See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/273387237

Emergence of Cognition from Action

Article  in  Cold Spring Harbor Symposia on Quantitative Biology · March 2015

DOI: 10.1101/sqb.2014.79.024679 · Source: PubMed

CITATIONS READS

33 453

3 authors:

Gyorgy Buzsáki Adrien Peyrache

NYU Langone Medical Center NYU Langone Medical Center
455 PUBLICATIONS   70,532 CITATIONS    38 PUBLICATIONS   1,415 CITATIONS   

SEE PROFILE SEE PROFILE

John L Kubie
State University of New York Downstate Medical Center
58 PUBLICATIONS   5,110 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Optoelectrode View project

Extracellular Action Potentials View project

All content following this page was uploaded by Adrien Peyrache on 01 April 2015.

The user has requested enhancement of the downloaded file.

 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

Emergence of Cognition from Action

GYORGY BUZSÁKI,1,2 ADRIEN PEYRACHE,1 AND JOHN KUBIE3
1
The Neuroscience Institute, New York University, School of Medicine, New York, New York 10016
2
Center for Neural Science, New York University, School of Medicine, New York, New York 10016
3
Department of Cell Biology, Downstate Medical Center, State University of New York
at Brooklyn, New York, New York 11203
Correspondence: [email protected]

Theories of brain function have evolved through multiple stages. The first proposition was that brain networks support a set of
reflex responses, with current sensory inputs producing immediate motor outputs. The behaviorist paradigm suggested that
actions can always be explained as a response to immediate external cues. In response to these views, the cognitive paradigm
argued that behavior cannot be understood simply as input– output functions because the hidden layers of brain generate
unpredictability. The central processing was termed “cognition.” Here we propose a neuroscience-based model of cognition.
Our core hypothesis is that cognition depends on internal models of the animal and its world, where internally generated
sequences can serve to perform “what if” scenarios and anticipate the possible consequences of alternative actions without
actually testing them, and aid in the decisions of overt actions. We support our hypotheses by several examples of recent
experimental findings and show how externally guided cell assembly sequences become internalized to support cognitive
functions.

Humans, and likely other animals, can imagine far into example is the hippocampal system. Lesion data in hu-
the future; we spend a great deal of waking time devoted mans provided evidence that bilateral removal of the hip-
to such covert activities (Killingsworth and Gilbert 2010; pocampi produced severe and irreversible amnesia
Immordino-Yang et al. 2012; Mar et al. 2014). The main (Scoville and Milner 1957). Single unit studies in the
thesis of this essay is that such ability of the brain arises hippocampus and entorhinal cortex of animals gave rise
from its disengagement from immediate overt actions as a to the prominent theory that the fundamental function of
result of experience. We hypothesize that brains, irrespec- the hippocampal system is supporting spatial navigation
tive of size, are predictive devices that exploit regularity (O’Keefe and Nadel 1978; Hafting et al. 2005; McNaugh-
and recurrence as a fundamental property of the surround- ton et al. 2006), assisted by inputs from the head-direction
ing world and apply effective heuristics to make such system (Ranck 1985; Taube 2007). In contrast, studying
predictions. Prediction is made possible by adaptive the collective behavior of hippocampal neurons by record-
mechanisms that are supported by learning rules that ei- ing the local field potential (LFP) offered the conclusion
ther apply across generations (evolutionary adaptation) or that hippocampal theta oscillations are an unmistakable
within the lifetime of the organism. As a result, organ- reflection of voluntary action (Vanderwolf 1969). These
isms can deal with a future occurrence of the same or independent ideas of hippocampal functions persisted in
similar situations more effectively. This is the fundamen- parallel for decades without true interactions, apart from
tal organization principle of any adaptive system. Here we occasional polite gestures and references to the competing
describe example organizational brain patterns, with in- frameworks (Buzsáki 2005). As discussed below, these
creasing predictive efficiency, and present a hypothesis of barriers are now being dismantled by the recognition
how internally generated neuronal network operations, that although neurons and networks in the memory/nav-
detached from outside inputs, can serve as the necessary igational/planning systems are primarily engaged in self-
substrate for cognition. organized activity by theta oscillations, they can robustly
One aspect of the argument we will present is that most respond to environmental or body-derived inputs (Frank
brain structures can have “dual use”: at times, connected et al. 2000; Wood et al. 2000; Fujisawa et al. 2008; Pas-
to immediate sensory input and output; at other times talkova et al. 2008).
entrained by internal signals and brain rhythms. For In “simple” nervous systems, the connection between
many years, researchers tended to associate preexisting output and input networks is direct and immediate and the
verbal terms, such as memory, planning, envisioning the consequences of their action are mediated by the environ-
future, volition, and decision-making, with different and ment, as illustrated in Figure 1A. More complex brains
distinct brain structures. Oftentimes, even studying the are organized in a “multiple loop” pattern; that is, a series
same structure or system generated seemingly contradic- of interacting parallel loops are imposed between input
tory hypotheses depending on the chosen approach, pre- and output, as illustrated in Figure 1B,C. For example, in
conception, or experimental method used. A striking mammals, new circuits are added onto the monosynaptic

Copyright # 2014 Cold Spring Harbor Laboratory Press; all rights reserved; doi: 10.1101/sqb.2014.79.024679
Cold Spring Harbor Symposia on Quantitative Biology, Volume LXXIX 1
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

2 BUZSÁKI ET AL.

Figure 1. Externally driven and self-organized cell assemblies. (A) Evolutionarily simple brains contain simple neural networks.
Sensory input from the body and the environment activates input neurons, which interact with output neurons to generate appropriate
reflex actions in a short time window. (B) In more complex brains, multiple interacting loops of increasing length improve prediction
of more elaborate events and can make predictions in more complex environments at longer timescales. (C ) After extensive training, the
loops can sustain self-organized, long-lasting neuronal sequences without reliance on external cues and can, therefore, support cognitive
operations such as memory, planning, and imagination. Progression of neuronal operations correlates with elapsed time (green arrows)
irrespective of whether the operations are driven externally or internally. (Reprinted, with permission, from Buzsáki 2013.)

spinal cord/brainstem afferent – efferent connections, are synonymous. Therefore, we separate the functioning
including the thalamocortical system, cortico-corticorti- of complex brains into two modes: First, an “engaged”
cal, corticocerebellar, corticostriatal, and corticolimbic mode, when the immediate environment controls and
loops. As in the simple system, the parallel loops operate feeds back on behavior even when the inputs are quite
to predict and secure effective outcomes of actions in ever complex and actions need to be delayed, illustrated in
more complex environments and at much longer time- Figure 1B. Second, a “disengaged” or detached mode,
scales. These operations require storage of large amounts when processing continues independent of environmental
of past experience and the ability to evaluate and weigh inputs and behavioral responses of the animal. This in-
the importance of simultaneously experienced modalities put/output disengagement in the presence of continued
and compare them with similar occurrences in the past computation of potential responses is what we refer to as
and heuristics responses of the organism. Many of these “cognition.” This framework may explain why damage to
operations require that brain activities continue even after multiple loops of the brain has no or relatively minor
transient or more durable disappearance of sensory inputs impact on operations when input dimensions are limited
and delaying the brain’s deliberations in terms of effector and input – output gaps are short, but their role becomes
outputs. We speculate that such a training process, a result evident when the organism needs to evaluate a large and
of both phylogenetic and ontogenetic accumulation of complex parameter space and execute a response most
experience, makes the longer, more complex neuronal beneficial to the organism’s needs.
loops “smarter” (i.e., more effective by enabling them There are costs and advantages to the disengaged
to interpolate and extrapolate events surrounding the or- mode. The cost is that it takes greater neural resources
ganism). Our second and critical assumption is that a self- to create the “virtual worlds”; in growing brains, these are
organized computation emerges as a result of training the necessary for processing internally generated patterns
brain, which can proceed to deliver abstract goals by without reliance on immediate sensory feedback cues.
disengaging the effector networks from overt (muscular In the past, Gibson and heirs developed ideas that have
or vegetative) actions. come to be known as “embodied cognition” (Gibson
Activity in any system, including the brain, can be 1977; O’Regan and Noë 2001; Thompson and Varela
maintained by only two mechanisms: externally applied 2001). These authors emphasize that the environment
energy or intrinsic resources. By definition, a system that of an animal affords physical and sensory constraints
maintains its activity without an external drive must be (i.e., reduces the need to compute all possible circum-
driven by “self-organized” mechanisms. “Self-orga- stances) that ease the computational load (Gibson’s
nized” is the term most often used in physics and engi- “affordances”). Without real-world constraints, disen-
neering that refers to autonomously generated patterns. In gaged processing requires that the brain creates a well-
neurophysiology, the term “spontaneous” activity has endowed model of the world. With an internalized “vir-
been in use for decades to refer to the same process. In tual world” in the brain, the animal is capable of cognitive
cognitive science, the terms “internally generated” or “in- vicarious behavior; that is, it can perform “what if” sce-
ternally organized” refer to the same process. These terms narios in its head and anticipate the consequences of
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

EMERGENCE OF COGNITION FROM ACTION

3
alternative actions without actually testing them. With single dimension, which is the head direction. Head-direc-
this mechanism at hand, the brain can create new knowl- tion neurons fire robustly when the animal’s head points in
edge not only by interacting with the outside world, but a specific direction (Ranck 1985; Taube et al 1990a,b;
also through “vicarious” (imagined) experience. Disen- Taube 2007; McNaughton et al. 2006). Thus, the one-di-
gaged processing, which we call “cognition,” provides mensional head-direction system offers an opportunity to
dramatic advantages in predicting the consequence of understand the nature of external and internal interactions
actual behavior in complex environments and at long experimentally. The head-direction system consists of a
timescales. largely serially connected brain network, which includes
There are numerous examples of disengagement of the brainstem, mammillary bodies, anterodorsal thalamic
brain activity from afferent inputs. A striking example nucleus, postsubiculum, and entorhinal cortex (Mc-
is the “internalization” of brain rhythms. The first orga- Naughton et al. 2006; Sargolini et al. 2006; Taube 2007).
nized neocortical pattern in the developing mammalian Numerous experiments have showed that head-direc-
brain is the emergence of transient periods of spindle- tion neurons can be controlled by peripheral inputs, main-
shaped rhythmic activity at 10 – 25 Hz amid seconds- or ly the vestibular, visual, and ancillary afferents and the
tens-of-seconds-long silent periods (“tracé alternant”; angular velocity of head rotation (Sharp et al. 2001;
Dreyfus-Brisac 1962; Anderson et al. 1985; Khazipov Taube 2007). How do these neurons behave in the ab-
et al. 2004). In the newborn rat and premature human sence of such inputs, when the brain is disengaged from
babies, all these early spindles are triggered by some the environment, as is the case during sleep? Computa-
kind of a movement, including isolated muscle twitches, tional models have long assumed that head-direction neu-
limb jerks, and whole body startles (Khazipov et al. rons with similar preferred directions fire together, while
2004). This organization is hypothesized to serve as a the other neurons are suppressed. According to these
mechanism to construct an internal map of the body in models, the temporally engaged group of head-direction
the somatosensory system (Khazipov et al. 2004; Buzsáki neurons (i.e., a “hill of activity” or “an activity packet”)
2006; Khazipov and Buzsáki 2010). Once the body map moves on a virtual ring as the animal turns its head
is formed (5 – 10 d postnatally in the rat and in the third (Fig. 2; Skaggs et al. 1995; Redish et al. 1996; Burak
trimester of pregnancy in humans), the dependence of and Fiete 2012; Knierim and Zhang 2012). However,
brain activity on action-triggered reafferentation is no experimental demonstration of the existence of internally
longer needed and spindles become part of sleep, the organized neuronal populations endowed with such prop-
default organizing mode of the brain. This is a simple erties had to wait until recording technologies matured to
example of how brain patterns that are initially correlated be able to record simultaneous population of neurons and
with sensorimotor activity can disengage from external compare their dynamic interactions during both waking
inputs, while sustaining their “spontaneous” activity pat- behavior and sleep. Recent experiments provide such ev-
terns. Below we discuss two more examples to illustrate idence (Peyrache et al. 2015).
our disengagement hypothesis of cognition. First, we il- Not only do head-direction neurons continue to be ac-
lustrate how neurons in the head direction system contin- tive during sleep but their temporal sequential activity
ue their coordinated activity during sleep. Second, we patterns in the waking mouse are preserved. At any time,
show how internally organized cell assembly sequences the neuronal population was characterized by the presence
can support mental travel of memory and action planning, of an activity packet that was consistent with its structure
and how these functions emerge from spatial navigation during waking: Neurons coding for nearby orientation
in the physical environment. continued to fire together. Therefore, a “virtual gaze”
(i.e., the direction the mouse was “looking”) could be re-
constructed from firing of neurons in the anterodorsal tha-
SELF-ORGANIZED ATTRACTORS
lamic nucleus and postsubiculum (Fig. 2). Although both
OF DIRECTION SENSE
firing rates and sequential firing patterns persisted in all
The relationship between stimulus-driven and self-or- brain states, the temporal dynamics of the drift of the ac-
ganized (or internally generated) activity is a recurring tivity packet varied. During REM sleep, when brain elec-
topic in neuroscience. It is often tacitly assumed that the trophysiology is strikingly similar to waking, the packet
information from the sensory world is “relayed” by the moved at the same speed as in the waking animal, essen-
various thalamic nuclei to the neocortex where it is com- tially recapitulating the situation when the mouse moved
bined with top-down, internally generated cortical ac- its head in different direction in the waking animal. During
tivity. How feed-forward, “labeled line” sensory signals non-REM sleep, the HD signal drifted at a 10 times faster
in subcortical, thalamic, and cortical networks interact speed, compared with waking and REM sleep, essentially
with self-generated (“spontaneous”) activity is not well- following the faster dynamics of non-REM sleep de-
understood. Part of the difficulty is technical, due largely scribed in cortical networks (Buzsáki 1989; Wilson and
to the high dimensionality of sensory signal attributes McNaughton 1994). Of course, no real head movement
(e.g., intensity, color, orientation in case of vision; Rao accompanied these changes in the sleeping animal; there-
and Ballard 1999; Engel et al. 2001; Varela et al. 2001; fore, the temporal organization had to rely on internally
Buzsáki 2006). generated mechanisms rather than external stimuli.
Perhaps the simplest representation animals have is the The observation that the neuronal code of the head-
sense of direction. It is simple in the sense that it has a direction sense is preserved during sleep raises the possi-
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

4 BUZSÁKI ET AL.

Figure 2. Persistence of head-direction cell assemblies in the anterodorsal nucleus of the thalamus during waking and sleep.
Waking. Simultaneously recorded head-direction cells with different direction preferences during exploration (right). Each line is a
neuron, increasing firing rates are represented by hot colors. Neurons are ordered according to their preferred head-direction. Note
sequential activity of head-direction neurons (from top to bottom). Bottom trace (white) is local field potential recording from the
hippocampus. (Left) Head-direction cell assembly activity displayed on a ring. As the head points to different directions, the hill of
spiking activity moves on the ring accordingly. REM, same arrangement as in waking. Note continued sequentially changing
activity of the head-direction cells despite the absence of head or body movement. (Reprinted, with permission, from Peyrache et al.
2015.)

bility that self-organized mechanisms continue to be at ner 1957; O’Keefe and Nadel 1978). It has been proposed
work also in the waking animal. Such an internal com- recently that neuronal mechanisms of memory and plan-
putation can amplify and enhance precision of the head- ning (“mental travel into the past and future”; Suddendorf
direction signal by filtering out, in the case of the head- and Corballis 1997, 2007) have evolved from mecha-
direction, all additional information that is not compatible nisms of navigation in the physical world (Buzsáki
with a one dimensional angular signal. In the exploring 2005; Buzsáki and Moser 2013). Below, we discuss that
animal, sensory inputs can be combined with the predic- the neuronal algorithms underlying navigation in real and
tion of the internally generated head direction and rapidly mental space are fundamentally the same. The key dis-
adapt to reconfigured environments (Zugaro et al. 2003). tinction is that during real world travel, external cues,
In case of ambiguous or conflicting signals, self-orga- such as landmarks, combined with self-motion cues aid
nized mechanisms may generate the brain’s “best guess” in driving cell assembly sequences; when travel is inter-
by interpolating across input signals or extrapolating the nal, the brain must supply the supporting cues for navi-
correct position of the head-direction vector from limited gating through a mental space.
or ambiguous sensory information. Overall, these exper- To support mental travel effectively, a neural sys-
imental findings in a simple but fundamentally important tem evolved for navigation must develop the ability
sense show that internally organized mechanisms are to self-generate temporally evolving cell assembly se-
continuously used to improve the brain’s interpretation quences and a capacity to store large quantities of seem-
of the external world (Fig. 1B). ingly unrelated episodes. It has been suggested that the
evolution of the entorhinal –hippocampal system illus-
trates the phylogenetic continuity of navigation and
FROM NAVIGATION IN THE PHYSICAL
memory and has the anatomical and physiological prop-
WORLD TO MENTAL TRAVEL
erties that make it especially suitable for meeting the
The hippocampal – entorhinal cortex has long been above requirements (Buzsáki 2005; Buzsáki and Moser
linked to both navigation and memory (Scoville and Mil- 2013).
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

EMERGENCE OF COGNITION FROM ACTION

5
SENSORY-DRIVEN ACTIVITY OF THE 1996). Moreover, spatial firing does not require a rich
HIPPOCAMPAL NAVIGATION SYSTEM set of landmarks. Landmarks can be impoverished, sub-
sets removed, or the lights turned out, and firing will
The explicit spatial coordinates of an animal are encod- remain stable and robust (Muller and Kubie 1987; Quirk
ed by a combination of a variety of neuron types, including et al. 1990). These findings indicate that the brain is
“head-direction cells” of the thalamus and postsubiculum capable of interpolating between cues and extrapolating
(Ranck 1985; Taube 2007), “place cells” in the hippo- (generalizing or “pattern-completing”) from the available
campus (O’Keefe and Nadel 1978), “grid cells” (Hafting external information.
et al. 2005), and “border cells” (Solstad et al. 2008) of the However, even small insects, whose nervous systems
entorhinal cortex and surrounding structures. The exact possess many fewer neurons and are made of simpler
functions of these cell types are yet to be determined, but circuits, are able to navigate using such strategies (Men-
it has been suggested that the periodically active fields of zel et al. 1998). In principle, a small-sized network
the grid cells provide a metric for the neural representa- should be sufficient for rodents as well if navigation in
tion of space, in the same way that head-direction cells a relatively fixed environment and range would be the
provide a directional reference frame, whereas border only goal. Indeed, in laboratory experiments, the explored
cells can assess allocentric distances in the discoverable environment can be mapped at centimeter precision by
environment by triangulation (Moser et al. 2014). just a dozen or so grid cells (Fyhn et al. 2007) or place
The “temporally evolving sequences” referred to above cells (Wilson and McNaughton 1993), What advantage,
are, in navigational terms, paths. The paths represent a then, do the complex networks of the entorhinal – hippo-
temporal sequence of neighboring locations on a map. campal system of mammals offer? One can speculate that
Each map represents a specific environment. “Remap- there are two advantages: first, to map specific features of
ping” refers to the finding that the organization of place multiple environments; second, that such self-organized
cells scramble from one environment to another (Muller networks (Fig. 1B) also have the ability to effectively
and Kubie 1987; Kubie and Muller 1991). Because place interpolate and extrapolate from the available cues and
cells that have neighboring firing fields will likely not hold information over long timescales, as well as use
have neighboring firing fields in a second environment, previous experience from similar situations in more com-
a single sequence of place cell firing not only codes a plex environments. The ability to rapidly acquire multiple
specific path, but also a specific environment. Recent representations, to maintain separation of these represen-
findings, however, refine this strict “orthogonalized” tations, to hold the representations over long timescales,
view of remapping. Although most neurons have single and to operate effectively on any, even in degraded con-
place fields in a given environment, a small minority can ditions, are several of the robust navigational capacities in
have multiple fields and this same minority of neurons mammalian brains compared with simpler brains of spe-
continues to fire in multiple environments. In fact, the cies where navigation is accomplished by much smaller
distribution of the number of place fields per hippocam- circuits.
pal neuron follows a lognormal distribution (Buzsáki and
Mizuseki 2014). This small subset is not only more active
in multiple environments, but their firing rates are higher,
SELF-ORGANIZED ACTIVATION OF THE
emit more spike bursts and their place fields are larger
HIPPOCAMPAL NAVIGATION SYSTEM
compared with those of the majority neurons. Important-
ly, the log-based rule organization applies to all cortical A second way of generating sequential firing of cell
regions and multiple neuronal functions from synaptic assemblies is by self-organization (Fig. 3), independent
strength distribution to macroscopic connectivity. The of cues from either the environment or the body. Such
ever-active minority provides the brain’s “best guess” internally generated neural assembly sequences have
and offers “good enough” solutions to get by in any sit- been long assumed to be a necessary substrate of cogni-
uation. On the other hand, the majority of less active tion for various mental operations such as memory recall,
neurons comprises a large reservoir that can be mobilized planning, and imagination (Hebb 1949). Indeed, perpet-
to precisely distinguish one situation from another and ually changing hippocampal assembly sequences can be
incorporate novel ones as distinct (Buzsáki and Mizuseki observed experimentally. For example, if a rat is trained
2014). to run in a wheel with approximately the same speed and
When the animal explores an environment, head-direc- facing in the same direction, the environmental and body-
tion neurons, border cells, grid cells, and place cells con- derived cues remain constant throughout the run. If such
tinuously and coherently change their firing activity. “clamped” behavior is part of a memory task to remem-
Traditionally, two sources of information have been con- ber a previously made choice for planning for the appro-
sidered as driving factors of the temporal evolution of cell priate future choice, hippocampal neurons display con-
assembly sequences (Fig. 2). The first one is the constel- tinuously changing assembly sequences (neuronal
lation of the landmarks available as the animal navigates “trajectories”; Fig. 4). Importantly, several measures of
(O’Keefe and Nadel 1978). The second may derive from the firing patterns of the internally organized sequences
the ideothetic (i.e., body-derived) cues from the vestibu- during the delay period, including the duration of the
lar and proprioceptic systems and self-motion-driven op- spiking activity of the neurons and the temporal relation-
tical flow from mostly local cues (McNaughton et al. ship of their spikes relative to the reference theta oscilla-
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

6 BUZSÁKI ET AL.

Figure 3. Cell assembly sequences. Sequential activa-

tion of neuronal assemblies neurons (1 –n) can be con-
trolled by the changing constellation of environmental
cues and/or proprioceptive information from the body
(top). During cognitive activity, sequential activation is
supported by self-organized patterning of assembly ac-
tivity (bottom). Not only first-order (neighbor) but also
higher-order (nonneighbor) connections can be repre-
sented in strongly connected recurrent networks.

tion cycle (O’Keefe and Recce 1993), are virtually iden- activity even in the absence of theta rhythms (Wang
tical with those of the place cells during translational et al. 2015). The cognitive relevance of self-organized
behavior (Pastalkova et al. 2008). The implication of sequential activity is emphasized by the observation
these observations is that the physiological mechanisms that identical initial conditions (e.g., a left choice was
that govern the progression of cell assembly sequences in rewarded) induces a similar assembly sequence each
the hippocampus during navigation and memory are quite time, whereas different conditions (i.e., different memo-
similar. Although theta oscillation is an obligatory mech- ries) can give rise to uniquely different neuronal trajecto-
anism for generating internal cell assembly sequences, ries, which accurately predicts upcoming choices in the
environmental cues can maintain sequential place cell maze, including erroneous turns (Fig. 4). In accordance

Figure 4. Internally generated assembly sequences during cognitive activity. (A) Sequential firing patterns of hippocampal neurons in
a memory task. Center: Color-coded spikes (dots) of simultaneously recorded hippocampal CA1 pyramidal neurons. The rat was
required to run in the wheel facing to the left during the delay between the runs in the maze. (Left) Normalized firing rate profiles of
neurons during wheel running, ordered by the latency of their peak firing rates during left trials (each line is a single cell). (Right)
Normalized firing rates of the same neurons during right trials. (B) Sequential firing patterns of prefrontal pyramidal cells in a working
memory task. (Middle) Cheese odor or chocolate odor in the start area signals the availability of cheese or chocolate reward in the left
or right goal area ( position 1), respectively. Travel trajectories were linearized (0 – 1). (Left) Neurons were ordered by the location of
their peak firing rates relative to the rat’s position in the maze during left trials. Each row represents the position-dependent normalized
firing rate of a single neuron. (Right) Normalized firing rates of the same neurons during right trials. (A and B are reprinted, with
permission, from Pastalkova et al. 2008 and Fujisawa et al. 2008, respectively.)
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

EMERGENCE OF COGNITION FROM ACTION

7
with experiments in rodents, single-unit studies in human 1992). Episodic memory, in contrast, allows learning
patients showed that the hippocampus and entorhinal and recalling self-referenced (first-person) experiences
cortex can generate numerous trajectories corresponding in the context of both space and subjective time (Tulving
to different memory episodes (Gelbard-Sagiv et al. 2008). and Thomson 1972) for planning actions (Tulving 2002;
Subjects first viewed cinematic episodes that depicted Buckner 2010), similar to linking together location se-
short clips of famous people and characters. Several neu- quences during exploration. Acquiring a spatial map re-
rons responded selectively to only one or a few of the quires repeated exploration of the landmarks by self-
episodes. In the critical part of the experiment, subjects referenced navigation. Similarly, semantic knowledge is
were asked to recall the characters and scenes without generated by repeated episodic encounters until the items
any cues. Remarkably, the neurons that fired slightly be- lose their spatiotemporal context (Fig. 5; Squire 1992;
fore free recall (i.e., a verbal report) corresponded to the Eichenbaum et al. 1999; Buzsáki 2005).
same neurons that were activated while watching the cin- On the one hand, declarative memories may be recalled
ematic episodes in the learning phase (Gelbard-Sagiv by supplying a variety of real world cues that induce an
et al. 2008). These results support the view that the neu- association. This is, roughly, equivalent to an animal ex-
ronal assembly sequences that encode episodes are the posed to a few cues from an environment recalling an
same that are active during the mental representation of entire environmental representation. On the other hand,
those episodes (Tulving and Thomson 1973). declarative memories may be recalled without real
Overall, these experiments illustrate that perpetually world cue triggers. For example, one may recall an epi-
changing cell assembly sequences can be induced in the sodic memory or a fact by association with another mem-
hippocampus independent of external cues (Fig. 1C), and ory. Such vicarious memory recall is analogous to
such disengagement can support cognitive performance. vicarious navigational recall. We propose they have iden-
Generation of neural sentences is not confined to the tical mechanisms.
hippocampal system. In the medial prefrontal cortex of
the rat, long neuronal sequences can also reliably differ-
entiate between right and left trajectories in the maze
INTERNALLY GENERATED NEURONAL
before making a choice, with individual neurons active
PATTERNS AND COVERT OUTPUTS IN THE
only for a short duration (Fig. 4; Fujisawa et al. 2008).
CEREBRAL CORTEX
The possibility exists that choice specific neuronal trajec-
tories in the hippocampus are selected by inputs from the We hypothesize that internally generated patterns are a
medial prefrontal cortex. Similar neuronal trajectories common feature of cortical organization which can func-
have been also been described in the parietal cortex (Har- tion as a neuronal substrate to bridge long delays between
vey et al. 2012). In summary, cortical circuits can produce inputs and action. Connectivity patterns found in neocor-
multitudes of unfolding assembly sequences in two dif- tex support this notion. Superficial cortical neuronal lay-
ferent ways: either by responding to environmental/idio- ers may be viewed as a substrate for inducing such covert
thetic stimuli, when such inputs are available, or by outputs. While deep layer cortical neurons project to ac-
generating them internally. tion systems of the thalamus, colliculi, brainstem, and
Analogous to spatial navigation, two forms of declar- even the spinal cord, the superficial layer can be viewed
ative memory can be distinguished (Squire 1992). Similar as an extra, added loop to the deep layer with subcortical
to the defining locations of the allocentric map, semantic outputs (Shepherd 2013). A novel neuron loop that is
memory explicitly defines living things, objects, facts, added to the sensorimotor short loop is the extensive
and events of the surrounding world independent from associational areas, particular the prefrontal cortex.
temporal context (Tulving and Thomson 1972; Squire Thus, perhaps it is not surprising that the anatomical

Figure 5. Place cell firing represents both episodic and semantic memory. Paths 1 and 2 are accompanied by cell assembly sequences
A – B – C and D– B – E. Internally generated firing of either of these sequences is equivalent to a specific episodic memory: movement
along Path 1 or Path 2. After a series of paths that activate cell assembly B, B alone is not associated with a particular path: it explicitly
represents a location. In this representation, the activation of cell assembly B, alone, is a semantic fact, location B.
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

8 BUZSÁKI ET AL.

afferent – efferent connectivity and internal organization 2012). Imaging experiments in humans have repeatedly
of the agranular motor cortex and the prefrontal cortex are found that imagining and planning invariably activated
so remarkably similar (Nimchinsky et al. 1995; Gabbott structures previously categorized as parts of the memory
et al. 2005). The notion is that there are two types of system (Buckner and Carroll 2007; Schacter and Addis
motor output: the familiar “engaged” output, where mo- 2007; Buckner 2010). The emerging field of “active sens-
tor cortical areas activate motor neurons and produce ing” (Katz 1989; Engel et al. 2013; Loeb and Fishel 2014)
behavior; and vicarious motor output, where an internal also emphasizes the primacy of action in acquiring mean-
analog of motor action is generated, with output remain- ing for sensation and perception (Schroeder et al. 2010).
ing entirely within the brain. If the brain contains a suf- These novel (or “revisited”) views break away from the
ficient model of the self and world, this type of imagined tradition of “boxing” man-invented terms into particular
or vicarious motor action can be examined internally for brain structures and strategies to identify physiological
potential consequences. This vicarious action is what we boundaries for preconceived mental boundaries (Buzsáki
call a “cognitive process.” Such vicarious action can be 2010). Instead, they emphasize interactions of brain sys-
used for navigation, as in the hippocampal navigational tems as the fundamental operations that give rise to
system described earlier, or for other forms of motor behavioral entities that are hard to categorize strictly as
behaviors. At the physiological level such cognitive op- memory or planning, overt or covert, conscious or “auto-
erations require that the self-organized neuronal assembly pilot.” Regarding the self-organized brain activity as the
sequences are propelled forward via an optimal path or by fundamental brain operation and viewing the brain’s re-
covert trial and error evaluating multiple possible out- sponses to external perturbations as secondary actions
comes, solving detours, and finding shortcuts (Gupta offers a new paradigm, in which “meaning” is acquired
et al. 2012), much the same way as we navigate through by matching preexisting neuronal patterns to action-per-
locomotion in the real world in the presence of environ- ception (Buzsáki 2006). In support of this conjecture,
mental feedback. experiments in developing ferrets show that the similarity
between spontaneous and evoked activity increases pro-
gressively with age and is specific to responses evoked by
natural scenes, suggesting that internal models result
MEMORY AND THE OPTIMIZATION
from the adaptation of brain states to the statistics of the
OF BEHAVIOR
surrounding world (Fiser et al. 2004). By extension, the
As discussed earlier, a fundamental goal of many neu- construction of an internal representation of the self as a
ronal circuits and the brain as a whole is to predict the special entity can be accomplished by the reprocessing of
future. However, predicting the future is possible only internal data simultaneously with information about the
after the organism has already experienced the conse- environment.
quences of its actions and stored the outcomes of those A critical issue that has to be explored in future exper-
successful and failed actions in memory. Indeed, a bar- iments is the significance of the duration of internally
rage of recent papers document the recognition that struc- organized sequences for cognition. There are numerous
tures that have been traditionally viewed as memory brain operations, especially in the motor system, which
systems are inseparable parts of planning, imagining, require sequential activity of neuronal firing, yet we are
and action systems (Buckner and Carroll 2007; Schacter not aware of the consequences of those operations and,
and Addis 2007; Lisman and Redish 2009; Buckner thus, they have not been considered as cognitive. Neuro-
2010). What appeared to be memory of the past in the nal events that are too short (,500 msec; Libet 2004)
sequential activity of cell assembly sequences, equally tend not to evoke awareness. These likely include the
well reflected the planned future action of the animal, short (,100 msec) sharp wave ripple events of the hip-
including commission errors, indicating that the action pocampus (Buzsáki 1989). Yet, these events may be re-
outcome is the consequence of the brain’s interpretation garded as “subconscious rehearsals” for future actions or
of the past experience (Fujisawa et al. 2008; Pastalkova priming for memory recall.
et al. 2008). Numerous sleep “replay” experiments in Although ample evidence exists to support the view
both rodents and humans show that sleep patterns are that internally organized cell assembly sequences are
far from random but, instead, are reminiscent of waking the fundamental basis of mental operations, the “clutch”
when the neuronal trajectories evolve under the influence operation is not well understood. However, currently ex-
of sensory inputs (Wilson and McNaughton 1994; Ná- isting and future technologies can be deployed to test the
dasdy et al. 1999; cf. Diekelmann and Born 2010). Trans- hypothesis that mental travel and planning occur in real
fer operations, such as the sharp wave ripples of the time, similar to navigation. By reading out brain signals
hippocampus, which were originally suggested to con- and connecting them to actuators, such as a navigating
solidate and move transient memories from the hippo- robot, thought control of the robot’s movements should
campus to neocortex (Buzsáki 1989; Wilson and be possible. Experiments performed within the brain –
McNaughton 1994), have been shown to generate spike machine interface paradigm (Donoghue 2002; Fetz
sequences that predict to-be-visited places and delayed 2007; Nicolelis and Lebedev 2009) show the feasibility
actions or a chaining of past and expected events (Diba of such outcomes. Furthermore, neuronal patterns that
and Buzsáki 2007; Davidson et al. 2009; Gupta et al. can be read out during sleep can also be exploited to
2010; cf., Carr et al. 2011; Buzsáki and Lopes da Silva compare them with waking operations and manipulate
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

EMERGENCE OF COGNITION FROM ACTION

9
them to establish whether internally generated patterns in Engel AK, Fries P, Singer W. 2001. Dynamic predictions: Os-
the absence of conscious awareness can reflect waking- cillations and synchrony in top-down processing. Nat Rev
Neurosci 2: 704 – 716.
related neuronal trajectories. This paper highlights recent Engel AK, Maye A, Kurthen M, König P. 2013. Where’s the
work in animal physiology that shows the implications of action? The pragmatic turn in cognitive science. Trends Cogn
self-organized, internally generated neuronal assembly Sci 17: 202– 209.
sequences. These findings support an emerging concep- Fetz EE. 2007. Volitional control of neural activity: Implications
tion of the widespread importance of internal models as for brain computer interfaces. J Physiol 579: 571 – 579.
Fiser J, Chiu C, Weliky M. 2004. Small modulation of ongoing
covert, brain-based testing grounds for efficient optimi- cortical dynamics by sensory input during natural vision. Na-
zation of behavior. ture 431: 573 – 578.
Frank LM, Brown EN, Wilson M. 2000. Trajectory encoding
in the hippocampus and entorhinal cortex. Neuron 27:
ACKNOWLEDGMENTS 169– 178.
Fujisawa S, Amarasingham A, Harrison MT, Buzsáki G. 2008.
This work was supported by the National Institute of Behavior-dependent short-term assembly dynamics in the
Health Grants MH54671 and NS074015, National Sci- medial prefrontal cortex. Nat Neurosci 11: 823 – 833.
ence Foundation Grant, SBE 0542013, The Mathers Fyhn M, Hafting T, Treves A, Moser MB, Moser EI. 2007.
Hippocampal remapping and grid realignment in entorhinal
Foundation, Human Frontier Science Program (G.B. cortex. Nature 446: 190 – 194.
and A.P.), and European Molecular Biology Organization Gabbott PL, Warner TA, Jays PR, Salway P, Busby SJ. 2005.
postdoctoral fellowship (A.P.). Prefrontal cortex in the rat: Projections to subcortical auto-
nomic, motor, and limbic centers. J Comp Neurol 492: 145 –
177.
REFERENCES Gelbard-Sagiv H, Mukamel R, Harel M, Malach R, Fried I. 2008.
Internally generated reactivation of single neurons in human
Anderson CM, Torres F, Faoro A. 1985. The EEG of the early hippocampus during free recall. Science 322: 96 –101.
premature. Electroencephalogr Clin Neurophysiol 60: 95 – Gibson J. 1977. The theory of affordances. In Perceiving, acting,
105. and knowing: Toward an ecological psychology (ed. Shaw R,
Buckner RL. 2010. The role of the hippocampus in prediction Bransford J), pp. 62 – 82. Lawrence Erlbaum Associates, Hill-
and imagination. Annu Rev Psychol 61: 27 – 48. sdale, NJ.
Buckner RL, Carroll DC. 2007. Self-projection and the brain. Gupta AS, van der Meer MA, Touretzky DS, Redish AD. 2010.
Trends Cogn Sci 11: 49 –57. Hippocampal replay is not a simple function of experience.
Burak Y, Fiete IR. 2012. Fundamental limits on persistent ac- Neuron 65: 695 – 705.
tivity in networks of noisy neurons. Proc Natl Acad Sci 109: Gupta AS, van der Meer MA, Touretzky DS, Redish AD. 2012.
17645 – 17650. Segmentation of spatial experience by hippocampal theta se-
Buzsáki G. 1989. Two-stage model of memory trace formation: quences. Nat Neurosci 15: 1032 – 1039.
A role for “noisy” brain states. Neuroscience 31: 551– 570. Hafting T, Fyhn M, Molden S, Moser MB, Moser EI. 2005.
Buzsáki G. 2005. u rhythm of navigation: Link between path Microstructure of a spatial map in the entorhinal cortex. Na-
integration and landmark navigation, episodic and semantic ture 436: 801 – 806.
memory. Hippocampus 15: 827 – 840. Harvey CD, Coen P, Tank DW. 2012. Choice-specific sequences
Buzsáki G. 2006. Rhythms of the brain. Oxford University Press, in parietal cortex during a virtual-navigation decision task.
New York. Nature 484: 62 – 68.
Buzsáki G. 2010. Neural syntax: Cell assemblies, synapsembles, Hebb DO. 1949. The organization of behavior. Wiley, New
and readers. Neuron 68: 362 –385. York.
Buzsáki G. 2013. Cognitive neuroscience: Time, space and mem- Immordino-Yang MH, Christodoulou JA, Singh V. 2012. Rest is
ory. Nature 497: 568 – 569. not idleness: Implications of the brain’s default mode for
Buzsáki G, Lopes da Silva FL. 2012. High frequency oscilla- human development and education. Perspect Psychol Sci 7:
tions in the intact brain. Prog Neurobiol 98: 241 – 249. 352– 364.
Buzsáki G, Mizuseki K. 2014. The log-dynamic brain: How Katz D. 1989. The world of touch. Lawrence Erlbaum Associ-
skewed distributions affect network operations. Nat Rev Neu- ates, Hillsdale, NJ.
rosci 15: 264 – 278. Khazipov R, Buzsáki G. 2010. Early patterns of electrical activ-
Buzsáki G, Moser EI. 2013. Memory, navigation and theta ity in the developing cortex. In Oxford handbook of develop-
rhythm in the hippocampal-entorhinal system. Nat Neurosci mental behavioral neuroscience (ed. Blumberg MS, Freeman
16: 130 – 138. JH, Robinson SR), pp. 161 – 177. Oxford University Press,
Carr MF, Jadhav SP, Frank LM. 2011. Hippocampal replay in Oxford.
the awake state: A potential substrate for memory consolida- Khazipov R, Sirota A, Leinekugel X, Holmes GL, Ben Ari Y,
tion and retrieval. Nat Neurosci 14: 147– 153. Buzsáki G. 2004. Early motor activity drives spindle bursts in
Davidson TJ, Kloosterman F, Wilson MA. 2009. Hippocampal the developing somatosensory cortex. Nature 432: 758– 761.
replay of extended experience. Neuron 63: 497 – 507. Killingsworth MA, Gilbert DT. 2010. A wandering mind is an
Diba K, Buzsáki G. 2007. Forward and reverse hippocampal unhappy mind. Science 330: 932.
place-cell sequences during ripples. Nat Neurosci 10: 1241 – Knierim JJ, Zhang K. 2012. Attractor dynamics of spatially
1242. correlated neural activity in the limbic system. Annu Rev Neu-
Diekelmann S, Born J. 2010. The memory function of sleep. Nat rosci 35: 267 – 285.
Rev Neurosci 11: 114 –126. Kubie JL, Muller RU. 1991. Multiple representations in the
Donoghue JP. 2002. Connecting cortex to machines: Recent hippocampus. Hippocampus 1: 240 – 242.
advances in brain interfaces. Nat Neurosci Suppl: 1085 – Libet B. 2004. Mind time. The temporal factor in consciousness.
1088. Harvard University Press, Cambridge, MA.
Dreyfus-Brisac C. 1962. The electroencephalogram of the pre- Lisman J, Redish AD. 2009. Prediction, sequences and the hip-
mature infant. World Neurol 3: 5 – 15. pocampus. Philos Trans R Soc Lond B Biol Sci 364: 1193 –
Eichenbaum H, Dudchenko P, Wood E, Shapiro M, Tanila H. 1201.
1999. The hippocampus, memory, and place cells: Is it spatial Loeb GE, Fishel JA. 2014. Bayesian action&perception: Repre-
memory or a memory space? Neuron 23: 209 – 226. senting the world in the brain. Front Neurosci 8: 341.
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

10 BUZSÁKI ET AL.

Mar RA, Mason MF, Litvack A. 2014. How daydreaming relates Scoville WB, Milner B. 1957. Loss of recent memory after
to life satisfaction, loneliness, and social support: The impor- bilateral hippocampal lesions. J Neurol Neurosurg Psychiatry
tance of gender and daydream content. Conscious Cogn 21: 20: 11 – 21.
401 – 407. Sharp PE, Blair HT, Cho J. 2001. The anatomical and compu-
McNaughton BL, Barnes CA, Gerrard JL, Gothard K, Jung MW, tational basis of the rat head-direction cell signal. Trends
Knierim JJ, Kudrimoti H, Qin Y, Skaggs WE, Suster M, et al. Neurosci 24: 289 –294.
1996. Deciphering the hippocampal polyglot: The hippo- Shepherd GM. 2013. Corticostriatal connectivity and its role in
campus as a path integration system. J Exp Biol 199: 173 – disease. Nat Rev Neurosci 14: 278 – 291.
185. Skaggs WE, Knierim JJ, Kudrimoti HS, McNaughton BL. 1995.
McNaughton BL, Battaglia FP, Jensen O, Moser EI, Moser MB. A model of the neural basis of the rat’s sense of direction. Adv
2006. Path integration and the neural basis of the ‘cognitive Neural Inf Process Syst 7: 173 – 180.
map’. Nat Rev Neurosci 7: 663 – 778. Solstad T, Boccara CN, Kropff E, Moser MB, Moser EI. 2008.
Menzel R, Geiger K, Joerges J, Muller U, Chittka L. 1998. Bees Representation of geometric borders in the entorhinal cortex.
travel novel homeward routes by integrating separately ac- Science 322: 1865– 1868.
quired vector memories. Anim Behav 55: 139 – 152. Squire LR. 1992. Memory and the hippocampus: A synthesis
Moser EI, Roudi Y, Witter MP, Kentros C, Bonhoeffer T, Moser from findings with rats, monkeys, and humans. Psychol Rev
MB. 2014. Grid cells and cortical representation. Nat Rev 99: 195 – 231.
Neurosci 15: 466 – 481. Suddendorf T, Corballis MC. 1997. Mental time travel and the
Muller RU, Kubie JL. 1987. The effects of changes in the en- evolution of the human mind. Genet Soc Gen Psychol Monogr
vironment on the spatial firing of hippocampal complex-spike 123: 133 – 167.
cells. J Neurosci 7: 1951 – 1968. Suddendorf T, Corballis MC. 2007. The evolution of foresight:
Nádasdy Z, Hirase H, Czurkó A, Csicsvari J, Buzsáki G. 1999. What is mental time travel, and is it unique to humans? Behav
Replay and time compression of recurring spike sequences in Brain Sci 30: 299 – 313.
the hippocampus. J Neurosci 19: 9497 – 9507. Taube JS. 2007. The head direction signal: Origins and sensory-
Nicolelis MA, Lebedev MA. 2009. Principles of neural ensem- motor integration. Annu Rev Neurosci 30: 181 – 207.
ble physiology underlying the operation of brain– machine Taube JS, Muller RU, Ranck JB Jr. 1990a. Head-direction cells
interfaces. Nat Rev Neurosci 10: 530 – 540. recorded from the postsubiculum in freely moving rats.
Nimchinsky EA, Vogt BA, Morrison JH, Hof PR. 1995. Spindle I. Description and quantitative analysis. J Neurosci 10: 420–
neurons of the human anterior cingulate cortex. J Comp Neu- 435.
rol 355: 27 – 37. Taube JS, Muller RU, Ranck JB Jr. 1990b. Head-direction cells
O’Keefe J, Nadel L. 1978. The hippocampus as a cognitive map. recorded from the postsubiculum in freely moving rats. II.
Oxford University Press, New York. Effects of environmental manipulations. J Neurosci 10:
O’Keefe J, Recce ML. 1993. Phase relationship between hippo- 436 – 447.
campal place units and the EEG theta rhythm. Hippocampus Thompson E, Varela F. 2001. Radical embodiment: Neural dy-
3: 317 –330. namics and consciousness. Trends Cogn Sci 5: 418 – 425.
O’Regan JK, Noë A. 2001. A sensorimotor account of vision and Tulving E. 2002. Chronesthesia: Conscious awareness of sub-
visual consciousness. Beh Brain Sci 25: 883– 975. jective time. In Principles of frontal lobe function (ed. Stuss
Pastalkova E, Itskov V, Amarasingham A, Buzsáki G. 2008. DT, Knight RC), pp. 311 – 325. Oxford University Press,
Internally generated cell assembly sequences in the rat hippo- New York.
campus. Science 321: 1322– 1327. Tulving E, Thomson D. 1973. Encoding specificity and retrieval
Peyrache A, Lacroix MM, Petersen PC, Buzsaki G. 2015. Inter- processes in episodic memory. Psychol Rev 80: 352– 373.
nally organized mechanisms of the head-direction sense. Nat Tulving E, Donaldson W, Bower GH. 1972. Organization of
Neurosci (in press). memory. Academic Press, New York.
Quirk GJ, Muller RU, Kubie JL. 1990. The firing of hippocam- Vanderwolf CH. 1969. Hippocampal electrical activity and vol-
pal place cells in the dark depends on the rat’s recent experi- untary movement in the rat. Electroencephalogr Clin Neuro-
ence. J Neurosci 10: 2008 – 2017. physiol 26: 407 – 418.
Ranck J. 1985. In Electrical activity of archicortex (ed. Buzsáki Varela F, Lachaux JP, Rodriguez E, Martinerie J. 2001. The
G, Vanderwolf CH). Akadémiai Kiadó, Budapest. brainweb: Phase synchronization and large-scale integration.
Rao RP, Ballard DH. 1999. Predictive coding in the visual cor- Nat Rev Neurosci 2: 229 – 239.
tex: A functional interpretation of some extra-classical recep- Wang Y, Romani S, Lustig B, Leonardo A, Pastalkova E. 2015.
tive-field effects. Nat Neurosci 2: 79 – 87. Theta sequences are essential for internally generated hippo-
Redish AD, Elga AN, Touretzky DS. 1996. A coupled attractor campal firing fields. Nat Neurosci 18: 282 – 288.
model of the rodent head direction system. Netw Comput Wilson MA, McNaughton BL. 1993. Dynamics of the hippo-
Neural Syst 7: 671 – 685. campal ensemble code for space. Science 261: 1055 – 1058.
Sargolini F, Fyhn M, Hafting T, McNaughton BL, Witter MP, Wilson MA, McNaughton BL. 1994. Reactivation of hippocam-
Moser MB, Moser EI. 2006. Conjunctive representation of pal ensemble memories during sleep. Science 265: 676 – 679.
position, direction, and velocity in entorhinal cortex. Science Wood ER, Dudchenko PA, Robitsek RJ, Eichenbaum H. 2000.
312: 758– 762. Hippocampal neurons encode information about different
Schacter DL, Addis DR. 2007. Constructive memory: The ghosts types of memory episodes occurring in the same location.
of past and future. Nature 445: 27. Neuron 27: 623– 633.
Schroeder CE, Wilson DA, Radman T, Scharfman H, Lakatos P. Zugaro MB, Arleo A, Berthoz A, Wiener SI. 2003. Rapid spatial
2010. Dynamics of active sensing and perceptual selection. reorientation and head direction cells. J Neurosci 23: 3478–
Curr Opin Neurobiol 20: 172 – 176. 3482.
 Downloaded from symposium.cshlp.org on March 11, 2015 - Published by Cold Spring Harbor Laboratory Press

Emergence of Cognition from Action

Gyorgy Buzsáki, Adrien Peyrache and John Kubie

Cold Spring Harb Symp Quant Biol published online March 9, 2015
Access the most recent version at doi:10.1101/sqb.2014.79.024679

P<P Published online March 9, 2015 in advance of the print volume.

Email alerting Receive free email alerts when new articles cite this article -
service sign up in the box at the top right corner of the article or click
here

Advance online articles have not yet appeared in the print volume. Citations to
Advance online articles must include the digital object identifier (DOI) and date
of initial publication.

To subscribe to Cold Spring Harbor Symposia on Quantitative Biology go to:

http://symposium.cshlp.org/subscriptions

Copyright © 2014 Cold Spring Harbor Laboratory Press; all rights reserved

View publication stats