Grief is a multifaceted response to loss, often (although not exclusively) following the death of a person with whom one shares a close bond. Grief is a human inevitability, arising out of our love for our family and friends. However, individuals vary considerably in how they respond to the death of a loved one. Although most people ultimately adapt with time, a small minority—about 10% of bereaved individuals [1]—continue to experience grief of an intensity that makes it difficult to function in daily life, even many years later. This lack of adaptation can increase suicidality and even medical consequences. This severe, chronic, and profound form of grief is known as complicated grief (CG) or prolonged grief disorder (PGD). Just as someone might develop complications after surgery that prevents physical healing, the terminology of “complicated” grief reflects a theory that there is something preventing healing and recovery from taking place. Given the significant distress, functional impairment, and negative health consequences associated with CG, this phenomenon was recently recognized in the DSM-5 as a psychological disorder meriting further study, termed “persistent complex bereavement disorder” [2]. In the following review, we will use CG to refer to all three criteria sets: CG, PGD, and persistent complex bereavement disorder (PCBD) . We assume (as much of the research to date has done) that these three are basically referring to a single phenomenon. However, we will return to this definitional issue at the end of the review. In contrast, we will refer to the typical pattern of healthy adaptation seen during bereavement as non-complicated grief (non-CG) . The period of time immediately following the loss, in which grief symptoms are expected to be most intense and severe, is referred to here as acute grief.

What methods might we use to understand grief, and the differences between the most common pattern of adaptation to the death of a loved one (i.e., resilience) and the lack of adaptation we see in CG? One method is to investigate the processes occurring in the brain as we go from the perception of the death event to experiencing the emotional consequences, and incorporating the event into our memories, our schemas, and our identity. Only a few studies have started to unravel this relationship between grief and neural processes.

Neurobiological measurement provides us with one lens through which to view affective processes and attempt to understand them. Emotional experiences such as grief are often assessed using self-report, behavioral observation, or behavioral tasks. These measurements often assess outcomes without clarifying the neural architecture contributing to those behaviors and symptoms (whether promoting or inhibiting). We know that there are several neural processes that can achieve the same behavioral or subjective outcome. For instance, two individuals with grief may display the same overt behavioral symptoms and report the same level of distress. One could imagine that the first person might have difficulty with information processing, making it difficult to incorporate the death event. A second person might have difficulty with repetitive thoughts, making it difficult to concentrate on other aspects of their life. These differences may not be evident at the level of self-report or behavioral observation. Therefore, we feel it is helpful to look at the mechanisms through which grief may become “complicated,” in concert with self-reported affect and behavior. Identifying the neurobiological mechanisms implicated in the grieving process may not only provide the ability to develop and test hypotheses at the basic science level, but also identify opportunities to intervene, optimally targeted to reduce the burden of suffering and functional impairment that can follow the death of a loved one.

Although a variety of methods can be used to measure neurobiology, functional magnetic resonance imaging (fMRI) is often the method chosen for investigating neurobiological aspects of grief. fMRI measures changes in cerebral blood flow as an indirect measure of neural activity. The accuracy or detail of the image is better than other methods such as electroencephalogram (EEG) , which makes fMRI advantageous for documenting activity in diverse brain regions. fMRI also allows investigators to examine how networks of brain regions interact during mental processes. Although there is not a one-to-one correspondence of neural activation to behavior or subjective experience, mental functions implicated in grief (such as memory and emotion regulation) are instantiated in different brain regions [3]. A number of these functions may act in concert to influence how people feel and act when bereaved. In addition, structural MRI can tell us about the neural scaffolding that supports the mental functions observed using fMRI.

Although emotions have been the focus of considerable work in recent decades, we know relatively little about grief, compared to emotions such as sadness or anger. CG is characterized by recurrent painful emotions, preoccupation with the death, intense yearning for the deceased, and difficulty accepting the reality of the death, among other symptoms that cause impairment in social and occupational functioning. CG as a disorder involves a number of transdiagnostic constructs, such as avoidance, intrusive thoughts, strong physiological response to stress, and difficulties regulating emotion. Research has also shown CG to be clearly distinct from other disorders such as depression and post-traumatic stress disorder (PTSD) [4].

Three previous papers have reviewed the results of fMRI studies of grief and CG in the past decade, the earliest of which was written by O’Connor in 2005, when only one study on neuroimaging and bereavement had been published [5]. The second was written by O’Connor in 2012, reviewing a total of four studies examining neural and immunological correlates of grief [6]. The most recent review, by Silva and colleagues, found five studies in this area [7]. While the number of neuroimaging studies of grief has increased considerably since 2005, there are still relatively few experimental paradigms designed specifically to elicit and capture the multifaceted response that is grief.

The elicitation of grief in a sterile scanner environment could be difficult—and it was for this reason that the grief elicitation paradigm was originally developed. The paradigm is a 2 × 2 design: photos of the deceased contrasted with photos of a stranger, and embedded into those photos, grief-related words contrasted with neutral words. This resulted in four possible conditions [8, 9]. Several variations of this task have been used. For a study of women who had recently terminated a pregnancy due to fetal malformation (compared to women who recently had a healthy baby), the task included photos of a happy baby compared to photos of an adult with a happy or neutral facial expression [10]. These three studies employed similar grief elicitation tasks, albeit with different bereaved samples. The first study included all bereaved participants [8], the second included CG and non-CG participants [9], and the third included bereaved and non-bereaved participants [10]. The grief elicitation paradigm was designed simply to tap the neural response to the deceased loved one compared to a stranger. However, we are now aware that this “response” comprises multiple mental functions. Gündel and colleagues reported that the grief condition evoked activity in regions implicated in affect processing, mentalizing, episodic memory retrieval, processing of familiar faces, visual imagery, autonomic regulation, and modulation/coordination of these functions.

A notable similarity across the three studies is that bereaved participants exhibited activity in brain regions previously implicated in the experience of pain, including the dorsal anterior cingulate cortex (ACC) , insula, and periaqueductal gray (PAG) , when viewing the spouse compared to a stranger. Non-bereaved participants demonstrated greater PAG activity compared to bereaved participants [10]; however, activation did not differ between participants with and without CG when these three regions were specifically examined [9], suggesting that these regions are involved in grief more broadly, rather than being specific to CG. This research is consistent with subjective reports of grief being an especially painful experience, both emotionally and even physically. Importantly, the neural evidence suggests that this is a common phenomenon across bereaved individuals and this neural pattern does not necessarily indicate CG.

Additionally, activation of the posterior cingulate cortex (PCC) and cuneus has been evidenced in bereaved vs. non-bereaved comparisons. However, these regions were not specifically investigated in the study of CG, limiting our ability to draw conclusions with regard to the implication of these regions in this disorder. Nonetheless, given the importance of the PCC in the processing of autobiographical emotional memories and the cuneus in visual processing, they are likely important regions for the grief elicitation task. In grief, adaptation necessarily involves emotional autobiographical memories, and evidence suggests that autobiographical memories function differently in CG [11, 12].

The PCC also functions as an important hub in the default mode network (DMN) . The DMN is a network of interconnected brain regions that exhibit activation during “resting state” (i.e., when the participant is instructed to rest quietly in the scanner without any particular task) and deactivation during cognitive or attentional tasks [13]. Given that much time is spent during grief in recalling autobiographical emotional memories, it makes sense that the DMN is related to self-reference [14], autobiographical memory [15], and rumination [16]. Perturbations in DMN connectivity have been thus associated with major depressive disorder [17], a condition that also involves deficits in certain of these domains.

Only one study has investigated DMN functioning in grief: Liu and colleagues examined a group of older Chinese adults who had experienced the death of their only child [18]. Bereaved participants (compared to non-bereaved) had decreased connectivity of brain hubs within the DMN. They also showed decreased connectivity in hubs of the central executive network, including the dorsolateral prefrontal cortex (DLPFC) . This region plays an important role in cognitive regulation of emotion. This was especially true in those bereaved participants with negative coping styles (e.g., avoidance, alcohol use), suggesting that differential responses to grief might be linked to differences in neural functioning.

The grief elicitation paradigm was designed simply to tap the response to the deceased loved one, using the response to a stranger for comparison. However, we are now aware that this “grief response” comprises multiple mental functions. Although the original grief elicitation task was very useful in determining what general regions might be involved in this individualized emotional response, and although some of the same regions appear in multiple studies, the task is perhaps too broad to usefully tap the critical, necessary, or sufficient neural activations in the grief response, and future research could benefit from parsing the mental functions in grief (and their concomitant neural activations) to determine how they might relate to the maintenance of complicated grief, how they might predict typical adaptation over time, or might correlate with current functioning.

People with both acute grief and CG report intrusive grief experiences, reminders of the loss that occur unbidden, and difficulty concentrating when they experience pangs of grief. Neuropsychological functioning during bereavement has been assessed, although reviewing all related studies is beyond the scope of this chapter. However, a report from the most comprehensive study (groups that included 150 with CG, 615 with non-CG, and 4700 non-bereaved) led to the following conclusions [19]. Participants with CG had lower processing speed and verbal fluency scores compared with non-bereaved and non-CG participants, and had lower Mini Mental State Examination (MMSE) scores than those with non-CG. No differences in performance on either a Stroop task or word-learning tests of immediate and delayed recall were observed between the CG and non-CG.

Cognitive difficulties during bereavement could also be related to the interaction of emotion and cognition. The emotional Stroop (eStroop) is designed to assess emotional interference and measure the extent to which the participant can disengage from the emotionally salient stimuli in order to remain focused on the task. There are several variants of the task that have been applied to bereaved populations, including stimuli that use the name of the deceased, idiographic grief-related stimuli, and categorical grief-related stimuli. In addition, participants can be asked to report on the color of the word or the number of words on the screen (i.e., the emotional counting Stroop). All versions of the eStroop, however, are designed to assess emotional interference and the extent to which the participant can disengage from the emotionally salient stimuli in order to remain focused on the task (i.e., reporting the color or number). This is measured by comparing the reaction time to grief words compared to neutral words. Slower reaction times to grief-related words indicate greater interference.

Two of the fMRI studies of grief reported to date have utilized the eStroop. The primary finding of the first study is that the magnitude of one’s attentional bias correlated with amygdala, insula, and dorsolateral prefrontal cortex (DLPFC) activity [20]. In addition, self-reported intrusiveness of grief-related thoughts correlated with ventral amygdala and rostral anterior cingulate (rACC) activation, while avoidance correlated with deactivation of dorsal amygdala and DLPFC.

The most recent study [21] employed the eStroop with a sample consisting of CG, non-CG, and non-bereaved participants. The primary finding was that participants with non-CG exhibited activity in the rostral ACC/orbitofrontal cortex, which was not observed in the non-bereaved control group. By contrast, the CG group did not show significant activation in any areas when compared to either non-CG or non-bereaved controls. Further, the CG group displayed no rACC activation even when examined alone using a region-of-interest (ROI) approach. This could be interpreted as a relative inability to recruit the regions necessary for successful completion of this emotional task in those with CG.

It is difficult to compare the two existing eStroop fMRI studies, because of the large differences between them. Methodologically important differences (e.g., pet vs. human, length of bereavement) between the studies might account for the disparate findings. Most importantly, Freed et al. [20] used reaction time as a covariate in the analyses, and thus are not just a contrast of grief and neutral stimuli (as reported in the study by Arizmendi and colleagues). The role of the amygdala in automatic responding to emotional stimuli may mean that this was the relevant mental function captured by this analysis, albeit in the context of grief. In the study by Freed and colleagues, the loss was more recent (average of 3 months as opposed to 3 years), and did not categorize groups by grief severity.

Three additional behavioral studies of the eStroop in bereaved samples (without a scanning component) can be found in the literature [22–24]. The fMRI study by O’Connor and colleagues is a subset of the group who participated in the behavioral study by O’Connor and Arizmendi [24], and then additionally had a neuroimaging scan. In all of the reported reaction times (both for behavioral and neuroimaging studies), grief words have a longer reaction time than neutral words for bereaved people (whether or not they had CG). This result is interpreted such that bereaved individuals attend more to grief cues and have a harder time disengaging from them, once they have grabbed their attention. Two of the three studies comparing CG and non-CG reported that the CG group had slower reaction times to grief words than the non-CG group [22, 24]. In contrast, the study by Mancini and Bonanno found that the CG group had faster reaction times to grief words than the non-CG group [23].