The Entangled Brain
Luiz Pessoa
抽象的
■ The Entangled Brain (Pessoa, L。, 2002. 与新闻界) promotes
the idea that we need to understand the brain as a complex,
entangled system. Why does the complex systems perspective,
one that entails emergent properties, matter for brain science?
实际上, many neuroscientists consider these ideas a distraction.
We discuss three principles of brain organization that inform
the question of the interactional complexity of the brain: (1)
massive combinatorial anatomical connectivity; (2) highly
distributed functional coordination; 和 (3) networks/circuits
as functional units. To motivate the challenges of mapping
structure and function, we discuss neural circuits illustrating
the high anatomical and functional interactional complexity typ-
ical in the brain. We discuss potential avenues for testing for
network-level properties, including those relying on distributed
computations across multiple regions. We discuss implications
for brain science, including the need to characterize decentra-
lized and heterarchical anatomical–functional organization. 这
view advocated has important implications for causation, 也,
because traditional accounts of causality provide poor candi-
dates for explanation in interactionally complex systems like
the brain given the distributed, 相互的, and reciprocal nature
of the interactions. 最终, to make progress understanding
how the brain supports complex mental functions, 我们需要
to dissolve boundaries within the brain—those suggested to
be associated with perception, 认识, 行动, 情感,
motivation—as well as outside the brain, as we bring down
the walls between biology, 心理学, mathematics, 电脑
科学, 哲学, 等等. ■
介绍
Neuroscience tends to study parts of the brain separately.
The Entangled Brain (Pessoa, 2022乙) promotes the idea
那, 反而, we need to understand the brain as a com-
丛, entangled system. 因此, the business of a
brain region needs to be situated in the context of multi-
region circuits: What does a brain region do “in combina-
tion” with other areas? 从某种意义上说, when one discusses
regions R1, ……, R4 as part of some function, the decision
to “not” discuss other areas is fairly arbitrary. We could
have discussed the roles of regions R5, R6, 等等.
One of the main reasons we don’t is due to the limitations
of the tools available to neuroscientists, which are ill-
suited to investigating large-scale, distributed systems
(although techniques are advancing fast). 因此, 我们
still do not know much about collective computations
involving larger numbers of gray matter components.
The word “entangled” conjures multiple interrelated
ideas but is not intended to suggest something like
threads that are mixed together but can be separated given
enough time. The meaning is closer to “integrated,” but
single words do not do justice to the general theme per-
meating the book—for example, cars are highly integrated
systems but are designed with parts with well-defined
功能. 反而, the sense of “entangled” is one in
which brain parts dynamically assemble into coalitions
that support complex cognitive–emotional behaviors,
University of Maryland
© 2022 麻省理工学院
coalitions composed of parts that jointly do their job.
因此, an entangled system is a deeply context-dependent
one in which the function of parts (such as a brain region,
or a population of cells within a region) must be under-
stood in terms of other parts: an interactionally complex
系统 (如下所述).
在这幅作品中, I summarize a few of the key themes of the
argument built in The Entangled Brain, including that
brain functions need to be understood as “emergent prop-
erties.” Of course, this is not a new idea. 然而, 神经的-
scientists still study and explain brain functions in a way
that does not heed this assertion. 更, new gen-
erations of students learn about the nervous system in a
piecemeal fashion as if processes were fairly localizable—
if not in areas, at least in relatively simple networks.
所以, revisiting these issues is valuable for students
of the brain at all levels of expertise. (N.B.: Citations in the
text that follows are only illustrative, and in no way seek to
be representative. It is my hope that specific work can be
given proper credit in the ensuing discussions initiated
by this piece.)
WHAT EMERGES?
The prevailing modus operandi of science can be summa-
rized as explaining phenomena by reducing them to an
interplay of elementary units that can be investigated inde-
pendently of one another (Von Bertalanffy, 1950). Such a
“reductionistic approach” reached its zenith, 也许,
认知神经科学杂志 35:3, PP. 349–360
https://doi.org/10.1162/jocn_a_01908
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
with the success of chemistry and particle physics in the
20th century. In the present century, its power is clearly
evidenced by dramatic progress in molecular biology
and genetics. At its root, this attitude to science “resolves
all natural phenomena into a play of elementary units,
the characteristics of which remain unaltered whether
they are investigated in isolation or in a complex” ( Von
Bertalanffy, 1950, p. 135).
当然, the reductionistic framework is not the only
game in town, as everyone knows that “the whole is
greater than the sum of its parts.” Scientists study objects
that have many components that interact in manifold
方法, so figuring out the parts is not enough—or so the
saying implies. 再次, in the words of one of early propo-
nents of complex systems, Von Bertalanffy, it is necessary
to investigate “not only parts but also relations of
organization resulting from a dynamic interaction” leading
to “the difference in behavior of parts in isolation and in
the whole organism” ( Von Bertalanffy, 1950, p. 135). 但
what does it mean to say that parts behave differently in
isolation relative to when they are part of a system?
Enter “emergence,” a term originally coined in the
1870s to describe instances in chemistry and physiology
where new and unpredictable properties appear that are
not clearly ascribable to the elements from which they
arise. 例如, when amino acids organize themselves
into a protein, the protein can carry out enzymatic func-
tions that the amino acids on their own cannot. 更多的
重要的是, they behave differently as part of the protein
than they would on their own. But it is actually more than
那. The dynamics of the system (the protein) closes off
some of the behaviors that would be open to the compo-
尼特 (amino acids) were they not captured by the overall
系统 ( Juarrero, 1999). Once folded up into a protein,
the amino acids find their activity regulated—one sense
in which they behave differently.
A possible definition of emergence is as follows: a novel,
collective property that is observed when multiple ele-
ments interact that is not readily reducible to the function
of the elements alone. Both scientifically and philosophi-
cally speaking, the friction caused by the idea of emer-
gence arises because it is actually unclear what precisely
emerges. 例如, what is it about amino acids as part
of proteins that differs from free floating ones? 这
question revolves around the exact status of emergent
特性. Philosophers refer to this question as the
“ontological” status of emergence, 那是, one concerning
the “proper existence” of the higher-level properties. 做
emergent properties point to the existence of new laws
that are not present at the lower level? Is something
fundamentally irreducible at stake? As the philosopher
Alicia Juarrero (1999) 说, it is particularly intriguing
when “systems exhibit organized and apparently novel
特性, seemingly emergent characteristics that
should be predictable in principle, but are not in fact”
(p. 6). These types of question remain by-and-large
unsolved and subject to vigorous intellectual battles (一个
excellent treatment is provided by Humphreys, 2016;
see also Juarrero, 1999).
幸运的是, we do not need to crack the problem and
can instead use “lower” and “higher” levels pragmatically
when they are epistemically useful—when the theoretical
stance advances knowledge. To provide an oversimplified
例子, we do not need to worry about the status be-
tween quarks and aerodynamics. Massive airplanes are of
course made of matter, which are agglomerations of
elementary particles such as quarks. But when engineers
design a new airplane, they consider the laws of aerody-
namics, the study of the motion of air, and particularly
the behavior of a solid object, such as an airplane wing,
in air—they need no training at all in particle physics!
所以, there is no need to agonize about the “true” relation-
ship between aerodynamics and particle physics (例如, 能
the former be reduced to the latter?). The practical thing
to do is simply to study the former.
One could object to this example because the inherent
levels of particle physics and aerodynamics are very far
已删除, one level too micro and the other too macro.
More interesting cases present themselves when the con-
stituent parts and the higher-level objects are closer to
彼此, 例如, the behavior of an individual
ant and the collective behavior of the ant colony, the flight
of a pelican and the V-shape pattern of the flock, or amino
acids and proteins. And of course, such is the case of the
脑.
THE BRAIN AS A COMPLEX SYSTEM?
Why does the complex systems perspective matter for
brain science? 实际上, many renowned neuroscientists
consider the issues above a distraction. 例如:
[A]lthough network properties of a system are a con-
venient explanation for complex responses, they tell
us little about how they actually work, and the concept
tends to stifle exploration for more parsimonious
explanations…[例如, 这] highly intercon-
nected nature of the central autonomic control system
has for many years served as an impediment to assign-
ing responsibility for specific autonomic patterns [到
particular groups of neurons]. (Saper, 2002, p. 460)
Under this view, treating the brain as a complex system
is not only a temporary distraction but also an actual
impediment to progress.
One scenario that justifies the quote’s stance is if we
consider the brain to be a “near-decomposable” system.
Herbert Simon (1962) proposed that scientists are fre-
quently interested in systems exhibiting “near-complete
decomposability” (see also Bechtel & 理查森, 2010),
where intrasystem interactions are much stronger than
extrasystem ones. Engineered systems work this way,
and much research in neuroscience—lesion work in neu-
ropsychology, systems neuroscience, fMRI research, 和
so on—proceeds from this vantage point. Such systems
350
认知神经科学杂志
体积 35, 数字 3
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
are “interactionally simple,” with parts interacting weakly
with anything considered beyond the system’s bound-
aries, under a given, “reasonable” decomposition. A trivial
example is a rock, where the atoms within the rock inter-
act strongly but weakly with atoms elsewhere. 相比之下,
a system is “interactionally complex” to the extent that its
元素 (under a certain decomposition) cross its
boundaries in important ways. 当然, systems exhibit
a spectrum of interactivity, from low to high (for in-depth
分析, see Wimsatt, 2007, Chap. 9).
Biologists of the brain, indeed academics across varied
学科, have been repeatedly turned off by some of
the putative mystical features of emergent properties.
With this in mind, it might be advantageous to switch
the language used to, hopefully, stimulate debate along
more productive directions centered around “system
interactivity.” The question of interest then shifts toward
dissecting the types of communication and interactivity
we find in the nervous system. Temporarily, 至少, 它
might be productive to have “emergence” and “complex
systems” recede into the background (hopefully in way
that does not amount to a pure semantic sleight of hand).
The question we face is thus the following: What kind
of interactional system is the brain?
PRINCIPLES OF BRAIN ORGANIZATION
To address this question, consider three principles of
brain organization: (1) massive combinatorial anatomical
连接性, (2) highly distributed functional coordina-
的, 和 (3) networks/circuits as functional units.
Massive Combinatorial Anatomical Connectivity
Anatomical pathways are dominated by short-distance
连接. 实际上, 70% of all the projections to a given
locus on the cortical sheet arise from within 1.5–2.5 mm
(马尔可夫等人。, 2011). Does this not dictate that processing
in the brain is local, or quasilocal?
Computational analyses of anatomical cortical pathways
gathered from a large number of studies inform this ques-
的. Studies initially suggested that the cortex operates
as a “small-world” (斯波恩斯 & Zwi, 2004), an organization
that supports enhanced signal propagation speed and syn-
chronizability between parts, among other properties
(巴拉巴斯 & 阿尔伯特, 1999; Watts & Strogatz, 1998). 在
small-world networks, though most of the connectivity is
当地的, a modest amount of long-range random connections
suffices to endow networks with these properties. Argu-
干练地, the most important insight of these analyses is not
that the brain really follows a small-world organization;
毕竟, biological systems would not be expected to
exhibit “random” nonlocal connections (that’s a mathe-
matical construct!). 反而, the key idea is that it’s possi-
ble for a system with mostly local physical connections,
but some mid- and long-range connections, to display
“unexpected” large-scale system properties.
的确, cortical organization is not small-world. 第一的,
nonlocal pathways are not random and instead target a
“core of regions.” Although different arrangements have
been proposed, they indicate that cortical signals flow
via a relatively small subset of richly interconnected and
integrated areas, at times called a “rich club.” For example,
Markov et al. (2013) identified a small subset of areas in
temporal cortex, parietal cortex, frontal cortex, and pFC
that are very highly connected structurally. It is thus likely
that brain communication relies heavily on signals being
communicated via a core (Figure 1A). 第二, and surpris-
英利, experiments indicate that cortical regions are con-
siderably more interconnected than previously believed.
Some estimates are that 60% of the possible connections
between pairs of areas are indeed observed in some corti-
cal patches (马尔可夫等人。, 2013)—clearly not a small-world
组织! The precise implications of these findings, 如果
confirmed, must also consider pathway strength (不仅
the existence vs. absence of a connection), which varies
over several orders of magnitude, because computational
work demonstrates that the type of network organization
(is it small-world?) strongly depends on the pattern of
pathway strengths (Gallos, Makse, & Sigman, 2012).
Cortical connectivity, although important, is only one
ingredient contributing to the anatomical organization
of the CNS. 实际上, the focus on cortical connections of
most of the computational work neglects major connec-
tional properties that shape the overall neuroarchitecture
(for a comprehensive treatment, see Nieuwenhuys,
Voogd, & van Huijzen, 2008). (1) The entire cortical sheet
projects to the striatum and loops back to the cortex
via the thalamus, forming the so-called BG–cortical loops
(Figure 1B). (2) The cortex and the thalamus are mas-
sively interconnected. Most of the thalamic volume is
involved in bidirectional circuits with the cortex via the
so-called higher-order regions (Sherman & Guillery,
2002). 例如, the pulvinar nucleus is bidirectionally
connected to the entire cortical sheet. (3) The hypothal-
amus is frequently viewed as a “descending” controller of
autonomic functions. 然而, the mammalian cerebral
数字 1. Combinatorial anatomical connectivity. (A) 计算型
analysis of cortical pathways suggests that a subgroup of regions works as
a “rich club” (orange circles in the middle): a set of highly interconnected
nodes that play a major role in determining the flow of signals across the
脑. (乙) The neuroarchitecture also includes multiple large-scale
connectional systems, such as via the BG, as illustrated here. 额外的
systems include those involving the thalamus, the hypothalamus, 这
BLA, and the cerebellum, 除其他外.
Pessoa
351
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
cortex and the hypothalamus share massive “bidirec-
tional” connections. 尤其, in rodents, 有
direct hypothalamic projections to all parts of the cortical
sheet (as well as multiple indirect connectivity systems
with cortex; Risold, 汤普森, & 斯旺森, 1997). In pri-
伙伴, the hypothalamus has widespread projections to
all sectors of the pFC, including lateral sectors. (4) 这
basolateral amygdala (BLA) is bidirectionally connected
with the entire cortical sheet; these connections are quite
substantial with parts of frontal and temporal cortex, 带领-
ing to the suggestion that this amygdala sector be called
the “frontotemporal amygdala” (斯旺森 & Petrovich,
1998). (5) The cerebellum not only receives inputs from
broad swaths of the cerebral cortex but also projects to
许多, if not all, of these areas. 尤其, a significant
portion of the output from the dentate nucleus of the cer-
ebellum projects to nonmotor areas, including regions of
pFC and posterior parietal cortex (Bostan, Dum, & 针织,
2013). Other major connectivity systems involve the
claustrum, the septum, and the brainstem (Nieuwenhuys
等人。, 2008).
清楚地, a more complete elucidation of the properties
of the connectional neuroarchitecture requires combining
both cortical and noncortical pathway systems. The over-
all picture is one of massive interconnectivity, leading
to “combinatorial” pathways between sectors. 其他
字, one can go from point A to point B in a multitude
of ways. We propose that, 合并的, connectivity systems
spanning the entire neuroaxis (cortical forebrain, subcor-
tical forebrain, midbrain, and hindbrain) provide the basis
for both broadcasting and integration of diverse signals
linked to the external and internal worlds. Such crisscross-
ing connectional systems support the interaction and
integration of signals that are typically associated with
standard mental domains, including emotion, motivation,
洞察力, 认识, and action (Pessoa, 2013) 但,
批判地, in a manner that does not abide by putative
boundaries between these categories (见下文). I pro-
pose that this general architecture supports a degree of
computational flexibility that enables animals to cope suc-
cessfully with complex and ever-changing environments.
The overall architecture may produce circuits with local
specificity while attaining large-scale sensitivity, a type of
“global-within-local design,” which likely contributes to
more sophisticated, plastic, and context-sensitive
behaviors (Pessoa, Medina, & Desfilis, 2022).
Highly Distributed Functional Coordination
The complexity of anatomical pathways allows signals to
flow across the brain in a staggeringly large set of ways.
Anatomy provides a backbone that constrains function,
but the structure–function relationship is anything but
simple when one considers the abundance of bidirectional
connections and loop-like organization (as in the BG),
combined with excitation, inhibition, and nonlinearities.
In this manner, the anatomy supports a large range of
“functional interactions,” namely, particular relationships
between signals in disparate parts of the brain (例如, 他们
might fire coherently). 对于一个, the anatomy will support
the efficient communication of signals, even when strong
direct pathways are absent, such as the functional coordi-
nation between signals in the amygdala and lateral pFC,
although the two are not strongly connected physically.
These ideas, 当然, are related to the notion of func-
tional connectivity, which in its most basic form can be
indexed via the correlation coefficient between two time
系列 (例如, of the amygdala and the lateral pFC).
As an illustration of functional interactions, consider an
experiment that acquired fMRI in monkeys when they
were not performing an explicit task (Grayson et al.,
2016). The study observed robust signal correlation
between the amygdala and several regions that are not
connected to it (as far as it is known). They asked, 也,
whether functional connectivity was more related to direct
(monosynaptic) pathways or multipath (polysynaptic)
connectivity by undertaking graph analysis. Are there effi-
cient routes of travel between regions even when they are
not directly connected? To address this question formally,
they estimated a graph measure called “communicability”
(related to the concept of “efficiency”) and found that
amygdala functional connectivity was more closely related
to communicability than would be expected by consider-
ing only monosynaptic pathways. Their finding illustrates
that the relationship between signals in disparate parts of
the brain is not determined by structural pathways in a
straightforward manner.
Networks/Circuits as Functional Units
The combination of the prior two principles leads to the
present one. In a highly interconnected system, to under-
stand function, we need to shift away from thinking in
terms of individual brain regions: The network itself is
the functional unit, not the brain area (图2A). Pro-
cesses that support behavior are not implemented by an
individual area but depend on the interaction of multiple
地区, which are dynamically recruited into multiregion
assemblies. Such functional networks are based on the
relationships between signals across disparate parts of
大脑.
But how are networks/circuits defined? Let us consider
here large-scale networks, such as those studied with
fMRI in humans and rodents (Grandjean et al., 2020; 杨
等人。, 2011). (Other examples of networks/circuits will
be discussed in the context of extinction learning below.)
The most popular partitioning schemes parse individual
元素 ( brain regions or parcels) into unique
groupings—a node belongs to one and exactly one com-
社区. (A community refers to a subdivision of a larger
网络, 即, a subnetwork. At times we will refer to
subnetworks as “networks,” as in “default network,” given
common usage in the literature.) Based on fMRI data in the
absence of a task, Yeo et al. (2011) described a seven-
352
认知神经科学杂志
体积 35, 数字 3
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
highly connected structurally. Regions that work as
“connector hubs” (Guimera & Nunes Amaral, 2005) 是
distinctly interesting because they have the potential to
integrate diverse types of signals (if they receive inputs
from disparate sources) and/or to distribute signals widely
(if they project to disparate targets). They are a good
reminder that communities are not islands; regions within
A (disjoint) community have connections both within and
outside the community.
Although there are many ways to operationalize over-
lapping networks, a simple way is to allow each brain
region to participate in all communities simultaneously
but in a graded fashion. 因此, if region A does not partic-
ipate in community C1, its membership value is 0; 骗局-
versely, a membership value of 1 indicates that it belongs
maximally to C1. It is also useful to conceive of member-
ship as a finite resource, such that it sums to 1. Applying
these notions formally, we found that functional brain net-
works based on fMRI data both when tasks are not
required and during task conditions are highly overlap-
平 (Najafi, McMenamin, 西蒙, & Pessoa, 2016). 在
也就是说, a considerable fraction of regions shared
their memberships across multiple communities. 的确,
overlapping organization has been detected via multiple
networks analysis techniques (Faskowitz, Esfahlani, Jo,
斯波恩斯, & 贝策尔, 2020; 杨, 克里宁, Chee, & 巴克纳,
2014).
The brain is a dynamic, constantly moving object, 所以
are its networks. Functional relationships between groups
of regions are constantly fluctuating based on cognitive,
emotional, and motivational demands. Paying attention
to a stimulus that is emotionally significant (说, paired
with mild shock in the past), increases functionally con-
nectivity between the visual cortex and the amygdala
(Lojowska, Ling, Roelofs, & Hermans, 2018). Performing
a challenging task in which an advance cue stimulus indi-
cates that participants may earn extra cash for performing
it correctly increases functional connectivity between the
parietal/frontal cortex (important for performing the task)
and the ventral striatum (important for reward-related
流程; Padmala & Pessoa, 2011).
A vast literature has documented such changes in
functional connectivity between pairs of regions, 但不-
tributed, large-scale changes have been observed, 也
(Cole et al., 2013). 例如, in the reward study just
描述的, the nucleus accumbens and the caudate each
increased their functional connectivity with nearly all cor-
tical regions engaged by the cue. Network analysis identi-
fied two communities, one cortical and another composed
mostly of subcortical regions (including the nucleus
accumbens and caudate). It also revealed a decrease of
modularity when potential reward cues were encoun-
tered, consistent with the notion that particular condi-
系统蒸发散 (在这种情况下, the possibility of reward) reorganize
large-scale functional organization (Kinnison, Padmala,
Choi, & Pessoa, 2012). In another study, we observed a
progression of network-level changes when participants
Pessoa
353
数字 2. From regions to networks. (A) The meaningful functional
unit is not the brain region (左边) but networks of brain regions that
aggregate and disassemble as a function of time (middle and right). (乙)
Illustration of network properties (功能) in a scenario in which
regions carry out well-defined “primitives” (B1) and when they do not
(B2); in the latter, the function in question needs to be understood in terms
a set of regions (blue contours). Note that in both cases, understanding the
circuit behavior (function F ) is not well approximated by considering the
individual functions, F. 反而, it is necessary to consider the coordinated
(emergent) circuit function. In B1, individual functions can be specified
based on single regions (例如, F(R1)), but in B2 depend on more than one
region in some cases (例如, F(R1, R2)). Boxes at the bottom indicate criteria
to determine network-level properties, as well as Type II networks (in B2).
community division of the entire cortex, where each local
patch of tissue was assigned to a single community. 在
也就是说, the overall space was broken into disjoint
社区. Their elegant work has been very influen-
提尔, and their seven-network partition has been adopted
as a sort of “canonical” division of the cortex. Whereas
discrete clusters simplify the description of a system, 做
they capture the underlying organization?
Multiple types of networks (社会的, 生物) exhibit
nontrivial “overlapping organization” (Palla, Derenyi,
Farkas, & 笑话, 2005). 例如, the study of chemi-
cal interactions reveals that a substantial fraction of pro-
teins interacts with several protein groups, indicating that
actual networks are made of interwoven sets of overlap-
ping communities. Another way to motivate overlapping
organization in networks is by considering “hub regions.”
Both structural and functional analyses of brain data have
revealed the existence of particularly well-connected
地区, called hubs. 例如, 如上所述,
Markov et al. (2013) described a set of areas in temporal
cortex, parietal cortex, frontal cortex, and pFC that are very
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
experienced threat, uncovering how network organiza-
tion unfolds across time during anxious apprehension
(McMenamin, Langeslag, Sirbu, Padmala, & Pessoa,
2014), a reminder that network functional organization
must be understood dynamically, as further illustrated
by studies of time-varying functional connectivity (Lurie
等人。, 2020).
The ideas of network overlap and dynamic organization
are related. If brain areas can belong to multiple networks,
what determines the strength of a region’s affiliation to a
specific network? 这里, context plays a pivotal role:
region A will participate strongly in network N1 during a
certain context C1 but will be more strongly linked with
network N2 during context C2. These ideas resonate with
the “flexible hub theory” (Cole et al., 2013), where some
regions are suggested to adjust their functional connectiv-
ity patterns as a function of task demands. This conceptu-
alization brings us back to the functions of brain regions:
The processes carried out by an area will depend on its
network affiliations (IE。, the regions it clusters with) 在一个
given time.
Overlap and dynamics promote a view in which net-
works do not consist of fixed collections of regions but
instead are made of coalitions that form and dissolve to
meet computational needs. 相比之下, in the literature,
networks frequently are described in terms of fixed sets
of nodes; 例如, the “salience network” might refer
to nine bilateral regions plus the dorsal ACC (Hermans
等人。, 2011). But conceptualizing networks in more
dynamic fashion is fruitful. 例如, at time t1, 地区
R1, R2, R7, and R9 might form a natural cluster; at a later
time t2, regions R2, R7, and R17 might coalesce. This shift
in perspective challenges the notion of a network as a sta-
ble unit, at least for longer periods of time, and raises new
问题. At what point does a coalition of regions
become something other than network N? 反过来,
can we think of the “salience network,“ 例如, 作为一个
set of regions that varies temporally (see Figure 2A).
THE INTERACTIONAL COMPLEXITY OF
FEAR EXTINCTION
We now discuss the neural circuits of fear extinction as
an example of the types of network studied by systems
neuroscientists, which also illustrates the challenges of
trying to unravel the mapping between structure and
function. When a conditioned stimulus no longer predicts
the unconditioned stimulus to which it was paired in the
过去的 (说, a sound no longer is followed by a shock),
the conditioned stimulus gradually stops eliciting the
conditioned response. This process is called “fear extinc-
tion.” Understanding it is of potentially enormous conse-
quence given the prevalence of anxiety and other related
disorders.
The medial pFC plays an important role in regulating the
amygdala during fear extinction (摩根, Romanski, &
LeDoux, 1993). Extinction critically depends on context,
也. 例如, a sound may no longer signal an aversive
事件, but not necessarily in a completely novel environ-
蒙特, and the hippocampus is thought to provide such
critical contextual information to the amygdala. 其他
region influencing extinction is the nucleus reuniens of
the thalamus, which allows discrimination of dangerous
from safe contexts (Ramanathan, Jin, Giustino, 佩恩, &
Maren, 2018). At first glance, fear extinction appears to
fit the scheme of separate contributions interacting to gen-
erate a new behavior: 认识 (tied to the medial pFC)
controlling emotion (tied to the amygdala) in a top–down
fashion, with additional contributions related to the con-
text of extinction and other factors (Figure 3A). 然而,
such characterization does not do justice to the behavioral
and neural richness of the phenomenon.
Let’s consider a few additional findings about fear
extinction (Figure 3B). Multiple cell groups in the BLA
actually project to the medial pFC whose outputs in turn
influence amygdala signals. Some studies even have sug-
gested that the BLA is upstream of the medial pFC,
because a population of extinction neurons in the BLA
(which project to the medial pFC) increase their activity
during extinction learning (Herry et al., 2008). The medial
pFC is also the target of the hippocampus, and this input
potentiates medial pFC signals during extinction. 更远-
更多的, the medial pFC receives substantial inputs from the
thalamus, itself a major subcortical–cortical connectivity
hub. Additional contributors to this circuit include the
ventral tegmental area, where dopamine neurons are
activated by the omission of the aversive unconditioned
stimulus during extinction (Salinas-Hernández et al.,
2018) and are suggested to influence the BLA (possibly
via indirect projections). Although the role of the medial
pFC is well established in extinction, it is likely that both
the OFC and the ventrolateral pFC are important for
behavioral regulation in the presence of aversive stimuli,
数字 3. Fear extinction circuits. (A) Basic extinction circuit centered on
the BLA. The contributions of a few key regions are labeled with their
putative functional contributions. (乙) Extended circuit, with a larger set of
brain regions believed to be involved (not intended to be comprehensive).
The blue arrows indicate indirect anatomical connectivity. The red arrows
indicate the extensive norepinephric projections of the locus coeruleus.
HIPP, hippocampus; LC, locus coeruleus; MPFC, medial pFC; THAL,
thalamus; VTA, ventral tegmental area.
354
认知神经科学杂志
体积 35, 数字 3
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
也 (Shiba, Santangelo, & 罗伯茨, 2016). 最后, the locus
coeruleus in the brainstem, which is a primary source of
forebrain norepinephrine, has important neuromodula-
tory effects on extinction. 实际上, stress-related engage-
ment of the locus coeruleus opposes extinction (Maren,
2022). (To simplify the discussion, we described the
medial pFC as a unit, but in rodents, the main contribu-
tions during extinction involve the ventral/infralimbic
成分, which probably corresponds to the ventro-
medial pFC in humans. 此外, complex microcircuits
exist within critical nodes of the circuit, such as the amyg-
dala [Whittle et al., 2021], but are not discussed here.)
现在, let us return to the initial scheme of Figure 3A.
This description of extinction instantiates a boxes-and-
arrows arrangement that is a mainstay of psychology and
神经科学, where semantic labels can be added to
some of the interactions when their interpretation can
be distilled to a convenient concept. 然而, the depic-
tion is fundamentally wanting not only because it lacks
some regions and connections but also because of the
implicit assumption that well-defined functions are imple-
mented by individual regions, with their outputs being
read by downstream regions. 例如, the hippocam-
pus determines context, and the medial pFC some kind of
appraisal that determines when the amygdala should be
downregulated; 因此, one can place these functions at
the regions. Their outputs are then read by the amygdala
to determine what to do given the inputs.
相比之下, an alternative mode of thinking considers
how multiple regions jointly and dynamically implement
key processes (Figure 3A). 例如, 正如所讨论的
多于, extinction neurons in the BLA are reciprocally
connected to the medial pFC (Herry et al., 2008). 因此,
whereas they actually could be thought to be “upstream”
of the medial pFC (thus flipping the typical way of thinking
about the two regions, as indicated previously), 这是
important to evaluate the possibility that the two regions
work in a coordinated fashion during extinction learning.
Another reason fear extinction should be considered a
circuit/network property is that extinction has to do with
processes that convert a fear-inducing stimulus back to a
status of neutrality—an “off” switch, if you will. 然而,
as animals navigate their environment, 有许多
stimuli that do exactly the opposite, and they can be
considered “on” switches. 因此, 去理解
complex behaviors, one needs to consider how defensive
behaviors are dynamically engaged and disengaged.
Even more broadly, the defensive circuits involved
intersect and interact heavily with those that promote
exploratory and appetitive behaviors. Should the animal
withdraw, stay, or approach?
任何状况之下, even under a more constrained conceptual-
化, the preceding discussion should help highlight
the interactional complexity of systems-level processes
neuroscientists often focus on. 对于一个, the circuit con-
tains both unidirectional and bidirectional connections,
as well as both excitatory and inhibitory components.
The case advanced here is that the field should in fact
embrace this level of interactivity, not attempt to side step
它. 否则, it will continue to be no surprise how little
progress has been made in ameliorating the debilitating
impacts of fear- and anxiety-related mental health condi-
tions in the lives of so many people. Broadly speaking,
extinction can be studied in the context of Pavlovian or
instrumental learning (an example of the latter is avoid-
ance learning where the animal makes a response to avoid
foot shock). In both laboratory animals and humans, 亲-
cedures to extinguish behavior (“instrumental extinction”)
elicit well-documented “side effects” (for a discussion, 看
Bouton, Maren, & McNally, 2021). Examples include tem-
porary increase of the very behavior being extinguished, A
return of other behaviors previously extinguished, 还有
as increased frequency of undesirable behaviors such as
aggression. The claim made here is that these examples
appear to be secondary consequences when regions are
considered as well-defined investigative units with puta-
tively specialized functions. 反而, they are exactly the
types of effects routinely found in nondecomposable,
interactionally complex systems, where cascades of inter-
actions generate “side effects.”
WHAT KIND OF NETWORK?
Let us consider two scenarios to further clarify what is
meant by “network properties.” In a Type I network, 脑
regions carry out (compute) fairly specific functions. 为了
例子, in the context of extinction, the hippocampus
determines contextual information, and the ventral teg-
mental area computes omission prediction errors. 在这个
scenario, a process of interest (说, fear extinction) is still
viewed as a network property that depends on the interac-
tions of the brain regions involved. That is to say, it is nec-
essary to investigate the orchestration of multiple regions
to understand how the regions, 集体地, carry out the
processes of interest. 重要的, 然而, the collective
properties of the system are not accessible, or predictable,
from the behavior of the individual regions alone: 这
multiregion function, F(R1, R2, ……, Rn), is poorly character-
ized from considering f(R1), F(R2), 等等.
Poorly characterized in what sense? In a near-
decomposable system, lesion of R1, 例如, will cause
a deficit to the network that is directly related to the puta-
tive function of R1. 然而, this is not the outcome in an
interactionally complex system. Consider multispecies
ecological systems in which the introduction of a new spe-
cies or the removal of an existing one causes completely
unexpected knock-on effects (莱文, Bascompte, 阿德勒,
& Allesina, 2017). The claim being made here is that, 在
many cases, we need to consider brain networks in much
the same way: A complex system that is not well approx-
imated by simple decompositions; F(R1, R2, ……, Rn) 将要
not be well approximated by considering f(R1), F(R2),
……, F(Rn) (Figure 2B1).
Pessoa
355
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
Now let us turn to Type II networks, where areas do not
instantiate specific functions. 反而, two or more regions
working together instantiate the basic function of interest,
such that its implementation is distributed across regions.
It is easy to provide an example of Type II networks if we
consider computational models where undifferentiated
units are trained together to perform a function of interest.
但, are there examples of this type of situation in the
脑? Multiarea functions are exemplified by reciprocal
dynamics between the FEFs and the lateral intraparietal
area in macaques supporting persistent activity during a
delayed oculomotor task (哈特 & Huk, 2020). Based,
除其他外, on the tight link between these areas at
the trial level, the authors suggested that the two areas be
viewed as a single functional unit (see Murray, Jaramillo, &
王, 2017, and Kang & Drukmann, 2020, for a computa-
tional model; see also Mejías & 王, 2022; Figure 2B2).
In rodents, motor preparation requires reciprocal
excitation across multiple brain areas (Guo et al., 2017).
Persistent preparatory activity cannot be sustained within
cortical circuits alone but in addition requires recurrent
excitation through a thalamocortical loop. Inactivation of
the parts of the thalamus reciprocally connected to the
frontal cortex results in strong inhibition of frontal cortex
神经元. 反过来, the frontal cortex contributes major
driving excitation to the higher-order thalamus in ques-
的. 更, persistent activity in frontal cortex also
requires activity in the cerebellum and vice versa (高
等人。, 2018), revealing that persistent activity during motor
planning is maintained by circuits that span multiple
地区. 索赔要求, 因此, is that persistent motor activity
is a circuit property that requires multiple brain regions.
In such case, one cannot point to a brain region (甚至
a sector) and label “working memory” as residing there.
It could be argued that, in the brain, the two types of
networks discussed here—with and without well-defined
node functions—are not really distinct and that what dif-
fers is the granularity of the function. 毕竟, if above one
could decompose the function “persistent motor activity”
into basic primitives, it is conceivable that they could be
carried out in separate regions. In such case, we would
revert back to the situation of networks with nodes that
compute well-defined functions. Put another way, a skep-
tic could quibble that, in the brain, a putative Type II net-
work is a reflection of our temporary state of ignorance.
The conjecture advanced here is that, in the brain, 这样的
reductive reasoning will fare poorly in the long run: 这是
not the case that one can develop a system of primitive
properties that, 一起, span the functions/processes
of interest. 在很多情况下, network properties are not
reducible to component interactions of well-defined
subfunctions—they are inexorably distributed.
SOME IMPLICATIONS FOR BRAIN SCIENCE
In the preceding sections, we discussed one of the major
themes of The Entangled Brain: Neural processes that
accompany behavior are profitably viewed through the
lens of complex, networked systems. 这里, we summarize
some of the implications of the framework to the general
goal of elucidating brain functions and how they relate to
brain parts.
Interactional Complexity
The brain is a system of interacting parts. At a local level,
say within a specific Brodmann’s region or subcortical
区域, populations of neurons interact. But interactions
are not only local. Massive anatomical connectivity pro-
vides the substrate for communication crisscrossing the
entire span of the neuroaxis (hindbrain, midbrain, 和
forebrain). This structural interactional complexity has
important implications for brain function: Simpler decom-
positions that insulate brain regions from one another will
capture only a slice of the contributions of the parts in
问题.
Anatomical interactional complexity implies that net-
work or circuits are the functional unit of interest. 这
结论, when considered in a broad sense, may seem
incontrovertible to many (最多?) neuroscientists. 这
question then is what kind of network/circuit are we study-
英? I contend that simply enlarging the functional unit
from an area to a standard, fixed network is only a modest
step. Networks should be considered inherently overlap-
ping and dynamic. Parts of the brain (说, populations of
neurons within areas) affiliate dynamically with other ele-
ments in a highly context dependent manner driven by the
current endogenous and exogenous demands and oppor-
tunities present to the animal. Critically, network proper-
ties are novel (with respect to that of individual regions),
and key functions are distributed across regions or neuro-
nal populations.
从这个角度来看, it is no surprise that neuroscien-
tists are constantly discovering that brain regions partici-
pate in novel and unexpected ways in previously studied
circuits and/or processes. Examples abound, but in the
context of extinction learning, 例如, a growing
number of critical contributions of the thalamus are being
discovered (Silva et al., 2021; Ramanathan et al., 2018).
最后, the inflexible nature of laboratory testing plays
no small role in the apparent low interactional complexity
of functional brain circuits (Paré & Quirk, 2017). 经过
restricting the conditions under which circuits are inter-
rogated, it appears that neuronal populations, 地区, 和
circuits are considerably more selective for the properties
and functions investigated.
Decentralization, Heterarchy, and Causation
In many systems, and the brain is no exception, it is instinc-
tive to think that many of its important functions depend
on centralized processes; 例如, the pFC may be
viewed as a convergence sector for multiple types of infor-
运动, allowing it to control behavior. The view advanced
356
认知神经科学杂志
体积 35, 数字 3
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
here favors PDP (Goldman-Rakic, 1988). Instead of infor-
mation flowing hierarchically to an “apex region” where
signals are integrated, information travels in multiple
directions without a strict hierarchy. An organization of
this sort is termed a “heterarchy” to emphasize the multi-
directional flow of information. As discussed previously,
this does not imply an absence of organization. The ana-
tomical backbone itself is highly structured, and several
cortical regions are important anatomical/functional core
地区 (马尔可夫等人。, 2013). Other noncortical areas
are also important hubs, including the thalamus, hypothal-
amus, BLA, and parts of the midbrain, including the supe-
rior colliculus.
The decentralized nature of processing should be
understood temporally, 也. When a novel stimulus
and/or context is encountered by an animal, signals might
flow first along the most direct and potent routes. 如何-
曾经, behavior evolves temporally, and signal flow will
progress in complex, decentralized ways. 实际上, the spiral-
ing pathways of the neuroarchitecture support communi-
cation and integration of signals across different spatial
extents. The processing of new stimuli always take place
against ongoing activity reflecting the immediate, 最近的
过去的 (and of course the more remote past), further decou-
pling functional states from what would be anticipated by
considering the most immediate anatomical pathways.
Investigating systems that are conceptualized as rela-
tively decentralized instigates different classes of research
questions about brain and behavior. If it is the coordina-
tion between the multiple parts that leads to the proper-
ties of interest, the object of scientific studies shifts to
unraveling how such interactions work. 例如,
instead of investigating how property P is encoded in area
A, the question becomes one of elucidating how the
property arises from decentralized coordination. 这
coordination framework also moves the goalpost away
from deciphering what information is passed from region
to region—or relatedly how a region decodes the signals
of other regions—to how the coordinated activity of
multiregion assemblies generates signals with specific
特性.
The view has important implications for causation, 也,
as the concept needs to be substantially reformulated.
Neuroscientists often operate with an implicit billiard ball
model of causation, a Newtonian scheme in which signals
in one region affect the response in another, 很像
billiard balls affect each other. 然而, Newtonian cau-
sality provides an extremely poor candidate for explana-
tion in interactionally complex systems like the brain
because of the distributed, 相互的, and reciprocal nature
of the causal contributions. This is not to promote a Lash-
leyan view of causal equipotentiality; the brain is clearly
very highly structured. 所以, how should one proceed?
A possible strategy to advance the understanding of
causation is to investigate circuit controllability (唐 &
Bassett, 2018; 刘 & 巴拉巴斯, 2016). By using tools from
network science and mathematical control theory, 一
数字 4. Network controllability. (A) Multiperturbation methods can
be used to activate and/or silence multiple brain regions simultaneously
(here R1 and R3). (乙) Such perturbations can be used to attempt to
steer the trajectory of the system from some initial state, 我, toward a
final target state. 这里, the state of the system can be represented as a
point in four dimensions, each corresponding to the activity level at an
individual region. The temporal evolution of the system along the state
space corresponds to a trajectory.
seeks to determine the extent to which certain network
nodes can steer the system into different states. 因此, 骗局-
trollability of future system states provides a promising
tool to understand the emergence of multipart properties.
An interesting concept from this field is the notion of
“pinning control,” where multiple inputs are applied
(“pinned”) and propagate through the system, 与
goal of defining the (未来) trajectory of the system
( 王 & 陈, 2002; 数字 4). Transplanting such rea-
soning to neuroscience, one could see how it could inform
perturbation experiments. The ability to determine
specific future states will depend on simultaneously
stimulating and/or silencing sets of regions, not a single
一, in particular ways.
This perspective offers avenues for testing network-
level properties, 也. To observe a certain function, F,
instantiated by a certain future circuit state requires
pinning multiple regions (or neuronal subpopulations);
pinning a single one is insufficient to attain the collective
state in question (see also Fakhar & 希尔格塔格, 2021, 为了
arguments that multiregion lesion experiments are neces-
sary). This type of approach also helps evaluate Type II
网络, where function is not well defined at the area
等级. 在这种情况下, manipulating two or more regions is
necessary for the function in question to be instantiated
(such as R1 and R3 in Figure 2B2). 更普遍, 神经的-
science will benefit from the development of mathematical
techniques to investigate causation in complex systems, 作为
in other areas such as weather prediction (Runge et al.,
2019) 和生态学 (Sugihara et al., 2012).
Mental Categories and the Entangled Brain
Categories such as perception, 认识, 行动, 情感,
and motivation organize how we understand and study
brain function. But are such mental domains consistent
with the framework described here? In a nutshell, 不.
The standard decomposition adopted by neuroscientists
Pessoa
357
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
requires an organization that is fairly modular, 这是
inconsistent with the principles of the anatomical and
functional neuroarchitecture discussed. 一般来说, 精神的
processes of interest cut across domains and do not
respect putative boundaries between traditional systems
(例如, 情感, 认识). 实际上, crisscrossing anatomical/
functional connectional systems dissolve potential lines of
划界.
More broadly, brains have evolved to provide adaptive
responses to problems faced by living beings, promoting
survival and reproduction. 在此背景下, even the mental
vocabulary of neuroscience (注意力, cognitive control,
ETC。), with origins disconnected from the study of animal
行为, provides problematic theoretical pillars. 反而,
approaches inspired by evolutionary considerations pro-
vide potentially better scaffolds to sort out the relation-
ships between brain structure and function (Cisek, 2022;
Pessoa et al., 2022).
FINAL THOUGHTS
Neuroscience strives to elucidate the neural underpin-
nings of behaviors and has done so in a preponderantly
reductionistic fashion for over a century and a half. 这
time is ripe for transitioning into a period when a truly
dynamic and networked view of the brain takes hold.
Future research will need to strive to make progress along
several fronts: dynamics, decentralized computation, 和,
是的, emergence.
Why do we need the perspective advocated here? 这
claim goes back to the interactional complexity of the
脑. If some of the ideas above are correct, 神经科学
needs to stop treating the brain as a near-decomposable
系统. Doing so distorts our view of the very system we’re
attempting to decipher. 例如, we will think we can
make progress in understanding fear and anxiety by focus-
ing on a few regions at a time, or even isolated circuits. 这
contention made here is that this strategy is deficient (看
Pessoa, 2022A).
How can the shift advocated be implemented? At least
部分地, current limitations stem from the neurotechniques
可用的. Novel neurotechniques will play a major role. 在
特别的, developments that allow recording over a larger
number of regions simultaneously, as well as accomplish-
ing multiregion perturbations. 同时, a science
of the mind–brain must stand on a solid foundation of
understanding behavior (Krakauer, Ghazanfar, Gomez-
Marin, MacIver, & Poeppel, 2017), while employing com-
putational and mathematical tools in an integral manner.
The field needs to take stock and invest on the develop-
ment of conceptual and theoretical pillars. Bigger and
shinier tools and techniques alone will not yield the neces-
sary progress; we run the risk of being able to measure
every cell (or subcellular component even) in the brain
in a theoretical vacuum. The current obsession in the field
with causation is equally problematic. Without conceptual
clarity—how should we even think of causation in highly
entangled systems?—causal explanations in fact might
miss the point.
最终, to explain the cognitive–emotional brain,
we need to dissolve boundaries within the brain—
洞察力, 认识, 行动, 情感, and motivation
—as well as outside the brain, as we bring down the walls
between biology, 心理学, ecology, mathematics, com-
puter science, 哲学, 等等.
致谢
The author is grateful for support from the National Institute of
Mental Health (MH071589 and MH112517) and Brad Postle for
constructive feedback on earlier versions of the article.
Reprint requests should be sent to Luiz Pessoa, 大学
Maryland, 学院公园, 医学博士 20742, or via e-mail: pessoa@umd.edu.
Funding Information
Luiz Pessoa, National Institute of Mental Health (https://
dx.doi.org/10.13039/100000025), grant numbers:
MH071589, MH112517.
Diversity in Citation Practices
Retrospective analysis of the citations in every article pub-
lished in this journal from 2010 到 2021 reveals a persistent
pattern of gender imbalance: Although the proportions of
authorship teams (categorized by estimated gender iden-
tification of first author/last author) publishing in the Jour-
nal of Cognitive Neuroscience ( JoCN) 在这段时期
were M(一个)/米= .407, 瓦(oman)/米= .32, M/ W = .115,
and W/ W = .159, the comparable proportions for the arti-
cles that these authorship teams cited were M/M = .549,
W/M = .257, M/ W = .109, and W/ W = .085 (Postle and
Fulvio, JoCN, 34:1, PP. 1–3). 最后, JoCN encour-
ages all authors to consider gender balance explicitly when
selecting which articles to cite and gives them the oppor-
tunity to report their article’s gender citation balance.
参考
巴拉巴斯, A. L。, & 阿尔伯特, 右. (1999). Emergence of scaling in
random networks. 科学, 286, 509–512. https://doi.org/10
.1126/science.286.5439.509, 考研: 10521342
Bechtel, W., & 理查森, 右. C. (2010). Discovering
复杂: Decomposition and localization as strategies in
scientific research. 剑桥, 嘛: 与新闻界. https://土井
.org/10.7551/mitpress/8328.001.0001
Bostan, A. C。, Dum, 右. P。, & 针织, 磷. L. (2013). Cerebellar
networks with the cerebral cortex and basal ganglia. Trends
in Cognitive Sciences, 17, 241–254. https://doi.org/10.1016/j
.tics.2013.03.003, 考研: 23579055
Bouton, 中号. E., Maren, S。, & McNally, G. 磷. (2021). Behavioral and
neurobiological mechanisms of Pavlovian and instrumental
extinction learning. Physiological Reviews, 101, 611–681.
https://doi.org/10.1152/physrev.00016.2020, 考研: 32970967
Cisek, 磷. (2022). Evolution of behavioural control from chordates
to primates. Philosophical Transactions of the Royal Society
of London, Series B: Biological Sciences, 377, 20200522.
https://doi.org/10.1098/rstb.2020.0522, 考研: 34957850
358
认知神经科学杂志
体积 35, 数字 3
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
Cole, 中号. W., Reynolds, J. R。, 力量, J. D ., Repovs, G。, Anticevic,
A。, & Braver, 时间. S. (2013). Multi-task connectivity reveals
flexible hubs for adaptive task control. 自然神经科学,
16, 1348–1355. https://doi.org/10.1038/nn.3470, 考研:
23892552
Fakhar, K., & 希尔格塔格, C. C. (2021). Systematic perturbation of
an artificial neural network: A step towards quantifying causal
contributions in the brain. bioRxiv. https://doi.org/10.1101
/2021.11.04.467251
Faskowitz, J。, Esfahlani, F. Z。, Jo, Y。, 斯波恩斯, 奥。, & 贝策尔, 右. F.
(2020). Edge-centric functional network representations of
human cerebral cortex reveal overlapping system-level
建筑学. 自然神经科学, 23, 1644–1654. https://
doi.org/10.1038/s41593-020-00719-y, 考研: 33077948
Gallos, L. K., Makse, H. A。, & Sigman, 中号. (2012). A small world
of weak ties provides optimal global integration of self-similar
modules in functional brain networks. 诉讼程序
美国国家科学院, 美国。, 109, 2825–2830.
https://doi.org/10.1073/pnas.1106612109, 考研: 22308319
高, Z。, 戴维斯, C。, 托马斯, A. M。, Economo, 中号. N。, Abrego,
A. M。, Svoboda, K., 等人. (2018). A cortico-cerebellar loop for
motor planning. 自然, 563, 113–116. https://doi.org/10
.1038/s41586-018-0633-x, 考研: 30333626
Goldman-Rakic, 磷. S. (1988). Topography of cognition: Parallel
distributed networks in primate association cortex. Annual
Review of Neuroscience, 11, 137–156. https://doi.org/10.1146
/annurev.ne.11.030188.001033, 考研: 3284439
Grandjean, J。, Canella, C。, Anckaerts, C。, Ayrancı, G。, Bougacha,
S。, Bienert, T。, 等人. (2020). Common functional networks in
the mouse brain revealed by multi-Centre resting-state fMRI
分析. Neuroimage, 205, 116278. https://doi.org/10.1016/j
.neuroimage.2019.116278, 考研: 31614221
Grayson, D. S。, Bliss-Moreau, E., Machado, C. J。, Bennett, J。,
沉, K., 授予, K. A。, 等人. (2016). The rhesus monkey
connectome predicts disrupted functional networks resulting
from Pharmacogenetic inactivation of the amygdala. 神经元,
91, 453–466. https://doi.org/10.1016/j.neuron.2016.06.005,
考研: 27477019
Guimera, R。, & Nunes Amaral, L. A. (2005). Functional cartography
of complex metabolic networks. 自然, 433, 895–900. https://
doi.org/10.1038/nature03288, 考研: 15729348
Guo, Z. 五、, Inagaki, H. K., Daie, K., Druckmann, S。, Gerfen, C. R。,
& Svoboda, K. (2017). Maintenance of persistent activity in a
frontal thalamocortical loop. 自然, 545, 181–186. https://土井
.org/10.1038/nature22324, 考研: 28467817
哈特, E., & Huk, A. C. (2020). Recurrent circuit dynamics
underlie persistent activity in the macaque frontoparietal
网络. 电子生活, 9, e52460. https://doi.org/10.7554/eLife.52460,
考研: 32379044
Hermans, 乙. J。, Van Marle, H. J。, Ossewaarde, L。, Henckens, 中号. J。,
Qin, S。, Van Kesteren, 中号. T。, 等人. (2011). Stress-related
noradrenergic activity prompts large-scale neural network
reconfiguration. 科学, 334, 1151–1153. https://doi.org/10
.1126/science.1209603, 考研: 22116887
Herry, C。, Ciocchi, S。, Senn, 五、, Demmou, L。, 穆勒, C。, & Lüthi,
A. (2008). Switching on and off fear by distinct neuronal
circuits. 自然, 454, 600–606. https://doi.org/10.1038
/nature07166, 考研: 18615015
Humphreys, 磷. (2016). Emergence: A philosophical account.
牛津大学出版社. https://doi.org/10.1093/acprof:oso
/9780190620325.001.0001
Juarrero, A. (1999). Dynamics in action: Intentional behavior
as a complex system. 剑桥, 嘛: 与新闻界. https://土井
.org/10.7551/mitpress/2528.001.0001
Kang, B., & Druckmann, S. (2020). Approaches to inferring
multi-regional interactions from simultaneous population
录音. 神经生物学的当前观点, 65, 108–119.
https://doi.org/10.1016/j.conb.2020.10.004, 考研: 33227602
Kinnison, J。, Padmala, S。, Choi, J. M。, & Pessoa, L. (2012).
Network analysis reveals increased integration during
emotional and motivational processing. 杂志
神经科学, 32, 8361–8372. https://doi.org/10.1523
/JNEUROSCI.0821-12.2012, 考研: 22699916
Krakauer, J. W., Ghazanfar, A. A。, Gomez-Marin, A。, MacIver,
中号. A。, & Poeppel, D. (2017). Neuroscience needs behavior:
Correcting a reductionist bias. 神经元, 93, 480–490. https://
doi.org/10.1016/j.neuron.2016.12.041, 考研: 28182904
莱文, J. M。, Bascompte, J。, 阿德勒, 磷. B., & Allesina, S. (2017).
Beyond pairwise mechanisms of species coexistence in
complex communities. 自然, 546, 56–64. https://doi.org/10
.1038/nature22898, 考研: 28569813
刘, 是. Y。, & 巴拉巴斯, A. L. (2016). Control principles of
complex systems. Reviews of Modern Physics, 88, 035006.
https://doi.org/10.1103/RevModPhys.88.035006
Lojowska, M。, Ling, S。, Roelofs, K., & Hermans, 乙. J. (2018).
Visuocortical changes during a freezing-like state in humans.
Neuroimage, 179, 313–325. https://doi.org/10.1016/j
.neuroimage.2018.06.013, 考研: 29883732
Lurie, D. J。, Kessler, D ., Bassett, D. S。, 贝策尔, 右. F。, Breakspear,
M。, Kheilholz, S。, 等人. (2020). Questions and controversies in
the study of time-varying functional connectivity in resting
功能磁共振成像. 网络神经科学, 4, 30–69. https://doi.org/10
.1162/netn_a_00116, 考研: 32043043
Maren, S. (2022). Unrelenting fear under stress: Neural circuits
and mechanisms for the immediate extinction deficit.
Frontiers in Systems Neuroscience, 39, 888461. https://土井
.org/10.3389/fnsys.2022.888461, 考研: 35520882
Markov, 氮. T。, Ercsey-Ravasz, M。, Van Essen, D. C。, Knoblauch,
K., Toroczkai, Z。, & 肯尼迪, H. (2013). Cortical high-density
Counterstream architectures. 科学, 342, 1238406. https://
doi.org/10.1126/science.1238406, 考研: 24179228
Markov, 氮. T。, Misery, P。, Falchier, A。, Lamy, C。, Vezoli, J。,
Quilodran, R。, 等人. (2011). Weight consistency specifies
regularities of macaque cortical networks. 大脑皮层,
21, 1254–1272. https://doi.org/10.1093/cercor/bhq201,
考研: 21045004
McMenamin, 乙. W., Langeslag, S. J。, Sirbu, M。, Padmala, S。, &
Pessoa, L. (2014). Network organization unfolds over
time during periods of anxious anticipation. 杂志
神经科学, 34, 11261–11273. https://doi.org/10.1523
/JNEUROSCI.1579-14.2014, 考研: 25143607
Mejías, J. F。, & 王, X. J. (2022). Mechanisms of distributed
working memory in a large-scale network of macaque
neocortex. 电子生活, 11, e72136. https://doi.org/10.7554/eLife
.72136, 考研: 35200137
摩根, 中号. A。, Romanski, L. M。, & LeDoux, J. 乙. (1993).
Extinction of emotional learning: Contribution of medial
前额皮质. Neuroscience Letters, 163, 109–113. https://
doi.org/10.1016/0304-3940(93)90241-C, 考研: 8295722
穆雷, J. D ., Jaramillo, J。, & 王, X. J. (2017). Working
memory and decision-making in a frontoparietal circuit
模型. 神经科学杂志, 37, 12167–12186. https://
doi.org/10.1523/JNEUROSCI.0343-17.2017, 考研:
29114071
Najafi, M。, McMenamin, 乙. W., 西蒙, J. Z。, & Pessoa, L. (2016).
Overlapping communities reveal rich structure in large-scale
brain networks during rest and task conditions. Neuroimage,
135, 92–106. https://doi.org/10.1016/j.neuroimage.2016.04
.054, 考研: 27129758
Nieuwenhuys, R。, Voogd, J。, & van Huijzen, C. (2008). 这
human central nervous system: A synopsis and atlas
(4第三版。). Springer Science and Business Media. https://土井
.org/10.1007/978-3-540-34686-9
Padmala, S。, & Pessoa, L. (2011). Reward reduces conflict by
enhancing attentional control and biasing visual cortical
加工. 认知神经科学杂志, 23,
Pessoa
359
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
3419–3432. https://doi.org/10.1162/jocn_a_00011, 考研:
21452938
Palla, G。, Derenyi, 我。, Farkas, 我。, & 笑话, 时间. (2005). Uncovering
the overlapping community structure of complex networks in
nature and society. 自然, 435, 814–818. https://doi.org/10
.1038/nature03607, 考研: 15944704
Paré, D ., & Quirk, G. J. (2017). When scientific paradigms lead
to tunnel vision: Lessons from the study of fear. NPJ Science
of Learning, 2, 1–8. https://doi.org/10.1038/s41539-017-0007-4,
考研: 30294453
Pessoa, L. (2013). The cognitive–emotional brain: 从
interactions to integration. 剑桥, 嘛: 与新闻界.
https://doi.org/10.7551/mitpress/9780262019569.001.0001
Pessoa, L. (2022A). How many brain regions are needed to
elucidate the neural bases of fear and anxiety? OSF Preprints.
Pessoa, L. (2022乙). The entangled brain: How perception,
认识, and emotion are woven together. 剑桥, 嘛:
与新闻界. https://doi.org/10.7551/mitpress/14636.001.0001
Pessoa, L。, Medina, L。, & Desfilis, 乙. (2022). Refocusing
神经科学: Moving away from mental categories and
towards complex behaviours. Philosophical Transactions of
the Royal Society of London, Series B: Biological Sciences,
377, 20200534. https://doi.org/10.1098/rstb.2020.0534,
考研: 34957851
Ramanathan, K. R。, Jin, J。, Giustino, 时间. F。, 佩恩, 中号. R。, & Maren,
S. (2018). Prefrontal projections to the thalamic nucleus
reuniens mediate fear extinction. Nature Communications,
9, 1–12. https://doi.org/10.1038/s41467-018-06970-z, 考研:
30375397
Risold, 磷. Y。, 汤普森, 右. H。, & 斯旺森, L. 瓦. (1997).
The structural organization of connections between
hypothalamus and cerebral cortex. Brain Research Reviews,
24, 197–254. https://doi.org/10.1016/S0165-0173(97)00007-6,
考研: 9385455
Runge, J。, Bathiany, S。, Bollt, E., Camps-Valls, G。, Coumou, D .,
Deyle, E., 等人. (2019). Inferring causation from time series
in earth system sciences. Nature Communications, 10,
1–13. https://doi.org/10.1038/s41467-019-10105-3, 考研:
31201306
Salinas-Hernández, X. 我。, 沃格尔, P。, Betz, S。, Kalisch, R。,
Sigurdsson, T。, & Duvarci, S. (2018). Dopamine neurons
drive fear extinction learning by signaling the omission of
expected aversive outcomes. 电子生活, 7, e38818. https://doi.org
/10.7554/eLife.38818, 考研: 30421719
Saper, C. 乙. (2002). The central autonomic nervous system:
Conscious visceral perception and autonomic pattern
一代. Annual Review of Neuroscience, 25, 433–469.
https://doi.org/10.1146/annurev.neuro.25.032502.111311,
考研: 12052916
Sherman, S. M。, & Guillery, 右. 瓦. (2002). The role of the
thalamus in the flow of information to the cortex.
Philosophical Transactions of the Royal Society of London,
Series B: Biological Sciences, 357, 1695–1708. https://doi.org
/10.1098/rstb.2002.1161, 考研: 12626004
Shiba, Y。, Santangelo, A. M。, & 罗伯茨, A. C. (2016). 超过
the medial regions of prefrontal cortex in the regulation
of fear and anxiety. Frontiers in Systems Neuroscience,
10, 12. https://doi.org/10.3389/fnsys.2016.00012, 考研:
26941618
席尔瓦, 乙. A。, Astori, S。, Burns, A. M。, Heiser, H。, van den Heuvel,
L。, Santoni, G。, 等人. (2021). A thalamo-amygdalar circuit
underlying the extinction of remote fear memories. 自然
神经科学, 24, 964–974. https://doi.org/10.1038/s41593
-021-00856-y, 考研: 34017129
西蒙, H. A. (1962). The architecture of complexity. 会议记录
of the American Philosophical Society, 106, 467–482.
斯波恩斯, 奥。, & Zwi, J. D. (2004). The small world of the cerebral
cortex. 神经信息学, 2, 145–162. https://doi.org/10
.1385/NI:2:2:145, 考研: 15319512
Sugihara, G。, 可能, R。, 叶, H。, Hsieh, C. H。, Deyle, E., Fogarty, M。,
等人. (2012). Detecting causality in complex ecosystems.
科学, 338, 496–500. https://doi.org/10.1126/science
.1227079, 考研: 22997134
斯旺森, L. W., & Petrovich, G. D. (1998). What is the
amygdala? Trends in Neurosciences, 21, 323–331. https://土井
.org/10.1016/S0166-2236(98)01265-X, 考研: 9720596
唐, E., & Bassett, D. S. (2018). Colloquium: Control of
dynamics in brain networks. Reviews of Modern Physics, 90,
031003. https://doi.org/10.1103/RevModPhys.90.031003
Von Bertalanffy, L. (1950). An outline of general system theory.
British Journal for the Philosophy of Science, 1, 134–165.
https://doi.org/10.1093/bjps/I.2.134
王, X. F。, & 陈, G. (2002). Pinning control of scale-free
dynamical networks. Physica A: Statistical Mechanics and its
应用领域, 310, 521–531. https://doi.org/10.1016/S0378
-4371(02)00772-0
Watts, D. J。, & Strogatz, S. H. (1998). Collective dynamics of
‘small-world’ networks. 自然, 393, 440–442. https://doi.org
/10.1038/30918, 考研: 9623998
Whittle, N。, Fadok, J。, MacPherson, K. P。, 阮, R。, Botta, P。,
Wolff, S. B., 等人. (2021). Central amygdala micro-circuits
mediate fear extinction. Nature Communications, 12, 1–11.
https://doi.org/10.1038/s41467-021-24068-x, 考研:
34230461
Wimsatt, 瓦. C. (2007). Re-engineering philosophy for limited
beings: Piecewise approximations to reality. 哈佛
大学出版社. https://doi.org/10.2307/j.ctv1pncnrh
杨, 乙. 时间. T。, 克里宁, F. M。, Chee, 中号. W., & 巴克纳, 右. L.
(2014). Estimates of segregation and overlap of functional
connectivity networks in the human cerebral cortex.
Neuroimage, 88, 212–227. https://doi.org/10.1016/j
.neuroimage.2013.10.046, 考研: 24185018
杨, 乙. 时间. T。, 克里宁, F. M。, 墓, J。, 肥皂, 中号. R。,
军队, D ., 霍林斯黑德, M。, 等人. (2011). The Organization
of the Human Cerebral Cortex Estimated by intrinsic
功能连接. 神经生理学杂志, 106,
1125–1165. https://doi.org/10.1152/jn.00338.2011, 考研:
21653723
360
认知神经科学杂志
体积 35, 数字 3
我
D
哦
w
n
哦
A
d
e
d
F
r
哦
米
H
t
t
p
:
/
/
d
我
r
e
C
t
.
米
我
t
.
e
d
你
/
j
/
哦
C
n
A
r
t
我
C
e
–
p
d
我
F
/
/
/
/
3
5
3
3
4
9
2
0
6
9
4
4
1
/
j
哦
C
n
_
A
_
0
1
9
0
8
p
d
.
F
乙
y
G
你
e
s
t
t
哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3
