Review

Brain Rhythms of Pain

Markus Ploner,1,* Christian Sorg,2 and Joachim Gross3
Pain is an integrative phenomenon that results from dynamic interactions
Trends
between sensory and contextual (i.e., cognitive, emotional, and motivational)
Pain is a vital phenomenon that
processes. In the brain the experience of pain is associated with neuronal depends on the dynamic integration
oscillations and synchrony at different frequencies. However, an overarching of sensory and contextual processes.
In chronic pain the adaptive integration
framework for the significance of oscillations for pain remains lacking. Recent of sensory and contextual processes is
concepts relate oscillations at different frequencies to the routing of information severely disturbed.
flow in the brain and the signaling of predictions and prediction errors. The
Neuronal oscillations and synchrony at
application of these concepts to pain promises insights into how flexible routing different frequencies provide evidence
of information flow coordinates diverse processes that merge into the experi- on information flow across brain areas.
The flexible relationship between oscil-
ence of pain. Such insights might have implications for the understanding and
lations at different frequencies and pain
treatment of chronic pain. indicates flexible routing of information
flow in the cerebral processing of pain.
How Can the Study of Brain Rhythms Advance Our Understanding of Pain?
The systematic assessment of oscilla-
Pain results from dynamic interactions between sensory and contextual (i.e., cognitive, emo- tions and synchrony in the processing of
tional, and motivational) processes [1]. Pain is thus essentially an integrative phenomenon. In pain provides insights into how sensory
recent years it has been shown that oscillations and synchrony serve integrative functions by and contextual processes are flexibly
flexibly routing information flow in the brain [2–6]. Thus, understanding the role of oscillations in integrated into a coherent percept and
into abnormalities of these processes in
the processing of pain promises insights into how functionally diverse processes dynamically chronic pain. Predictive coding frame-
merge into the experience of pain in health and disease. works might help us understand these
integration processes.
Here we review recent evidence on the role of neuronal oscillations and synchrony in the
processing of pain. We begin with a brief discussion of the peculiarities of pain and its processing
in the brain. We then summarize recent insights into the significance of neuronal oscillations and
synchrony for the routing of information flow in the brain. On this basis we review evidence on the
role of oscillations in the processing of pain. We specifically discuss how oscillations and synchrony
serve the flexible routing of information flow in the integration of sensory and contextual factors into
a coherent percept. Moreover, we review and discuss the role of these processes in pathological
abnormalities of the pain experience in chronic pain. Finally, we consider perspectives and future
directions for the study of the role of neuronal oscillations in the cerebral processing of pain.

Pain
Pain is an unpleasant sensory and emotional experience that signals threat and promotes
behavior to protect the individual. Commonly, the underlying process is that a noxious stimulus 1
Department of Neurology and
induces physiological processes, referred to as nociception (see Glossary), that translate into TUMNeuroimaging Center, Technische
pain [1]. This translation process is influenced by a broad variety of contextual factors. We Universität München, Munich,
Germany
routinely make use of this influence; for example, when comforting an injured child or 2
Departments of Neuroradiology and
when harnessing placebo effects for pain therapy. Pain thus results from the integration of Psychiatry and TUMNeuroimaging
nociceptive and contextual information mediated by feedforward and feedback processes in Center, Technische Universität
München, Munich, Germany
the human brain [7]. This integration process is not static but has to be dynamically adjusted to 3
Institute of Neuroscience and
the continuously changing demands of everyday life. For example, the same noxious input can Psychology, University of Glasgow,
yield no pain when a competing goal has to be achieved (e.g., during a long-distance run) but Glasgow, UK
under other contextual conditions can result in strong pain (e.g., when a severe disease is
feared). Thus, the dynamic integration of sensory and contextual processes plays a preeminent
*Correspondence:
role in pain that probably exceeds its role in other modalities. [email protected] (M. Ploner).

100 Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2 http://dx.doi.org/10.1016/j.tics.2016.12.001
© 2016 Elsevier Ltd. All rights reserved.
In chronic pain states, pain often persists without objective threat to the body. Chronic pain is a Glossary
disease in its own right that affects about a fifth of the adult population in the Western world [8,9], Default mode, salience, and
imposes an enormous economic burden on society [10,11], and causes severe suffering to sensorimotor networks: important
individuals. In chronic pain the relationship between nociception and pain is often weak or lost intrinsic brain networks that are
particularly active during rest and
[12] indicating abnormal integration of nociceptive and contextual information. In particular, there during the detection of and
is a close and mutual relationship between contextual factors and chronic pain [13]. For orientation to salient events and
example, certain psychological factors such as passive coping strategies predispose to the sensorimotor processes, respectively.
development of chronic pain and, conversely, chronic pain yields severe cognitive, affective, and These networks have been shown to
be involved in the processing of pain.
functional deficits [13]. Thus, the adaptive integration of nociceptive and contextual processes is Granger causality: a measure of the
severely disturbed in chronic pain. causal relationship between two time
series. In EEG, MEG, and intracranial
recordings, it is often taken as a
Pain and the Brain
measure of the causal relationship
Pain is associated with the activation of an extended network of brain areas including the between neural signals originating
somatosensory, insular, cingulate, and prefrontal cortices, the thalamus, subcortical areas, and from different locations in the brain.
the brainstem [14]. These areas do not constitute a dedicated pain system but belong to different Infra- and supragranular layers:
the largest part of the cerebral
functional systems of the brain that are transiently orchestrated in the processing of pain. None
cortex, termed the isocortex, is
of these areas exclusively processes or singularly determines the experience of pain [15] (see characterized by a six-layered
[16–19] for an ongoing discussion of this topic). It is thus likely that the integration of neuronal structure. The different layers contain
activity across brain areas eventually determines pain. Structural connections represent the characteristic distributions of neuronal
cell types and connections. From
anatomical basis for this integration process. However, to continuously adjust pain to the
outside to inside, the layers are
momentary behavioral demands, the integration process has to be highly flexible. This flexibility numbered with Roman numerals
requires dynamic changes of neuronal integration at timescales that can be provided not by from I to VI. Layer IV is termed the
changes of structural connections but rather by dynamic changes of functional connections. internal granular layer. Layers I–III are
summarized as the supragranular
Such dynamic changes of functional connections in the processing of pain have recently been layers and V and VI as the
conceptualized as the dynamic pain connectome [20]. This concept does not conceive the infragranular layers. The supra- and
cerebral processing of pain as a static phenomenon but emphasizes that the dynamics of infragranular layers differ in their
functional connections flexibly determine the experience of pain. patterns of feedforward and feedback
connections.
Intracranial recordings:
Pain is associated not only with a spatially extended network of dynamically recruited brain areas neurophysiological recordings of brain
but also with complex temporal–spectral patterns of brain activity. In particular, pain-related activity obtained either from
electrodes placed on the cortex
neuronal oscillations at frequencies ranging from infraslow fluctuations below 0.1 Hz (Box 1) via
(electrocorticography) or from
theta (4–7 Hz), alpha (8–13 Hz), and beta (14–29 Hz) to gamma (30–100 Hz) oscillations [21–31] electrodes inserted in the brain
have been observed. These oscillations have been recorded during different contextual con- (LFPs). In humans intracranial
ditions and at different timescales. However, an overarching framework for the significance of recordings can be obtained during
surgery or after surgical implantation
these oscillations for pain remains lacking.
of electrodes.
Intrinsic brain networks: sets of
The close relationship between chronic pain and psychological factors [13] and the substantial brain areas that exhibit synchronous
comorbidity of chronic pain and mental disorders [32] indicates that brain dysfunction plays a activity in fMRI recordings during the
resting state.
central role in the development and maintenance of chronic pain. Recent neurobiological
Nociception: the neural process of
investigations corroborate the crucial role of the brain in chronic pain by showing substantial encoding noxious stimuli.
structural and metabolic changes of the brain in chronic pain [12,33]. Moreover, neurophysio- Nociceptive: related to noxious
logical and functional imaging studies found abnormalities of the frequency spectrum of brain stimuli.
Slow waves: slow oscillations at
activity ranging from infraslow fluctuations (Box 1) to gamma oscillations in patients with chronic frequencies below 3 Hz observed in
pain [26,27,34,35]. Most recent evidence indicates that some of these changes are causally EEG and LFP recordings. Slow
involved in the development and maintenance of chronic pain [36,37]. waves are mostly observed during
sleep and are likely to play an
important role for memory
Neuronal Oscillations and Synchrony consolidation.
Brain rhythms or brain oscillations refer to rhythmic fluctuations of neural mass signals recorded
by local field potentials (LFPs), electroencephalography (EEG), or magnetoencephalography
(MEG) [38]. Brain oscillations are most prominent at frequencies between 1 and 100 Hz [39].
They originate from the dynamic interplay of excitation and inhibition of neuronal populations
leading to periodic synchronization of action potentials. In addition, infraslow fluctuations of brain

Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2 101

 Box 1. Infraslow Fluctuations and Pain
In resting-state fMRI data, infraslow fluctuations of blood oxygenation level-dependent (BOLD) signals below 0.1 Hz have
been observed [108]. The synchronization of these infraslow fluctuations across brain areas defines intrinsic brain
networks [108]. Accumulating evidence indicates that particularly the activities of the default mode, salience, and
sensorimotor networks are changed in chronic pain [26–29]. The findings show changes of connectivity not only within
but also across [109] these networks in chronic pain. Moreover, changes of the frequency pattern [26,27] and the
variability [110] of infraslow fluctuations have been observed. Thus, the analysis of infraslow fluctuations is a promising
approach to advance our understanding of the brain mechanisms of chronic pain. However, changes of infraslow
fluctuations of BOLD signals are found not only in pain but in many neuropsychiatric disorders [111] so that the disease
and symptom specificity of the observed changes remains an important question.

Two physiological aspects of intrinsic brain networks might be of particular interest in the processing of pain. First,
intrinsic brain networks relate to slowly propagating waves of neuronal activity at a timescale of about 1 s for the whole
cortex [112,113]. These waves provide subthreshold depolarization to individual neurons indicating modulation of
neuronal excitability [114]. Changes of infraslow BOLD fluctuations in chronic pain might therefore indicate impairment of
slowly propagating waves. As slowly propagating waves are intimately linked to slow waves during sleep and memory
consolidation [115], changes of infraslow BOLD fluctuations might relate to abnormal learning and memory consolidation
processes in chronic pain [13]. Second, infraslow fluctuations are linked to fluctuations of faster oscillations at alpha and
gamma frequencies [116,117] indicating cross-frequency coupling with infraslow fluctuations shaping faster oscillations.
In chronic pain, abnormal infraslow fluctuations, particularly in the salience and default mode networks [26–29], might
interfere with faster oscillations and the flexible routing of information flow in the brain.

In summary, infraslow fluctuations provide a mechanistic framework that connects slow wave propagation and learning
processes to oscillations at higher frequencies and flexible cerebral information flow. Abnormalities of infraslow
fluctuations and their synchronization might signal abnormal slow wave propagation and related learning processes
and contribute to abnormal routing of information flow in chronic pain.

activity are observed at frequencies below 0.1 Hz by functional magnetic resonance imaging
(fMRI) (Box 1). At any frequency the synchronization of brain activity can occur both within and
between brain areas [40,41]. Brain oscillations have been observed in association with a broad
variety of perceptual, cognitive, and behavioral functions. The interpretation of their functional
significance therefore varies substantially between tasks and backgrounds. It is only recently that
these different interpretations have been complemented by a unified physiological framework
indicating that brain oscillations are mechanistically involved in the dynamic routing of information
flow [2–6].

This framework is based on a convergence of anatomical and functional findings in animals and
humans. First, in the visual system anatomical connections have been differentiated into
feedforward (bottom-up) and feedback (top-down) connections [42,43]. This anatomical differ-
entiation is apparent in distinct distributions of both types of connections across the various
layers of the cortex. Feedforward projections typically start in supragranular layers and
terminate in layer IV. Feedback projections predominantly start in infragranular layers and
terminate in layers other than layer IV. Second, the non-homogeneous distribution of feedfor-
ward and feedback connections is complemented by a non-homogeneous distribution of brain
oscillations across cortical layers. Several studies demonstrate that oscillations at alpha and beta
frequencies (8–29 Hz) are stronger in infragranular layers than in supragranular layers. By
contrast, oscillations in the gamma frequency band (30–100 Hz) are stronger in supra- than
in infragranular layers of the cortex [44–47]. In light of the aforementioned laminar distribution of
anatomical connections, this suggests a link between feedforward signaling and gamma
oscillations and feedback signaling and alpha/beta oscillations. A recent study provided direct
evidence for these associations. The study characterized the information flow in human visual
areas based on MEG data [48]. Specifically, measures of directed connectivity (such as Granger
causality) indicated stronger connectivity in the gamma band from lower towards higher
hierarchical areas (feedforward signal) whereas directed connectivity in the opposite direction
(from higher to lower areas) is stronger in alpha/beta frequencies.

102 Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2

 Box 2. Predictive Coding
Predictive coding is a framework of brain function that states that the brain is not a passive recipient of information but
generates and optimizes predictions about the environment [118,119]. These predictions are continuously compared
against sensory evidence and discrepancies produce prediction errors that serve to optimize future predictions. In this
way the brain efficiently allocates resources to events that are behaviorally relevant and useful for updating predictions (i.
e., learning processes). These processes are implemented in a hierarchical processing model in which predictions and
prediction errors are passed in feedback and feedforward directions, respectively. At every level of the hierarchy,
prediction errors are minimized by optimizing predictions. In this optimization process, the influences of predictions and
sensory evidence are weighted according to their precision.

In the cortex predictive coding processes are likely to be implemented by various neuronal populations [44]. In particular,
predictions have been related to infragranular neurons and feedback connections while prediction errors have been
related to supragranular neurons and feedforward connections [44]. In light of the predominance of alpha/beta and
gamma oscillations in infra- and supragranular layers, respectively, the relation of predictions and prediction errors to
neuronal oscillations at alpha/beta and gamma frequencies, respectively, is obvious [44,94]. Recent studies have
provided the first experimental evidence for such relationships [106,107].

Taken together these findings indicate that neuronal oscillations and synchrony in distinct
frequency bands serve the dynamic routing of information flow in the brain. Previously seemingly
independent strands of research converged on the notion that alpha/beta oscillations mediate
feedback signals whereas gamma oscillations mediate feedforward signals. In predictive coding
frameworks of brain function, this might correspond to the signaling of predictions and predic-
tion errors, respectively (Box 2). The involvement of neuronal oscillations in the flexible routing of
information flow has been largely demonstrated and developed in the visual system. In the
following section, we apply this concept to findings on neuronal oscillations and synchrony in the
processing of pain.

Neuronal Oscillations and the Experience of Pain

Most studies on the cerebral processing of pain have investigated responses to phasic pain
stimuli in the range of milliseconds to seconds. These results are likely to apply to acute pain
events that signal threat and promote protective behavior. Fewer studies have investigated the
brain mechanisms of longer-lasting pain of months and years as a key feature of pathological
chronic pain conditions. Furthermore, experimental studies on longer-lasting pain in the range of
minutes (tonic pain) have investigated pain at timescales between those of phasic and chronic
pain and are intended to represent an experimental approach towards chronic pain.

Phasic Pain
EEG and MEG studies showed that brief noxious stimuli induce a complex spectral–temporal–
spatial pattern of neuronal responses with at least three different components. First, pain stimuli
evoke increased neural activity at frequencies below 10 Hz. These increases occur between 150
and 400 ms after stimulus application. They originate from the sensorimotor cortex and the
frontoparietal operculum including the insula and secondary somatosensory cortex as well as
from the mid-/anterior cingulate cortex. They correspond to the well-investigated pain-related
evoked potentials [49,50]. Second, phasic pain stimuli transiently suppress oscillations at alpha
and beta frequencies [23,24,51,52]. These suppressions are observed at latencies between
about 300 and 1000 ms in the sensorimotor cortex and occipital areas [24,51]. Third, phasic
pain stimuli induce oscillations at gamma frequencies over the sensorimotor cortex at latencies
of between 150 and 350 ms [21,22,25].

The functional significance of the different components of pain-related brain activity is not yet fully
understood. So far the evidence indicates that the components are differentially sensitive to
different modulations of pain. Bottom-up modulations of pain by varying stimulus intensity (i.e.,
nociceptive information) influences all three components [21,22,25,53,54]. Similarly, top-down
modulations by varying attention affect all components [22,51,52,54,55]. However, during

Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2 103

spontaneous fluctuations of pain [21], pain modulations by music and music therapy [56], and
repetitive painful stimulation [25] gamma oscillations are more closely related to pain intensity
than the other components. By contrast, when pain is modulated by varying the expectation
about the upcoming stimulus in the form of a placebo manipulation, evoked potentials and alpha
suppressions are more closely related to pain than gamma oscillations [53]. Hence, bottom-up
modulations affect all components of pain-related brain activity whereas different top-down
modulations selectively modulate certain components. The available evidence does not yet allow
more precise assignment of the different components to the manifold modulations of pain.

The findings, however, indicate that brain activity at different frequencies provides different
and complementary information about pain. Moreover, they indicate that there is no one-to-
one correspondence between any frequency component of brain activity and pain, which
extends the lack of specificity of brain activity for pain [15] to the frequency domain. Instead,
the relationship between pain and brain activity is variable and context dependent. In the
context of an involvement of oscillations in the flexible routing of information flow, the findings
suggest that different contextual modulations of pain differentially change the information
flow between the involved brain areas (Figure 1, Key Figure). For example, when pain is
mostly driven by nociceptive processing, gamma oscillations in the somatosensory cortex
may serve the feedforward signaling of sensory information to other brain areas involved in
pain processing and behavioral responses to pain [57]. By contrast, when other processes
such as affect or, evaluation dominate, the information flow is changed with gamma
oscillations and feedforward signaling from the somatosensory cortex playing a rather minor
role.

The assessment of brain responses to phasic painful stimuli shows the impact of contextual
modulations on stimulus processing but not the mechanisms of the modulations. A straightfor-
ward approach to the disentangling of contextual processes from stimulus processing is the
assessment of prestimulus activity, which cannot be contaminated by any stimulus-related
processes. The few studies on this topic with respect to pain [58–60] suggest that ongoing
oscillations play an important role in shaping pain perception. Specifically, the amplitude of
prestimulus alpha oscillations over the sensorimotor cortex is negatively correlated with pain
perception [59,60]. Correspondingly, attention to pain [52] and the expectation of analgesia [61]
are associated with changes of alpha oscillations in the sensorimotor and prefrontal cortex,
respectively. In addition, the amplitudes of prestimulus gamma oscillations are correlated with
pain perception [58,59], although the direction of the effect differed. Intriguingly, alpha and
gamma oscillations together have a stronger predictive value than each component alone [59],
which supports the view that they provide different and complementary information about
feedforward and feedback signaling in pain processing.

Studies using intracranial recordings in a few patients with epilepsy investigated the signifi-
cance not only of prestimulus oscillations but also of prestimulus connectivity between brain
areas for pain. The results indicate that attention to pain changes the connectivity between pain-
relevant brain areas at alpha and beta frequencies [62–64]. Intriguingly, the analysis of directed
functional (or effective) connectivity indicates that the information flow is flexibly changed by
attention. Specifically, during attention to a painful stimulus the medial prefrontal cortex exerted
causal influences on the primary sensorimotor cortex whereas during distraction the causal
influences were reversed. These findings provide evidence for the context-dependent routing of
information flow in the processing of pain (Figure 1). Moreover, the findings are well compatible
with a role for synchrony at alpha and/or beta frequencies in the top-down signaling of contextual
factors and/or, in a predictive coding framework (Box 2), predictions of pain. However, these
promising findings originate from three patients and need replication and elaboration in further
studies.

104 Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2

Key Figure
Flexible Routing of Information Flow in the Processing of Pain

(A) Pain (B) Pain (C) Pain Key:

Gamma
feedforward

Alpha/beta
feedback

ContextA Smulus ContextB ContextA Smulus ContextB ContextA Smulus ContextB

Figure 1. Schematic representation of three brain areas in the processing of pain under three different conditions. (A) Pain is mainly driven by stimulus processing. A brain
area associated with stimulus processing sends feedforward information to other brain areas implicated in pain processing. The sending of feedforward information is
associated with gamma oscillations and gamma synchrony across brain areas. (B,C) Pain is mainly driven by contextual processes (e.g., attention, expectation, emotion).
Brain areas associated with the respective contextual factor send feedback information to other brain areas. This is associated with alpha/beta oscillations and alpha/beta
synchrony across brain areas.

Tonic Pain
The above-reviewed evidence relates to the processing of brief experimental pain stimuli. It is,
however, unclear how these findings relate to the brain mechanisms of longer-lasting pain of
months and years, which is the key feature of chronic pain. Experimental studies using longer-
lasting tonic experimental pain stimuli in the range of minutes represent a step further in that
direction. These studies have shown that tonic pain is associated with suppression of oscillations
at alpha frequencies [65–75]. However, as most mental processes suppress alpha oscillations,
the specificity of this effect is unclear. Some studies have claimed it to be pain specific based on
covariation of alpha oscillations and pain intensity [70,71,74]. Another recent study showed that
the suppression of alpha and beta oscillations during tonic pain is more closely related to
stimulus intensity as a proxy for nociception than to the perceived pain intensity [73] indicating
that these suppressions reflect stimulus processing rather than perception. In addition, several
studies have recorded gamma oscillations during tonic pain [72,73,76]. Intriguingly, during tonic
pain gamma oscillations encoded pain rather than nociception [73]. Moreover, in contrast to
phasic pain, they were not recorded over sensorimotor areas but over the medial prefrontal
cortex [73].

Thus, during a few minutes of painful stimulation the encoding of pain shifts from gamma
oscillations over brain areas encoding sensory processes to gamma oscillations over brain areas
encoding emotional–motivational phenomena. These findings indicate that pain-related infor-
mation flow might change not only with the behavioral context but also with the duration of pain.
In the current framework of flexible routing of information flow (Figure 1), these findings suggest
that during longer-lasting pain, signals from brain areas encoding emotional–motivational
processes rather than from sensory brain areas dominate the processing and perception of
pain. In a predictive coding framework (Box 2), this might indicate that longer-lasting pain does
not generate prediction errors at the level of sensory processing but rather at the level of
emotional–motivational processing.

Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2 105

Chronic Pain Outstanding Questions
The analysis of oscillations and synchrony is conceptually promising and methodologically well Recent studies discuss the significance
suited for the investigation of ongoing processes such as chronic pain. However, remarkably few of interactions of oscillations at different
frequencies (i.e., cross-frequency cou-
studies have addressed this topic and the results are not fully consistent (see [77] for a recent pling). What is the role of cross-fre-
review). The most-noticed abnormality is an increase of theta oscillations in chronic pain patients quency coupling in pain? In particular,
(e.g., [34,35]). This phenomenon has been embedded in the framework of thalamocortical how do infraslow fluctuations observed
by fMRI relate to oscillations at higher
dysrhythmia [78,79]. This theory posits that abnormal thalamic theta oscillations play a crucial
frequencies in the processing of pain?
role in various neuropsychiatric disorders. In neuropathic pain deafferentation might cause these
thalamic theta oscillations, which in turn entrain thalamocortical loops. At the cortical level, the Pain modulations can be harnessed for
abnormal theta oscillations are supposed to reduce lateral inhibition, which might result in pain therapy. However, a systematic
abnormal gamma oscillations. Eventually, these abnormal gamma oscillations have been pro- understanding of pain modulations is
so far lacking. Can the assessment of
posed to result in positive neurological and psychiatric symptoms including ongoing pain. The
oscillations and patterns of cerebral
appeal of this framework is its internal coherence and there is some clinical and experimental information flow help to establish a
evidence in favor of the concept [34,35]. However, other studies did not observe abnormal theta brain-based taxonomy of pain
oscillations in chronic pain [80,81]. Moreover, as slowing of the peak alpha frequency in chronic modulations?

pain [34,35,82–85] has also been observed, abnormal amplitudes of theta oscillations might
It is tempting to relate the interaction
basically represent the unspecific slowing of EEG activity observed in many acute [86] and
between sensory input, contextual
chronic [87] neuropsychiatric disorders. information, and pain to predictive cod-
ing frameworks of brain function. How
A less-noticed finding is an increase of oscillations at alpha and beta frequencies [30,34,35,85]. can this relationship be specified and
experimentally tested? What are the
This is in line with studies in animal models of chronic pain that showed broadband increases of consequences for the understanding
oscillations from theta to beta frequencies in the primary somatosensory and medial prefrontal of pain and chronic pain?
cortex [88–90]. In particular, increases of beta oscillations were observed in frontal brain areas
[34,35,85,91,92]. Considering that beta oscillations are likely to serve feedback signaling [48,93] The analysis of oscillations is concep-
and/or the signaling of predictions [94], this would be compatible with abnormal predictions tually and methodologically well suited
for the investigation of the brain mech-
playing a crucial role in chronic pain [95]. anisms of chronic pain. However, evi-
dence on the role of oscillations and
In summary, the data show mostly changes of theta and beta oscillations in chronic pain, the synchrony in chronic pain is remarkably
limited. Can timely network analyses of
latter particularly in frontal brain areas. Considering disturbed integration of nociceptive and
EEG, MEG, and fMRI data specify
contextual processes in chronic pain, an abnormal balance of feedforward and feedback abnormalities of oscillations and syn-
signaling and thereby an abnormal balance of oscillations at different frequencies might play chrony underlying chronic pain?
an important role in chronic pain. However, the role of neuronal oscillations and synchrony in
chronic pain is a largely unexplored field and the emerging concepts await further empirical Subcortical areas including the ventral
striatum, amygdala, and hippocampus
testing.
play an important role in chronic pain.
Although neuronal oscillations from
Concluding Remarks and Future Perspectives these areas are well known they have
Recent evidence has shown that oscillations and synchrony play a crucial role in the flexible not so far been investigated during
pain. How can subcortical oscillations
routing of information flow in the brain. In particular, oscillations at gamma and alpha/beta
be recorded and how do they integrate
frequencies have been shown to serve feedforward and feedback processing, respectively. in patterns of cerebral information
The flexible routing of information flow might be particularly relevant in the processing of pain flow?
where the dynamic integration of sensory and contextual processes plays a crucial role (Figure 1).
The results available so far are compatible with these concepts. It has been shown that there is Recent studies discuss the use of pat-
terns of brain activity as markers of
no one-to-one correspondence between oscillations at any frequency or location and the pain. Can patterns of neuronal oscilla-
subjective experience of pain, which extends evidence on the lack of specificity of pain-related tions and synchrony serve as diagnos-
brain activity [15] to the frequency domain. Instead, different modulations of pain are associated tic and/or prognostic markers of pain?
with distinct changes of neuronal oscillations indicating flexible, context-dependent routing of Can we target neuronal oscillations and
synchrony for pain therapy using phar-
information flow. The available evidence does not so far allow more systematic mapping of the macological, behavioral, neuromodula-
relationship between oscillations, cerebral information flow, and the experience of pain and its tory, or neurofeedback approaches?
modulations. This lack of evidence is at least partly due to a lack of a systematic understanding of
pain modulations [96]. Conversely, the systematic assessment of brain oscillations might
represent a promising approach to establish a taxonomy [96] or ontology [97] of different types
of pain modulation based on patterns of oscillations and cerebral information flow.

106 Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2

Chronic pain appears to be associated with abnormal oscillations at theta and beta frequencies.
Although the specificity of these findings has remained unclear, at least part of them would be
compatible with abnormal contextual feedback processes playing a central role in the pathology
of chronic pain. More standardized approaches, larger patient samples, data-sharing initiatives,
and more sophisticated and timely analysis strategies such as graph theory-based network
analyses [98] are needed to further our understanding of the role of neuronal oscillations and
flexible cerebral information flow in chronic pain. A better understanding of these processes
might eventually help in the diagnosis and treatment of chronic pain. In particular, the assess-
ment of oscillations and cerebral information flow might help to establish brain-based diagnostic
markers of pain [99,100]. Moreover, the frequency-selective modulation of neuronal oscillations
by brain stimulation techniques [101,102] can determine causal influences between oscillations
and behavior and might represent an option for the treatment of pain.

Furthermore, considering the preeminent role of the integration of contextual and sensory
information in the processing of pain, an application of predictive coding frameworks (Box 2)
to the processing of pain is obvious. In such a framework, contextual and nociceptive informa-
tion might be conceptualized as predictions and sensory evidence, respectively. Pain thereby
results from the comparison and adjustment of predictions, sensory evidence, and prediction
errors rather than directly from nociceptive information. Accordingly, it has recently been
proposed that predictive coding represents a suitable and testable model of pain processing
[7,103–105]. Paradigmatically, a predictive coding model for pain and placebo analgesia has
been presented [103]. In this model placebo-induced treatment expectations were conceptu-
alized as feedback-mediated predictions, which modulate pain by changing the balance of
feedback and feedforward processes at different levels of a neural processing hierarchy. It will be
intriguing to extend this model to other modulations of pain. Moreover, considering the rela-
tionship of predictions and prediction errors with alpha/beta and gamma oscillations, respec-
tively [44,94,106,107], the assessment of oscillations could provide novel insights into predictive
coding processes related to pain. This is even more appealing as abnormally precise predictions
[95] and/or abnormal updating of predictions [7] might play an important role in the pathology of
chronic pain.

Thus, based on recent progress in our understanding of neuronal oscillations, their systematic
assessment might provide a unique window onto the dynamics of cerebral information flow and
related predictive coding processes underlying the experience of pain in health and disease.

Acknowledgments
M.P. is supported by the Deutsche Forschungsgemeinschaft (PL 321/11-1). J.G. is supported by the Wellcome Trust
(098433). The authors thank Ulrike Bingel, Henrik Heitmann, Elisabeth May, Moritz Nickel, Son Ta Dinh, Laura Tiemann, and
Katja Wiech for helpful discussions and comments on the manuscript.

References
1. Melzack, R. and Casey, K.L. (1968) Sensory, motivational, and 6. Mejias, J.F. et al. (2016) Feedforward and feedback frequency-
central control determinants of pain: a new conceptual model in dependent interactions in a large-scale laminar network of the
pain. In The Skin Senses (Kenshalo, D.R.J., ed.), pp. 423–439, primate cortex. Sci. Adv. 2, e1601335
Charles C. Thomas 7. Wiech, K. (2016) Deconstructing the sensation of pain: the
2. Schnitzler, A. and Gross, J. (2005) Normal and pathological influence of cognitive processes on pain perception. Science
oscillatory communication in the brain. Nat. Rev. Neurosci. 6, 354, 584–587
285–296 8. Breivik, H. et al. (2006) Survey of chronic pain in Europe: preva-
3. Akam, T. and Kullmann, D.M. (2014) Oscillatory multiplexing of lence, impact on daily life, and treatment. Eur. J. Pain 10, 287–333
population codes for selective communication in the mammalian 9. Kennedy, J. et al. (2014) Prevalence of persistent pain in the U.S.
brain. Nat. Rev. Neurosci. 15, 111–122 adult population: new data from the 2010 national health inter-
4. Fries, P. (2015) Rhythms for cognition: communication through view survey. J. Pain 15, 979–984
coherence. Neuron 88, 220–235 10. Breivik, H. et al. (2013) The individual and societal burden of
5. Saalmann, Y.B. et al. (2012) The pulvinar regulates information chronic pain in Europe: the case for strategic prioritisation and
transmission between cortical areas based on attention action to improve knowledge and availability of appropriate care.
demands. Science 337, 753–756 BMC Public Health 13, 1229

Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2 107

11. Gaskin, D.J. and Richard, P. (2012) The economic costs of pain 38. Panzeri, S. et al. (2015) Neural population coding: combining
in the United States. J. Pain 13, 715–724 insights from microscopic and mass signals. Trends Cogn. Sci.
12. Baliki, M.N. and Apkarian, A.V. (2015) Nociception, pain, nega- 19, 162–172
tive moods, and behavior selection. Neuron 87, 474–491 39. Buzsaki, G. and Draguhn, A. (2004) Neuronal oscillations in
13. Flor, H. and Turk, D.C. (2011) Chronic Pain, IASP Press cortical networks. Science 304, 1926–1929

14. Apkarian, A.V. et al. (2005) Human brain mechanisms of pain 40. Siegel, M. et al. (2012) Spectral fingerprints of large-scale neu-
perception and regulation in health and disease. Eur. J. Pain 9, ronal interactions. Nat. Rev. Neurosci. 13, 121–134
463–484 41. Gross, J. (2016) Let the rhythm guide you: non-invasive tracking
15. Legrain, V. et al. (2011) The pain matrix reloaded: a salience of cortical communication channels. Neuron 89, 244–247
detection system for the body. Prog. Neurobiol. 93, 111–124 42. Felleman, D.J. and Van Essen, D.C. (1991) Distributed hierarchi-
16. Segerdahl, A.R. et al. (2015) The dorsal posterior insula cal processing in the primate cerebral cortex. Cereb. Cortex 1,
subserves a fundamental role in human pain. Nat. Neurosci. 1–47
18, 499–500 43. Markov, N.T. et al. (2014) Anatomy of hierarchy: feedforward and
17. Davis, K.D. et al. (2015) Evidence against pain specificity in the feedback pathways in macaque visual cortex. J. Comp. Neurol.
dorsal posterior insula. F1000Res. 4, 362 522, 225–259
18. Lieberman, M.D. and Eisenberger, N.I. (2015) The dorsal anterior 44. Bastos, A.M. et al. (2012) Canonical microcircuits for predictive
cingulate cortex is selective for pain: results from large-scale coding. Neuron 76, 695–711
reverse inference. Proc. Natl. Acad. Sci. U.S.A. 112, 15250– 45. Wang, X.J. (2010) Neurophysiological and computational
15255 principles of cortical rhythms in cognition. Physiol. Rev. 90,
19. Wager, T.D. et al. (2016) Pain in the ACC? Proc. Natl. Acad. Sci. 1195–1268
U.S.A. 113, E2474–E2475 46. van Kerkoerle, T. et al. (2014) Alpha and gamma oscillations
20. Kucyi, A. and Davis, K.D. (2015) The dynamic pain connectome. characterize feedback and feedforward processing in monkey
Trends Neurosci. 38, 86–95 visual cortex. Proc. Natl. Acad. Sci. U.S.A. 111, 14332–14341
21. Gross, J. et al. (2007) Gamma oscillations in human primary 47. Scheeringa, R. et al. (2016) The relationship between oscillatory
somatosensory cortex reflect pain perception. PLoS Biol. 5, EEG activity and the laminar-specific BOLD signal. Proc. Natl.
e133 Acad. Sci. U.S.A. 113, 6761–6766
22. Hauck, M. et al. (2007) Attention to painful stimulation enhances 48. Michalareas, G. et al. (2016) Alpha–beta and gamma rhythms
gamma-band activity and synchronization in human sensorimo- subserve feedback and feedforward influences among human
tor cortex. J. Neurosci. 27, 9270–9277 visual cortical areas. Neuron 89, 384–397

23. Mouraux, A. et al. (2003) Non-phase locked electroencephalo- 49. Garcia-Larrea, L. et al. (2003) Brain generators of laser-evoked
gram (EEG) responses to CO2 laser skin stimulations may reflect potentials: from dipoles to functional significance. Neurophysiol.
central interactions between A@- and C-fibre afferent volleys. Clin. Clin. 33, 279–292
Neurophysiol. 114, 710–722 50. Lorenz, J. and Garcia-Larrea, L. (2003) Contribution of attentional
24. Ploner, M. et al. (2006) Pain suppresses spontaneous brain and cognitive factors to laser evoked brain potentials. Neuro-
rhythms. Cereb. Cortex 16, 537–540 physiol. Clin. 33, 293–301

25. Zhang, Z.G. et al. (2012) Gamma-band oscillations in the primary 51. Hu, L. et al. (2013) Functional features of nociceptive-induced
somatosensory cortex – a direct and obligatory correlate of suppression of alpha band electroencephalographic oscillations.
subjective pain intensity. J. Neurosci. 32, 7429–7438 J. Pain 14, 89–99

26. Baliki, M.N. et al. (2011) The cortical rhythms of chronic back 52. May, E.S. et al. (2012) Pre- and post-stimulus alpha activity
pain. J. Neurosci. 31, 13981–13990 shows differential modulation with spatial attention during the
processing of pain. Neuroimage 62, 1965–1974
27. Malinen, S. et al. (2010) Aberrant temporal and spatial brain
activity during rest in patients with chronic pain. Proc. Natl. Acad. 53. Tiemann, L. et al. (2015) Differential neurophysiological correlates
Sci. U.S.A. 107, 6493–6497 of bottom-up and top-down modulations of pain. Pain 156,
289–296
28. Napadow, V. et al. (2012) Decreased intrinsic brain connectivity is
associated with reduced clinical pain in fibromyalgia. Arthritis 54. Hauck, M. et al. (2015) Top-down and bottom-up modulation of
Rheum. 64, 2398–2403 pain-induced oscillations. Front. Hum. Neurosci. 9, 375

29. Alshelh, Z. et al. (2016) Chronic neuropathic pain: it's about the 55. Tiemann, L. et al. (2010) Gamma oscillations as a neuronal
rhythm. J. Neurosci. 36, 1008–1018 correlate of the attentional effects of pain. Pain 150, 302–308
30. Green, A.L. et al. (2009) Neural signatures in patients with neu- 56. Hauck, M. et al. (2013) The influence of music and music therapy
ropathic pain. Neurology 72, 569–571 on pain-induced neuronal oscillations measured by magneten-
cephalography. Pain 154, 539–547
31. Huang, Y. et al. (2016) Characteristics of local field potentials
correlate with pain relief by deep brain stimulation. Clin. Neuro- 57. Schulz, E. et al. (2012) g Oscillations are involved in the sensori-
physiol. 127, 2573–2580 motor transformation of pain. J. Neurophysiol. 108, 1025–1031

32. Fishbain, D.A. (2013) Psychiatric pain-associated co-morbidities. 58. Taesler, P. and Rose, M. (2016) Prestimulus theta oscillations
In Wall and Melzack's Textbook of Pain (McMahon, S.B. et al., and connectivity modulate pain perception. J. Neurosci. 36,
eds), pp. 273–282, Elsevier 5026–5033

33. Rauschecker, J.P. et al. (2015) Frontostriatal gating of tinnitus 59. Tu, Y. et al. (2016) Alpha and gamma oscillation amplitudes
and chronic pain. Trends Cogn. Sci. 19, 567–578 synergistically predict the perception of forthcoming nociceptive
stimuli. Hum. Brain Mapp. 37, 501–514
34. Sarnthein, J. et al. (2006) Increased EEG power and slowed
dominant frequency in patients with neurogenic pain. Brain 60. Babiloni, C. et al. (2006) Anticipatory electroencephalography
129, 55–64 alpha rhythm predicts subjective perception of pain intensity. J.
Pain 7, 709–717
35. Stern, J. et al. (2006) Persistent EEG overactivation in the cortical
pain matrix of neurogenic pain patients. Neuroimage 31, 61. Huneke, N.T. et al. (2013) Experimental placebo analgesia
721–731 changes resting-state alpha oscillations. PLoS ONE 8, e78278

36. Baliki, M.N. et al. (2012) Corticostriatal functional connectivity 62. Ohara, S. et al. (2006) Analysis of synchrony demonstrates ‘pain
predicts transition to chronic back pain. Nat. Neurosci. 15, 1117– networks’ defined by rapidly switching, task-specific, functional
1119 connectivity between pain-related cortical structures. Pain 123,
244–253
37. Vachon-Presseau, E. et al. (2016) Corticolimbic anatomical
characteristics predetermine risk for chronic pain. Brain 139, 63. Liu, C.C. et al. (2011) Attention to painful cutaneous laser stimuli
1958–1970 evokes directed functional interactions between human sensory

108 Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2

 and modulatory pain-related cortical areas. Pain 152, 2781– 87. Fonseca, L.C. et al. (2013) Comparison of quantitative EEG
2791 between patients with Alzheimer's disease and those with Par-
64. Liu, C.C. et al. (2011) Attention to painful cutaneous laser stimuli kinson's disease dementia. Clin. Neurophysiol. 124, 1970–1974
evokes directed functional connectivity between activity 88. LeBlanc, B.W. et al. (2016) Electroencephalographic signatures
recorded directly from human pain-related cortical structures. of pain and analgesia in rats. Pain 157, 2330–2340
Pain 152, 664–675 89. LeBlanc, B.W. et al. (2016) T-type calcium channel blocker Z944
65. Chang, P.F. et al. (2002) Differential cerebral responses to aver- restores cortical synchrony and thalamocortical connectivity in a
sive auditory arousal versus muscle pain: specific EEG patterns rat model of neuropathic pain. Pain 157, 255–263
are associated with human pain processing. Exp. Brain Res. 147, 90. Leblanc, B.W. et al. (2014) Cortical theta is increased while
387–393 thalamocortical coherence is decreased in rat models of acute
66. Chang, P.F. et al. (2004) Comparative EEG activation to skin pain and chronic pain. Pain 155, 773–782
and muscle pain induced by capsaicin injection. Int. J. Psycho- 91. Gonzalez-Roldan, A.M. et al. (2016) Altered dynamic of EEG
physiol. 51, 117–126 oscillations in fibromyalgia patients at rest. Pain Med. 17,
67. Chen, A.C. and Rappelsberger, P. (1994) Brain and human pain: 1058–1068
topographic EEG amplitude and coherence mapping. Brain Top- 92. Hargrove, J.B. et al. (2010) Quantitative electroencephalographic
ogr. 7, 129–140 abnormalities in fibromyalgia patients. Clin. EEG Neurosci. 41,
68. Dowman, R. et al. (2008) EEG indices of tonic pain-related activity 132–139
in the somatosensory cortices. Clin. Neurophysiol. 119, 1201– 93. Bastos, A.M. et al. (2015) Visual areas exert feedforward and
1212 feedback influences through distinct frequency channels. Neuron
69. Giehl, J. et al. (2014) Responses to tonic heat pain in the ongoing 85, 390–401
EEG under conditions of controlled attention. Somatosens. Mot. 94. Arnal, L.H. and Giraud, A.L. (2012) Cortical oscillations and
Res. 31, 40–48 sensory predictions. Trends Cogn. Sci. 16, 390–398
70. Li, L. et al. (2016) Placebo analgesia changes alpha oscilla- 95. Edwards, M.J. et al. (2012) A Bayesian account of ‘hysteria’.
tions induced by tonic muscle pain: EEG frequency analysis Brain 135, 3495–3512
including data during pain evaluation. Front. Comput. Neuro-
96. Ploner, M. et al. (2015) Towards a taxonomy of pain modulations.
sci. 10, 45
Trends Cogn. Sci. 19, 180–182
71. Nir, R.R. et al. (2012) Tonic pain and continuous EEG: prediction
97. Poldrack, R.A. and Yarkoni, T. (2016) From brain maps to cog-
of subjective pain perception by alpha-1 power during stimulation
nitive ontologies: informatics and the search for mental structure.
and at rest. Clin. Neurophysiol. 123, 605–612
Annu. Rev. Psychol. 67, 587–612
72. Peng, W. et al. (2014) Changes of spontaneous oscillatory activ-
98. Bullmore, E. and Sporns, O. (2009) Complex brain networks:
ity to tonic heat pain. PLoS ONE 9, e91052
graph theoretical analysis of structural and functional systems.
73. Schulz, E. et al. (2015) Prefrontal gamma oscillations encode Nat. Rev. Neurosci. 10, 186–198
tonic pain in humans. Cereb. Cortex 25, 4407–4414
99. Schulz, E. et al. (2012) Decoding an individual's sensitivity to pain
74. Shao, S. et al. (2012) Frequency-domain EEG source analysis for from the multivariate analysis of EEG data. Cereb. Cortex 22,
acute tonic cold pain perception. Clin. Neurophysiol. 123, 2042– 1118–1123
2049
100. Kuo, P.C. et al. (2016) Decoding the perception of endoge-
75. Huishi Zhang, C. et al. (2016) Spectral and spatial changes of nous pain from resting-state MEG. Neuroimage. Published
brain rhythmic activity in response to the sustained thermal pain online October 14, 2016. http://dx.doi.org/10.1016/j.
stimulation. Hum. Brain Mapp. 37, 2976–2991 neuroimage.2016.09.040
76. Li, L. et al. (2016) Changes of gamma-band oscillatory activity to 101. Herrmann, C.S. et al. (2016) EEG oscillations: from correlation to
tonic muscle pain. Neurosci. Lett. 627, 126–131 causality. Int. J. Psychophysiol. 103, 12–21
77. Pinheiro, E.S. et al. (2016) Electroencephalographic patterns in 102. Romei, V. et al. (2016) Information-based approaches of non-
chronic pain: a systematic review of the literature. PLoS ONE 11, invasive transcranial brain stimulation. Trends Neurosci. 39,
e0149085 782–795
78. Llinas, R.R. et al. (1999) Thalamocortical dysrhythmia: a neu- 103. Buchel, C. et al. (2014) Placebo analgesia: a predictive coding
rological and neuropsychiatric syndrome characterized by perspective. Neuron 81, 1223–1239
magnetoencephalography. Proc. Natl. Acad. Sci. U.S.A. 96,
104. Morrison, I. et al. (2013) Facets and mechanisms of adaptive pain
15222–15227
behavior: predictive regulation and action. Front. Hum. Neurosci.
79. Llinas, R. et al. (2005) Rhythmic and dysrhythmic thalamocortical 7, 755
dynamics: GABA systems and the edge effect. Trends Neurosci.
105. Seymour, B. and Dolan, R.J. (2013) Emotion, motivation and
28, 325–333
pain. In Textbook of Pain (McMahon, S.B. et al., eds), pp. 248–
80. Jensen, M.P. et al. (2013) Brain EEG activity correlates of chronic 255, Elsevier
pain in persons with spinal cord injury: clinical implications. Spinal
106. Brodski, A. et al. (2015) The faces of predictive coding. J. Neuro-
Cord 51, 55–58
sci. 35, 8997–9006
81. Schmidt, S. et al. (2012) Pain ratings, psychological functioning
107. Bauer, M. et al. (2014) Attentional modulation of alpha/beta and
and quantitative EEG in a controlled study of chronic back pain
gamma oscillations reflect functionally distinct processes. J.
patients. PLoS ONE 7, e31138
Neurosci. 34, 16117–16125
82. Wydenkeller, S. et al. (2009) Neuropathic pain in spinal cord
108. Raichle, M.E. (2015) The restless brain: how intrinsic activity
injury: significance of clinical and electrophysiological measures.
organizes brain function. Philos. Trans. R. Soc. Lond. B Biol.
Eur. J. Neurosci. 30, 91–99
Sci. Published online May 19, 2015. http://dx.doi.org/10.1016/
83. de Vries, M. et al. (2013) Altered resting state EEG in chronic 10.1098/rstb.2014.0172
pancreatitis patients: toward a marker for chronic pain. J. Pain
109. Hemington, K.S. et al. (2016) Abnormal cross-network functional
Res. 6, 815–824
connectivity in chronic pain and its association with clinical
84. Boord, P. et al. (2008) Electroencephalographic slowing and symptoms. Brain Struct. Funct. 221, 4203–4219
reduced reactivity in neuropathic pain following spinal cord injury.
110. Rogachov, A. et al. (2016) Regional brain signal variability: a novel
Spinal Cord 46, 118–123
indicator of pain sensitivity and coping. Pain 157, 2483–2492
85. Lim, M. et al. (2016) Increased low- and high-frequency oscil-
111. Greicius, M. (2008) Resting-state functional connectivity in neu-
latory activity in the prefrontal cortex of fibromyalgia patients.
ropsychiatric disorders. Curr. Opin. Neurol. 21, 424–430
Front. Hum. Neurosci. 10, 111
112. Mitra, A. et al. (2015) Lag threads organize the brain's intrinsic
86. Sutter, R. and Kaplan, P.W. (2013) Neuroimaging correlates of
activity. Proc. Natl. Acad. Sci. U.S.A. 112, E2235–E2244
acute encephalopathy. J. Clin. Neurophysiol. 30, 517–525

Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2 109

113. Matsui, T. et al. (2016) Transient neuronal coactivations embed- 117. Brookes, M.J. et al. (2011) Investigating the electrophysiological
ded in globally propagating waves underlie resting-state func- basis of resting state networks using magnetoencephalography.
tional connectivity. Proc. Natl. Acad. Sci. U.S.A. 113, 6556–6561 Proc. Natl. Acad. Sci. U.S.A. 108, 16783–16788
114. Wu, J.Y. et al. (2008) Propagating waves of activity in the neo- 118. Huang, Y. and Rao, R.P. (2011) Predictive coding. Wiley Inter-
cortex: what they are, what they do. Neuroscientist 14, 487–502 discip. Rev. Cogn. Sci. 2, 580–593
115. Mitra, A. et al. (2015) Propagated infra-slow intrinsic brain activity 119. Clark, A. (2013) Whatever next? Predictive brains, situated
reorganizes across wake and slow wave sleep. Elife 4, 10781 agents, and the future of cognitive science. Behav. Brain Sci.
116. Mantini, D. et al. (2007) Electrophysiological signatures of resting 36, 181–204
state networks in the human brain. Proc. Natl. Acad. Sci. U.S.A.
104, 13170–13175

110 Trends in Cognitive Sciences, February 2017, Vol. 21, No. 2