A total of 47 participants took part on the research (see Table 1). They were recruited from the Clinical Unit at the Faculty of Psychology in Granada (Spain), the Palliative Care Unit of San Cecilio Clinical Hospital (Granada – Spain), and from two associations supporting grief and bereavement: Alma y Vida (Jaén- Spain) and Talitha (Albacete – Spain). The inclusion criteria were: (a) having experienced the death of a loved one in the last 6 months and (b) being 18 years-old or more. The exclusion criteria were: (a) having writing and/or reading problems, (b) having a previous diagnosis of psychiatric illness, and (c) an extreme score (>75th Percentile) on the variables of depression and/or PTSD. Three participants were excluded due to criterion number three. The final sample had 44 participants (Table 1).

The Inventory of Complicated Grief was used to divide the sample in two groups (N = 24 for the CG group, and N = 20 for the non-CG group). A score of 25 or higher was the criterion for CG (Prigerson et al., 1995). Both groups showed no statistical difference in gender, months since death, years of education and the type of relationship with their loved one. However, as expected, they showed differences in their scores on measures of depression, CG, and PTSD (see Table 1).

Participants that accepted to take part in the study, and met inclusion criteria, completed a large number of neuropsychological tasks and self-report measures. Participation in the study was voluntary, and participants did not receive financial or other compensation. They were contacted through the database of patients from San Cecilio Clinical Hospital, and Associations Talitha and Alma y Vida. Each participant performed a total of three sessions, each of them of one and half hours duration. In the first session they completed a sheet with all the self-report questionnaires. During the second session they completed the IAPS and the FEEST, which were presented in the same order to all participants. The last session was used to assess other neuropsychological variables. All sessions were performed in a quiet location, administered by the same researcher, and under homogeneous conditions.

The research was approved by the Ethical Committee of Human Research at the University of Granada (Spain). All participants were informed of the objectives of the study and were given an information sheet. Written informed consent was collected from all participants before the start of the study.

First, mean and standard deviation were calculated for each group (CG versus Non-CG), based on the total scores of each task. Second, to test if there were significant differences between groups, t-tests for independent samples and bifactorial MANOVAS were carried out, followed by post hoc comparisons when the results were statistically significant. Cohen’s d and partial-eta squared values were calculated as a measure of effect size. Holm–Bonferroni correction was used for controlling multiple comparisons and p-values were set below 0.02.

There were no differences between groups in the total number of emotional faces recognized on the FEEST, t(41) = 0.436, p = 0.66, d = 0.13 or in any of the discrete emotions (anger, surprise, sadness, happiness, disgust, and fear).

To analyze the different emotional categories from the IAPS, three bifactorial MANOVAS were completed, having group as an independent variable (CG versus Non-CG) and the scores on each dimension (valence, arousal and dominance) as dependent variables. There were statistically significant differences for the valence dimension, F(5,38) = 3.27, p = 0.015, ηp2 = 0.301, Wilks’ Lambda = 0.699, and marginally significant differences for the arousal dimension, F(5,38) = 2.39, p = 0.055, ηp2 = 0.240, Wilks’ Lambda = 0.760. Scores on dominance were not different between groups, F(5,38) = 1.26, p = 0.30, ηp2 = 0.142, Wilks’ Lambda = 0.858.

Post hoc comparisons showed statistically significant differences for the valence of the unpleasant non-arousing pictures (UNAs), t(42) = 2.75, p = 0.009, d = 0.86 and the pleasant arousing pictures (PA), t(42) = 2.70, p = 0.01, d = 0.84. Results indicated that the CG group perceived the images of these categories as more unpleasant than the Non-CG group (see Figure 1).

The three main dimensions of emotional psychopathology assessed through the SCL-90-R were the Global Severity Index (GSI), the Positive Symptom Total (PST), and the Positive Symptom Distress Index (PSDI). The GSI values were statistically significant [t(42) = 3.2, p = 0.003, d = 1], as well as the PST [t(42) = 3.22, p = 0.002, d = 1] and the PSDI [t(42) = 2.03, p = 0.049, d = 0.62]. Participants from the CG group showed higher scores on the three indices of global psychopathology. Post hoc analysis showed that scores on eight of the nine sub-scales were significantly different between groups, indicating very substantial number of emotional symptoms for those with CG. Five of nine dimensions (obsession, depression, inferiority, hostility, and psychoticism) had p values below 0.02 (see Figure 2).

Finally, levels of PTSD are presented on Table 1. Those with CG had a high number and intensity of symptoms of PTSD, on the dimensions of re-experience (d = 1.02), avoidance-affective numbness (d = 1.04), and hyper-arousal (d = 1.1). A total of 12 participants (50%) from the CG group fulfilled the criteria for PTSD, by comparison with only one participant (5%) from the non-CG group [χ²(1) = 10.71, p = 0.001].

Regarding the recognition of emotion, none of the six Ekman emotion categories differed significantly between groups. Previous studies have reported face recognition as useful in identifying psychopathologies such as depression (Kohler et al., 2011; Jaworska et al., 2012) or PTSD (Felmingham et al., 2003). However, the present results suggest that face recognition tasks in general may be of limited use in identifying patients with CG. Nevertheless, some tasks have produced significant effects using emotional faces. For example, studies using a photograph of the deceased relative triggered loss-reaction responses in those with CG (O’Connor et al., 2008; Eisma et al., 2014). Presumably this is because they are related to the internal emotional experience of participants (Salas et al., 2012) and perhaps because none of these experiments directly asked about the knowledge of the emotions that were displayed (Adolphs and Damasio, 2000). Also, although it seems that the facial expression of emotion may be altered in CG (Diminich and Bonanno, 2014), our results indicate that accurate recognition of emotions might not be disrupted.

The findings were strikingly different for the subjective experience of emotion. Previous studies on CG have used pictures from the IAPS (Gupta and Bonanno, 2011), but the present is the first to systematically use it to measure emotional experience, controlling for the range of valence and arousal, and assessing the five classical conditions of the emotional space (Lang and Bradley, 2010). The CG group showed substantially different scores on the dimension of valence on two conditions, which included both pleasant and unpleasant pictures, with medium and large effect sizes. These results may be important for understanding the subjective experience in those with CG. Previous studies have reported difficulties in both the enhancement and suppression of emotion in people diagnosed with CG (Gupta and Bonanno, 2011), suggesting a lack of emotional flexibility (Bonanno et al., 2004; Bonanno and Burton, 2013). It is possible that symptoms of CG may lead to an hyperactivation of the aversive motivational system, which make it difficult to regulate both positive and negative emotions (Bradley and Lang, 2000; Lang and Bradley, 2010).

A question that remains unclear is why the differences in valence were so striking on two specific conditions on the IAPS. Following Peter Lang’s theory of emotion, it is interesting to notice that the UNA and PA pictures are opposed categories in the emotional space, both in the American and the Spanish version (Moltó et al., 1999; Vila et al., 2001). In addition, UNA pictures are the most difficult pictures to find; due to the complication in obtaining unpleasant emotional stimulus that do not imply substantial arousal (Moltó et al., 1999, 2013; Lang and Bradley, 2010). Future studies might employ a large number of pictures from each condition, in order to replicate current findings and establish their implications.

The content of the pictures may also have an important role. It is interesting to note that the ratings differed on pictures that were related to sadness and sickness (UNA pictures) and to joyful and arousing situations (PA pictures). We could argue that these two conditions were both linked to loss-related and to restoration-related feelings, as suggested by the Dual Process Model of Coping with Bereavement (Stroebe and Schut, 2010). From this point of view, CG is characterized by a difficulty in oscillating between the loss-oriented and restoration-oriented responses. It is possible that the two conditions on the IAPS reflected this distinction, between loss and restoration affective stimuli (Stroebe and Schut, 2010; Fasse and Zech, 2015). Again, further research is needed to test this hypothesis, using distinct pictures for each category.

The differences in subjective emotional experience reported by those with CG may be linked to data from neuroimaging and affective neuroscience. Firstly, the importance of the subjective clearly emphasizes the critical role of the felt aspects of emotion (Solms and Turnbull, 2002; Panksepp, 2010). From this point of view, the unpleasant experience of both UNA and PA pictures may be an indication of a disturbance on the level of affect or feelings (Panksepp, 2010). In line with this hypothesis, a wide range of subcortical structures (associated by Panksepp with primary-process networks) have a critical role on the experience of CG. When participants with and without CG were compared (on an emotional task that used stimuli related and unrelated to the deceased) the nucleus accumbens was the critical brain structure that differentiates between groups (O’Connor et al., 2008). Other emotional areas, such as the amygdala or the orbitofrontal cortex, also involved in the evaluation of the reward, have also shown increased activation on those with CG (O’Connor and Arizmendi, 2015).

Secondly, some of the previously discussed brain areas are an integral part of the dopaminergic reward system, which has been suggested to be of central importance for understanding CG. The emotional experience of loss-related stimuli seems to hyper-activate this wanting or SEEKING system (O’Connor et al., 2008; Panksepp and Watt, 2011; Zellner et al., 2011). This pattern is distinct from the one presented in depression, which is characterized by an activation of the PANIC/GRIEF system (mediating separation-distress) and decreased activation on the SEEKING system – producing a decrease in arousal, and dysphoric symptoms (Panksepp, 2010; Zellner et al., 2011). Many clinicians will be unaware of this important behavioral distinction (and its neural basis), but the finding may well have implications for the treatment of CG as compared to other forms of depression – from both a psychotherapeutic and a pharmacotheraputic perspective.

Future researches, including neuroimaging studies, are clearly needed to assess the specific role of CG symptoms on these emotional networks, as well as any similarities and differences from depression. One possible way of addressing this objective may be to compare the neural activation of participants diagnosed with only with CG, versus those that also fulfill the criteria of clinical depression. This kind of research may clarify the role of SEEKING and PANIC/GRIEF system on grief.

Although previous studies have investigated psychopathology in CG (Simon et al., 2011; Boelen, 2013) the current research sought to identify the general profile of symptomatology that characterized this population. Emotion-related psychopathology was assessed through the SCL-90-R. Participants with CG showed higher scores on all subscales, in line with previous research (Simon et al., 2007; Golden and Dalgleish, 2010; Boelen, 2013). These results suggest, perhaps unsurprisingly, that people with CG are at greater risk of a range of emotional problems. In addition, their symptom profile may thus be easily confounded not only with well-known psychopathologies such as depression, anxiety or PTSD (Bryant, 2014), but also with symptoms that represent obsessive behavior, feelings of inferiority, outbursts of hostility, or even frankly psychotic features.

A consequence of this complex symptom profile is that those with CG may be inappropriately diagnosed and treated, perhaps to control the intense emotions in the early stages of bereavement (García et al., 2013; García and Landa, 2015), leading to later difficulties with grief resolution. This is an under-investigated topic, though in one qualitative study of bereaved mothers, all showed remorse that their medication schedule did not allow them to fully experience loss-related feelings (Bosquet-del Moral et al., 2012). In addition, this misidentification may lead to over-treatment and excessive diagnosis of both depression and CG, due to the new diagnostic criteria for depression of the DSM-5 (Bryant, 2013; Bandini, 2015).

Data from the present study, supporting the key role of subjective emotional experience in CG, will be of special interest for clinicians – clearly suggesting that psychotherapeutic intervention should focus on internally focussed elements of the client’s experience. This is in line with many current clinical approaches, mainly focused on the use of specific emotional techniques to promote elaboration and integration of the loss, developing new coping strategies and interpretations of the inner emotional narrative (Boelen et al., 2006; Mancini et al., 2012; Alves et al., 2014). Indeed, there may be a role for strategies such as holding and supporting negative emotional experiences (Winnicott, 1960), which may help to integrate the intense feelings of yearning, sadness and shock into the biographical self (Maccallum and Bryant, 2013). In addition, because of the differences in the role of dopaminergic (SEEKING) systems, there are also implications for the pharmacotherapy of CG, by comparison with other forms of depression (as discussed above). Clearly, this is an important topic for future research.

There were both strengths and limitations in the study. The principal limitation was the small sample size, which may not have allowed enough statistical power to detect a difference in emotion conditions. However, studies involving patients with CG usually have a range between 15 and 30 participants (Robinaugh and McNally, 2013; O’Connor and Arizmendi, 2014). In addition, effect sizes (independent of sample size) were reported in the present study, and are of a reasonable magnitude. A second issue relates to gender. There is evidence from previous research that men and women may experience and cope with loss in different ways (Chiu et al., 2011; Kersting et al., 2011), and no previous studies have directly addressed gender differences in the study of emotional processing on CG. Due to the limited sample size of the present study it was not feasible to satisfactorily study gender effects. Finally, current research on the subjective aspects of emotion involves taking into account the specificity of an emotional schema related to the death (Martí-García et al., 2016). Future studies might be aimed at identifying whether these differences in subjective emotional processing are related to the activation of this death-related schema.