In this research we found that patients with chronic psychosis featured a higher prevalence of insecure attachment. The symptomatology of patients with insecure attachment was more severe. Also, a high prevalence of childhood trauma related to attachment figures was observed in patients with psychosis compared to our control sample. A large majority of subjects in both groups believed in a spiritual figure that functioned like an attachment figure. Amongst them, a compensation process was observed in some subjects, i.e. they showed a stable attachment to a spiritual figure in the context of a primary insecure attachment towards caregivers.
Insecure attachment style and psychosis
The first aim of this study was to explore attachment models among patients with psychosis. They featured a high prevalence of insecure attachment (mostly dismissing/avoidant) as compared with the control group. The results for the latter group can be compared with those of the meta-analysis of van IJzendoorn and Bakermans-Kranenburg  on a group of 4392 non-clinical subjects showing a 57 % rate of secure attachment. Indeed most studies to now have shown a higher prevalence of insecure attachment (mostly dismissing/avoidant) in patients with psychosis [7, 9, 30–37] as compared with non-clinical samples .
Insecure dismissing attachment displays strong avoidant cognitive, emotional and behavioral attitudes intended to avoid activating the attachment system. The discrepancy between the episodic and the semantic memory highlighted by the AAI illustrates this specific psychological mechanism accurately. When an emotionally challenging life event related to an attachment figure enhances suffering, episodic memory tends to dysfunction (often in an adaptive way), hence allowing subjects to construct a corrected representation of their primary attachment relations and maintain the bond. Benedetti  as well as others emphasized the presence of dysfunctional ways of communication in the families of psychotic patients. This is often related to experiences of intrusion, rejection and extreme invalidation of children’s feelings in the context of their relation to a primary caregiver. In such a context, where the deep cognitive and emotional distortions required are useful to the child in maintaining the bond with his/her attachment figures, it seems logical to consider that, when growing up, individuals could be prone to develop an affect regulation mechanism that may contribute at some point to a rupture with socially shared reality. Therefore, repetitive experiences of intrusion, rejection and invalidation of one’s feelings in association with genetic vulnerability might significantly increase the risk of psychosis as demonstrated in epidemiological surveys (e.g. ). This may be reflected in more recent work such as the one conducted on mentalization , which relates historically but also conceptually to the concept of attachment. The capacity to mentalize, defined as “… the activity of understanding behavior in relation to mental states such as thoughts and feelings” , develops optimally in the context of a relationship with a stable attachment. Lack of mentalizing is found in various psychiatric disorders such as autism (in this case probably related to a neurobiological disorder), borderline personality disorder and schizophrenia . Therefore a history of an unstable relationship with caregivers may lead to insecure attachment and later on to altered mentalizing capacities, i.e. an important limitation on the lives of people suffering from psychosis. In this perspective, MacBeth et al.  showed that, independently from symptomatology, reflective function (RF, a measure of mentalization) was higher in psychotic patients with secure attachment than in those with insecure attachment.
As mentioned above, we cannot state that an experience of insecure attachment may represent per se a specific causal factor for psychosis. First, the prevalence of insecure attachment in psychosis was the same as that reported in other clinical populations . Second, there are patients with psychosis who do not report a history of unreliable attachment. Third, some patients, both in MacBeth et al.  study, Gumley et al.  study, and in our data, feature a current secure attachment. Fourth, research shows that a history of unstable relationships may lead to various psychiatric conditions, depending on other factors in the array of biology, psychology or social context. Therefore, unreliable attachment affects the illness trajectory and process and is not a cause of the illness.
Attachment and symptoms
In the two researches conducted on first-episode patients in Scottland [8, 9] no association was identified between attachment categories and symptoms in psychosis, whereas, in our study, secure attachment was associated with less positive symptoms, anxiety/depression, and agitation/mania. This discrepancy could be explained by the nature of the samples (first-episode patients vs. patients with chronic psychosis). Indeed, the process of dealing with psychosis involves cognitive and emotional schemas which may become more visible over time. Therefore, an association between symptoms and a coping process could be more evident in chronic patients rather than first-episode patients.
Some studies code the AAI along two dimensions by the Q-sort method; i.e. security/anxiety and repression/preoccupation. With this procedure, no associations were found between baseline clinical severity and attachment for 31 patients at-risk of psychosis ; whereas, for 40 chronic patients, insecure attachment was associated with more positive symptoms, especially delusion, hallucination and suspiciousness . Data from several studies using auto-questionnaires to assess attachment dimensions provide evidence for a link between avoidant attachment and positive and negative symptoms .
The association of insecure attachment with positive symptoms warrants some comments, beyond the fact that persistent positive symptoms may per se lead to some “insecure attachment”, or conversely that insecure attachment would lead to social or cognitive problems leading to positive symptoms, in the context of an altered process of affect regulation . In the field of diabetes, Ciechanowski et al.  showed that dismissing attachment in the setting of poor patient-provider communication was associated with poorer treatment adherence in patients with diabetes. The differences between groups found by these authors, in medication adherence and glucose monitoring suggested a disengagement from treatment by patients who exhibit dismissing attachment, particularly in the absence of good patient-provider communication. Translated into the field of psychosis treatment, both in its pharmacological and psychosocial dimensions, this work may suggest some hypotheses. Indeed, in addition to other features related to their symptoms, patients with psychosis featuring insecure attachment would have more problems with compliance, due to a poor quality of patient-provider relationship. Hence, a worst profile for positive symptoms may be due to a poorer adherence to treatment, amongst other causes. This hypothesis is in line with research showing that insecure attachment with mental health services may alter the course of patients with psychosis . These authors conclude an interesting clinical implication: security of attachment should be assessed in order to identify patients who might experience difficulties in engaging with services and who may need increased input on this issue.
To note, we did not find an association between negative symptoms and attachment categories. That may be related to a “floor effect”, i.e. to the fact that patients included in the research were stabilized. That means that they were supposed to feature no or little positive symptoms, depending on the magnitude of the effect of their neuroleptic treatment. In association, negative symptoms may appear quite stable in the lower range hence not allowing a clear distinction between categories of subjects for this parameter (our data show almost same results for all groups, with a quite small DS). An alternative explanation would be that attachment style may not influence negative symptoms per se. However, this hypothesis would be in contradiction with literature showing an association between attachment categories and negative symptoms (see Gumley’s et al. review ). Indeed, it is unlikely that the process leading for example to dismissing/avoidant attachment would not entail some behaviors and emotional features which would not be part of negative symptoms. Hence insecure attachment may be related to the deactivation of positive and negative affect . Also, it is unlikely that “primary” negative symptoms would not affect the relationships hence altering the perception of attachment in its assessment.
Attachment and trauma
Concerning the association between psychosis and a high prevalence of childhood trauma related to attachment figures, the small number of subjects in our research hinder us from drawing conclusions, beyond the fact that our results illustrate (rather than demonstrate) the fact that many patients suffering from psychosis report a history of trauma such as abuse, parental neglect and others. This clinical observation is confirmed by recent epidemiological surveys such as the one of Read et al. , or Shevlin et al. , the latter showing an additive effect of multiple traumatic experiences. Indeed, in this research, whereas a single trauma type did not appear to increase the risk of psychosis, experiencing two or more types of trauma significantly increased this risk, with dramatic increases associated with experiencing all types of trauma. Picken et al.  showed in 110 patients with psychosis and substance misuse, that anxious attachment was associated with a number of interpersonal traumas and post traumatic symptoms, like in other clinical populations.
The issue of the influence of trauma on attachment style in this particular population of patients with psychosis was described in the recent meta-analysis of Gumley et al. . Berry et al.  found higher levels of attachment anxiety in subjects who had experienced trauma from caregivers during childhood. Picken et al.  found that attachment anxiety was associated with the total number of traumatic events. Keeping in mind our small sample size for this kind of statistics, we note that this literature is in accordance with our results showing for the whole sample an association between insecure attachment and separation from first figure of attachment and multiple traumatic experiences.
Attachment: clinical implications
The fact that some patients with psychosis feature a secure attachment or an insecure yet preoccupied attachment is another important issue arising from these data. Actually about one out of three patients falls into these two categories. This means that, beyond the stigmatizing view that patients with psychosis may feature an “autistic” relationship (see Laing , or classical literature on first rank symptoms, e.g. Andreasen and Black ), a third of them may be able to have significant and rewarding relationships (through a secure attachment) or a significant investment to others (although to some extent problematic, i.e. in the cases of preoccupied attachment). This may represent for these patients an important goal in their lives, which should be kept in mind by clinicians who often deny them the actual possibility to build high quality relationship such as friendship or romantic bounds . Indeed, this involves that clinicians should encourage patients to build relationships far beyond the social skills training which is offered in many places (e.g. ). That may involve specific psychotherapeutical intervention. In this perspective, targeting mentalization, known to be altered in this population [9, 40] may improve to some extent cognitive, emotional and behavioral attitudes related to attachment. In research on the particular realm of patient-therapist relationship, AAI classification has been shown to predict the kind of collaboration involved in the treatment process . These authors showed that secure and preoccupied patients were more likely, as compared to dismissing patients, to seek emotional closeness with the therapist, dismissing patients were more likely to avoid proximity, and preoccupied patients were more likely to resist the therapist’s support or connection. Even this study was done with patients likely to suffer from disorders other than psychosis, there remains to be assessed whether or not patients with psychosis may feature such pattern. At least this should be considered when treating this particular population of patients with psychosis, taking into account the fact that the quality of social relationships impacts on the disorder’s outcome . In this perspective, Korver-Nieberg et al.  associated the category of attachment to the recovery style and suggested the need to improve attachment security in a context of therapeutic relationship, before encouraging people to explore their experiences of psychosis.
Attachment and spiritual figures
The use of a semi-structured interview and a qualitative coding grid based on attachment theory as well as on the spiritual coping grid developed previously in our research group  allowed us to investigate more deeply the process of spiritual coping in relation to spiritual figures. Firstly, we showed that in the great majority of cases spiritual figures functioned like attachment figures. Having a secure attachment to a spiritual figure (even when there was no secure attachment towards primary caregivers) was associated with lower levels of symptoms such as suspiciousness and anxiety and better coping strategy with regard to self-esteem, depression, hope, relation to others and giving a meaning to life. Hence, some aspects of religion may promote “earned security”, because God is perceived as a loving attachment figure . Furthermore, our results showed that even for the 12 participants (10 patients and two controls) for whom the spiritual figure worked like an insecure attachment figure, this characteristic appeared to be associated with a better symptomatic and/or coping profile. Having an insecure spiritual attachment figure still appears to be more useful than having none.
Even if this study, due to its cross-sectional design, is not designed to address causality, in terms of attachment to religious figure’s style vs. level of symptoms, this issue needs to be discussed. Indeed this question is important in two ways. First, one may question whether stable attachment to a spiritual figure may improve symptoms or if conversely a better symptoms profile may allow patients to “build” a stable attachment to a religious figure. Our results show indeed such an association, although for 8 symptoms quite heterogeneous in their nature. Some of those symptoms, mostly in the “emotional” field, may give support to the first formulation (e.g. a “stable attachment to God” may relieve anxiety and depression); conversely some symptoms, more in the “cognitive” domain such as guilt, suspiciousness and conceptual disorganization, may hinder building such a favorable relation with a spiritual figure.
Also it is worth discussing the fact that the presence of psychotic symptoms such as delusions with religious content may alter the way patients report their attachment to spiritual figures. Indeed some studies  showed that 15 % of stabilized patients with psychosis feature delusions with religious content. Amongst those, some felt having a relationship with God involving some influence of Him, in a delusional way (e.g. being controlled or even persecuted by Him). In the present study, this kind of phenomenon has been considered as a negative form of coping and the sign of an insecure attachment to the religious figure. Yet we should keep in mind that this may be state dependent, hence a reversible condition possibly related to an increase of positive symptoms. Only repeated assessments may possibly address this issue.
The correspondence and compensation hypothesis could explain some aspects of religiosity from a developmental perspective: the “relational part” of religiosity could be either 1) the result of a secure attachment to a spiritual figure, based on the previous development of a secure attachment towards parents or 2) a secure attachment toward a spiritual attachment figure based on a compensation process following a primary insecure attachment towards caregivers. Such an arrangement was described by Granqvist & Hagekull  who reported that parental insensitivity predicted increased importance of the relationship with God.
Attachment models are described as generally stable over lifespan . However, attachment theorists consider that, at some point during lifespan, it is possible to experience different kinds of relationships, these other relations sometimes taking the form of emotionally significant experiences able to transform these IWMs . Therefore, these authors emphasized the need to think in terms of a hierarchy of IWMs over lifespan. This is consistent with our finding that correspondence is more prevalent than transformation in IWMs. The transformations we observed in participants’ IWM in their relation to spiritual figures could be the result of two different processes. Firstly, the relationship could be enhanced by an emotionally significant change experienced by subjects in relation to their beliefs concerning their current relation to fantasized spiritual figure. Secondly, it could be the result of an emotionally significant change that subjects experienced with another relation such as a friend, which was then projected on the representation of the spiritual figure. Our data do not allow us to find out which of these processes is involved. However, Bowlby  insisted that only sensitive experiences were able to change IWMs. From this perspective, our qualitative data indicated that all seven participants who modified their IWM in relation to their spiritual figure reported that they had felt sensitive experiences in their bodies and had experienced specific events which were interpreted as the manifestation of a spiritual figure. Hence, patients’ spiritual interpretations of sensitive experiences may represent what Bowlby described as sensitive experiences in the relation to an attachment figure.
The strength of our research is definitely the use of clinically powerful instruments (such as the AAI, the semi-structured interview for the assessment of spirituality and the SCID for the confirmation of the diagnosis). The length of the investigation process as well as the sensitivity of the AAI led to some drop outs however. Therefore the small size of our sample definitely represents a limit to the external validity of this research. Some limits are based on conceptual issues. Firstly, as discussed above, the fact that psychosis is associated with a higher rate of insecure avoidant attachment is not per se the proof of a causality of attachment disruptions on the later onset of psychosis. This issue pervades the literature on attachment and adult psychopathology (e.g. Dozier et al. ). Only prospective studies could provide support for this connection. Until then, it cannot be excluded that insecure avoidant attachment could be one feature of a severe form of psychosis. The same kind of issue arises when considering the correspondence and the compensation hypotheses. We cannot exclude the possibility that those who are able to compensate are also able to develop a global improvement in their ability to cope with interpersonal issues, because of a milder form of psychotic disorder. This could explain why they feature less persistent symptoms.
Finally the fact that patients were to some extend stabilized may not bring about a sufficient symptom range allowing fully efficient statistical analyses: overall 19 patients (68 %) featured at least one moderate or severe psychotic positive symptom in BPRS.