There has been a long controversy concerning whether the amygdala’s response to emotional stimuli is automatic or dependent on
attentional load. Using magnoencephalography and an advanced beamformer source localization technique, we found that amygdala
automaticity was a function of time: while early amygdala responding to emotional stimuli (40–140 ms) was unaffected by attentional
the amygdala response to emotional stimuli. Rodent data indi-
cate that the amygdala responds to emotional stimulation
through a fast but coarse subcortical (thalamus–amygdala) pro-
cessing stream in parallel to a slower, more detail-oriented, cor-
proposed that these dual routes are present in humans too, and
because of the existence of the subcortical route, the amygdala’s
response to emotional stimuli is largely automatic and can occur
2003; Whalen et al., 2004; for review, see Dolan and Vuilleumier,
2003; Johnson, 2005). Others have suggested that it is unlikely
resentational competition and that increased attention to non-
emotional stimulus features should significantly reduce the
2002; Bishop et al., 2007; Blair et al., 2007; Mitchell et al., 2008;
Pessoa, 2009; for review, see Pessoa and Ungerleider, 2004). It is
possible, however, that this debate has been fostered by the lim-
ited temporal resolution of blood oxygenation level-dependent
(BOLD) functional magnetic resonance imaging (fMRI) data.
stimuli, possibly via the subcortical route, that is contaminated
by a slower, attentionally modulated response. If this is true, it
can be predicted that there is an early amygdala response that is
response to the same emotional stimuli that is subject to signifi-
cant attentional modulation.
To test this hypothesis, it is necessary to use techniques with
good temporal and spatial resolution. Magnetoencephalography
(MEG), combined with the advanced source analysis technique
synthetic aperture magnetometry (SAM) and a sliding window
analysis, is ideal for this purpose (for details, see supplemental
Methods, available at www.jneurosci.org as supplemental mate-
identify activity in subcortical nuclei (Vrba and Robinson, 2001,
2002), previous work from a variety of laboratories has demon-
strated the ability of MEG to detect a signal from deep structures
such as hippocampus (Rogers et al., 1991; Ioannides et al., 1995)
and amygdala (Ioannides et al., 1995; Streit et al., 2003) even
when using evoked field methods. Moreover, SAM uses the
second-order covariance between channels rather than single-
channel averages and thus is sensitive to spatially correlated ac-
and Robinson, 2001, 2002). Indeed, recent studies using SAM
2007; Luo et al., 2007, 2009).
We focused on gamma band oscillations as they are consid-
ered to be of particular importance for cognition (Singer, 1999;
(Keil et al., 2001; Oya et al., 2002; Luo et al., 2007; Desmond et al.,
2008) and have a demonstrable relationship with the BOLD re-
sponse (Logothetis et al. 2001; Niessing et al., 2005). An adapted
was used in the present study.
Sixteen volunteers (eight males) aged between 22 and 41 years partici-
pated. All gave written informed consent and the study was approved by
the National Institute of Mental Health Institutional Review Board.
This paradigm, adapted from Erthal et al. (2005), involved a 2 (task
load: high, low) ? 2 (distracter: fearful, neutral) design. Each trial in-
volved the participant responding to a stimulus array containing a face
(fearful or neutral) bracketed between two lines by judging, via button
press, whether the lines were parallel or not. Task load was manipulated
(52 trials in each of the four conditions). The face stimuli (50% male)
org/resource.htm). We cut away the hair and transformed the photos
ms blank screen before the response window was presented for 1500 ms.
Mental Health, National Institutes of Health, 15K North Drive, Mail Stop Code 2670, Bethesda, MD 20892-2670.
TheJournalofNeuroscience,April28,2010 • 30(17):5825–5829 • 5825
They responded by pressing either the left (S:
same) or the right (D: different) button after
seeing the response cue (S D). The response
window was followed by a second blank
within Presentation software (Neurobehav-
channel CTF whole-head MEG system in a
shielded environment. The CTF MEG system
is equipped with synthetic third gradient bal-
ancing, an active noise cancellation technique
that uses a set of reference channels to subtract
background interference. The resulting noise
floor is on the order of 5–7 fT above 1 Hz. At
the beginning and end of each measurement,
the participant’s head position was registered
with localization coils that were placed at the
nasion and the bilateral preauricular points. It
was required that head movements did not ex-
MEG data could be superimposed on the individual anatomical images
with an accuracy of a few millimeters. High-resolution anatomical MRI
images were acquired using a T1-weighted, three-dimensional, spoiled
GRASS (gradient-recalled acquisition in steady state) imaging sequence
(1 ? 1 ? 1.5 mm3) with a 1.5 tesla GE scanner.
Preprocessing. The VSM/CTF software and software developed at the
National Institute of Mental Health (NIMH) MEG Core Facility
(Bethesda, MD) together with AFNI (http://afni.nimh.nih.gov/afni/)
were used for data processing. Before doing SAM analysis, the data were
marked according to the four trial types. A multisphere head model was
created for each participant based the anatomical image of each partici-
that in the former each sphere (one per MEG sensor) is fitted to a small
patch of the head model (directly under the sensor) to better model the
local return currents.
Time–frequency analysis. Time–frequency results provide a view of
neuromagnetic signals represented over frequency across time. The re-
sults would help us to determine the specific frequency range of our
software developed by the NIMH MEG Core Facility (http://kurage.
nimh.nih.gov/meglab/Meg/Ctf2st) was adopted. Ctf2st performs Stock-
well time–frequency analysis within Matlab. To be consistent with later
SAM analyses, the control window was ?150 to 0 ms, and the active
window was 0 to 500 ms. This was done on the averaged data on all the
participants. The group time–frequency results were then contrasted
with zero and thresholded at p ? 0.05.
Sliding window SAM analysis. In the present analysis, SAM was then
used to analyze task-related activation differences in the gamma fre-
poral development of the brain’s activity, a sliding window analysis was
used in combination with SAM. With a window length of 150 ms and a
step of 10 ms, we estimated the signal power in each voxel by using
ms, etc. The dual-state SAM output was the contrast between the active
state and the control state. With sliding window SAM, we could obtain
information regarding when significant gamma band oscillations
emerged, peaked, and offset. For example, if significant gamma band
seen in the “?100 to 50 ms” window, then we could infer that the onset
dual-state SAM imaging analyses were performed with a spatial resolu-
tion of 7 mm. The output results were then concatenated, enabling us to
obtain a time course in combination with spatial activation maps across
all the time points starting from 150 ms before the stimulus to 500 ms
for analysis. The high-performance computational capabilities of the
NIH Biowulf PC/Linux cluster (http://biowulf.nih.gov) was used to per-
form the above computation-intensive tasks.
participants were also normalized (transformed to z-score) and regis-
for each of the 50 time windows was performed using a random effects
2 ? 2 ANOVA model in AFNI, which generated the gamma band oscil-
lation results. Oscillations in the gamma band of p ? 0.005 were consid-
ered statistically significant.
A 2 ? 2 ANOVA was conducted on the response time (RT) and
error rate data. This revealed a significant effect of task load
Fearful expression, high task load; NH. neutral expression, high task load; FL, fearful expression, low task load; NL, neutral
Time–frequency results. Time–frequency change from 0 to100 Hz (the y-axis)
5826 • J.Neurosci.,April28,2010 • 30(17):5825–5829Luoetal.•EmotionalAutomaticity
[F(1,15)? 16.103, p ? 0.001]; the participants showed signifi-
cantly longer RTs for the high-load relative to the low-load trials
(423 and 370 ms, respectively). There was no significant effect of
The results on error rates were similar: a significant effect of task
load [F(1, 15)? 25.340, p ? 0.001], more errors for the high-load
condition relative to the low-load condition (17 and 1%, respec-
0.000, p ? 1) or distracter emotionality by task load interaction
[F(1, 15)? 0.044, p ? 0.837].
The result showed that within the gamma band there was strong
power at 30–50 Hz (Fig. 2). This range was therefore used in the
are to date the most established frequencies associated with cog-
nition (Singer, 1999; Varela et al., 2001). (2) Our previous MEG
et al., 2007, 2009).
to emotional stimuli is under different levels of attentional con-
participants showed an early amygdala response to emotional
stimuli that is independent of attention, i.e., that is not modu-
lated by task load. Consistent with our hypothesis, a significant
main effect of distracter was seen within the left amygdala; in-
creased gamma band activity was seen in response to fearful rel-
ative to neutral expressions very rapidly after stimulus onset
(30–40 ms) and lasted until 50–60 ms ( p ? 0.005; at a more
lenient threshold of p ? 0.05 this effect lasted until 140 ms).
There was no significant task load or task load-by-distracter in-
teraction within the amygdala for this time period; i.e., task load
did not modulate the early amygdala response (Fig. 3a,c).
Our second goal was to determine whether later amygdala
responding is modulated by task load. Consistent with our hy-
within the right amygdala starting at 280–290 ms and lasting
until 330–340 ms ( p ? 0.005; at p ? 0.05 this effect lasted until
410 ms). Notably this interaction became significant after a sig-
nificant main effect of task load was seen within regions impli-
cated in top-down attentional control (Desimone and Duncan,
high-load conditions, there was no significant effect of distracter
emotionality on the amygdala response. In contrast, under
low-load conditions, there was significantly greater gamma
band activity within the amygdala to fearful relative to neutral
expressions ( p ? 0.005) (Fig. 3b,d).
Our results demonstrated an early (40–140 ms) amygdala re-
sponse to emotional information that was independent of atten-
tional modulation and a later (290–410 ms) amygdala response
that showed significant attentional modulation. In short, these
data indicate that the degree of automaticity of the amygdala
response is a function of time.
amygdala in response to emotional expressions (Luo et al., 2007,
Luoetal.•EmotionalAutomaticityJ.Neurosci.,April28,2010 • 30(17):5825–5829 • 5827
2009), and work with rodents has indicated that auditory fearful
signals can reach the amygdala at a short latency of 12 ms (Quirk
et al., 1995; LeDoux, 1998). The current data suggest that this
early amygdala response is not modulated by attentional load
frontal and parietal cortices. Later amygdala activity, which oc-
curred after task related activity within these regions, showed a
significant emotion-by-task load interaction. We hypothesize
that this interaction reflects the impact of top-down attentional
mechanisms. Specifically, stronger representation of task-relevant
information during high-load trials via top-down attention
should result in reduced representation of emotional distracter
information caused by representational competition (Desimone
and Duncan, 1995; Kastner and Ungerleider, 2000; Marois et al.,
2004). We believe that the limited temporal resolution of BOLD
data results in the contamination of the rapid, automatic amyg-
dala response to emotional stimuli by the slower, attentionally
dual-route model to amygdala activation (LeDoux, 2000; de
Gelder et al., 2003; Dolan and Vuilleumier, 2003; Ohman et al.,
2007). Considerable animal work has demonstrated that condi-
tioned stimulus (CS) inputs can reach the amygdala via this
over, the human amygdala has been reported to show activity
changes during conditioning to a masked CS that correlate with
activity in the thalamus but not the cortex (Morris et al., 1999).
Early findings of automatic amygdala activity that was not re-
duced by reduced attention were thought to reflect activity via
this subcortical route (Vuilleumier et al., 2001; Anderson et al.,
tion to nonemotional stimulus features did significantly reduce
the amygdala’s response to emotional stimulus features (e.g.,
Pessoa et al., 2002; Bishop et al., 2007; Blair et al., 2007; Mitchell
et al., 2008; Pessoa, 2009). It is possible that the early amygdala
activity seen in the current study that was independent of atten-
the subcortical route. The later amygdala activity that showed
significant attentional modulation may reflect the interaction of
tional control priming competing task-relevant representations
within the cortical route.
that was independent of attentional modulation, while the right
amygdala showed a later response to emotional information that
was significantly modulated by attention. However, it should be
noted that, at a more lenient significance threshold ( p ? 0.05),
the left amygdala also showed a later response to emotional in-
formation that was significantly modulated by attention. This
started at 180–190 ms and lasted until 210–220 ms. While there
was no indication of an early right amygdala response to emo-
previous work (Luo et al., 2007). In short, on the basis of the
our findings reflects genuine processing/connectivity differences
between the left and right amygdala.
for emotional expressions or whether they might generalize to
other stimulus categories. Most work considering the degree of
attentional modulation of amygdala activity and the potential
Whalen et al., 2004; Mitchell et al., 2008), although not all has
done so (Blair et al., 2007). However, the potential evolutionary
cessing of potential threats) (Ohman et al., 2007) might suggest
threat stimuli (e.g., snakes, spiders, and snarling animals), al-
though not for evolutionarily more recent threats (e.g., pointed
guns). In this regard, it is interesting to note recent findings in-
dicating altered processing of biologically relevant threat stimuli
in a patient with pulvinar damage (Ward et al., 2007). Our cur-
rent work explores these predictions.
In summary, our results suggest that the early amygdala re-
the representational strength of task-relevant stimuli and thus,
following representational competition, reduce the representa-
dala response is modulated by task load because frontoparietal
cortex has had sufficient time to augment the representation of
task-relevant information. In short, amygdala automaticity is a
matter of timing.
Anderson AK, Christoff K, Panitz D, De Rosa E, Gabrieli JD (2003) Neural
correlates of the automatic processing of threat facial signals. J Neurosci
Bichot N, Rossi A, Desimone R (2005) Parallel and serial neural mecha-
nisms for visual search in macaque area V4. Science 308:529–534.
Bishop SJ, Jenkins R, Lawrence AD (2007) Neural processing of fearful
faces: effects of anxiety are gated by perceptual capacity limitations.
Cereb Cortex 17:1595–1603.
RJ (2007) Modulationofemotionbycognitionandcognitionbyemotion.
(2007) Neural responses to auditory stimulus deviance under threat of
electric shock revealed by spatially-filtered magnetoencephalography.
deGelderB,FrissenI,BartonJ,HadjikhaniN (2003) Amodulatoryrolefor
facial expressions in prosopagnosia. Proc Natl Acad Sci U S A
Desimone R, Duncan J (1995) Neural mechanisms of selective visual atten-
tion. Annual Rev. Neurosci 18:193–222.
Newman MG, Nitschke J (2008) Worry, generalized anxiety disorder,
and emotion: evidence from the EEG gamma band. Biol Psychol
Dolan RJ, Vuilleumier P (2003) Amygdala automaticity in emotional pro-
cessing. Ann N Y Acad Sci 985:348–355.
Erthal FS, de Oliveira L, Mocaiber I, Pereira MG, Machado-Pinheiro W,
Volchan E, Pessoa L (2005) Load-dependent modulation of affective
picture processing. Cogn Affect Behav Neurosci 5:388–395.
Fries P, Nikolic ´ D, Singer W (2007) The gamma cycle. Trends Neurosci
Ioannides AA, Liu MJ, Liu LC, Bamidis PD, Hellstrand E, Stephan KM
ples of ‘real-time mapping’ of averaged and single trial MEG signals. Int
J Psychophysiol 20:161–175.
Johnson MH (2005) Sub-cortical face processing. Nat Rev Neurosci
Kastner S, Ungerleider L (2000) Mechanisms of visual attention in the hu-
man cortex. Annu Rev Neurosci 23:315–341.
Keil A, Mu ¨ller MM, Gruber T, Wienbruch C, Stolarova M, Elbert T (2001)
Effects of emotional arousal in the cerebral hemispheres: a study of oscil-
latory brain activity and event-related potentials. Clin Neurophysiol
LeDoux JE (1998) The emotional brain. New York: Sïmon and Schuster.
5828 • J.Neurosci.,April28,2010 • 30(17):5825–5829Luoetal.•EmotionalAutomaticity
LeDoux JE (2000) Emotion circuits in the brain. Annu Rev Neurosci
Logothetis NK, Pauls J, Augath M, Trinath T, Oeltermann A (2001) Neuro-
physiological investigation of the basis of the fMRI signal. Nature
Luo Q, Holroyd T, Jones M., Hendler T, Blair J (2007) Neural dynamics for
MEG. Neuroimage 34:839–847.
Luo Q, Mitchell D, Cheng X, Mondillo K, Mccaffrey D, Holroyd T, Carver F,
Coppola R, Blair RJ (2009) Visual awareness, emotion, and gamma
band synchronization. Cereb Cortex 19:1896–1904.
MaroisR,ChunMM,GoreJC (2004) Acommonparieto-frontalnetworkis
recruited under both low visibility and high perceptual interference con-
ditions. J Neurophysiol 92:2985–2992.
Mitchell D, Luo Q, Mondillo K, Vythilingam M, Finger E, Blair RJ (2008)
antagonistic relationship between the amygdala and frontoparietal corti-
ces. Neuroimage 40: 859–868.
Morris JS, Ohman A, and Dolan RJ (1999) A subcortical pathway to the
right amygdala mediating “unseen” fear. Proc Natl Acad Sci U S A
Hemodynamic signals correlate tightly with synchronized gamma oscilla-
Ohman A, Carlsson K, Lundqvist D, Ingvar M (2007) On the unconscious
subcortical origin of human fear. Physiol Behav 10:180–185.
Oya H, Kawasaki H, Howard III MA, Adolphs R (2002) Electrophysiologi-
cal responses in the human amygdala discriminate emotion categories of
complex visual stimuli. J Neurosci 22:9502–9512.
Pessoa L (2009) How do emotion and motivation direct executive control?
Trends Cogn Sci 13:160–166.
PessoaL,UngerleiderLG (2004) Neuroimagingstudiesofattentionandthe
processing of emotion-laden stimuli. Prog Brain Res 144:171–182.
PessoaL,McKennaM,GutierrezE,UngerleiderLG (2002) Neuralprocess-
ing of emotional faces requires attention. Proc Natl Acad Sci U S A
Quirk GJ, Repa C, LeDoux JE (1995) Fear conditioning enhances short-
latency auditory responses of lateral amygdala neurons: parallel record-
ings in the freely behaving rat. Neuron 15:1029–1039.
QuirkGJ,ArmonyJL,LeDouxJE (1997) Fearconditioningenhancesdiffer-
and lateral amygdala. Neuron 19:613–624.
Rogers RL, Papanicolaou AC, Baumann SB, Bourbon TW, Alagarsamy S,
Eisenberg HM (1991) Localization of P3 sources using magnetoen-
cephalography and magnetic resonance imaging. Electroencephalogr
Clin Neurophysiol 79:308–321.
SingerW (1999) Neurobiology:strivingforcoherence.Nature397:391–393.
Streit M, Dammers J, Simsek-Kraues S, Brinkmeyer J, Wolwer W, Ioannides
AA (2003) Time course of regional brain activations during facial emo-
tion recognition in humans, Neurosci Lett 342:101–104.
Varela F, Lachaux J, Rodriguez E, Martinerie J (2001) The brain web: phase
VrbaJ,RobinsonSE (2001) Signalprocessinginmagnetoencephalography.
Methods Mol Med 25:249–271.
Vrba J, Robinson SE (2002) SQUID sensor array configurations for
magnetoencephalography applications. Supercond Sci Technol 15:
Vuilleumier P, Armony JL, Driver J, Dolan RJ (2001) Effects of attention
and emotion on face processing in the human brain: an event-related
fMRI study. Neuron 30:829–841.
Vuilleumier P, Armony JL, Driver J, Dolan RJ (2003) Distinct spatial fre-
quency sensitivities for processing faces and emotional expressions. Nat
Ward R, Calder AJ, Parker M, Arend I (2007) Emotion recognition follow-
ing human pulvinar damage. Neuropsychologia 45:1973–1978.
Whalen PJ, Kagan J, Cook RG, Davis FC, Kim H, Polis S, McLaren DG,
Somerville LH, McLean AA, Maxwell JS, Johnstone T (2004) Human
amygdala responsivity to masked fearful eye whites. Science 306:2061.
Luoetal.•EmotionalAutomaticity J.Neurosci.,April28,2010 • 30(17):5825–5829 • 5829