Vous êtes sur la page 1sur 12


Critchley et al. reported functional MRI

(fMRI) and morphometric data indicating
a specific role for the right AIC in heart-
How do you feel now? The anterior beat awareness3, an interoceptive measure
that correlates with individual subjective
insula and human awareness emotional awareness4. A PET study of non-
painful gastric distension reported that a
subjective sense of fullness was associated
A. D. (Bud) Craig with activation peaks in the bilateral dorsal
posterior insula, the left mid-insula, the left
Abstract | The anterior insular cortex (AIC) is implicated in a wide range of AIC and the ACC5; these results conform to
conditions and behaviours, from bowel distension and orgasm, to cigarette craving the posterior-to-mid-to-anterior pattern of
and maternal love, to decision making and sudden insight. Its function in the integration mentioned above, but the left-
re-representation of interoception offers one possible basis for its involvement in sided asymmetry seems to correlate with
all subjective feelings. New findings suggest a fundamental role for the AIC (and the fact that this stimulus activates primarily
vagal (parasympathetic) afferents (BOX 3).
the von Economo neurons it contains) in awareness, and thus it needs to be
Finally, activation was observed in the bilat-
considered as a potential neural correlate of consciousness. eral AIC (right side (R)>left side (L)) during
non-painful oesophageal distention or view-
In a 2002 Perspective article I discussed the jointly activated, consistent with the idea that ing of fearful faces, and such activation dis-
phylogenetically new primate lamina I spino- they serve as complementary limbic sensory played synergistic enhancement when these
thalamocortical pathway, which provides a and motor regions that work together, similar stimuli were delivered simultaneously 6, sug-
primary interoceptive representation of the to the somatosensory and motor cortices gesting that emotional states are integrated
physiological condition of the body in the (BOX 1). The studies that reported activation with interoceptive states in the representation
posterior insular cortex 1. The evidence at that of the AIC but not the ACC are explicitly of the subjective feelings of the moment.
time indicated that the anterior insular cortex noted. In addition, activation in the AIC Pain is a significant interoceptive feeling,
(AIC) (FIG. 1) contains interoceptive re- often spills over into the frontal operculum and four recent pain studies are noteworthy
representations that substantialize (that and the neighbouring inferior frontal gyrus here. One study reported that noxious heat
is, provide the basis for) all subjective feel- (IFG), and it overlies the distribution of stimulation of the left or right hand activated
ings from the body and perhaps emotional von Economo neurons (VENs) in this the contralateral dorsal posterior insula, the
awareness, consistent with the essence of fronto-insular junction (BOX 2). bilateral mid-insula (R>L) and, when
the JamesLange theory of emotion and the subject attended to the stimulus, the
Damasios somatic marker hypothesis. Interoception. Interoceptive stimuli that right AIC (R>>L)7. Another study reported
Studies in diverse fields now offer a wealth of have been shown to activate the AIC include that the objective intensity of a heat pain
convergent data that support and extend these thirst, dyspnea, air hunger, the Valsalva stimulus correlated with activation in the
proposals. In this Perspective, I discuss recent manoeuvre, sensual touch, itch, penile stim- posterior insula, whereas the subjective eval-
functional-imaging reports that describe ulation, sexual arousal, coolness, warmth, uation of heat pain correlated with activation
activation of the AIC and highlight those that exercise, heartbeat, wine-tasting (in sommel- in the bilateral AIC (R>L)8. A different study
seem the most important for understand- iers), and distension of the bladder, stomach, found that empathic feelings for a loved one
ing its role. The available evidence suggests rectum or oesophagus (see Supplementary receiving painful simulation were associated
strongly that the AIC has a fundamental role information S1 (table)). Our original with activation of the bilateral AIC but not
in human awareness. After discussing this positron emission tomography (PET) study the posterior insula9. Finally, investigators
concept, I describe a model that could explain showed that objective cool temperatures injected volunteers with hypertonic saline
how the AIC might play this role. are represented linearly in the contralateral in the arm and the leg to produce a painful
dorsal posterior insula, whereas subjective stimulation of muscle or the overlying skin,
Recent findings of AIC activation ratings of these stimuli correlate with activa- and they observed distinct, neighbouring
The recent imaging studies that report acti- tion of the contralateral mid-insula and then sites of activation in the AIC that were
vation of the AIC are here categorized in most strongly with the AIC and the adjacent somatotopically arranged10.
different fields of inquiry (FIG. 2). Selected orbitofrontal cortex on the right side2, sug-
studies are highlighted and others are listed gestive of a posterior-to-mid-to-anterior Awareness of body movement. Two
in Supplementary information S1 (table). pattern of integration of interoceptive infor- reports11,12 showed that the feeling of agency
In most but not all of these studies, the AIC mation. Here, I highlight three reports that or awareness of body control during hand
and the anterior cingulate cortex (ACC) are substantiate and expand these findings. movements is associated with activation



in well-known melodies were mis-timed)16.

A recent fMRI study compared activation
induced by listening to pleasant and unpleas-
ant music that the subjects selected individu-
ally and found that subjectively pleasant
music induced activation of the AIC that was
greater on the left than on the right side17.
SPS One review presented evidence that the AIC
is activated bilaterally during overt speech
ms ps
as and singing and, strikingly, that the AIC acti-
ac al H
APS vation is asymmetric during covert singing:
pl activity in the right AIC was higher during
slow tempos (<3 Hz); activity in the left AIC
IPS was higher during fast tempos (>3 Hz)18.

Emotional awareness. Almost all recent

imaging studies of emotion report joint acti-
vation of the AIC and the ACC in subjects
experiencing emotional feelings, including
Figure 1 | Anatomy of the insula. A photograph of the left insular cortex of a human patient. For a maternal and romantic love, anger, fear,
comparison with the insulae of other patients, see REF. 114. The human insular cortex is a distinct but sadness, happiness, sexual arousal, disgust,
hidden lobe of the brain. It is disproportionately (~30%) enlarged in the human relative to the macaque aversion, unfairness, inequity, indigna-
monkey109. It has 57 oblique gyri, but its morphology is quite variable, even between the two sides114118.
tion, uncertainty, disbelief, social exclusion,
A comprehensive hodological description in the macaque is lacking, and few connectivity analyses of
the insula have been made in humans. Primary interoceptive representations are located in the dorsal trust, empathy, sculptural beauty, a state of
posterior insula and re-represented in a polymodal integrative zone in the mid-insula and again in the union with God, and a hallucinogenic state
anterior insular cortex (AIC)2,7,119,120. The primary interoceptive, gustatory and vagal representations (induced by ayahuasca). Thus, the AIC is
extend to the anterior limit of the insula in macaques but only to the middle of the insula in humans98,121124, activated not just in association with subjec-
which suggests that the AIC of humans has no equivalent in the monkey. The most anterior and ventral tive feelings from the body, but apparently
(inferior) portion of the human insula that adjoins the frontal operculum is probably the most recently with all subjective feelings. One noteworthy
evolved, because this part (as well as the anterior cingulate cortex; BOX 1) contains von Economo neurons study examined resting-state functional
(BOX 2). as, anterior short insular gyrus; al, anterior long insular gyrus; ac, accessory gyrus; APS, anterior connectivity and found two networks: an
peri-insular sulcus; H, Heschls gyrus; IPS, inferior peri-insular sulcus; ms, middle short insular gyrus; ps, executive control network that included
posterior short insular gyrus; pl, posterior long insular gyrus; SPS, superior peri-insular sulcus. Photograph
the dorsolateral prefrontal cortex (DLPFC)
is courtesy of Professor Thomas P. Naidich, Mount Sinai Medical Center, New York.
and parietal areas, and an emotional sali-
ence network that included the bilateral
in the bilateral mid-insula. Another study most recent study 15 contrasted brain activa- AIC, the ACC, the amygdala and the hypo-
reported activation in the right mid-insula tion in subjects viewing photos of their own thalamus19. Significantly, small regions in
during the rubber-hand illusion, in which face or body with activation when these the left AIC and the medial prefrontal cortex
the subject was not actually moving but felt subjects viewed photos of a close colleague (MPFC) near the ACC were included in
like the moving hand was their own13. The or scrambled images. During both types of both networks, suggesting a basis for both
authors suggested that the mid-insula activa- visual self-recognition the authors found emotional awareness of cognitive functions
tion during movement might represent a selective activation of the AIC and the adja- and the influence of emotion on cognition.
sense of body ownership rather than agency. cent IFG and ACC, all on the right side. In addition, several studies (for example,
No activation was observed in the ACC in They suggested that the right AIC and ACC REFS 20,21) found selective activation in
these studies. Finally, a recent study reported could give rise to an abstract representation overlapping regions in the AIC during an
that simply hearing the piano notes that a of oneself that could possibly participate in empathic feeling and a comparable subjec-
subject has just learned to play activates the maintaining a sense of self . tive feeling (for example, seeing disgust
same mid-insular region14. Based on a meta- expressed on anothers face and smelling
analysis that they performed of prior studies, Vocalization and music. There are differing a disgusting odour, respectively), with the
the authors proposed that the insula contains views on the involvement of the insula in former located anterior to the latter (consist-
a sensorimotor map that represents move- speech (some authors include the insula ent with the posterior-to-anterior gradient
ments. My interpretation of these studies is in Brocas area), as summarized in the towards greater behavioural complexity in
that the insular cortex contains a somatotopic reviews cited in Supplementary information the frontal cortex 22). Notably, similar loca-
representation of the subjective feelings of S1 (table), but there is consistent evidence for tions in the AIC may be active during quite
ones current movements as part of a the involvement of the AIC in music. Of note, different emotions and behaviours20,21,23;
representation of all feelings from the body. a PET study reported findings that differenti- across all studies, a main determinant of
ated brain regions associated with familiarity, location in the AIC seems to be the region
Self-recognition. The imaging studies cited pitch, rhythm and timbre in music listening; of somatic association, with face representa-
in Supplementary information S1 (table) the authors reported robust activation spe- tions located most anterior and hand and
reported that the act of seeing ones own cifically in the left AIC (not the ACC) during foot representations located more posterior
image produced activation in the AIC. The the rhythm task (in which occasional notes (see also REF. 14).

60 | JANUARY 2009 | VOLUME 10 www.nature.com/reviews/neuro


Error awareness Feeling of knowing Subjective cooling Pleasant music

Attention to heat pain Rhythm

Free wont

Heartbeat awareness

Moment of recognition Maternal affiliation

Inspection time

Learned pain now

Happy voices

Decision making

Anterior insula
Self recognition

Seeing or making a smile

Anterior insula Time perception

Figure 2 | Activation of the AIC. A summary of imaging results showing activation of the left AIC was observed in subjects experiencing joy128; and
activation of the anterior insular cortex (AIC) during particular tasks and emo- selective activation of the left AIC in females was found that correlated with
tions highlighted in the text. The two right-hand columns contain images from self-reported orgasm ratings129. ACC, anterior cingulate cortex. Images are
different studies that found predominantly unilateral activation; images in the reproduced, with permission, from (in order from top to bottom and left to
two left-hand columns show mostly bilateral activation. Stimuli that activate right): (first column) REF. 23 (2007) Elsevier, REF. 45 (2007) Society for
the right AIC are generally arousing to the body (for example, pain). The left Neuroscience, REF. 39 (2007) Society for Neuroscience, REF. 40 (2007)
AIC is activated mainly by positive and affiliative emotional feelings (BOX 3). Society for Neuroscience; (second column), REF. 41 (2002) Cell Press, REF. 31
For example, activation of the left AIC was reported in mothers viewing pho- (2004) Elsevier, REF. 34 (2007) Elsevier; (third column) REF. 2 (2000)
tos of their own child125; greater activation of the left than of the right AIC was Macmillan Publishers Ltd (all rights reserved), REF. 7 (2002) Elsevier, REF. 3
associated with both maternal and romantic love126; activation of the left AIC (2004) Macmillan Publishers Ltd (all rights reserved), REF. 24 (2004)
was reported while subjects were either seeing or making a smile21; activation Macmillan Publishers Ltd (all rights reserved), REF. 15 (2007) Elsevier; (fourth
of the left AIC was found while subjects attended to happy voices127; activa- column) REF. 17 (2006) Wiley, REF. 16 (1997) Oxford Journals, REF. 125
tion of the left AIC was associated with hearing pleasant music17; selective (2004) Elsevier, REF. 127 (2006) Oxford Journals, REF. 20 (2005) Elsevier.



Box 1 | The co-activation of the AIC and the ACC

It is understandably mystifying that a region of a Cingulate sulcus
the ventrolateral prefrontal cortex (the anterior
insular cortex (AIC)) and a region of the medial
prefrontal cortex (the anterior cingulate cortex
(ACC)) are co-active in so many behaviours,
because such widely separated regions in the
cortex generally have distinct roles. The
insular cortex was long regarded simply as a
visceral sensory region, based on findings by
Penfield, Mesulam and Saper8486, whereas the
medial prefrontal cortex has been associated
with conflict responses, impulsive behaviour
and autonomic activity. In an earlier article I Hippocampus
suggested that the insula and the ACC be
regarded as limbic sensory and motor cortices Rhinal sulcus Collateral sulcus
that respectively engender the feeling and the
motivation (agency) that constitute any Laterobasal cortical amygdala
emotion1. This suggestion was based on the
dual lamina I spinothalamocortical projection
to both the insula and the ACC, the b
co-activation of these areas in virtually all
studies of emotion, their respective descending
projections to sensory (parabrachial nucleus)
and motor (periaqueductal grey) brainstem
regions, the overall anatomical organization of Olfactory
the frontal cortex into sensory and motor bulb
networks, and the evolutionarily ancient limbic
role of the cingulate cortex in integrated Olfactory
behavioural control. I regarded the AIC as the peduncle
probable site for awareness on the basis of its Insula
afferent representation of the feelings from Olfactory tract
the body, and the ACC as the probable site Olfactory cortex Anterior olfactory nucleus
for the initiation of behaviours. A recent Diagonal band
review87 offered support for this view and an Basolateral amygdala
explanation for the anatomical separation of
the insula and the ACC. The ACC evolved first Hippocampus
as a motor-control region aligned with the
sensory integration, in the hippocampus and Entorhinal area
the amygdala, of olfactory-guided group
behaviour in mammals. The insula evolved later
for cortical processing of homeostatic sensory
activity in the individual animal. The two
regions naturally became linked for integrative
autonomic control, and in mammalian evolution
the insula grew as limbic behavioural activity
became aligned more with autonomic activity
than with olfactory activity. This anatomical
perspective87 is illustrated in the ventral view of
the brain in the figure, which reveals the
common relationship of these structures to the
olfactory epithelium. Figure is reproduced, with
permission, from REF. 87 (1999) Elsevier.

Risk, uncertainty and anticipation. Of the and the risk-prediction error signals speci- bistable percepts must be explicitly noted
articles listed in Supplementary information fied in the model were present only in the here2729, but I highlight three additional
S1 (table), I note particularly one report that bilateral AIC25. No activation was observed in studies. The first is a PET study which found
emotional value in the immediate present, the ACC in this study. Activation in the AIC that the right AIC and the ACC are sensitive
as specified in a second-order temporal- is also correlated with feelings of anticipated to cross-modal sensory time synchroniza-
difference model of learning, correlated value during purchase and sales decisions26. tion and display a graded response to a
selectively with activity in the right AIC24; mismatch in timing between auditory and
another study (in which this learning model Visual and auditory awareness of the visual stimuli that should normally be syn-
was adapted for neuroeconomic studies) moment. Several reports that associated acti- chronous (for example, a speaking mouth)30.
reported that both the risk-prediction signals vation in the AIC with awareness of sensory The second is an fMRI study that examined

62 | JANUARY 2009 | VOLUME 10 www.nature.com/reviews/neuro


Box 2 | VENs and awareness example, see REF. 35), and some have argued
that attention and consciousness must be
An extraordinary morphological characteristic of the anterior insular cortex (AIC) and the anterior different processes36. Nevertheless, two
cingulate cortex (ACC) in hominoid primates is the unique concentration of clusters of large recent studies of attention described activa-
spindle-shaped neurons among the pyramidal neurons in layer 5, called von Economo neurons
tion in the AIC/IFG37,38. The authors of one
(VENs) after an early neuroanatomist88,89. Their connections are not known, but I have proposed78
study reported decreased activation in the
that VENs are the substrate for fast interconnections between the physically separated advanced
limbic sensory (the AIC) and motor (the ACC) cortices. Analogous to the tight interconnections right IFG, ACC and middle frontal gyrus
between the contiguous somatosensory and motor cortices (so-called U-fibres) needed for manual (DLPFC) just before lapses of attention dur-
dexterity (for example, for playing a musical instrument), the VENs might enable fast, highly ing a monotonous selective-attention task
integrated representations of emotional moments and behaviours. They may underlie the joint (in which lapses of attention were marked
activity in the AIC and the ACC reported in most studies. The loss of emotional awareness and by increased reaction times), and they
self-conscious behaviours in patients with frontotemporal dementia that correlates with the observed increased activation after such
degeneration of VENs directly supports this notion75,76,90,91. lapses in approximately the same regions
Hof, Allman and colleagues reported that many VENs are present in aged humans, but (bilateral AIC, ACC, DLPFC, occipital and
progressively fewer are found in infants, gorillas, bonobos and chimpanzees, and they are not
parietal visual regions), possibly correspond-
present at all in macaque monkeys88,89. This clear phylogenetic progression parallels the results of
ing to renewed attention37. They suggested
the mirror test for self-awareness2,55. VENs have recently been reported in elephants92 and
whales93. The possibility that this implies sentience in these animals received support recently that the AIC/IFG is involved in the stimulus-
when elephants were reported to pass the mirror test94. Interestingly, there are reports that triggered reorienting of attention and that
elephants sing or make music communally (as captured in a film on the National Geographic the ACC is involved in the detection and/or
website), which would be consistent with the present model for awareness, in which music (viewed resolution of processing conflicts; however,
as a rhythmic temporal progression of emotionally laden moments) is an emergent property of if the AIC/IFG and the ACC are regarded
awareness. By contrast, although most people feel that particular dogs and cats have some sense as complementary limbic sensory and
of awareness, the cortex of these animals receives an integrated brainstem homeostatic pathway motor regions, respectively (BOX 1), then this
and only a primordial homologue of the ascending lamina I pathway that underlies the emergence response profile would also be consistent
of interoceptive re-representations in the AIC of humanoid primates1. Whether that is a sufficient
with the interpretation that target awareness
basis for sentience is unresolved: cats and dogs do not pass the mirror test, and a lesion of the
is engendered in the AIC/IFG and control of
presumptive homologous pathway in cats does not produce the same devastating effects on
interoception that it produces in humans1. Rats and lizards apparently do not have a homologous directed effort is engendered in the ACC. In
anatomical substrate at all1. another study, a well-practised behavioural
task was used to encourage mind-wandering
(stimulus-independent thought), and the
inspection time using a briefly displayed processed input from a parietal network authors found activation associated with
asymmetric visual test stimulus (a one-sided and requires a finite period of time to self-reported mind-wandering periods
fork) followed immediately by an ambiguous transfer information into working memory in several regions, particularly the insula
stimulus (so-called backward masking)31. (in the DLPFC). (middle and posterior) and the ACC38.
The authors reported that subjects perform- They interpreted the insular activity with
ance in detecting the asymmetry decreased Time perception. Joint activation of the reference to interoception and emotional
progressively from 100% to chance levels for AIC and the ACC or nearby areas has been awareness.
presentation times shorter than 150 ms, reported in many studies of mental time-
yet activation in the AIC and the ACC keeping and interval estimation across the Perceptual decision making. I highlight
selectively and progressively increased with range of seconds to sub-seconds, but little three seminal reports in this category. First,
shorter presentation times. The authors explanation has been given for this activity Ploran et al.39 tracked brain activation in
inferred from these data an effort-related (for example, see REF. 33). In a recent report, subjects watching a screen on which an
process that guides goal-directed attention, task difficulty was manipulated in order image was slowly being revealed. They
which they related to psychometric intelli- to isolate time estimation from other task- found a gradual increase in activation in
gence; however, this observation can also be related cognitive demands34. The authors brain regions that are involved in object
interpreted as evidence for a role for the AIC found three small regions that seem to be identification, but a sudden burst of activ-
and ACC in heightened awareness of the crucial for time perception: one in the dorsal ity in the AIC and the ACC at the moment
immediate moment. The third study 32 used putamen (bilaterally), another in the left of recognition (that is, coinciding with the
the attentional-blink paradigm, in which a inferior parietal cortex and another at the awareness of the percept itself). Second,
second target cannot be perceived if it occurs junction of the AIC and the IFG bilaterally Thielscher and Pessoa40 examined a two-
too soon after a primary target in a rapid (with no ACC activation). They suggested choice perceptual task using a graded series
series of visual stimuli. The authors reported that the AIC/IFG focus must be of central of morphed emotional faces (an experimen-
joint activation of the AIC/IFG and the importance in time perception. tally generated bi-state percept). They found
ACC (plus activation of the DLPFC and an inverted-U-shaped correlation between
parietal regions) when the second target was Attention. Activation of the AIC (and the perceptual choice and both reaction time
correctly detected at the shortest intervals ACC) is reported by most studies of goal- (which represented the decision-making
(100200 ms) but not when it was not per- directed attention, but these studies often process) and bilateral AIC/IFG (R>L) and
ceived. They suggested that the AIC/IFG and lack comments regarding the role of this ACC activation. They concluded that the
the ACC might contain a conceptual short- activation. One influential model for the AIC may have been important in the
term memory buffer prone to decay and neural correlates of the executive control of actual generation of the perceptual choice.
replacement by other stimuli that receives attention does not include the insula (for Finally, Kikyo et al.41 examined the



Box 3 | Forebrain asymmetry of emotion first set consisted of the bilateral AIC and the
ACC, which they suggested act as a perform-
There are two sides of the brain, and so if there are separate time-based representations of the ance monitor. Another set of brain regions
sentient self that subserve awareness in the anterior insular cortex (AIC) (one in each side), how they identified consisted of the right AIC/IFG
are they coordinated to generate one unified self? The homeostatic model of awareness
and an MPFC region near the ACC, which
presented in this and prior articles1,78 suggests that there is an energy efficiency-optimal process
they associated with the evaluation of task
that is based on the coordinated opponency of the autonomic system that activity in the right
side of the forebrain is associated with energy expenditure, sympathetic activity, arousal, costs and the maintenance of task effort. In
withdrawal (aversive) behaviour and individual-oriented (survival) emotions, and activity in the the latter regions, activation declined just
left side is associated with energy nourishment, parasympathetic activity, relaxation, approach before an error occurred and increased after
(appetitive) behaviour and group-oriented (affiliative) emotions. This proposal is described in an error had occurred, similar to the pattern
detail in prior articles78,95. Briefly, the origin of this asymmetry can be related to the asymmetric of attention-related activity in the study of
autonomic innervation of the heart, and its evolutionary development would have been lapses of attention described above37. These
compelled by the need for energy optimization in the brain (which consumes 25% of the bodys results are therefore also consistent with the
energy). This model fits with an accumulating body of psychophysical literature which indicates alternative interpretation that the AIC activ-
that the left and right forebrain halves are differentially associated with positive and negative
ity represents awareness and that the ACC
affect96 and, importantly, with the anatomical and functional asymmetry in the homeostatic
activity represents the control of directed
afferent input to the insular cortex5,97,98 and in forebrain cardiac control99. It can explain why
under particular conditions the AIC is more active on one side (FIG. 2) but also why in most effort. A recent study performed several
conditions the two sides display joint activity, which mirrors the coordinated sympathetic and sophisticated connectivity and correlation
parasympathetic control of the heart. The model offers explanations for why positive emotions analyses of attentional transitions and con-
can reduce or block negative emotions (and vice versa), why the left (affiliative, vagal) side firmed that the AIC and the ACC act as a
controls deictic pointing and verbal communication, and how increased parasympathetic activity cognitive-control network and, further, that
(for example, activation of vagal afferents by gastric distension, slow breathing or elecrical the right AIC in particular plays a critical
stimulation) can reduce negative emotions (for example, pain). Given the many asymmetries in and causal role in switching between the
the activations of the AIC noted in this article, investigations addressing this model in split-brain central executive network and the
patients could be enlightening, particularly if performed with those rare patients in whom both
default-mode or self-reflective network48.
the corpus callosum and the anterior commissure are sectioned100,101.
Finally, a study by Klein et al.23 may pro-
vide direct evidence that the AIC engenders
awareness. The authors used an antisaccade
feeling of knowing, which is the subjective sensory processing 44. A study that used a task with a distractor (in which subjects were
sense of knowing a word before recalling it. stop-signal paradigm in which the subjects instructed to shift their gaze in the opposite
(The feeling of knowing has been inter- voluntarily initiated a hand movement but direction to a briefly displayed indicator but
preted as a subjective feeling that represents on signalled trials inhibited the movement were occasionally misdirected by an inter-
the retrievability, accessibility or familiarity at the last possible subjective instant found vening cue that indicated the other direc-
of the particular target word, which corre- that the bilateral AIC (L>R) and a small tion), which produced occasional erroneous
lates with the subsequent ease of recall41,42.) region in the MPFC near the ACC region saccades that the subjects were either aware
Using a well-tested set of general-knowledge were specifically involved in the active, of (and signalled with a button press) or
questions without prior exposure or prim- willed inhibition of a motor act (which the unaware of. The subjects error-awareness
ing, they found pronounced activation in authors called free wont, as opposed to reports were corroborated by slowed reac-
the bilateral AIC/IFG and the ACC that cor- free will)45. The authors associated the AIC tion times and improved performance (for
related parametrically with the strength of activation with the affective consequence of most subjects) on trials following aware
the feeling of knowing. The bilateral AIC/ cancelling a motor intention (frustration), errors but not following unaware errors.
IFG was not recruited during the successful and other investigators in similar studies The fMRI data on error trials revealed acti-
recall process itself, which implied to the associated such activation with autonomic vation of the bilateral AIC and three small
authors a particular role in meta-memory arousal46. However, an alternative inter- MPFC regions near the ACC. By contrasting
processing for this area. pretation is that the occurrence of the stop brain activity during aware errors with that
signal naturally elicited an immediately during unaware errors, the authors found
Cognitive control and performance moni- heightened awareness such heightened activation only in the left inferior AIC (simi-
toring. Several reports in this major field awareness would certainly have increased lar activation in the right AIC was just sub-
of study described strong activation in the with the rarity of the stop signal, as the threshold). Significantly, activation near the
AIC. One described functional-connectivity investigators found for the AIC46 and ACC was not associated with error aware-
analyses that were performed on brain acti- the MPFC region45. In another study, per- ness. The authors suggested that the activity
vation during a combined working-memory formance errors were examined in subjects of the AIC during aware errors might be
and target-switching task; a cognitive- engaged in a demanding repetitive task, explained by interoceptive awareness of
control network was identified that similar to that used in the lapse-of-attention greater autonomic responses to aware errors
includes the bilateral AIC and the ACC43. study described above37, and then (which have been documented previously),
In another report, brain activation was ana- independent component analyses and back- but they also recognized that the AIC activ-
lysed during various cognitive-control tasks ward deconvolutions were performed to ity and error awareness might precede the
and the authors concluded that the bilateral identify brain-activation patterns that pre- autonomic response23. They recommended
AIC and the ACC form a highly intercon- ceded errors47. The authors found four sets the use of electroencephalographic record-
nected core system for task-dependent of brain regions in which activity patterns ings to resolve this uncertainty, because the
control of goal-directed behaviour and predicted the commission of an error; the temporal resolution of fMRI is too low.

64 | JANUARY 2009 | VOLUME 10 www.nature.com/reviews/neuro


Box 4 | Alternative perspectives damage to right parietal cortex), an impair-

ment in the movement of ones limbs (ano-
There are many alternative views of the neural substrates that are involved in human awareness or sognosia for hemiplegia following damage to
consciousness. Two particular viewpoints suggest that the anterior cingulate cortex (ACC) has a the mid-insula) or ones entire autobiographi-
singular role in the representation of the self and emotional awareness of internal processes102,103,
cal history and yet maintain phenomenal
but neither has incorporated the anatomical perspective that the anterior insular cortex (AIC) and
awareness of feelings and existence (for
the ACC are complementary limbic regions or the evidence reviewed here.
Some investigators suggest that the posteromedial cortex (including the cingulate and the example, the patient with only visceral
precuneus) provides the basis for self-awareness (for example, see REF. 104). However, this area is feelings remaining and an intact anterior
part of the so-called default network, which seems to be involved in self-reflective episodic insula51, or patient R.B., who lives in a mov-
memory retrieval, and activity in this area is inversely correlated in functional-imaging studies with ing 40-second window of present time51,52). A
the activation in the AIC that is associated with awareness and task-related attention44,90. The further proposal is that a reflective awareness
default network also seems to be present in the macaque monkey105. of oneself across time that can compare the
There are numerous proposals for the neural basis of the various forms of the clinical effects of ones actions now, in the past and in
syndrome anosognosia (a lack of awareness of a functional impairment), but these are the future was required for the evolution of
complicated by the multiple levels of integration that are required for each perceptual
deliberate social signalling (intentional emo-
capacity and by confounding issues in the clinical documentation67,106. Recent clinical and
tional interaction between individuals)49,53.
anatomical correlations in patients with anosognosia for hemiplegia and hemianaesthesia have
focused on the insula66,67. Prior studies implicated the right inferior parietal cortex and the Some investigators view this capacity as a
angular gyrus, but recent authors agree with the idea that this region processes a higher-order level of awareness that might
representation of extrapersonal space that incorporates self-generated movements, and provide the introspective feeling of subjec-
that this module precedes the integration of information about the self in the AIC66,67,106, tivity 51,53 (see BOX 4 and further discussion
although this idea remains to be tested. below). The mirror test for self-recognition
Several authors suggest the existence of different levels of awareness or consciousness51,70,107. has been used as an operational test for self-
Although this view is not incompatible with the present hypothesis, because more than one awareness, and although objections to this
conceptual level of awareness might be instantiated in the AIC (including a comparator that could test have been raised, I concur with the view
feel like an observer), other investigators might regard it as unlikely, because they prefer the
that it has validity as a sufficient but not nec-
notion that consciousness depends on recurrent activity in a distributed network across the entire
essary criterion5457; that is, a feeling of owner-
brain (for example, see REF. 77). For example, one author suggested that the orbitofrontal cortex is
part of a global workspace that underlies consciousness108, although its role in hedonic valuation ship of and identification with movements
(especially for feeding behaviour) is well-developed in all mammals and it is not disproportionately and emotional gestures reflected in a mirror
enlarged in humans109. Nevertheless, the AIC certainly does not operate autonomously and the is possible only with a mental representation
evidence reviewed here indicates that it is involved in more than one functional network; further, of a sentient self. Recently, it has also been
there are probably numerous modules in the AIC, the complexity of which might increase from the proposed that awareness in humans could be
posterior to the anterior, that could accommodate higher levels of abstraction21,22,83,110. For measured with post-decision wagering 58.
example, although the hurt that is associated with social exclusion was reported to overlap with
the AIC region that is activated during physical pain, a close examination of the imaging data The implications of the imaging data.
reveals that in fact the activation lies considerably more anterior111.
The results highlighted in the categories
Interoception, Awareness of body move-
ment and Emotional awareness indicate
The role of the AIC What is awareness? Until we know more that the AIC provides a unique neural sub-
This brief review reveals that an astonish- about how brains work, only a working strate that instantiates all subjective feelings
ing number of recent studies from a broad definition is possible. I regard awareness as from the body and feelings of emotion in the
range of fields reported activation of the knowing that one exists (the feeling that I immediate present (now). The anatomical
AIC. These studies associate the AIC not am); an organism must be able to experience posterior-to-mid-to-anterior progression of
just with all subjective feelings but also its own existence as a sentient being before integration from the primary interoceptive
with attention, cognitive choices and it can experience the existence and salience representations to the middle integration
intentions, music, time perception and, of anything else in the environment. One zone to the ultimate representation of all
unmistakably, awareness of sensations and proposal holds that awareness of any object of ones feelings (that is, the sentient self)
movements2,3,1114, of visual and auditory requires, first, a mental representation of is now well documented and supports the
percepts2729,31,32, of the visual image of the oneself as a feeling (sentient) entity; second, proposition that subjective awareness is built
self 15, of the reliability of sensory images30 a mental representation of that object; and on homeostasis1,4,51,61. The noted studies on
and subjective expectations24,25, and of the third, a mental representation of the salient the perception of bistable percepts2729,
trustworthiness of other individuals (see interrelationship between oneself and that time synchronization30, inspection time31,
REF. 1 for further detail). In several key object in the immediate moment (now)49,50. the attentional blink32 and perceptual deci-
experiments, the AIC was activated without As in Damasios neural self (REF. 51), this sion making 39,40 all imply directly that the
apparent activation in the ACC1114,16,2325,34. formulation inherently creates a subjective AIC supports awareness of the immediate
No other region of the brain is activated in (personal) perspective that differentiates moment with a coherent representation of
all of these tasks, and the only feature that inner and outer realms, because the inner my feelings about that thing. The dem-
is common to all of these tasks is that they feelings that underlie ones representation of onstration that the AIC activates during
engage the awareness of the subjects. Thus, oneself as a sentient being are accessible only self-recognition satisfies the criterion of
in my opinion, the accumulated evidence from ones own brain51,52. In this view, one instantiation of self-awareness, equivalent to
compels the hypothesis that the AIC can lose the ability to perceive a portion of the mirror test 15, because it reveals a subjec-
engenders human awareness. extrapersonal space (for example, following tive feeling of enhanced emotional salience



in the representation of the sentient self Clinical observations. Clinical observa- demonstrating deficits in emotional
during self-identification. Further, the dem- tions offer corroborative evidence for the awareness in adult humans who have
onstration that AIC activation is correlated hypothesis that the AIC engenders human well-established behavioural patterns.
with the feeling of knowing (REF. 41) sug- awareness. Supplementary information S1
gests that the AIC also engenders awareness (table) lists clinical reports that indicate Summary. Thus, the available data provide
of feelings that are associated with mental involvement of the AIC and the mid-insula compelling support for the concept that the
constructs and operations59,60. with conditions that include anosognosia, AIC contains the anatomical substrate for
The highlighted findings in the catego- anergia, anxiety, alexithymia, depression, the evolved capacity of humans to be aware
ries Risk, uncertainty and anticipation, aphasia, amusia, ageusia, drug craving, eat- of themselves, others and the environment.
Time perception and Cognitive control ing disorders, conduct disorder, panic dis- In my opinion, these data suggest that the
and performance monitoring are consist- order, mood disorders, post-traumatic stress AIC uniquely fulfils the requirements to be
ent with the notion that the AIC contains a disorder, schizophrenia, SmithMagenis the neural correlate of awareness. The brain
representation of the sentient self not only in syndrome, cardiac arrhythmia, vertigo and is well organized into networks that distrib-
the immediate moment but at each moment frontotemporal dementia (FTD). Many of ute functionality across multiple sites, and
across a finite period of time. The AIC is a these reports are quite narrowly focused and the localization of awareness of feelings
central component of a neural substrate that do not describe comprehensive behavioural and existence in a single substrate might seem
represents the passage of time34, it is sensitive testing, whereas the evidence reviewed unlikely 67,77. Nevertheless, the evidence sug-
to time synchronization30 and it is uniquely above suggests that most patients with insu- gests that the AIC and the adjoining frontal
involved in automatic comparisons of feel- lar damage could have several significant operculum (on both the left and right sides)
ings in the present moment with those in neurological deficits. contains an ultimate representation of the
the past and the future25. These findings also Several clinical studies indicate that sentient self in humans (and perhaps homi-
indicate that the AIC incorporates a buffer, damage to or abnormal activation or noid primates, elephants and whales; BOX 2).
or comparator, that is used for such com- development of the AIC is associated with These data recommend a discussion of the
parisons. These observations suggest that the altered awareness. Two studies reported possibility that the AIC is a neural correlate of
AIC fulfils a requirement of the proposed49,53 that anosognosia for hemiplegia is specifi- consciousness (REF. 77) and of the question
evaluative and predictive role of awareness cally associated with damage in the right of how the AIC might engender awareness.
in the evolution of emotional communica- mid-insula66,67; adolescents with conduct
tion, and that it affords an integral mecha- disorder were found to have a significant A model for awareness in the AIC
nism for decision making. (Interestingly, decrease in grey matter volume in the In a recently published book chapter 78, I
a stable comparator might be perceived bilateral AIC that correlated with a lack of expanded on the ideas presented in the 2002
introspectively as an ephemeral observer, or empathy and with aggressive behaviour 68; Perspective article1 by outlining a theoreti-
a Cartesian theater, that nonetheless cannot patients with borderline personality dis- cal model for the structural instantiation
see itself, as described61 for consciousness.) order who are incapable of cooperating of awareness in the AIC. The evidence
Significantly, the evidence that the AIC is do not display the graded activation of the described above of a role for the AIC in
also associated with predictions of future AIC that is associated with awareness of awareness is consistent with this model,
feelings can explain its involvement in the social gestures in normal subjects69; high- and thus I elaborate the model here and
distorted interoceptive predictions that functioning people with autism displayed incorporate these recent findings.
are associated with anxiety and functional increased alexithymia and decreased empa- Briefly, in this model the cortical basis
somatic disorders62,63. thy, both of which were correlated with for awareness is an ordered set of repre-
The findings I highlighted on perceptual reduced activation in the AIC (in a task sentations of all feelings at each immediate
decision making (choice)40 and cognitive in which the subjects assessed their feel- moment extending across a finite period of
control (free wont (REF. 45), preparation for ings about unpleasant images)70 (however, time. The key to the cortical (that is, men-
error commission47 and attentional transi- see REFS 71,72); hyperactivity in the right tal) representation of the sentient self is the
tions48) imply that the AIC has a role in the AIC was selectively associated with anxi- integration of salience across all relevant
subjective guidance of mental and physical ety 62; infarcts of the AIC were reported to conditions at each moment. The salience of
behaviour. This inference is consistent with produce anergia, or complete listlessness73; any factor is determined by its significance
the introspective feeling that I am not sim- and congenital malformation of the bilat- for the maintenance and advancement of
ply a passive observer (whether this is an eral insula (SmithMagenis syndrome) the individual and the species. At the most
illusion or not; see REF. 64). Such involvement produces childhood mental retardation and fundamental level, this means the energy-
does not contradict the idea that the ACC the disruption of self-guided behaviour 74. efficient maintenance of the health of the
subserves volitional agency (BOX 1), an idea Most significantly, patients with FTD asso- physical body (and the brain) in other
that is supported by the ACCs joint (with the ciated with degenerate fronto-insular and words, homeostasis. In this view, the neural
AIC) activation in almost all of these studies cingulate cortices display a selective loss basis for awareness is the neural representa-
and by its association with action and the of self-conscious behaviours and a loss of tion of the physiological condition of the
descending limbic motor system1,65. emotional awareness of self and others75; body, and the homeostatic neural construct
Finally, the highlighted findings on the in fact, Seeley et al. reported the landmark for a feeling from the body is the
activation of the AIC during musical enjoy- finding that the loss of subjective emotional foundation for the encoding of all feelings1,78.
ment and heightened awareness point to awareness in patients with FTD is specifi- The phylogenetically new homeostatic
specific characteristics of the AIC that need cally associated with the degeneration of afferent pathway from lamina I and the
to be explained by any model of its role in VENs76. These reports are especially nota- solitary nucleus in primates provides
awareness. ble in light of the obvious difficulties in the basis for the sense of the physiological

66 | JANUARY 2009 | VOLUME 10 www.nature.com/reviews/neuro


condition of the entire body in the posterior global emotional moments might be engen- of VENs still need to be identified. I believe
insular cortex 1; this includes numerous dered by clusters of VENs in the AIC, and that each successive stage of integration in
individually mapped and distinct feelings that these are interconnected with similar this model could have provided an evolu-
from the body. These neural constructs are clusters of VENs in the ACC and probably tionary advantage, in that it would improve
then re-represented in the mid-insula and in the AIC and ACC on the opposite side emotional communication between con-
again in the AIC (on the left or right side or (BOX 1), but the connections and functions specifics crucial for hominoid primates80
both, depending on the source of the activ-
ity; BOX 3). The mid-insula integrates these
homeostatic re-representations with activity a
Posterior insula Anterior insula
that is associated with emotionally salient
environmental stimuli of many sensory
modalities, probably by way of input from
higher-order sensory regions, the temporal
pole and the amygdala. Recent functional
connectivity analyses indicate that the mid-
insula is also modulated directly by the
ventral striatum (the nucleus accumbens)79,
which provides an important incentive,
or hedonic, signal for the integration of Primary Homeostatic Environmental Hedonic conditions Motivational, social
interoceptive motor function conditions (nucleus accumbens, and cognitive
salience. Thus, this posterior-to-anterior representation (hypothalamus (entorhinal and and orbitofrontal conditions (ACC,
progression which is consistent with the and amygdala) temporal poles) cortex) VMPFC and DLPFC)
general processing gradient for increasing
complexity in the frontal cortex 22,65 and with b Normal time passage Future
the enormous expansion of the anterior
insula across hominoid primates provides
a substrate for the sequential integration of
homeostatic conditions with the sensory
environment and with motivational, hedonic
and social conditions represented in other
parts of the brain, and this substrate is con-
structed on the foundation provided by the
feelings from the body (FIG. 3a).
I propose that the integration of sali-
ence across all of these factors culminates
in a unified final meta-representation of the
global emotional moment near the junction
of the anterior insula and the frontal oper-
culum. This processing stage is key, because
it generates an image of the material me (or
the sentient self) at one moment in time Subjective time passage
now. An anatomical repetition of this fun-
damental unit, indexed by an endogenous
timebase, is all that is required to generate a
set of repeated meta-representations of glo-
bal emotional moments that extends across
a finite period of time, and this anatomical
structure (a meta-memory) provides the
When salient moments occur rapidly, the
basis for the continuity of subjective emo- number of global emotional moments
tional awareness in a finite present 78 (FIG. 3b). increases during that time and, as a consequence,
The recent data emphasize that storage buff- subjective time dilates
ers for individual global emotional moments Figure 3 | A proposed model of awareness. Cartoons illustrating features of the proposed structural
must be present to enable comparisons of model of awareness. a | The posited integration of salient activity, progressing from the posterior insula
past, present and future feelings; this would (left) to the anterior insula (right). The primary interoceptive representations of feelings from the body
instantiate a reflexive observer, as noted provide a somatotopic foundation that is anchored by the associated homeostatic effects on cardio-
above. The anticipatory global emotional respiratory function, as indicated by the focus of the colours in the chest. The integration successively
includes homeostatic, environmental, hedonic, motivational, social and cognitive activity to produce
moments must be influenced by stored rep-
a global emotional moment, which represents the sentient self at one moment of time. b | The top
resentations of expectations that are based cartoon shows how a series of global emotional moments can produce a cinemascopic image of the
on acquired internal models of ones own sentient self across time. The lower cartoon shows how the proposed model can produce a subjective
and others behaviour. A straightforward, dilation of time during a period of high emotional salience, when global emotional moments are rap-
although speculative, anatomical inference idly filled up. ACC, anterior cingulate cortex; DLPFC, dorsolateral prefrontal cortex; VMPFC,
of this model is that meta-representations of ventromedial prefrontal cortex.



Box 5 | Future research questions following from the model the JamesLange theory 51, namely that the
theory did not allow for feelings of internally
What are the anatomical connections of the anterior insular cortex (AIC)? Modern imaging studies generated emotion. In the model presented
using diffusion spectrum imaging and functional and effective connectivity analyses are needed in this article, all stimuli, incentives, inten-
that incorporate the AICs dynamic connectivity83. Comprehensive analysis of the hodology of the
tions and cognitions that have salience are
macaque insular cortex using classic tract-tracing methods is also needed.
represented by feelings, a crucial neuropsy-
How do the two sides of the AIC differ, and how do they interact? Does one side lead48 and the
other side monitor23,48 at different times? Direct statistical comparisons of laterality in activation chological construct composed of nested sets
are needed. Advanced analyses based on demonstrated asymmetries (FIG. 2) can be designed; for of integrative associations that are elaborated
example, one could examine changes in activation patterns and effective connectivity during the on an interoceptive template and endowed
modulation of pain by pleasant music112. with characteristic homeostatic sequelae
Is there one somatotopic map that provides a common-resource pool of global emotional (and thus, all feelings move the heart). In
moments, represents all feelings and is dynamically accessed at each moment? Or are there this model, feelings are the computational
numerous modules in the AIC that are coordinated to represent different feelings at each global common currency of awareness59,60.
emotional moment? For example, evidence suggests that the eyes and the face are represented in Finally, this model includes the possibility
the anterior ventral AIC and that the hand and the foot are represented more posteriorly, which is
that emotional behaviours can occur without
consistent with the presence of one map; however, the somatotopy in the one study that included
awareness (that is, by activation of the ACC
stimulation of multiple body parts is reversed10.
Similarly, is there only one module in each AIC that represents time, as suggested by the tiny without integration in the AIC), it implies
areas of activation in the AIC found in one study34? And does the activity in these areas correlate that animals without these structures are not
with subjective time dilation? Are these areas asymmetric on the two sides with respect to the aware in the same way that we are (BOX 2)
contrasting effects of arousal and pleasant mood on time perception113? and it provides a ready basis for the inclusion
How are comparisons made between feelings at different points in time? Where are the of a module that would assign responsibil-
intermediate meta-memory buffers and how are they interconnected? ity for the behavioural agents actions to the
The neuropsychological construct called a feeling is crucial. Why is it important that this sentient self 64.
construct has a homeostatic basis? What do feelings from the body and feelings about objects or It is important to note this models cur-
people or cognitions have in common? Do all feelings have autonomic sequelae that is, do they
rent limitations. It does not explain how
move the heart? Can every thought be regarded as a feeling? Are feelings indeed the common
a feeling is constructed, the nature of the
currency of awareness?
Does emotional congruence between individuals reflect homeostatic resonance? timebase and its relation to homeostatic
Do elephants really make infrasonic music together? What about whales? activity (such as ones heartbeat), the mecha-
Infants do not pass the mirror test or show self-conscious behaviours until they are ~18 months nism for shifting moments across time, the
old50,53. Does the connectivity of their AIC and anterior cingulate cortex (ACC) show a similar neural metric of salience, the differentiation
developmental threshold? of emotions, the necessary dynamic con-
The activation of the AIC or the ACC has been used for biofeedback training. Could this nectivity of global emotional moments83 or
application be refined in light of the considerations set out in this article? the integration between the two sides of the
AIC (BOX 3). It does not specify the location
of memories of past feelings or the loca-
and that each stage would have been capacity of the neural instantiation of a glo- tion of the internal behavioural models that
easily realizable by evolutionary bal emotional moment must be finite. Thus, produce anticipatory feelings. The evidence
neuroanatomical modifications. the endogenous timebase would effectively reviewed above suggests that the final repre-
speed up during such a period (FIG. 3b) and, sentation of the sentient self in the AIC may
Implications and future directions consequently, time would appear to stand consist of one coherent somatotopic map
It is fascinating that the proposed model for still to the observer. Psychophysical data that has sufficiently global characteristics
awareness in the AIC produces an emer- from rapid-visual-search studies suggest and dynamic connectivity to encompass
gent basis for the uniquely human faculty that the maximal rate of passage of indi- all possible feelings, but few studies have
of music: viewed as the rhythmic temporal vidual moments is ~8 Hz31. A similar pro- compared different emotions and different
progression of emotionally laden moments, cess for the recruitment of global emotional body-part associations. It is even difficult
music arises as a natural concomitant of this moments could provide the basis for height- to systematically predict how lesions of the
structural model. The engagement of the ened awareness of the immediate moment AIC and the ACC should differ in terms of
AIC by emotional feelings, by the feeling of and for the enhanced activation of the AIC their effects considering their intimate con-
movement and by the sense of time encour- that is associated with such moments. nectivity (for example, see the contrasting
ages the view that this model can explain the In this model, the close integration results in the literature on alexithymia7072). I
observed activation of the AIC by rhythm16 between the AIC and the ACC implies that look forward particularly to the future iden-
and by musical enjoyment17, as well as the activity in the AIC can incorporate the urges tification of the various functional modules
primal emotional effects of communal of the volitional agent that is represented in the AIC and the characteristics and the
music-making, because music would in the ACC, and also that feelings in the language of VENs. There are many other
inherently involve the core of awareness. AIC can be modulated by that agent 82, so questions that need to be addressed, some of
This model also suggests a mechanism that each global emotional moment com- which are listed in BOX 5.
for the subjective time dilation, or slowing of prises both feelings and motivations (BOX 1). A. D. (Bud) Craig is at the Atkinson Research
time, that occurs during an intensely emo- Indeed, the representation in the sentient Laboratory, Barrow Neurological Institute, Phoenix,
tional period81. A high rate of salience accu- self of the active behavioural agent (the I) Arizona 85013, USA.

mulation would fill up global emotional fills a gap in the somatic marker hypoth- e-mail: bcraig@chw.edu
moments quickly, because the information esis51 and challenges a main criticism of doi:10.1038/nrn2555

68 | JANUARY 2009 | VOLUME 10 www.nature.com/reviews/neuro


1. Craig, A. D. How do you feel? Interoception: the sense 30. Bushara, K. O., Grafman, J. & Hallett, M. Neural 59. Ruys, K. I. & Stapel, D. A. The secret life of emotions.
of the physiological condition of the body. Nature Rev. correlates of auditory-visual stimulus onset Psychol. Sci. 19, 385391 (2008).
Neurosci. 3, 655666 (2002). asynchrony detection. J. Neurosci. 21, 300304 60. Duncan, S. & Barrett, L. F. Affect is a form of cognition:
2. Craig, A. D., Chen, K., Bandy, D. & Reiman, E. M. (2001). a neurobiological analysis. Cogn. Emot. 21,
Thermosensory activation of insular cortex. Nature 31. Deary, I. J. et al. The functional anatomy of inspection 11841211 (2007).
Neurosci. 3, 184190 (2000). time: an event-related fMRI study. Neuroimage 22, 61. James, W. The Principles of Psychology. Classics in the
3. Critchley, H. D., Wiens, S., Rotshtein, P., Ohman, A. & 14661479 (2004). History of Psychology [online], http://psychclassics.
Dolan, R. J. Neural systems supporting interoceptive 32. Kranczioch, C., Debener, S., Schwarzbach, J., Goebel, yorku.ca/James/Principles/ (1890).
awareness. Nature Neurosci. 7, 189195 (2004). R. & Engel, A. K. Neural correlates of conscious 62. Paulus, M. P. & Stein, M. B. An insular view of anxiety.
4. Barrett, L. F., Quigley, K. S., Bliss-Moreau, E. & perception in the attentional blink. Neuroimage 24, Biol. Psychiatry 60, 383387 (2006).
Aronson, K. R. Interoceptive sensitivity and self- 704714 (2005). 63. Mayer, E. A., Naliboff, B. D. & Craig, A. D.
reports of emotional experience. J. Pers. Soc. Psychol. 33. Coull, J. T. fMRI studies of temporal attention: Neuroimaging of the brain-gut axis: from basic
87, 684697 (2004). allocating attention within, or towards, time. Brain understanding to treatment of functional GI disorders.
5. Stephan, E. et al. Functional neuroimaging of gastric Res. Cogn. Brain Res. 21, 216226 (2004). Gastroenterology 131, 19251942 (2006).
distention. J. Gastrointest. Surg. 7, 740749 (2003). 34. Livesey, A. C., Wall, M. B. & Smith, A. T. Time 64. Wegner, D. M. The Illusion of Conscious Will (MIT
6. Phillips, M. L. et al. The effect of negative emotional perception: manipulation of task difficulty dissociates Press, Cambridge, Massachusetts, 2008).
context on neural and behavioural responses to clock functions from other cognitive demands. 65. Amodio, D. M. & Frith, C. D. Meeting of minds: the
oesophageal stimulation. Brain 126, 669684 Neuropsychologia 45, 321331 (2007). medial frontal cortex and social cognition. Nature Rev.
(2003). 35. Fan, J. et al. The relation of brain oscillations to Neurosci. 7, 268277 (2006).
7. Brooks, J. C., Nurmikko, T. J., Bimson, W. E., Singh, attentional networks. J. Neurosci. 27, 61976206 66. Karnath, H. O., Baier, B. & Nagele, T. Awareness of the
K. D. & Roberts, N. fMRI of thermal pain: effects of (2007). functioning of ones own limbs mediated by the insular
stimulus laterality and attention. Neuroimage 15, 36. Koch, C. & Tsuchiya, N. Attention and consciousness: cortex? J. Neurosci. 25, 71347138 (2005).
293301 (2002). two distinct brain processes. Trends Cogn Sci. 11, 67. Spinazzola, L., Pia, L., Folegatti, A., Marchetti, C. &
8. Kong, J. et al. Using fMRI to dissociate sensory 1622 (2007). Berti, A. Modular structure of awareness for
encoding from cognitive evaluation of heat pain 37. Weissman, D. H., Roberts, K. C., Visscher, K. M. & sensorimotor disorders: evidence from anosognosia
intensity. Hum. Brain Mapp. 27, 715721 (2006). Woldorff, M. G. The neural bases of momentary for hemiplegia and anosognosia for hemianaesthesia.
9. Singer, T. et al. Empathy for pain involves the affective lapses in attention. Nature Neurosci. 9, 971978 Neuropsychologia 46, 915926 (2008).
but not sensory components of pain. Science 303, (2006). 68. Sterzer, P., Stadler, C., Poustka, F. & Kleinschmidt, A.
11571162 (2004). 38. Mason, M. F. et al. Wandering minds: the default A structural neural deficit in adolescents with conduct
10. Henderson, L. A., Gandevia, S. C. & Macefield, V. G. network and stimulus-independent thought. Science disorder and its association with lack of empathy.
Somatotopic organization of the processing of muscle 315, 393395 (2007). Neuroimage 37, 335342 (2007).
and cutaneous pain in the left and right insula cortex: 39. Ploran, E. J. et al. Evidence accumulation and the 69. King-Casas, B. et al. The rupture and repair of
a single-trial fMRI study. Pain 128, 2030 (2007). moment of recognition: dissociating perceptual cooperation in borderline personality disorder. Science
11. Farrer, C. & Frith, C. D. Experiencing oneself vs recognition processes using fMRI. J. Neurosci. 27, 321, 806810 (2008).
another person as being the cause of an action: the 1191211924 (2007). 70. Silani, G. et al. Levels of emotional awareness and
neural correlates of the experience of agency. 40. Thielscher, A. & Pessoa, L. Neural correlates of autism: an fMRI study. Soc. Neurosci. 3, 97112
Neuroimage 15, 596603 (2002). perceptual choice and decision making during fear- (2008).
12. Farrer, C. et al. Modulating the experience of agency: disgust discrimination. J. Neurosci. 27, 29082917 71. Moriguchi, Y. et al. Empathy and judging others pain:
a positron emission tomography study. Neuroimage (2007). an fMRI study of alexithymia. Cereb. Cortex 17,
18, 324333 (2003). 41. Kikyo, H., Ohki, K. & Miyashita, Y. Neural correlates 22232234 (2007).
13. Tsakiris, M., Hesse, M. D., Boy, C., Haggard, P. & Fink, for feeling-of-knowing: an fMRI parametric analysis. 72. Kano, M. et al. Specific brain processing of facial
G. R. Neural signatures of body ownership: a sensory Neuron 36, 177186 (2002). expressions in people with alexithymia: an H215O-PET
network for bodily self-consciousness. Cereb. Cortex 42. Yonelinas, A. P., Otten, L. J., Shaw, K. N. & Rugg, study. Brain 126, 14741484 (2003).
17, 22352244 (2007). M. D. Separating the brain regions involved in 73. Manes, F., Paradiso, S. & Robinson, R. G.
14. Mutschler, I. et al. A rapid sound-action association effect recollection and familiarity in recognition memory. Neuropsychiatric effects of insular stroke. J. Nerv.
in human insular cortex. PLoS ONE 2, e259 (2007). J. Neurosci. 25, 30023008 (2005). Ment. Dis. 187, 707712 (1999).
15. Devue, C. et al. Here I am: the cortical correlates of 43. Cole, M. W. & Schneider, W. The cognitive control 74. Boddaert, N. et al. Anatomical and functional brain
visual self-recognition. Brain Res. 1143, 169182 network: integrated cortical regions with dissociable imaging evidence of lenticulo-insular anomalies in
(2007). functions. Neuroimage 37, 343360 (2007). Smith Magenis syndrome. Neuroimage 21,
16. Platel, H. et al. The structural components of music 44. Dosenbach, N. U. et al. Distinct brain networks for 10211025 (2004).
perception. A functional anatomical study. Brain 120, adaptive and stable task control in humans. Proc. Natl 75. Sturm, V. E., Rosen, H. J., Allison, S., Miller, B. L. &
229243 (1997). Acad. Sci. USA 104, 1107311078 (2007). Levenson, R. W. Self-conscious emotion deficits in
17. Koelsch, S., Fritz, T., v. Cramon, D. Y., Muller, K. & 45. Brass, M. & Haggard, P. To do or not to do: the neural frontotemporal lobar degeneration. Brain 129,
Friederici, A. D. Investigating emotion with music: an signature of self-control. J. Neurosci. 27, 91419145 25082516 (2006).
fMRI study. Hum. Brain Mapp. 27, 239250 (2006). (2007). 76. Seeley, W. W. et al. Early frontotemporal dementia
18. Ackermann, H. & Riecker, A. The contribution of the 46. Ramautar, J. R., Slagter, H. A., Kok, A. & Ridderinkhof, targets neurons unique to apes and humans. Ann.
insula to motor aspects of speech production: a review K. R. Probability effects in the stop-signal paradigm: Neurol. 60, 660667 (2006).
and a hypothesis. Brain Lang. 89, 320328 (2004). the insula and the significance of failed inhibition. 77. Christensen, M. S., Ramsoy, T. Z., Lund, T. E., Madsen,
19. Seeley, W. W. et al. Dissociable intrinsic connectivity Brain Res. 1105, 143154 (2006). K. H. & Rowe, J. B. An fMRI study of the neural
networks for salience processing and executive 47. Eichele, T. et al. Prediction of human errors by correlates of graded visual perception. Neuroimage
control. J. Neurosci. 27, 23492356 (2007). maladaptive changes in event-related brain networks. 31, 17111725 (2006).
20. Hennenlotter, A. et al. A common neural basis for Proc. Natl Acad. Sci. USA 105, 61736178 (2008). 78. Craig, A. D. in Handbook of Emotions 3rd edn (eds
receptive and expressive communication of pleasant 48. Sridharan, D., Levitin, D. J. & Menon, V. A critical role Lewis, M., Haviland-Jones, J. M. & Barrett, L. F.)
facial affect. Neuroimage 26, 581591 (2005). for the right fronto-insular cortex in switching between 272288 (Guilford, New York, 2008).
21. Jabbi, M., Swart, M. & Keysers, C. Empathy for central-executive and default-mode networks. Proc. 79. Menon, V. & Levitin, D. J. The rewards of music
positive and negative emotions in the gustatory Natl Acad. Sci. USA 105, 1256912574 (2008). listening: response and physiological connectivity of
cortex. Neuroimage 34, 17441753 (2007). 49. Frith, C., Perry, R. & Lumer, E. The neural correlates of the mesolimbic system. Neuroimage 28, 175184
22. Koechlin, E. & Jubault, T. Brocas area and the conscious experience: an experimental framework. (2005).
hierarchical organization of human behavior. Neuron Trends Cogn. Sci. 3, 105114 (1999). 80. Dunbar, R. I. & Shultz, S. Evolution in the social brain.
50, 963974 (2006). 50. Lewis, M. in Handbook of Emotions 3rd edn (eds Science 317, 13441347 (2007).
23. Klein, T. A. et al. Neural correlates of error awareness. Lewis, M., Haviland-Jones, J. M. & Barrett, L. F.) 81. Tse, P. U., Intriligator, J., Rivest, J. & Cavanagh, P.
Neuroimage 34, 17741781 (2007). 304319 (Guilford, New York, 2008). Attention and the subjective expansion of time.
24. Seymour, B. et al. Temporal difference models describe 51. Damasio, A. R. Descartes Error: Emotion, Reason, Percept. Psychophys. 66, 11711189 (2004).
higher-order learning in humans. Nature 429, and the Human Brain (Putnam, New York, 1993). 82. Sarinopoulos, I., Dixon, G. E., Short, S. J., Davidson,
664667 (2004). 52. Churchland, P. S. Self-representation in nervous R. J. & Nitschke, J. B. Brain mechanisms of
25. Preuschoff, K., Quartz, S. R. & Bossaerts, P. Human systems. Science 296, 308310 (2002). expectation associated with insula and amygdala
insula activation reflects risk prediction errors as well 53. Frith, C. D. & Frith, U. Social cognition in humans. response to aversive taste: implications for placebo.
as risk. J. Neurosci. 28, 27452752 (2008). Curr. Biol. 17, R724R732 (2007). Brain Behav. Immun. 20, 120132 (2006).
26. Knutson, B., Rick, S., Wimmer, G. E., Prelec, D. & 54. Macphail, E. M. The Evolution of Consciousness 83. Jabbi, M., Bastiaansen, J. & Keysers, C. A common
Loewenstein, G. Neural predictors of purchases. (Oxford Univ. Press, Oxford, 1998). anterior insula representation of disgust observation,
Neuron 53, 147156 (2007). 55. de Waal, F. B. The thief in the mirror. PLoS Biol.6, experience and imagination shows divergent functional
27. Kondo, H. M. & Kashino, M. Neural mechanisms of e201 (2008). connectivity pathways. PLoS ONE 3, e2939 (2008).
auditory awareness underlying verbal transformations. 56. Reiss, D. & Marino, L. Mirror self-recognition in the 84. Penfield, W. & Faulk, M. E. The insula. Brain 78,
Neuroimage 36, 123130 (2007). bottlenose dolphin: a case of cognitive convergence. 30470 (1955).
28. Pressnitzer, D. & Hupe, J. M. Temporal dynamics of Proc. Natl Acad. Sci. USA 98, 59375942 (2001). 85. Mesulam, M.-M. & Mufson, E. J. Insula of the old world
auditory and visual bistability reveal common 57. Prior, H., Schwarz, A. & Gunturkun, O. Mirror-induced monkey. III: Efferent cortical output and comments on
principles of perceptual organization. Curr. Biol. 16, behavior in the magpie (Pica pica): evidence of self- function. J. Comp. Neurol. 212, 3852 (1982).
13511357 (2006). recognition. PLoS Biol.6, e202 (2008). 86. Saper, C. B. The central autonomic nervous system:
29. Sterzer, P. & Kleinschmidt, A. A neural basis for 58. Persaud, N., McLeod, P. & Cowey, A. Post-decision conscious visceral perception and autonomic pattern
inference in perceptual ambiguity. Proc. Natl Acad. wagering objectively measures awareness. Nature generation. Annu. Rev. Neurosci. 25, 433469
Sci. USA 104, 323328 (2007). Neurosci. 10, 257261 (2007). (2002).



87. Heimer, L. & Van Hoesen, G. W. The limbic lobe and its 113. Droit-Volet, S. & Meck, W. H. How emotions colour our 125. Leibenluft, E., Gobbini, M. I., Harrison, T. & Haxby,
output channels: implications for emotional functions perception of time. Trends Cogn. Sci. 11, 504513 J. V. Mothers neural activation in response to pictures
and adaptive behavior. Neurosci. Biobehav. Rev. 30, (2007). of their children and other children. Biol. Psychiatry
126147 (2006). 114. Naidich, T. P. et al. The insula: anatomic study and MR 56, 225232 (2004).
88. Nimchinsky, E. A. et al. A neuronal morphologic type imaging display at 1.5 T. AJNR Am. J. Neuroradiol. 126. Bartels, A. & Zeki, S. The neural correlates of maternal
unique to humans and great apes. Proc. Natl Acad. 25, 222232 (2004). and romantic love. Neuroimage 21, 11551166
Sci. USA 96, 52685273 (1999). 115. Augustine, J. R. The insular lobe in primates including (2004).
89. Allman, J. M., Watson, K. K., Tetreault, N. A. & humans. Neurol. Res. 7, 210 (1985). 127. Johnstone, T., van Reekum, C. M., Oakes, T. R. &
Hakeem, A. Y. Intuition and autism: a possible role for 116. Augustine, J. R. Circuitry and functional aspects of the Davidson, R. J. The voice of emotion: an fMRI study of
Von Economo neurons. Trends Cogn. Sci. 9, 367373 insular lobe in primates including humans. Brain Res. neural responses to angry and happy vocal expressions.
(2005). Brain Res. Rev. 22, 229244 (1996). Soc. Cogn. Affect. Neurosci. 1, 242249 (2006).
90. Seeley, W. W. et al. Dissociable intrinsic connectivity 117. Mufson, E. J., Sobreviela, T. & Kordower, J. H. in 128. Takahashi, H. et al. Brain activations during judgments
networks for salience processing and executive Handbook of Chemical Neuroanatomy, Vol. 13: The of positive self-conscious emotion and positive basic
control. J. Neurosci. 27, 23492356 (2007). Primate Nervous System, Part I. (eds Bloom, F. E., emotion: pride and joy. Cereb. Cortex 18, 898903
91. Seeley, W. W. et al. Divergent social functioning in Bjorklund, A. & Hokfelt, T.) 377454 (Elsevier (2008).
behavioral variant frontotemporal dementia and Science, 1997). 129. Ortigue, S., Grafton, S. T. & Bianchi-Demicheli, F.
Alzheimer disease: reciprocal networks and neuronal 118. Ture, U., Yasargil, D. C., Al Mefty, O. & Yasargil, M. G. Correlation between insula activation and self-
evolution. Alzheimer Dis. Assoc. Disord. 21, S50S57 Topographic anatomy of the insular region. reported quality of orgasm in women. Neuroimage 37,
(2007). J. Neurosurg. 90, 720733 (1999). 551560 (2007).
92. Hakeem, A. Y. et al. Von Economo neurons in the 119. Olausson, H. et al. Unmyelinated tactile afferents
elephant brain. Anat. Rec. (in the press). signal touch and project to insular cortex. Nature Acknowledgements
93. Hof, P. R. & Van der Gucht, E. Structure of the cerebral Neurosci. 5, 900904 (2002). I thank J. Allman, L. Feldman Barrett, P. Churchland, M.
cortex of the humpback whale, Megaptera 120. Olausson, H. et al. Feelings of warmth correlate with Paulus and I. Strigo for their comments on the manuscript,
novaeangliae (Cetacea, Mysticeti, Balaenopteridae). neural activity in right anterior insular cortex. and Professor T. P. Naidich for supplying the photograph in
Anat. Rec. (Hoboken) 290, 131 (2007). Neurosci. Lett. 389, 15 (2005). figure 1. I am grateful for funds from the James S. McDonnell
94. Plotnik, J. M., de Waal, F. B. & Reiss, D. Self- 121. Small, D. M., Zatorre, R. J., Dagher, A., Evans, A. C. & Foundation and for continuous support from the Barrow
recognition in an Asian elephant. Proc. Natl Acad. Sci. Jones-Gotman, M. Changes in brain activity related to Neurological Foundation.
USA 103, 1705317057 (2006). eating chocolate: from pleasure to aversion. Brain
95. Craig, A. D. Forebrain emotional asymmetry: a 124, 17201733 (2001).
neuroanatomical basis? Trends Cogn. Sci. 9, 566571 122. Veldhuizen, M. G., Bender, G., Constable, R. T. & FURTHER INFORMATION
(2005). Small, D. M. Trying to detect taste in a tasteless A. D. (Bud) Craigs homepage: http://www.thebarrow.org/
96. Davidson, R. J. Well-being and affective style: neural solution: modulation of early gustatory cortex by Research_and_Clinical_Trials/Basic_Research_
substrates and biobehavioural correlates. attention to taste. Chem. Senses 32, 569581 Laboratories/Atkinson_Pain_Research_Lab/index.htm
Philos. Trans. R. Soc. Lond. B Biol. Sci. 359, (2007). National Geographic film on elephant song: http://ngm.
13951411 (2004). 123. Frank, G. K. et al. Sucrose activates human taste nationalgeographic.com/ngm/0510/feature5/audio.html
97. Rosenkranz, M. A. et al. Neural circuitry underlying pathways differently from artificial sweetener.
the interaction between emotion and asthma Neuroimage 39, 15591569 (2008).
See online article: S1 (table)
symptom exacerbation. Proc. Natl Acad. Sci. USA 124. Wang, G. J. et al. Gastric distention activates satiety
102, 1331913324 (2005). circuitry in the human brain. Neuroimage 39, ALL LINKS ARE ACTIVE IN THE ONLINE PDF
98. Henderson, L. A. et al. Neural responses during 18241831 (2008).
Valsalva maneuvers in obstructive sleep apnea
syndrome. J. Appl. Physiol. 94, 10631074 (2003).
99. Oppenheimer, S. M., Gelb, A., Girvin, J. P. &
Hachinski, V. C. Cardiovascular effects of human insular
cortex stimulation. Neurology 42, 17271732
100. Gazzaniga, M. S. Forty-five years of split-brain OPINION
research and still going strong. Nature Rev. Neurosci.
6, 653659 (2005).
101. Uddin, L. Q., Rayman, J. & Zaidel, E. Split-brain
reveals separate but equal self-recognition in the two
cerebral hemispheres. Conscious. Cogn. 14, 633640
Brain banking: opportunities,
102. Lane, R. D. Neural substrates of implicit and explicit
emotional processes: a unifying framework for
challenges and meaning for the future
psychosomatic medicine. Psychosom. Med. 70,
214231 (2008).
103. Lieberman, M. D. Social cognitive neuroscience: a Hans Kretzschmar
review of core processes. Annu. Rev. Psychol. 58,
259289 (2007). Abstract | Brain banks collect post-mortem human brains to foster research into
104. Parvizi, J., Van Hoesen, G. W., Buckwalter, J. &
Damasio, A. Neural connections of the posteromedial human CNS function and disease. They have been indispensable for uncovering the
cortex in the macaque. Proc. Natl Acad. Sci. USA 103,
15631568 (2006). secrets of many diseases, including Alzheimers and Parkinsons. At a time when
105. Vincent, J. L. et al. Intrinsic functional architecture in
the anaesthetized monkey brain. Nature 447, 8386
there are so many open questions in neuroscience and the incidence of brain
106. Orfei, M. D. et al. Anosognosia for hemiplegia after
diseases continues to increase in parallel with the aging of the population,
stroke is a multifaceted phenomenon: a systematic brain banking remains at the heart of brain research. However, the major source
review of the literature. Brain 130, 30753090
(2007). of brain banks, the clinical autopsy, is rapidly falling into limbo. New strategies,
107. Gray, M. A., Harrison, N. A., Wiens, S. & Critchley, H.
D. Modulation of emotional appraisal by false including donor programmes, medico-legal autopsies and banking in networks, as
physiological feedback during fMRI. PLoS ONE 2,
e546 (2007).
well as fresh considerations of the ethics and public relations, are required.
108. Kringelbach, M. L. The human orbitofrontal cortex:
linking reward to hedonic experience. Nature Rev.
Neurosci. 6, 691702 (2005).
When, in 1906, Alois Alzheimer first brain tissue that had been neuropathologi-
109. Semendeferi, K. & Damasio, H. The brain and its main observed the characteristics of the disease cally well characterized and frozen unfixed
anatomical subdivisions in living hominoids using
magnetic resonance imaging. J. Hum. Evol. 38,
that was to bear his name, he did so using after autopsy 2,3. A spate of biochemical and
317332 (2000). silver-impregnated tissue sections of a molecular-biological work led to the dis-
110. Schweinhardt, P. et al. An fMRI study of cerebral
processing of brush-evoked allodynia in neuropathic
formalin-fixed brain1. He described argyro- covery of amyloid precursor protein (APP),
pain patients. Neuroimage 32, 256265 (2006). philic senile plaques and neurofibrillary tan- APP mutations associated with familial
111. Eisenberger, N. I., Lieberman, M. D. & Williams, K. D.
Does rejection hurt? An fMRI study of social exclusion.
gles (NFTs). The biochemical nature of these Alzheimers disease (AD), and the metabo-
Science 302, 290292 (2003). structures remained enigmatic for decades, lism of APP, as well as to the creation of
112. Roy, M., Peretz, I. & Rainville, P. Emotional valence
contributes to music-induced analgesia. Pain 134,
until amyloid-B (AB), the major component transgenic animal models for AD and the
140147 (2008). of senile plaques, was isolated from human development of therapeutic strategies. Over

70 | JANUARY 2009 | VOLUME 10 www.nature.com/reviews/neuro