Cognitive conflicts in major depression: between desired change and personal coherence.

OBJECTIVES: The notion of intrapsychic conflict has been present in psychopathology for more than a century within different theoretical orientations. However, internal conflicts have not received enough empirical attention, nor has their importance in depression been fully elaborated. This study is based on the notion of cognitive conflict, understood as implicative dilemma (ID), and on a new way of identifying these conflicts by means of the Repertory Grid Technique. Our aim was to explore the relevance of cognitive conflicts among depressive patients.
DESIGN: Comparison between persons with a diagnosis of major depressive disorder and community controls.
METHODS: A total of 161 patients with major depression and 110 non-depressed participants were assessed for presence of IDs and level of symptom severity. The content of these cognitive conflicts was also analysed.
RESULTS: Repertory grid analysis indicated conflict (presence of ID/s) in a greater proportion of depressive patients than in controls. Taking only those grids with conflict, the average number of IDs per person was higher in the depression group. In addition, participants with cognitive conflicts displayed higher symptom severity. Within the clinical sample, patients with IDs presented lower levels of global functioning and a more frequent history of suicide attempts.
CONCLUSIONS: Cognitive conflicts were more prevalent in depressive patients and were associated with clinical severity. Conflict assessment at pre-therapy could aid in treatment planning to fit patient characteristics.

Background

Recent research has produced growing evidence for the role of cognitive factors (e.g., automatic thoughts, dysfunctional schemas, attribution bias) in the onset and maintenance of depression (e.g., Abramson et al., 2002). However, internal conflicts have not been considered in the cognitive model of depression even when they have been quite a common notion in psychology for more than a century. Indeed, various psychological theories have underlined the relevance of intrapsychic conflicts leading people to internal struggles that might block their development, giving rise to suffering and symptoms. Foremost, the importance of intrapsychic conflict is highlighted in virtually all psychoanalytic theories (e.g., Horowitz, 1988). Furthermore, a classical notion addressing internal conflict is that of cognitive dissonance, but we can find other more clinically oriented approaches such as those of Grawe’s (2004) incongruence theory, and widely used concepts of resistance and ambivalence in psychotherapy (e.g., Engle & Arkowitz, 2006). Another approach, Perceptual Control Theory (Powers, 2005) proposes a major role of goal conflict and influenced Grawe’s (2004) work. More recently, it has led to the development of method of levels, a transdiagnostic cognitive therapy which focuses specifically on conflict formulation (e.g., Carey, 2008; Mansell, Carey, & Tai, 2012). These authors consider conflicts as common processes underlying psychological distress and so aim at resolving them through reorganizing conflictual cognitive control systems. Unfortunately, and in contrast with the relative abundance of clinical literature, intrapersonal conflict is yet an underresearched topic within contemporary clinical psychology. Empirical research about this topic is needed to clarify to what extent cognitive conflicts might play a role in the onset and maintenance of psychological disorders, and whether they could be important in explaining some patients' difficulties for change (Michalak, Heidenreich, & Hoyer, 2011).

The growing literature around intrapersonal conflict and its role in psychological distress has supported the idea that, although conflict might be quite a common phenomenon, some forms of unresolved inner conflict lies at the core of many psychopathological manifestations (e.g., Carey, 2008; Lauterbach, 1996; Mansell, 2005). Broadly, two kinds of methods have been used to investigate the role of intrapersonal conflicts. On the one hand, various approaches created specific methods for measuring implicit conflicts. For instance, psychodynamic theories developed operationalized systems for assessing conflict as a crucial diagnostic axe (e.g., Cierpka, Rudolf, Grande, & Stasch, 2007; Luborsky, 1977; Perry, 1990). Lauterbach (1996; Lauterbach & Newman, 1999) developed the intrapersonal conflict test for quantifying conflict based on Heider’s (1946) balance theory and, thus, defined conflict as inconsistency between attitudes and beliefs. Studies using these methods have reported correlations between conflicts and symptom severity (e.g., Renner & Leibetseder, 2000) and reduction of conflict after psychotherapy (e.g., Hoyer, Fecht, Lauterbach, & Schneider, 2001). A common feature of these methods is that the elements involved in these conflicts are defined a priori by the researcher.

On the other hand, a second cluster of approaches have developed methods for quantifying conflict among ideographically generated personal goals (e.g., Emmons & King, 1988; Little, Salmera-Aro, & Phillips, 2007). Research on this topic has shown that goal conflict influenced subjective well-being and life satisfaction (see Riediger, 2007, for a review). High levels of goal conflicts were associated with increased levels of negative affect, depression, neuroticism, and psychosomatic complains (see Michalak et al., 2011, for a review). These methods are based mainly on motivational conflicts (e.g., approach-avoidance) and restricted to conscious interference between goals (as reported by participants). Therefore, they may be more sensitive to social desirability and self-presentation effects which may limit their applicability in psychotherapy. This vein of research has shown, nonetheless, how conflict, as a motivational factor, influences experience and behaviour generating behaviour inhibition, motivational deficits, and difficulties in action control (e.g., Emmons, King, & Sheldon, 1993). It is reasonable to assume that these difficulties can be contributing to a variety of depressive symptoms (e.g., reduced activity), to their persistence over time and, especially, to the engagement of clients in the process of therapeutic change. But further empirical studies are warranted to verify the extent and nature of the influence of conflict on depression and other disorders.

In this article, we build on existing support for the relevance of conflicts by studying the role of a particular variety of cognitive conflict, called implicative dilemma (ID) (see Feixas, Saúl, & Ávila, 2009), in a sample of patients diagnosed with major depression. Our study is based on Kelly’s personal construct theory (1955) which provides a suitable conceptual and methodological framework for the empirical study of internal conflicts related to the construction of the self. In brief, Kelly’s theory explores the subjective way in which people construct their experience by analysing their personal constructs, which are bipolar dimensions of personal meanings (e.g., being depressed vs. happy). This theory holds an agentive (Bandura, 2001) and proactive vision of human beings, and so asserts that individuals regulate motivational, emotional processes, and actions on the basis of the congruence or discrepancy between the construction of the ‘self’ and the ‘ideal self’ (coinciding with Carver & Scheier, 1998; Cervone & Shoda, 1999; Higgins, 1987). But self-ideal discrepancy is not necessarily a conflict. For conceptualizing conflicts, personal construct theory acknowledges that humans may employ a variety of constructions which are inferentially incompatible with each other (Kelly’s fragmentation corollary). From this perspective, it is likely that dilemmas arise when a person has to reconcile the self with personally held values. For instance, Rowe (1971) described the case of a chronic depressive patient who faced the dilemma between staying depressed (associated in her construct system with ‘being human’) or change, and become a ‘destructive’ or ‘unpleasant’ person (according to her own vision). This was conflict stemming from the particular configuration of implications of her construct system.

Kelly’s Repertory Grid Technique (RGT; Feixas & Cornejo, 2002; Fransella, Bell, & Bannister, 2004) allows for an individually tailored assessment of conflictual configurations of personal constructs within the client’s cognitive system. Specifically, the notion of ID (Feixas & Saúl, 2004; Feixas et al., 2009) makes reference to those conflicts in which a desired change (e.g., stop being depressed) implies an unwished change (e.g., becoming unpleasant). In this example, we see that a change in specific, symptom level aspects implies a change in identity (i.e., becoming a different kind of person). Operationally, two types of personal constructs are involved in an ID. On the one hand, discrepant constructs are those in which the person perceives a significant discrepancy between the ‘present self’ and the ‘ideal self’ so that one pole of the construct describes the present and the other pole the ideal self. They typically signify areas of malaise, such as symptoms in which change from one pole to the opposite one is desired. On the other hand, congruent constructs represent areas of self-satisfaction (as indicated by the similarity between the present and the ideal self, both described by one construct pole) which might be connected to personal values or beliefs. In the example of Figure1, the participant considered herself as being the kind of person who ‘does not love herself’ (left pole) but she would like to start ‘loving herself’ (right pole of the discrepant construct). At the same time, in congruency with her ideal self, she considered herself as being ‘protective’ (left pole) and did not want to become ‘unemotional’ (right pole of the congruent construct; note that all these constructs are personal, i.e., her own words). The RGT allows for the calculation of correlations among all the constructs elicited from the interviewee. So, whenever an association is found between the desired pole of the discrepant construct (‘love herself’ in Figure1) and the undesirable pole of the congruent construct (‘unemotional’ in Figure1), an ID is identified. Therefore, discrepant constructs per se do not represent a conflict but just a discrepancy, a goal that should be attained. Rather, it is the conflictive association between a discrepant and a congruent construct which causes conflict. In these cases, the need for change (she wants to love herself) might be hindered by the need for self-ideal congruency (continue being protective). What an ID tells us is that the need for change expressed by the discrepant construct is in conflict with the need for coherence expressed by the congruent construct. Thus, the patient unwittingly hesitates in taking a clear course of action because striving for loving herself has negative implications for her identity. In the view of such a dilemma, change may be less likely to occur because abandoning the symptoms would result in invalidation of core aspects of the self.

Figure 1

Example of an implicative dilemma of a depressed patient from the clinical sample.

The measure of IDs is both standardized and quantifiable in structure, and idiographic in content. It is also relevant for explaining both symptom maintenance and ambivalence towards change. In addition, it lessens the effect of social desirability given that conflicting constructs are detected from grid data using a structured computerized procedure (see below) which is not evident to the subject and does not employ any explicit question about conflicts. Thus, this method using IDs could constitute an integration of the advantages of the aforementioned approaches considering that it assesses conflict in a way which is not based in the explicit wording of a contradiction or conflict by the interviewees but by means of more implicit associations among their specific self-generated goals (personal constructs).

Early studies within the Multicenter Dilemma Project (Feixas & Saúl, 2004; www.usal.es/tcp) showed the relevance of IDs across various clinical samples over other types of cognitive conflicts. For instance, Feixas et al. (2009) found that more than half of a group of patients seeking psychotherapy (n = 284) presented this type of cognitive conflict. By contrast, less than a third of the non-clinical group (n = 322) did so. This significant difference was not found with other existing procedures to identify conflicts such as unbalanced triads (e.g., Sheehan, 1981). In a subsequent study with 87 patients it was found that most patients who presented conflicts at the initial psychotherapy assessment did not have those conflicts at post-therapy. Also, resolution of IDs was associated with symptom improvement (Feixas, Saúl, Winter, & Watson, 2008). In fact, in those cases in which conflicts were not resolved had a poorer outcome. These studies suggest that IDs could be targeted for psychotherapeutic intervention and their resolution could be pursued as a way out from suffering, blockage, and recurrence.

Although several treatments have proved their effectiveness in the treatment of depression, more than 20% of all patients with a major depressive episode develop a chronic course (Angst, Gamma, Rössler, Ajdacic, & Klein, 2009; Rubio et al., 2011). To date, there are no conclusive data about risk factors for chronic depression (Hölzel, Harter, Reese, & Kriston, 2011) neither consistent markers nor diagnostic tests to predict recovery (Solomon et al., 2008). Furthermore, duration of illness and intervals between episodes are highly variable from one patient to another which limits clinical decision making. In this sense, cognitive conflicts might help expand cognitive models of depression and provide a theoretical explanation as well as practical procedures to approach relapse and chronicity. The notion of ID seems particularly suitable for that because it encompasses both the patient’s need for change and his or her need for continuity.

Preliminary evidence for increased levels of conflict in depressed patients comes from studies showing significant associations between emotional distress and intergoal interference or attitude inconsistencies. In regard to goal conflicts, although two studies found that conflict was associated with higher levels of depression, two studies failed to replicate the findings in undergraduate and in outpatient samples (see Michalak et al., 2011, for a review). In addition, a clinical study (Püschel, Schulte, & Michalak, 2011) associated motive-goal discrepancies with higher levels of depression and reduced sense of coherence. Research using Lauterbach’s assessment of conflict found inverse relationships between conflict and mood in non-clinical samples. Furthermore, a recent study comparing depressed inpatients and controls showed large differences in conflictual constellations (Stangier, Ukrow, Schermelleh-Engel, Grabe, & Lauterbach, 2007). Finally, two preliminary studies, one with a sample of mixed depressive disorders (Feixas, Montesano, Erazo-Caicedo, Compañ, & Pucurull, 2014) and another with dysthymic patients (Montesano et al., 2014), found higher percentages of clinical participants presenting with IDs as compared to controls. However, to our knowledge, a systematic evaluation of cognitive conflicts in major depression has not yet been conducted.

This study explored the relevance of IDs to major depression by testing three hypotheses. First, based upon previous findings, we hypothesized that IDs would appear in greater frequency and proportion in depressed than non-depressed participants. The results of the above mentioned previous studies were promising but had important methodological limitations such as heterogeneity of diagnosis, small sample size, or inadequate control groups. To increase the generalizability of those findings, this study investigates the role of IDs in major depression by (1) using more careful sampling processes, (2) increasing the number of participants, and (3) refining the assessment methodology.

Our second hypothesis proposed that presence and number of IDs would be associated with different clinically relevant variables. Concretely, we expected that participants presenting with IDs would display higher levels of symptom severity, poorer global functioning, and higher prevalence of recurrent episodes. Finally, we explored the content of constructs forming IDs. In accordance with previous findings (Montesano et al., 2014) and the theoretical structure represented in Figure1, we expected that discrepant constructs would reflect predominantly emotional content connected to depressive symptoms (e.g., happy vs. depressed), whereas congruent constructs would be mostly of moral nature (e.g., good vs. bad) usually considered typical of core constructs, values, and beliefs. We predicted that this thematic configuration would differ from that of non-depressed participants.

Methods

Participants

The overall sample of the study consisted of 271 participants of both genders distributed in two different groups, 161 depressed patients and 110 controls. For the major depression sample inclusion criteria were as follows: (1) 18–70 years of age, (2) meeting DSM-IV-TR (APA, 2000) criteria for major depression, and (3) a score of more than 19 on the Beck Depression Inventory, second edition (BDI-II; Beck, Brown, & Steer, 1996). Exclusion criteria were (1) presence of bipolar, schizophrenia or schizoaffective disorders, (2) presence of psychotic symptoms, (3) current substance abuse, and (4) organic mental disorder, brain dysfunction or pervasive developmental delay. These criteria were adopted following the example of similar studies. In regard to patients' recruitment, 233 patients were referred to the study but 47 were initially excluded due to several reasons (not meeting age inclusion criterion, declining to participate, or not answering the phone). Of the 186 who were assessed, 19 were excluded because they did not meet diagnostic criteria and six rejected further participation in the study. Finally, the clinical sample was composed of 161 patients (78.3% females) recruited from several community health care centres of the city of Barcelona (Spain) and its surrounding area. The presence of other comorbid diagnoses was not criterion of exclusion, but it was recorded as a variable and controlled in statistical analyses. About 40% of participants presented with another Axis I diagnosis: panic disorder, 18.6%; panic disorder with agoraphobia, 8.7%; dysthymic disorder, 6.2%; post-traumatic stress disorder, 3.7%; anxiety disorder not specified, 1.9%; anorexia nervosa, 0.6%. Patients meeting the inclusion criteria and willing to participate in the study gave written informed consent on forms approved by local research ethics committees.

The non-clinical sample was composed of 110 participants (71.8% females). The inclusion criteria were as follows: (1) 18–65 years of age, (2) to score less than 14 on the BDI-II (to ensure that they did not present significant depressive symptomatology), and (3) no history of psychiatric or neurological illness.

The demographic and clinical characteristics of participants are summarized in Table 2002. No significant differences were found between the two groups in gender, age, and years of education whereas, as expected, samples differed significantly in BDI-II scores and number of constructs (e.g., Feixas, Erazo, Harter, & Bach, 2008; Feixas, Saúl, et al., 2008).

Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I; First, Spitzer, Gibbon, & Williams, 1996)

The SCID-I is a semistructured interview that includes modules designed to assess either the lifetime or current experience of categorically defined DSM-IV-TR Axis I psychiatric disorders. SCID-I diagnoses display adequate test–retest reliability for major depressive disorder in clinical samples (K = .66; Lobbestael, Leurgans, & Arntz, 2011). Information regarding diagnosis, global assessment scale scores, and history of suicide attempts was also gathered for each depressed participant.

Beck Depression Inventory-II (BDI-II; Beck et al., 1996)

The BDI-II is a self-report measure of depression that has demonstrated strong internal consistency in psychiatric outpatients (.92). It was used to assess participants' symptom severity. Jacobson and Truax (1991) formula to distinguish functional and dysfunctional population yielded a cut-off of 16.92 using Spanish normative data (Sanz & Vazquez, 2011). Having that score as a reference, we limited to 14 the maximum score for the control group and a minimum of 19 for the clinical sample as a way to control the symptom level within groups.

Repertory Grid Technique

The RGT is a constructivist assessment procedure originally proposed by Kelly (1955). It is a semistructured interview in which the interviewer first elicits elements (present self, ideal self, and significant others identified by the participant). Then, these elements are considered in dyads (two elements at a time) to elicit personal constructs by asking for similarities and differences between each pair of elements (e.g., both mother and sister are ‘friendly’). For each similarity or difference described by the participant, an opposite construct pole is also elicited (e.g., ‘friendly vs. unconcerned’). Elicitation continues until the person is unable to generate additional constructs and, then, the interviewer asks the participant to rate each element on each of the elicited personal constructs. These ratings use a 7-point Likert scale ranging from very much like the left pole of the construct to very much like the contrasting right pole. This provides a matrix of ratings for each participant’s grid, with columns representing important people of her or his interpersonal world (elements) and rows representing construct dimensions (see Figure2 for an example). Several indexes can be obtained through the mathematical analysis of the grid matrix; for the purpose of this study, two measures related to IDs were used:

Figure 2

Example of the repertory grid of a patient of the clinical sample.

Presence of IDs: According to Feixas and Saúl (2004) a construct is classified as discrepant whenever the ratings assigned to the element ‘present self’ and the element ‘ideal self’ differ on 4 or more points within the 7-point scale used. Conversely, those constructs in which the score given to ‘present self’ and ‘ideal self’ coincide (or there is no more than 1-point difference) are identified as congruent constructs. Whenever either of the two elements is rated as 4 (the middle point), then it is excluded from this classification. An ID is detected whenever the correlation between the ratings given to a discrepant construct and those given to a congruent construct is .35 or higher so that the desired pole of the discrepant construct is associated with the undesired pole of the congruent construct (see Figure1). It is important to note that the desirability of the poles is regarded by the score that participants give to their element ‘ideal-self’ for each construct. The cut-off of .35 is based on Cohen’s guidelines (1988) indicating that a correlation equal or higher than .30 implies a medium strength of the relationship for which we have summed a conservative add-on of .05 points.

Percentage of implicative dilemmas (PID): This percentage reveals the number of IDs in a participant’s grid taking into account its size. Since the number of constructs elicited during RGT administration varies across participants (thus influencing the possible number of IDs within a grid) this measure is calculated by dividing the number of dilemmas of a grid by the total number of possible combinations of constructs of such grid taking two at a time. The resulting proportion is multiplied by 100 to find the percentage.

The reliability of the RGT has been estimated with test–retest studies providing stability scores of 71–77% for the elements, and 47.7–69% for the elicited constructs. With respect to the measures derived from the RGT, studies provide test–retest correlations ranging from .61 to .95 (Feixas, López Moliner, Navarro Montes, Tudela Marí, & Neimeyer, 1992; see Feixas & Cornejo, 2002, for a review).

Classification System for Personal Constructs (CSPC; Feixas, Geldschläger, & Neimeyer, 2002)

The CSPC is a system designed to analyse and classify the content of personal constructs derived from constructivist assessment procedures, and particularly from the RGT. The CSPC is composed of 45 exclusive and mutually excluding categories arranged into six thematic areas (moral, emotional, relational, personal, intellectual, specific values, and interests). Feixas et al. (2002) reported a high reliability index for the CSPC (K = .93). It has already been applied to personal constructs of depressed (Montesano, Feixas, & Varlotta, 2009) and fibromyalgia patients (Compañ et al., 2011).

Procedure

Participants in the depressed group were recruited through medical referrals from several health care centres from the city of Barcelona from 2008 to 2011. After referral they were contacted for assessment sessions in their respective centres. Two independent and specifically trained master-level students administered the instruments in face-to-face interviews. These evaluators used the SCID-I, the BDI-II and other supplemental information to determine if the patient met inclusion criteria. Control group participants were recruited through agreements with cultural and civic associations to which free psycho-educational talks were offered in compensation for their participation in the study. Also, a call was made among graduate and undergraduate students to refer non-clinical participants (friends and relatives) for the study. Non-depressed participants were also administered the BDI-II to rule out depressive symptoms and to ensure the fulfilment of inclusion criteria. All participants provided informed consent. A second appointment was then scheduled to complete the assessment process with the RGT. Conflicts were analysed post hoc with the GRIDCOR v. 4.0 (Feixas et al., 2002) software and exported into the Statistical Package for Social Sciences version 20.0 (IBM Corporation, Armonk, New York, NY, USA) datasheet for further analyses. Finally, two independent well-trained graduate and undergraduate students codified with the CSCP the content of the 907 constructs forming participants' IDs (K = .98; p < .001). Raters did not know whether the grids corresponded to the clinical or the comparison group nor the hypotheses of the study.

Results

Presence and number of implicative dilemmas

With the purpose of checking whether the proportion of participants presenting at least one ID was different across samples, a chi-square test for independence was performed. The results indicated a statistically significant association between presence of IDs and type of sample – χ2(1, N = 271) = 30.08; p < .001 – with a moderate effect size (φ= −.33). As shown in Table 2000, IDs were found in more than two-thirds of the depression sample in contrast to about one-third of controls. To examine the influence of gender, the proportion of male and female participants with and without IDs was compared within both samples. Similar proportions of participants with and without IDs were found in the clinical (females = 69 vs. 31%; males = 65.7 vs. 34.3%) and the non-clinical samples (females = 34.2 vs. 65.8%; males = 35.5 vs. 64.5%). As expected, there were no statistical differences between gender and presence of IDs in the depression group – χ2(1) = 0.14; p = .71 – neither in the comparison group – χ2(1) = 0.17; p = .89. The influence of presenting a comorbid diagnosis within the clinical sample was also tested, with results indicating no differences – χ2(1) = 0.20; p = .88 – in the presence nor in the number of IDs (U = 3,008; p = .69) between clinical participants with (M = 2.60; SD = 3.76) and without (M = 2.71; SD = 3.65) another Axis I diagnosis.

With respect to the number of IDs, an independent samples t-test was conducted to compare the number of IDs between participants presenting IDs of both groups. There was a significant difference – t(94.37) = −3.27; p = .002 – in PID scores for depressed (M = 3.90; SD = 3.88) and control participants (M = 2.04; SD = 2.65). The magnitude of the differences (mean difference = −1.85, 95% CI: −2.98 to −.73) was large (d = .67).

Implicative dilemmas and clinical measures

Several analyses were performed to test the relationship of IDs with various clinically relevant variables. Significant differences were found – t(249) = −5.47, p < .001 (two-tailed) η2 = .10 (medium effect) – in BDI-II scores for participants with (M = 28.55; SD = 15.53; n = 148) and without IDs (M = 17.58; SD = 17.13; n = 123). To explore the relationship between symptom severity and number of IDs (in participants presenting IDs), BDI-II and PID correlations were calculated using Spearman’s coefficient. There was a medium positive correlation between the two variables (ρ =.29, n = 148, p = .001). PID scores explained 8.41% of the variance in respondent’s scores on BDI-II.

The relationship between IDs and chronicity was inspected attending to the recurrence of major depression episodes. In regard to presence of IDs and type of diagnosis (single episode vs. recurrent), no significant difference was found – χ2(1, n = 161) = 0.93; p = .33. However, it is worth highlighting that of the patients with a diagnosis of recurrent depression (n = 80), 72.5% presented IDs, and 27.5% did not. Furthermore, comparing the number of dilemmas for those participants presenting with IDs, a Mann–Whitney test (U = 1,140; z = −2.20, p = .03, r = .21, small effect) indicated a significant difference in the PID of single episode (M = 3.24; SD = 3.86; n = 52) and recurrent depression patients (M = 4.49; SD = 3.84; n = 58).

The level of global functioning of depressed participants was examined in relation to the presence and number of IDs. Significant differences in the global assessment scale scores were found for participants with IDs (M = 56.3; SD = 7.3; n = 110) and without IDs – M = 59.7; SD = 7.3; n = 51, t(159) = 2.76, p = .006; η2 = .046, medium effect. In addition, Spearman’s correlation coefficient revealed a moderate negative correlation of global functioning with the PID, ρ = −.26, n = 161, p = .001. Another clinical issue observed was the relationship between IDs and suicide attempts. Within the clinical sample, 22 depressive patients (13.7%) had tried to commit suicide at least once. Among these, 19 (86%) presented IDs while three did not. The BDI-II mean of these participants was examined to check out whether this difference was attributable to the level of symptomatology. Patients with IDs showed a lower score (M = 36.68; SD = 11.41) than participants without conflicts (M = 45; SD = 4). Small sample size precluded further statistical comparisons between these groups.

Content analysis of implicative dilemmas

Since previous results indicated that IDs might be a relevant structure in the cognition of major depressive patients, a content analysis of its components was performed. We observed statistically significant differences when comparing depressive patients and controls with respect to the content of congruent – χ2(5, n = 546) = 19.5; p = .002; Cramer’s V = .189 – and discrepant constructs – χ2(5, n = 332) = 15.76; p = .008; Cramer’s V = .218. With regard to congruent constructs, moral (e.g., responsible vs. irresponsible) and relational (e.g., tolerant vs. authoritarian) areas were more prevalent in the clinical sample, and congruent constructs were more emotional (visceral vs. rational) and personal (organized vs. disorganized) in the comparison group. On the other hand, discrepant constructs were more frequently coded as emotional for the clinical sample and as intellectual for non-depressed participants, according to the adjusted standardized residual values (cut-off point of 1.96 for an α = .05, see Table 2009).

We also tested whether discrepant and congruent constructs presented different areas of content within the clinical sample and which categories were more relevant for each type of construct. Results showed that congruent constructs differed significantly (medium effect size) from discrepant constructs in the type of content coded – χ2(5, n = 684) = 121.15; p < .001; Cramer’s V = .421. Congruent constructs belonged mostly to the moral and relational areas, whereas discrepant constructs were predominantly of emotional and personal content. For congruent constructs, the most frequent categories within the moral area were ‘altruist–selfish’ (30.5%), ‘good person–bad person’ (24%), and ‘responsible–irresponsible’ (16.2%). Categories within the relational area were ‘pleasant–unpleasant’ (17.7%), ‘tolerant–authoritarian’ (16.9%), and ‘extroverted–introverted’ (15.4%). On the other hand, the most frequent categories for discrepant constructs within the emotional area were ‘balanced–unbalanced’ (32.7%), ‘optimistic–pessimistic’ (23.5%), and specific emotions (22.4%). In the personal area only the category ‘strong–weak’ (20%) stood out.

Discussion

This study provided promising evidence for the relevance of IDs in major depression. Indeed, inverse patterns were observed between samples: while over two thirds of depressed patients presented at least one ID in their repertory grids, when assessed in the control group this proportion was only one third. This difference was underlined by the fact that, among those with at least one ID, depressed participants doubled the number of IDs of controls. So, results indicated differences between depressed and non-depressed participants regarding both prevalence of subjects with conflicts and individual frequency of conflicts. Results from this study not only confirm those of a preliminary study of Feixas et al. (2014) with depressive patients but they are also consistent with previous findings pointing out that IDs are more frequent in a variety of clinical samples (e.g., Compañ et al., 2011; Dada, Feixas, Compañ, & Montesano, 2012; Melis et al., 2011) including dysthymia (Montesano et al., 2014). Arguably, this type of cognitive conflict is not specific to depression but might cut across several diagnoses. Nevertheless, the proportion of participants with IDs in our study was higher than in other diagnostic entities, which lead us to consider that IDs play a significant role in depression.

The relevance of cognitive conflicts in major depression becomes more noticeable by observing the relationship between IDs and some clinical measures. In regard to symptom severity, higher levels of symptoms were associated with both presence and number of IDs in the sample as a whole. However, the correlation between BDI-II and PID explained 8.4% of the variance which is only a moderate support for the hypothesis that the number of these cognitive conflicts is associated with higher level of symptoms. It might also be the case that the relationship between symptom level and number of IDs is not linear. It would be informative to explore with a prospective design whether these cognitive conflicts are associated with higher rates of relapse and recurrence. In this study, we used a retrospective measure of chronicity comparing patients with a single episode to patients with a recurrent diagnosis. Our results indicated no significant difference between groups in relation to the presence of IDs, but patients with recurrent depression showed a tendency towards a higher number of IDs. Future research should include also more refined assessments of the course of the disorder taking into account, for instance, age of onset, number of episodes, and time between them.

An unexpected finding of the study was that clinical participants with IDs might have a higher risk of committing suicide. Although the low number of cases precluded performing statistical analyses, an inspection of BDI-II scores allowed us to discard attributing this effect to symptom severity. In fact, there were only three patients who did not present any cognitive conflicts of the 22 who tried to commit suicide. These three participants showed BDI-II scores much higher than the mean of the clinical sample. By contrast, the remaining 19 showed almost the same symptom severity than the whole clinical group. These preliminary results warrant further research to elucidate whether presence of identity-related conflicts, and the blockage associated with them, could be considered as a marker of suicidal attempts and constitute a mediating factor in the cognitive processes involved in suicidal behaviour of depressed patients.

Differences between groups were also found in relation to the global assessment scale. Among depressed participants, those who presented one or more IDs tended to display lower levels of global functioning. Likewise, number of dilemmas was inversely associated with global functioning scores indicating that patients with many IDs in their grids had a poorer level of functioning. Epidemiological studies point out that most major depressive patients tend to have, at some point in their evolution, substantial impairment of family, work, or social life (Kessler et al., 2003). This study suggests that those patients with more IDs might be contributing the most to those unfortunate effects.

Content analysis of the patients’ personal constructs offers a clearer image of how IDs might be involved in symptom maintenance. As shown above, these identity-related conflicts in depressed patients involve the association of symptom improvement with undesired qualities such as negative moral values or undesired relational positions. This reinforces the hypothesis that change may be less likely to occur (unless these dilemmas are dealt with in the therapy context) because eliminating the symptoms would imply also abandoning core constructs, values, or beliefs. To a large extent, our results concur with Rowe’s observation (1983) about a typical depressive dilemma ‘I'd rather be good than happy’ (p. 87).

Limitations and clinical implications

While the results of this study have clarified the role of IDs in major depression, there are limitations which need to be considered when interpreting the results and looking at clinical implications. First of all, it is important to note that the cross-sectional design of this study precluded causal analysis of reported associations. Second, as mentioned above, prospective design research is needed to determine the influence of IDs over the clinical course and rates of relapse and recurrence in depression. Third, in this study, the role of negative self- views of depressed participant’s was not controlled. Arguably, the higher intensity and prevalence of conflicts may respond to increased negativity of depressed participants' grid-ratings. However, it is important to note that the measure of IDs involves not only negative but also positive self-constructions. For every discrepant construct there is another congruent. In the light of our results, future research should elucidate whether the self of depressed patients is predominantly conflictual or negative. Fourth, we did not control the use of medication. Typically, drug treatments vary both within and between subjects, and this might have influenced the findings regarding recurrence and severity of symptoms. Finally, another limitation affecting generalizability is that our samples were selected so to include those individuals with high and low levels of symptom severity for clinical and control groups respectively. Although this selection helped to better depict the differences between samples, it is known that the distribution of depressive symptoms in clinical (diagnosis of major depression) and general populations overlap.

Our findings nonetheless lead us to some preliminary clinical implications. First, the concept of ID might help explain the difficulties that therapists frequently encounter in fully engaging depressed patients into therapeutic change. For instance, a prototypical intervention in cognitive therapy for depression consists of identifying negative self-appraisals and directly challenging patients’ beliefs about themselves. However, such intervention would increase its efficiency by taking into account the possible positive attributes related to symptoms and negative self-perceptions. In doing so, the intervention could better help in decreasing the chance of symptom persistence, which may manifest itself in ‘resistance’ to change or relapse, due to the lack of compatibility of the proposed change with the self-system. In this sense, it is important to remark that the concept of ID takes into account not only the need for change (discrepant constructs) but also the need for continuity and identity coherence (congruent constructs). Second, therapeutic strategies aiming at resolving cognitive conflicts should be helpful in alleviating depressive symptomatology. For instance, psychological interventions could specifically target dilemmas insofar as they could constitute a focus of therapeutic work and allow individualizing the design of the treatment to fit patient subjective experience. It is noteworthy that although the concept of ID arose from a particular theoretical framework (personal construct theory), a dilemma-focused intervention could constitute an ‘add-on’ for a broad range of therapeutic models as it is being tested out in combination with CBT (Feixas et al., 2013). Thirdly, IDs can be detected by means of the RGT. Thus, treatment procedures for preventing future relapses may benefit from screening for the presence of such conflicts and targeting those for intervention in case they appear in the patient’s grid. Certainly, knowledge of cognitive conflicts hindering change processes could shed light on the personal factors that maintain or worsen the disorder.

As said, there are already some therapeutic models that pay special attention to conflicts and ambivalent appraisals in relation to change processes. For example, the motivational interviewing model upholds that depressed patients tend to be ambivalent towards change. Decisional balance becomes then a preferred intervention (Arkowitz & Miller, 2008). Coherence therapy (e.g., Ecker & Hulley, 2008) also pointed out that symptom maintenance is linked to the need of preserving the coherence of self-identity. The aforementioned method of levels therapy focuses exclusively on addressing conflict. Our findings could be seen as empirical support for these models addressing ambivalence and conflict in the psychotherapeutic process. Regardless of the actual model adopted, it seems reasonable to assume that identity-related conflicts play a relevant role in depressed patients' cognitive system and change processes. Furthermore, it raises the need for the study of the effectiveness of a dilemma-focused intervention for major depression (Feixas et al., 2013), which should mainly set the agenda for our future research.

Table 1

Demographic characteristics of depressed patients and non-clinical participants

Characteristics	Depression group	Control group	p value
Gender (female:male)	126:35	79:31	.29
Age (M; SD)	47.1; 11.3	44.5; 14.4	.11
Years of education (M; SD)	11.9; 2.4	12.2; 3.8	.52
No. of constructs (M; SD)	18.8; 5.1	23.2; 7.4	<.001
BDI-II (M; SD)	36.1; 9.8	5.3; 4.4	<.001

Note

M = mean; SD = standard deviation; BDI-II = Beck Depression Inventory II.

Table 2

Presence of implicative dilemmas

	Presence of implicative dilemmas
No	Yes
Depression, n = 161
Fr	51	110
%	31.7	68.3
Non-clinical, n = 110
Fr	72	38
%	65.5	34.5

Note

Fr = frequency.

Table 3

Differences between groups in the content of congruent and discrepant constructs forming implicative dilemmas

	Moral	Emotional	Relational	Personal	Intellectual	Specific interests
CC
Depressed
%	36.7	11.2	31	13.6	1.7	6
Asr	2.2	−2.3	2.1	−2.8	−1.5	0.2
Non-clinical
%	26.2	19	21.4	23.8	4	5.6
Asr	−2.2	2.3	−2.1	2.8	1.5	−0.2
DC
Depressed
%	11	37.1	21.6	26.5	2.7	1.1
Asr	1.3	2.1	−0.9	−0.5	−2.8	−1.8
Non-clinical
%	5.9	23.5	26.5	29.4	10.3	4.4
Asr	−1.3	−2.1	0.9	0.5	2.8	1.8

Note

CC = congruent constructs; DC = discrepant constructs; Asr = adjusted standardized residuals.