlation of excitation andinhibition during ongoing and sensory-evoked activities. Nat. Neurosci. 11,535–537.

Olshausen, B.A., and Field, D.J. (2004). Sparse coding of sensory inputs. Curr.Opin. Neurobiol. 14, 481–487.

Pare, D., Shink, E., Gaudreau, H., Destexhe, A., and Lang, E.J. (1998). Impactof spontaneous synaptic activity on the resting properties of cat neocorticalpyramidal neurons In vivo. J. Neurophysiol. 79, 1450–1460.

Parga, N., and Abbott, L.F. (2007). Network model of spontaneous activity ex-hibiting synchronous transitions between up and down States. Front Neurosci1, 57–66.

Penttonen, M., Kamondi, A., Acsady, L., and Buzsaki, G. (1998). Gammafrequency oscillation in the hippocampus of the rat: intracellular analysisin vivo. Eur. J. Neurosci. 10, 718–728.

Pesaran, B., Pezaris, J.S., Sahani, M., Mitra, P.P., and Andersen, R.A. (2002).Temporal structure in neuronal activity during working memory in macaqueparietal cortex. Nat. Neurosci. 5, 805–811.

Pesaran, B., Nelson, M.J., and Andersen, R.A. (2008). Free choice activatesa decision circuit between frontal and parietal cortex. Nature 453, 406–409.

Peters, A. (1984). Chandelier Cells. In Cerebral Cortex. Cellular Components ofthe Cerebral Cortex, A. Peters and E.G. Jones, eds. (New York: Plenum Press),pp. 361–379.

Petersen, C.C., Hahn, T.T., Mehta, M., Grinvald, A., and Sakmann, B. (2003).Interaction of sensory responses with spontaneous depolarization in layer2/3 barrel cortex. Proc. Natl. Acad. Sci. USA 100, 13638–13643.

Petreanu, L., Huber, D., Sobczyk, A., and Svoboda, K. (2007). Channelrhodop-sin-2-assisted circuit mapping of long-range callosal projections. Nat. Neuro-sci. 10, 663–668.

Petreanu, L., Mao, T., Sternson, S.M., and Svoboda, K. (2009). The subcellularorganization of neocortical excitatory connections. Nature 457, 1142–1145.

Piwkowska, Z., Pospischil, M., Brette, R., Sliwa, J., Rudolph-Lilith, M., Bal, T.,and Destexhe, A. (2008). Characterizing synaptic conductance fluctuations incortical neurons and their influence on spike generation. J. Neurosci. Methods169, 302–322.

Pouille, F., and Scanziani, M. (2001). Enforcement of temporal fidelity in pyra-midal cells by somatic feed-forward inhibition. Science 293, 1159–1163.

Poulet, J.F., and Petersen, C.C. (2008). Internal brain state regulatesmembrane potential synchrony in barrel cortex of behaving mice. Nature454, 881–885.

Prescott, S.A., and De Koninck, Y. (2003). Gain control of firing rate by shuntinginhibition: roles of synaptic noise and dendritic saturation. Proc. Natl. Acad.Sci. USA 100, 2076–2081.

Priebe, N.J., and Ferster, D. (2008). Inhibition, spike threshold, and stimulusselectivity in primary visual cortex. Neuron 57, 482–497.

Priebe, N.J., Mechler, F., Carandini, M., and Ferster, D. (2004). The contribu-tion of spike threshold to the dichotomy of cortical simple and complex cells.Nat. Neurosci. 7, 1113–1122.

Reig, R., Gallego, R., Nowak, L.G., and Sanchez-Vives, M.V. (2006). Impact ofcortical network activity on short-term synaptic depression. Cereb. Cortex 16,688–695.

188 Neuron 62, April 30, 2009 ª2009 Elsevier Inc.

Reynolds, J.H., and Chelazzi, L. (2004). Attentional modulation of visualprocessing. Annu. Rev. Neurosci. 27, 611–647.

Reynolds, J.H., and Heeger, D.J. (2009). The normalization model of attention.Neuron 61, 168–185.

Reynolds, J.H., Pasternak, T., and Desimone, R. (2000). Attention increasessensitivity of V4 neurons. Neuron 26, 703–714.

Rudolph, M., and Destexhe, A. (2003). A fast-conducting, stochastic integra-tive mode for neocortical neurons in vivo. J. Neurosci. 23, 2466–2476.

Rudolph, M., Pelletier, J.G., Pare, D., and Destexhe, A. (2005). Characteriza-tion of synaptic conductances and integrative properties during electricallyinduced EEG-activated states in neocortical neurons in vivo. J. Neurophysiol.94, 2805–2821.

Rudolph, M., Pospischil, M., Timofeev, I., and Destexhe, A. (2007). Inhibitiondetermines membrane potential dynamics and controls action potential gener-ation in awake and sleeping cat cortex. J. Neurosci. 27, 5280–5290.

Rutishauser, U., and Douglas, R.J. (2008). State-dependent computationusing coupled recurrent networks. Neural Comput., in press. Published onlineFebruary 11, 2009. 10.1162/neco.2008.03-08-734.

Sachdev, R.N., Ebner, F.F., and Wilson, C.J. (2004). Effect of subthreshold upand down states on the whisker-evoked response in somatosensory cortex.J. Neurophysiol. 92, 3511–3521.

Sakmann, B. (2006). Patch pipettes are more useful than initially thought:simultaneous pre- and postsynaptic recording from mammalian CNSsynapses in vitro and in vivo. Pflugers Arch. 453, 249–259.

Salinas, E., and Abbott, L.F. (1996). A model of multiplicative neural responsesin parietal cortex. Proc. Natl. Acad. Sci. USA 93, 11956–11961.

Salinas, E., and Sejnowski, T.J. (2001). Correlated neuronal activity and theflow of neural information. Nat. Rev. Neurosci. 2, 539–550.

Sanchez-Vives, M.V., and McCormick, D.A. (2000). Cellular and networkmechanisms of rhythmic recurrent activity in neocortex. Nat. Neurosci. 3,1027–1034.

Sanchez-Vives, M.V., Nowak, L.G., and McCormick, D.A. (2000). Membranemechanisms underlying contrast adaptation in cat area 17 in vivo. J. Neurosci.20, 4267–4285.

Sato, T.R., Gray, N.W., Mainen, Z.F., and Svoboda, K. (2007). The functionalmicroarchitecture of the mouse barrel cortex. PLoS Biol. 5, e189.

Sawatari, A., and Callaway, E.M. (2000). Diversity and cell type specificity oflocal excitatory connections to neurons in layer 3B of monkey primary visualcortex. Neuron 25, 459–471.

Shadlen, M.N., and Newsome, W.T. (1998). The variable discharge of corticalneurons: implications for connectivity, computation, and information coding.J. Neurosci. 18, 3870–3896.

Shepherd, G.M., and Svoboda, K. (2005). Laminar and columnar organizationof ascending excitatory projections to layer 2/3 pyramidal neurons in rat barrelcortex. J. Neurosci. 25, 5670–5679.

Shoham, S., O’Connor, D.H., and Segev, R. (2006). How silent is the brain: isthere a ‘‘dark matter’’ problem in neuroscience? J. Comp. Physiol. A Neuroe-thol. Sens. Neural. Behav. Physiol. 192, 777–784.

Shu, Y., Hasenstaub, A., Badoual, M., Bal, T., and McCormick, D.A. (2003a).Barrages of synaptic activity control the gain and sensitivity of corticalneurons. J. Neurosci. 23, 10388–10401.

Shu, Y., Hasenstaub, A., and McCormick, D.A. (2003b). Turning on and offrecurrent balanced cortical activity. Nature 423, 288–293.

Shu, Y., Hasenstaub, A., Duque, A., Yu, Y., and McCormick, D.A. (2006).Modulation of intracortical synaptic potentials by presynaptic somaticmembrane potential. Nature 441, 761–765.

Shu, Y., Duque, A., Yu, Y., Haider, B., and McCormick, D.A. (2007). Propertiesof action-potential initiation in neocortical pyramidal cells: evidence fromwhole cell axon recordings. J. Neurophysiol. 97, 746–760.

Neuron

Review

Silberberg, G., and Markram, H. (2007). Disynaptic inhibition between neocor-tical pyramidal cells mediated by Martinotti cells. Neuron 53, 735–746.

Singer, W., and Gray, C.M. (1995). Visual feature integration and the temporalcorrelation hypothesis. Annu. Rev. Neurosci. 18, 555–586.

Sirota,A., Montgomery,S.,Fujisawa,S., Isomura,Y., Zugaro,M., andBuzsaki,G.(2008). Entrainment of neocortical neurons and gamma oscillations by thehippocampal theta rhythm. Neuron 60, 683–697.

Sjostrom, P.J., Rancz, E.A., Roth, A., and Hausser, M. (2008). Dendritic excit-ability and synaptic plasticity. Physiol. Rev. 88, 769–840.

Softky, W.R., and Koch, C. (1993). The highly irregular firing of cortical cellsis inconsistent with temporal integration of random EPSPs. J. Neurosci. 13,334–350.

Sohya, K., Kameyama, K., Yanagawa, Y., Obata, K., and Tsumoto, T. (2007).GABAergic neurons are less selective to stimulus orientation than excitatoryneurons in layer II/III of visual cortex, as revealed by in vivo functional Ca2+imaging in transgenic mice. J. Neurosci. 27, 2145–2149.

Song, S., Sjostrom, P.J., Reigl, M., Nelson, S., and Chklovskii, D.B. (2005).Highly nonrandom features of synaptic connectivity in local cortical circuits.PLoS Biol. 3, e68.

Steriade, M., Nunez, A., and Amzica, F. (1993a). Intracellular analysis of rela-tions between the slow (< 1 Hz) neocortical oscillation and other sleep rhythmsof the electroencephalogram. J. Neurosci. 13, 3266–3283.

Steriade, M., Nunez, A., and Amzica, F. (1993b). A novel slow (< 1 Hz) oscilla-tion of neocortical neurons in vivo: depolarizing and hyperpolarizing compo-nents. J. Neurosci. 13, 3252–3265.

Steriade, M., Timofeev, I., and Grenier, F. (2001). Natural waking and sleepstates: a view from inside neocortical neurons. J. Neurophysiol. 85, 1969–1985.

Stern, E.A., Kincaid, A.E., and Wilson, C.J. (1997). Spontaneous subthresholdmembrane potential fluctuations and action potential variability of rat cortico-striatal and striatal neurons in vivo. J. Neurophysiol. 77, 1697–1715.

Stuart, G., Schiller, J., and Sakmann, B. (1997). Action potential initiation andpropagation in rat neocortical pyramidal neurons. J. Physiol. 505, 617–632.

Sundberg, K.A., Mitchell, J.F., and Reynolds, J.H. (2009). Spatial attentionmodulates center-surround interactions in macaque visual area v4. Neuron61, 952–963.

Svoboda, K., and Yasuda, R. (2006). Principles of two-photon excitationmicroscopy and its applications to neuroscience. Neuron 50, 823–839.

Swadlow, H.A. (2002). Thalamocortical control of feed-forward inhibition inawake somatosensory ‘barrel’ cortex. Philos. Trans. R. Soc. Lond. B Biol.Sci. 357, 1717–1727.

Swadlow, H.A., Beloozerova, I.N., and Sirota, M.G. (1998). Sharp, localsynchrony among putative feed-forward inhibitory interneurons of rabbitsomatosensory cortex. J. Neurophysiol. 79, 567–582.

Tahvildari, B., Alonso, A.A., and Bourque, C.W. (2008). Ionic basis of ON andOFF persistent activity in layer III lateral entorhinal cortical principal neurons.J. Neurophysiol. 99, 2006–2011.

Thomson, A.M., Bannister, A.P., Mercer, A., and Morris, O.T. (2002). Targetand temporal pattern selection at neocortical synapses. Philos. Trans. R.Soc. Lond. B Biol. Sci. 357, 1781–1791.

Thurley, K., Senn, W., and Luscher, H.R. (2008). Dopamine increases the gainof the input-output response of rat prefrontal pyramidal neurons. J. Neurophy-siol. 99, 2985–2997.

Tiesinga, P., Fellous, J.M., and Sejnowski, T.J. (2008). Regulation of spiketiming in visual cortical circuits. Nat. Rev. Neurosci. 9, 97–107.

Timofeev, I., Grenier, F., Bazhenov, M., Sejnowski, T.J., and Steriade, M.(2000). Origin of slow cortical oscillations in deafferented cortical slabs. Cereb.Cortex 10, 1185–1199.

Timofeev, I., Grenier, F., and Steriade, M. (2001). Disfacilitation and active inhi-bition in the neocortex during the natural sleep-wake cycle: an intracellularstudy. Proc. Natl. Acad. Sci. USA 98, 1924–1929.

Treue, S., and Martinez Trujillo, J.C. (1999). Feature-based attention influencesmotion processing gain in macaque visual cortex. Nature 399, 575–579.

Tsujimoto, S., Genovesio, A., and Wise, S.P. (2008). Transient neuronal corre-lations underlying goal selection and maintenance in prefrontal cortex. Cereb.Cortex 18, 2748–2761.

van Vreeswijk, C., and Sompolinsky, H. (1996). Chaos in neuronal networkswith balanced excitatory and inhibitory activity. Science 274, 1724–1726.

Varela, J.A., Song, S., Turrigiano, G.G., and Nelson, S.B. (1999). Differentialdepression at excitatory and inhibitory synapses in visual cortex. J. Neurosci.19, 4293–4304.

Vogels, T.P., and Abbott, L.F. (2009). Gating multiple signlas through detailedbalance of excitation and inhibition in spiking networks. Nat. Neurosci. 12,483–491.

Vogels, T.P., Rajan, K., and Abbott, L.F. (2005). Neural network dynamics.Annu. Rev. Neurosci. 28, 357–376.

Volgushev, M., Chauvette, S., Mukovski, M., and Timofeev, I. (2006). Preciselong-range synchronization of activity and silence in neocortical neuronsduring slow-wave oscillations. J. Neurosci. 26, 5665–5672.

Wang, X.J. (2008). Decision making in recurrent neuronal circuits. Neuron 60,215–234.

Waters, J., and Helmchen, F. (2006). Background synaptic activity is sparse inneocortex. J. Neurosci. 26, 8267–8277.

Wehr, M., and Zador, A.M. (2003). Balanced inhibition underlies tuning andsharpens spike timing in auditory cortex. Nature 426, 442–446.

Wilent, W.B., and Contreras, D. (2005). Dynamics of excitation and inhibitionunderlying stimulus selectivity in rat somatosensory cortex. Nat. Neurosci. 8,1364–1370.

Williams, S.R. (2004). Spatial compartmentalization and functional impact ofconductance in pyramidal neurons. Nat. Neurosci. 7, 961–967.

Williford, T., and Maunsell, J.H. (2006). Effects of spatial attention on contrastresponse functions in macaque area V4. J. Neurophysiol. 96, 40–54.

Wilson, C.J., and Kawaguchi, Y. (1996). The origins of two-state spontaneousmembrane potential fluctuations of neostriatal spiny neurons. J. Neurosci. 16,2397–2410.

Womelsdorf, T., Schoffelen, J.M., Oostenveld, R., Singer, W., Desimone, R.,Engel, A.K., and Fries, P. (2007). Modulation of neuronal interactions throughneuronal synchronization. Science 316, 1609–1612.

Wong, K.F., Huk, A.C., Shadlen, M.N., and Wang, X.J. (2007). Neural circuitdynamics underlying accumulation of time-varying evidence during perceptualdecision making. Front. Comput. Neurosci. 1, 6.

Worgotter, F., Suder, K., Zhao, Y., Kerscher, N., Eysel, U.T., and Funke, K.(1998). State-dependent receptive-field restructuring in the visual cortex.Nature 396, 165–168.

Yoshimura, Y., Dantzker, J.L., and Callaway, E.M. (2005). Excitatory corticalneurons form fine-scale functional networks. Nature 433, 868–873.

Yu, Y., Shu, Y., and McCormick, D.A. (2008). Cortical action potential back-propagation explains spike threshold variability and rapid-onset kinetics.J. Neurosci. 28, 7260–7272.

Zemelman, B.V., Lee, G.A., and Miesenbock, G. (2002). Selective photostimu-lation of genetically chARGed neurons. Neuron 33, 15–22.

Zhang, F., Aravanis, A.M., Adamantidis, A., de Lecea, L., and Deisseroth, K.(2007). Circuit-breakers: optical technologies for probing neural signals andsystems. Nat. Rev. Neurosci. 8, 577–581.

Neuron 62, April 30, 2009 ª2009 Elsevier Inc. 189