Content uploaded by Mary-Frances O'Connor
Author content
All content in this area was uploaded by Mary-Frances O'Connor on Aug 16, 2020
Content may be subject to copyright.
Craving love? Enduring grief activates brain's reward center
Mary-Frances O'Connor
a,
⁎, David K. Wellisch
b
, Annette L. Stanton
a,b,c
, Naomi I. Eisenberger
c
,
Michael R. Irwin
a
, Matthew D. Lieberman
c
a
Cousins Center for Psychoneuroimmunology, Semel Institute for Neuroscience and Human Behavior, University of California at Los Angeles, USA
b
Department of Psychiatry, Semel Institute for Neuroscience, University of California at Los Angeles, USA
c
Department of Psychology, University of California at Los Angeles, USA
abstractarticle info
Article history:
Received 10 January 2008
Revised 8 April 2008
Accepted 29 April 2008
Available online 10 May 2008
Complicated Grief (CG) occurs when an individual experiences prolonged, unabated grief. The neural
mechanisms distinguishing CG from Noncomplicated Grief (NCG) are unclear, but hypothesized mechanisms
include both pain-related activity (related to the social pain of loss) and reward-related activity (related to
attachment behavior). Bereaved women (11 CG, 12 NCG) participated in an event-related functional magnetic
resonance imaging scan, during grief elicitation with idiographic stimuli. Analyses revealed that whereas both
CG and NCG participants showed pain-related neural activity in response to reminders of the deceased, only
those with CG showed reward-related activity in the nucleus accumbens (NA). This NA cluster was positively
correlatedwith self-reported yearning,but not with time since death, participantage, or positive/negativeaffect.
This study supports the hypothesis that attachment activates reward pathways. For those with CG, reminders of
the deceased still activate neural reward activity, which may interfere with adapting to the loss in the present.
© 2008 Elsevier Inc. All rights reserved.
There is no pain so great as the memory of joy in present grief.
~Aeschylus, founder of Greek tragedy
Introduction
Grief is one of life's most painful experiences. When
suffering the loss of a loved one, one can feel as if the attendant
sadness and longing will last indefinitely. Although successful
adaptation to the loss is the most frequent response (Bonnano
et al., 2002), grief does not abate in a substantial minority;
rather, it develops into Complicated Grief (CG) (Ott et al., 2007).
CG, previously known as chronic, pathological or traumatic
grief, includes debilitating recurrent pangs of painful emo-
tions, with intense yearning, longing and searching for the
deceased, and preoccupation with thoughts of the loved one.
This syndrome has now been defined by an empirically-
derived set of criteria (Boelen and van den Bout, 2005) and is
being considered for inclusion in the DSM-V.
The neurocognitive mechanisms involved in CG are currently
unknown. Some have hypothesized that attachment may
activate reward pathways and that this neural response may
have addiction-like properties (Insel, 2003; Panksepp et al.,
2002). We hypothesize that a major neurocognitive difference
between CG and Noncomplicated Grief (NCG) is that reminders
ofthedeceasedmaystillactivateneuralrewardsforthosewith
CG. This reward activation may interfere with adapting to the
loss in the present. Supporting this idea are subjective reports
from CG patients indicating pleasurable reveries about the lost
love (during which the reality of the loss is ignored) in addition to
painful yearning (Shear et al., 2005). These self-report findings
suggest the involvement of both reward and pain networks.
The nucleus accumbens (NA) is the region most commonly
associated with reward (Knutson et al., 2001). NA has also been
shown to play a role in social attachment, such as sibling and
maternal affiliation, via neurotransmitters and peptides (e.g.,
dopamine and oxytocin) (You ng e t al., 20 01). The pain network,
by contrast, including the dorsal anterior cingulate cortex (dACC),
insula, and periaqueductal gray (PAG) has been previously
implicated in both physical (Rainville, 2002) and social pain
(Eisenberger et al., 2003; Rainville, 2002). The present analysis
specifically investigates whether the CG group had greater
activity occur in the brain's reward or pain networks than the
NCG group.
Methods and materials
Participants
Women (11 CG, 12 NCG) who had experienced the death of
a mother or sister to breast cancer in the past 5 years were
NeuroImage 42 (2008) 969–972
⁎Corresponding author. Fax: (310) 794-9247.
E-mail address: mfoconnor@mednet.ucla.edu (M.-F. O'Connor).
1053-8119/$ –see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.neuroimage.2008.04.256
Contents lists available at ScienceDirect
NeuroImage
journal homepage: www.elsevier.com/locate/ynimg
recruited to participate in an event-related fMRI study. Par-
ticipants were recruited from a clinic for women at familial risk
of breast cancer and from the community. Sixty-four persons
were screened by telephone, and 35 had an initial interview.
Participants were excluded for current Axis I disorder (includ-
ing major depression), as evidenced by a structured clinical
interview (Spitzer et al., 1994). CG was diagnosed in a struc-
tured clinical interview (Prigerson and Jacobs, 2001) using
empirically-derived criteria. The mean age of participants was
43.70 (SD= 10.00), they had an average of 16.57 (SD= 2.35)
years of school and 78.26% were white. The two groups did not
differ on any demographic characteristics as assessed by
ANOVA and chi–square analyses. The Institutional Review
Board at UCLA approved the study and all participants gave
written informed consent.
Procedure
Each participant provided a photograph of her deceased
loved one and these photos were matched with photos of a
stranger. Participants also shared an autobiographical narra-
tive of the death event and 15 grief-related idiographic words
were chosen from it. These words were matched for part of
speech, number of letters and frequency of usage in the En-
glish language with 15 neutral words. These photos and words
were made into 60 composites (Fig. 1), each consisting of
one picture (deceased or stranger) and one word (grief-related
or neutral). Empirical support for the grief-eliciting task has
been shown previously using both skin conductance metho-
dology and participants' subjective reports (Gündel et al.,
2003). Participants viewed composites through goggles in
randomized order. Each participant was given the same set of
instructions: “Focus on any thoughts, feelings or memories
you have to the combination of the picture and the word on
the slide. Let yourself respond emotionally to each slide by
being aware of your feelings without trying to alter them”.
fMRI methodology
Data were acquired on a Siemens Allegra 3T scanner. For
each participant, a high-resolution structural T2-weighted
echo-planar image (spin-echo; TR = 5000 ms; TE = 33 ms;
matrix size 128× 128; 36 axial slices; FOV= 20 cm; 3-mm
thick, skip 1-mm) was acquired coplanar with the functional
scan. A functional scan, lasting 5 min and 30 s, was acquired
(echo-planar T2⁎-weighted gradient-echo, TR=2500 ms, TE =
25 ms, flip angle= 90, matrix size 64 × 64, 36 axial slices, FOV =
20 cm; 3-mm thick, skip 1-mm).
The imaging data were analyzed using statistical parametric
mapping (SPM'99; Wellcome Department of Cognitive Neu-
rology, Institute of Neurology, London, UK). Images were re-
aligned, normalized, and smoothed with an 8 mm Gaussian
kernel, full width at half maximum. Analysis used the general
linear model in an event-related analysis. Effects at each voxel
were estimated using linear contrasts to compare regionally
specific effects. The contrasts for the individual subjects were
aggregated for single-group analysis and for between-group
analysis according to the random effect model in SPM. All
group comparison analyses (e.g., CGNNCG) were thresholded
using an uncorrected p-value of .005 combined with a cluster
size threshold of 10 voxels. Where indicated, parameter esti-
mates were extracted from group comparisons and reported
for each group separately. All coordinates are reported in MNI
format.
Results
Reward activation
Greater reward-related activity occurred in the CG relative
to the NCG individuals. Notably, the NA (x=10, y= 20, z=−6;
Fig. 2A) was the only region of the brain that was more active in
response to grief-related words than neutral words among
those with CG compared to NCG (t= 3.51, pb0.001, 15 voxels).
Parameter estimates were extracted from this region and
reported separately for each group as shown in Fig. 2B. CG
individuals produced increased NA activity in response to grief
words (t=3.36, pb.01), whereas the NCG individuals did not
(t=−1.82, pb.10). A comparison of the deceased and stranger
photos did not produce NA activity. This may have been due to
habituation effects as 15 different grief words were used, but
only a single deceased photograph was shown repeatedly. To
examine this possibility, we compared deceased and stranger
photos from the first third of the trials only in order to isolate
pre-habituation activity. Here, as in the grief word comparison,
there was greater activation in the NA (x=−14, y=4,z=−16) for
the CG group than the NCG group (t=3.68, pb.001, 18 voxels).
Yearning
To clarify the functional role of NA activity in grief, cor-
relational analyses were performed with parameter estimates
of activity from the observed cluster (10, 20, −6). Greater NA
activation in the grief word comparison correlated significantly
with more self-reported yearning for the deceased (r=.42,
pb.05) (Fig. 3). No correlation was observed between NA acti-
vation and length of time since the death, participant age, or
general positive or negative affect. Similarly, the NA activation
observed during the early trials of the deceased photo
comparison was also associated with yearning (r= .36, pb.05),
and was not correlated with length of time since the death,
participant age, or general positive or negative affect.
Pain network activation
To examine pain–related neural activity, we examined: a)
whether each group showed neural activity in three regions
previously linked to pain processes (anterior cingulate cortex
Fig. 1. Example of picture/word composites presented to subject.
970 M.-F. O'Connor et al. / NeuroImage 42 (2008) 969–972
(ACC), insula, periaqueductal gray (PAG) (Peyron et al., 2000));
and b) whether there were significant differences between
the groups in pain–related neural activity in those regions
(pb.005, 10 voxels). If one group showed significant activation
in a region (pb.005, 10 voxels), we reported the activation from
that same coordinate in the other group (Table 1). Each group
showed significant activity in pain-related regions (dACC, in-
sula, PAG) in the comparisons of the deceased vs. stranger
pictures and grief vs. neutral words (Table 1), as has been pre-
viously shown in NCG (Gündel et al., 2003). The only activity in
pain-related regions that differed between groups was in the
left insula (x=−34, y=14,z= 18), which was more active in the
NCG group during the viewing of grief-related vs. neutral
words (t=4.20, pb0.001, 87 voxels).
Discussion
Two models of grief have been hypothesized: a detachment
model and a reunion model (Bowlby,1980). In the detachment
model, the grief emotion is believed to play a role in the
acceptance of the reality of the death and therefore assist in
recovery from the loss. In the reunion model, the grief is a form
of protest against the separation from the deceased, and serves
to promote reunion with the lost person, not detachment. After
the death, cues of the deceased (such as memories, photos,
etc.) persist, triggering yearning and grief. Coping with these
salient cues is a major task in adjusting to the death (Freed and
Mann, 2007). Freed and Mann hypothesize that if the detach-
ment model is correct, the pangs of grief would occur with
reduced NA activity over time, as the salience of the cues
Fig. 3. Positive correlation between self-reported yearning and BOLD activity in the NA
(10, 20, −6; N=23) for each subject.
Table 1
Effects in three regions of the pain network for both CG (N=11) and NCG (N=12) in
response to pictures of the deceased and grief-related words
Region Group MNI coordinates (x,y,z)Tp-value (voxels)
Grief-relatedNneutral (words):
ACC
NCG −2 34 8 5.12 0.001 (140)
CG −2 34 8 2.74 0.02
Insula
NCG −36 18 18 4.73 0.001 (44)
CG –––
PAG
NCG –––
CG –––
Deceased Nstranger (pictures)
dACC
NCG −2 24 24 4.17 0.001 (309)
CG −2 24 24 1.87 0.05
Insula
NCG −40 8 −2 4.76 0.001 (235)
48 10 0 4.47 0.001 (277)
CG −40 8 −2 1.70 0.02
48 10 0 2.31 0.03
PAG
NCG −10 −18 −16 9.74 0.00 01 (55)
CG −10 −18 −16 2.37 0.02
Significant activations in one group (pb005, 10 voxels) were examined and reported for
the other group. Tand p-values are listed for each group separately. Cluster sizes were
determined at the threshold of .005, 10 voxels (if the significance did not reach this
threshold, cluster sizes were not comparable and therefore not listed). dACC = dorsal
anterior cingulate cortex, PAG = periaqueductal gray.
Fig. 2. A) Nucleus accumbens activity (10, 20, −6) in response to grief-related vs. neutral
words that was significantly greater in the Complicated Grief group compared to the
Noncomplicated Grief group (pictured at pb.05). B) Bar graph showing nucleus
accumbens activity (10, 20, −6) in response to grief-related vs. neutral words for those
with Complicated and Noncomplicated Grief.
971M.-F. O'Connor et al. / NeuroImage 42 (2008) 969–972
decreases and acceptance of the reality leads to detachment. If
the reunion model is correct, then the pangs of grief would
continue to occur with NA activity, with reward activity in
response to the cues motivating reunion with the deceased.
This study demonstrates that each of these models may be
accurate for distinct subgroups of bereaved individuals: relative
to non-grief-eliciting stimuli, those with CG produced sig-
nificant NA activations, whereas those with NCG produced NA
reductions during grief elicitation.
Findings suggest that both CG and NCG groups may have
felt pain upon presentation of grief-related stimuli, but only
those with CG also activated an area important for reward
processing when viewing cues of the deceased. The associa-
tion between NA activity and yearning, but not time since
death, participant age, or positive/negative affect provides
convergent and discriminative validity. The addiction-rele-
vant aspect of this neural response (Knutson et al., 2001) may
help to explain why it is hard to resist engaging in pleasurable
reveries about the deceased even though engaging in these
reveries may prevent those with CG from adjusting to the
realities of the present. Many who suffer from addiction-like
disorders experience them as afflictions; similarly we are not
suggesting that reveries about the deceased are emotionally
satisfying, but rather may serve as craving responses that may
make adapting to the reality of the loss more difficult.
Prior work has demonstrated a relationship between auto-
nomic arousal and posterior cingulate (PCC) activity during
grief elicitation (O'Connor et al., 2007). O'Connor et al. hypo-
thesized a role for PCC in those with CG, given that those with
CG may have greater arousal. However, there were no sig-
nificant differences in PCC activity between the NCG and CG
groups in the present study. Future work is needed to more
carefully explore whether the PCC plays a role in CG in a larger
sample. In addition, connectivity analyses in the prior study
showed that PCC activation was associated with subgenual ACC
activation. Similarly, future studies should investigate connec-
tivity between NA and prefrontal or limbic regions important in
emotion regulation.
Understanding the reward processes activated in those with
CG could substantially change treatment of this disorder.
Therapies such as behavioral interventions that target reward
processes may confer benefit and preferentially aid in adapting
to the loss (Shear et al., 2005). Likewise, dopaminergic inter-
ventions that alter reward sensitivity could theoretically be
more effective in treating CG than serotonergic interventions,
which have failed to alter grief intensity (Zygmont et al., 1998).
Addressing the continued craving of past relationships may
assist those with CG in adapting to the loss.
Acknowledgments
We would like to thank the UCLA Brain Mapping Center for
their assistance. This research was supported by funds from
the California Breast Cancer Research Program Grant Number
10IB-0048. This work was also supported in part bygrant T32-
MH19925, the Cousins Center for Psychoneuroimmunology
and the Friends of the Semel Institute for Neuroscience and
Human Behavior.
References
Boelen, P.A., van den Bout, J., 2005. Complicated grief, depression, and anxiety as
distinct postloss syndromes: a confirmatory factor analysis study. Am. J. Psychiatry
162 (11), 2175–2177.
Bonnano, G.A., Wortman, C.B., Lehman, D.R., Tweed, R.G., Haring, M., Sonnega, J., et al.,
2002. Resilience to loss and chronic grief: a prospective study from preloss to 18-
months postloss. J. Pers. Soc. Psychol. 83 (5), 1150–1164.
Bowlby, J.,1980. Attachment and loss. Loss, Sadness and Depression, vol 3. Basic Books,
New York.
Eisenberger, N.I., Lieberman, M.D., Williams, K.D., 2003. Does rejection hurt? An fMRI
study of social exclusion. Science 302 (5643), 290–292.
Freed, P.J.,Mann, J.J., 20 07. Sadness and loss: Toward a neurobiopsychosocialmodel. Am.
J. Psych. 164 (1), 28–34 .
Gündel,H.,O'Connor,M.-F.,Littrell,L.,Fort,C.,Lane,R.D.,2003.Functional
neuroanatomy of grief: an fMRI study. Am. J. Psychiatry 160 (11), 1946–1953.
Insel, T.R., 2003. Is social attachment an addictive disorder? Physiol. Behav. 79 (3), 351.
Knutson, B., Adams, C.M., Fong, G.W., Hommer, D., 2001. Anticipation of increasing
monetary reward selectively recruits nucleus accumbens. J. Neurosci. 21 (16) RC159.
O'Connor, M.-F., Gündel, H., McRae, K., Lane, R.D., 2007. Functional neuroanatomical
correlates of emotion al arousal during grief. Neuropsychopharmacology 32,
2184–2189.
Ott, C.H., Lueger, R.J., Kelber, S.T., Prigerson, H.G., 2007. Spousal bereavement in older
adults: common, resilient, and chronic grief with defining characteristics. J. Nerv.
Ment. Disord. 195 (4), 332–341.
Panksepp, J., Knutson, B., Burgdorf, J., 2002. The role of brain emotional systems in
addictions: a neuro-evolutionary perspective and new “self-report”animal model.
Addiction 97 (4), 459–469.
Peyron, R., Laurent, B., Garcia-Larrea, L., 2000. Functional imaging of brain responses to
pain. A review and meta-analysis. Clin. Neurophysiol. 30 (5), 263–288.
Prigerson, H.G., Jacobs, S.C., 2001. Traumatic grief as a distinct disorder: a rationale, con-
sensuscriteria, and preliminaryempirical test.In: Stroebe,M.S., Hansson,R.O., Stroebe,
W., Schut, H. (Eds.), Handbook of Bereavement Research: Consequences, Coping and
Care. American Psychological Association, Washington, D.C., pp. 613–645.
Rainville, P., 2002. Brain mechanisms of pain affect and pain modulation. Curr. Opin.
Neurobiol. 12 (2), 195.
Shear, K., Frank, E., Houck, P.R., ReynoldsIII, C.F., 2005. Treatment of complicated grief: a
randomized controlled trial. J. Am. Med. Assoc. 293 (21), 2601–2608.
Spitzer, R.L., Williams, J.B.W., Gibbons, M., First, M.D.,1994. Structured Clinical Interview
of the DSM-IV. American Psychiatric Press, Washington, DC.
Young, L.J., Lim, M.M., Gingrich, B., Insel, T.R., 2001. Cellular mechanisms of social
attachment. Horm. Behav. 40 (2), 133–138.
Zygmont, M., Prigerson, H., Houck, P., Miller, M., Shear, M., Jacobs, S., et al., 1998. A post
hoc comparison of paroxetine and nortriptyline for symptoms of traumatic grief.
J. Clin. Psychiatry 59 (5), 241–245.
972 M.-F. O'Connor et al. / NeuroImage 42 (2008) 969–972
