How secure is the secure base? Romantic attachment, emotion regulation, and psychotherapists’ role as secure base figures

Abstract

In 1988, Bowlby posited that the emotional availability of psychotherapists in establishing a secure base environment is influenced by their personal relational history. Despite the acknowledged influence of the therapist’s attachment on therapeutic processes and outcomes, the therapist’s role as a secure base figure has received insufficient attention. This study delves into the connection between psychotherapists’ attachment organization and their self-perceived roles as secure base figures within clinical contexts. Additionally, we explore the mediating role of emotion regulation processes in this context. The dataset comprises self-reports from 384 psychotherapists with diverse theoretical orientations. Our analysis reveals both direct and indirect effects of psychotherapists’ attachment on the provision of a secure base. Emotion regulation, specifically through the dimension of clarity, emerges as a significant mediator in this relationship. This study offers a distinctive contribution to deepening our understanding of the relational dynamics inherent to psychotherapeutic practice. It sheds light on the nuanced relation between attachment and emotion regulation, influencing the psychotherapist’s role as a secure base figure in psychotherapy. The discussion of the results additionally emphasizes key implications for clinical practice and therapists’ training.

Introduction

Contrasting with Bowlby’s (1988) disappointment concerning the scarce use of the attachment framework in clinical settings, attachment is nowadays considered an important dimension of the psychotherapeutic process and outcome, with the potential to either facilitate or inhibit psychotherapy effectiveness (Holmes, 2011; Slade & Holmes, 2019). Although with less enthusiasm when compared with the client’s attachment literature, research has been addressing therapists’ attachment as a significant predictor of the therapeutic encounter (Daniel, 2006; Degnan et al., 2016; Heinonen & Nissen-Lie, 2020; Lingiardi et al., 2018; Ryan et al., 2023; Steel et al., 2018, for reviews). The operationalization of attachment varies across studies, with some researchers using interviews to evaluate attachment state-of-mind, while others relied on self-report measures, which encompassed assessments of both romantic attachment and general attachment orientations across diverse relationship contexts. Secure therapists tend to exhibit less negative countertransference behaviors, such as being critical or rejecting (Ligiéro & Gelso, 2002), are more attentive to the clients’ underlying needs, and provide relational experiences that challenge pre-existing schemas or models of the world (Bennett, 2008; Dozier et al., 1994; Tyrrell et al., 1999), and tend to cope better with clients who report higher levels of distress during therapy (Mikulincer et al., 2013). A longitudinal study conducted by Sauer et al. (2003) has shown that anxious-attached psychotherapists developed stronger working alliances at the beginning of the therapy compared to less anxious-attached ones, but this pattern seems to be reverted during the therapeutic process. If therapists who exhibit secure attachment status have been considered more responsive to their clients than their more insecure counterparts, the underlying mechanism of this process is still largely unknown (Carvalho & Matos, 2021; Mikulincer & Shaver, 2007; Talia et al., 2017; Talia et al., 2020). Considering the potential effects that attachment might carry for psychotherapy, shedding light on the underlying processes that lead therapists to develop helpful relationships and work as secure base figures for their clients could be particularly valuable for training and supervision (Farber & Metzger, 2009; Pistole, 1999; Talia et al., 2019).

Therapists as secure base figures

The bridge between early relational experiences and beliefs and expectations in late relationships has been paved by central concepts in attachment theory, such as internal working models of self, others, and the social world (Thompson, 2008). However, it is in the secure base concept that Bowlby saw the greatest potential of attachment theory for counseling and psychotherapy. In fact, for Bowlby (1977, 1988), the therapist’s role as a secure base could provide the emotional conditions from which the clients could explore themselves and the world and begin to revise their representational models of themselves and others. Later, research has shown that client representations of the therapist as a secure base figure seem to be a critical predictor of positive outcomes (Goodwin et al., 2003; Levy et al., 2011; Mallinckrodt et al., 2017). Considering that psychotherapist-client relationships share several features of attachment relationships (Bowlby, 1988; Dozier et al., 1994; Levy & Johnson, 2019; Slade & Holmes, 2019) can offer meaningful contributions to understanding responsiveness in psychotherapy, such as empathic concern or personal distress (Fabi et al., 2019). Therapists’ representations of attachment can therefore inform secure base responses (Romano et al., 2008; Slade & Holmes, 2019; Talia et al., 2020), their mentalizing abilities, and nurturing and responsive behaviors in therapy (Talia et al., 2020).

Studies on the mentalizing stance of the therapist found that attachment affects the therapist’s mental states, which in turn are involved in alliance variations between dyads and over time (Barreto & Matos, 2022). However, there is still a lack of research operationalizing the therapeutic relationship from an attachment perspective (Steel et al., 2018). A critical dimension of secure base behavior was more recently explored by Talia et al. (2020). The authors found that psychotherapy’s distinctive patterns of attunement towards the patient were highly associated with the therapist’s attachment classification as assessed by a transcript-based instrument on therapeutic sessions (Talia et al., 2020). One interesting finding concerns the implications of therapists’ secure attachment status on their ability to be open-minded and deal with ambiguity and doubt (Talia et al., 2020).

In another study, Carvalho and Matos (2021) also presented the development of a self-report questionnaire for psychotherapists’ ratings of secure base behaviors in psychotherapy in four distinctive dimensions [Secure Base Questionnaire (SBQ) - sensitivity, encouragement of exploration, compulsive caregiving, and avoidance of uncertainty]. They found that the therapist’s self-reported romantic attachment was associated with distinct representations of care and the client’s needs in therapy. Insecure dimensions of attachment, such as dependence and ambivalence, were associated with more insecure patterns of secure base provision, namely with a higher prevalence of compulsive caregiving behaviors and more difficulties regarding tolerance for uncertainty in the psychotherapeutic process (Carvalho & Matos, 2021). Both of these studies provide additional proof of the therapist’s attachment role when facing more self-challenging experiences in psychotherapy. Emotional arousal could be triggered by emotionally distressing moments in psychotherapy, namely the ones that seem to question the psychotherapist’s competence in managing boundaries, ruptures, and dealing with uncertainty (Barreto et al., 2020; Carvalho & Matos, 202; Mikulincer et al., 2013; Miller-Bottome et al., 2018; Rubino et al., 2000; Schauenburg et al., 2010).

The role of emotion regulation

Emotion regulation is a core construct in attachment theory (Bowlby, 1977; Cassidy, 1994; Mikulincer & Shaver, 2019) and refers to the processes that influence how and when emotions are experienced and expressed (Gross, 1988). More insecurely attached individuals face additional challenges when accessing emotions or experiencing heightened negative affect (Cassidy, 1994; Parada-Fernández et al., 2021; Shaver & Mikulincer, 2014). More specifically, avoidant people tend to block emotions that could threaten or promote vulnerability states, even if this implies dismissing their own emotional reactions. Differently, anxiously attached individuals’ hyperactivating strategies tend to exaggerate the appraisal process, heightening the threatful features of events and developing hypervigilant and pessimistic evaluations of their self-ability to manage distress (Mikulincer & Shaver, 2013, 2019).

In clinical settings, experiencing complex, diversifying, and chaotic feelings is inherent to the psychotherapist’s life (Najavits, 2000). From an attachment theoretical perspective, working with clients’ emotions can elicit discomfort in insecurely attached therapists, enhance withdrawal in more avoidant therapists, and increase over-involvement in anxiously attached ones (Degnan et al., 2016). Research found that a therapist’s attachment status could indirectly affect relational therapeutic dimensions and outcomes. Cologon et al. (2017) found that although attachment did not predict a therapist’s effectiveness, there was an important interaction effect with the psychotherapist’s reflective function.

Attachment security seems to compensate for low levels of reflective functioning, while high levels of reflective functioning seem to compensate for a more insecure attachment (Colognon et al., 2017). A similar finding was found with attachment and emotion regulation processes on relational dimensions, namely the working alliance (Ruiz-Aranda et al., 2021). Although there was no significant predicted effect of attachment on the therapeutic alliance, moderation analysis yielded an interaction effect between attachment and emotion regulation (Ruiz-Aranda et al., 2021). In this study, a secure attachment did not necessarily predict better alliances since this relation depended on the therapists’ ability to understand and manage their emotions (Ruiz-Aranda et al., 2021). For example, when more insecure attached therapists were able to attend to and regulate their emotions, the therapeutic relationship was not affected by the quality of the therapist’s attachment (Ruiz-Aranda et al., 2021). It is also possible that when therapists feel emotionally overwhelmed, attachment security could hinder emotional triggers elicited by the moment. For instance, Fuertes et al. (2019) found that therapists with higher levels of anxiety and avoidance in attachment dimensions reported more difficulties in developing a genuine relationship with their clients. In this sense, emotion regulation processes can be considered necessary linking mechanisms to understand the complex dynamics between attachment and the provision of a secure base in the psychotherapeutic encounter.

Current study

Considering the above-mentioned literature, the main aim of this study is to analyze how current attachment dimensions are associated with how therapists represent themselves as secure base figures for their clients. We will also analyze the potential mediating role of emotion regulation in the previous association.

It is expected that secure attachment dimensions will be positively related to dimensions more closely associated with a more secure script of the “secure base” (i.e., with SBQ’s dimensions of sensitivity and encouragement of exploration).

Similarly, it is expected that insecure attachment dimensions will be associated with over-involvement and difficulties in dealing with uncertainty during therapy (i.e., with SBQ’s dimensions of compulsive caregiving and avoidance of uncertainty) (Carvalho & Matos, 2016; Romano et al., 2008; Slade & Holmes, 2019; Talia et al., 2020). Emotional regulation processes are also expected to mediate therapists’ attachment representations and secure base provision in therapy (Degnan et al., 2016; Fuertes et al., 2019; Mikulincer & Shaver, 2014, 2019; Ruiz-Aranda et al., 2021).

Methods

Procedure

The sample was recruited over a period of 6 months, both online and offline. First, all societies and schools of psychotherapy in Portugal were reached to present the study and ask for collaboration on the data collection process. Questionnaires were administered in loco or sent with a pre-stamped envelope. From a total of 512 questionnaires delivered, 230 were sent back (a return rate of 44.92%). Second, a snowball sampling technique was also used. A link to the online questionnaire (allocated by the faculty web services, in compliance with the legal and ethical requirements, including the full anonymity of the respondents), was sent to a pool of psychologists and psychiatrists from all over the country available online (universities, health services, private practices), asking for their participation as well as to forward the email to other colleagues. A total of 154 questionnaires were collected through this method. The final sample included 40.1% of questionnaires in web-based format and 59.9% in paper-and-pencil format. Participation was voluntary and anonymous, and no financial compensation was involved. The study followed all ethical requirements and was vetted and approved by the Scientific Committee of the Faculty of Psychology and Education Sciences of the University of Porto.

Participants

Other than being a psychotherapist, no inclusion/exclusion criteria were applied, including age, years of training, profession/position held, or experience. The sample used in the current study has been used in a previous study for addressing the psychometric properties of a self-report scale (Carvalho & Matos, 2021). The sample consisted of 384 psychotherapists, 73 males, and 311 females, ranging from 22 to 68 years old [mean (M)=33.3; standard deviation (SD)=8.05]. Years of experience ranged from 6 months to 40 years (M=7.8; SD=6.22). Regarding professional background, 358 were psychologists, 11 were psychiatrists, 8 were physicians with psychotherapeutic training, and 7 considered themselves psychotherapists but did not specify their professional background. The strong imbalance between males and females is not completely surprising, given that the majority of the sample is composed of psychologists, who are overwhelmingly women (in the Portuguese context). However, the imbalance between professional backgrounds, namely between psychologists and psychiatrists, is more difficult to explain, given the absence, to the best of our knowledge, of an authoritative source regarding the characterization of Portuguese psychotherapists.

Therapists were not forced to self-identify with one single theoretical orientation; rather, they were asked to rate (on a 1-5 Likert scale) the degree to which they identified with several theoretical orientations. For the descriptive analysis, we consider participants who responded 4 or 5 on the scale (i.e., “I identify myself with this orientation” and “I completely identify myself with this orientation”, respectively) as having a salient theoretical identification with that orientation. Consequently, percentages do add up to more than 100%. Most psychotherapists reported being influenced by two or more orientations (84.9%). The percentages for the most salient theoretical orientation reported were as follows: 48.9% of the therapists rated 4 or 5 for cognitive-behavioral, 50% for constructivist, 48.2% for humanistic/existential, 38.8% for integrative, 29.7% for psychoanalytic/psychodynamic, and 64.3% for systemic. The implications of theoretical orientations on SBQ dimensions could be consulted elsewhere (Carvalho & Matos, 2021).

Clinical practices were mainly developed with adults (68.2%), followed by adolescents/young adults (44.8%), children (35.7%), and seniors (6%). Again, categories were not mutually exclusive.

Measures

Demographics and professional

Participants completed a demographic questionnaire addressing age, gender, and professional dimensions, including questions on professional background, years of experience, and theoretical orientations. For assessing theoretical orientations, therapists were asked to indicate on a 5-point Likert-type scale to what extent they identified themselves with each of the following six (not mutually exclusive) categories: cognitive-behavioral, constructivist, humanistic/existential, integrative, psychoanalytic, systemic, and other.

Attachment

For assessing the therapists’ attachment, the brief version of the Romantic Attachment Questionnaire (RAQ) (Matos & Costa, 2001), regarding the current or longest-lasting romantic relationship, was used. RAQ is a self-report measure designed to evaluate adult representations of romantic attachment. Principal component analysis and confirmatory factor analysis, using Portuguese independent samples (Ávila et al., 2012; Vieira et al., 2012), evidenced a reliable 4-factor structure. The brief version is composed of a total of 25 items, divided into four dimensions: i) trust [5 items, e.g, “I know that I can count on my partner whenever I need him/her” (α=.81)]; ii) ambivalence [8 items, e.g., “Sometimes I think that he/she is critical in my life, other times I don’t” (α=.78)]; iii) dependence [6 items, e.g., “When I cannot be with my partner, I feel abandoned” (α=.73)]; iv) avoidance [6 items, e.g., “When I have a problem, I prefer being alone instead of being with my partner” (α=.72)]. Expected associations between RAQ dimensions and other constructs, including subscale scores of the Experiences in Close Relationships Scale, were found to provide evidence for its construct validity (for a review, see Ávila et al., 2012).

Importantly, the RAQ was initially validated in Portuguese samples (Matos & Costa, 2001) and has since been used in several studies, particularly within the Portuguese context (including in the current study). Given its consistent record regarding psychometric properties, the RAQ is arguably the most suitable measure for assessing attachment in a sample of Portuguese therapists.

Emotion regulation

To assess the psychotherapist’s ability to attend to, discriminate against, and regulate emotions, the Portuguese version (Cabral et al., 2021) of the Trait-Meta Mood Scale (TMMS) (Salovey et al., 1995) was used. The original (30-item) 3-factor solution proposed by Salovey et al. (1995) demonstrated evidence of convergent and discriminant validity, as well as internal consistency (Cronbach’s α ranging from .82 to .88.). A subsequent study (Palmer et al., 2003) replicated the 3-factor structure through both exploratory and confirmatory factor analysis and reported good internal consistency indicators (Cronbach’s α of .87 for clarity, .84 for attention, and .71 for repair). In the current study, we were able to achieve satisfactory values in confirmatory factor analysis, but only after adjustments: χ2/degrees of freedom (df)=4.00, comparative fit index (CFI)=.93, root mean square error of approximation (RMSEA)=.080 [confidence interval (CI)=.060-.107].

Adjustments included the deletion of 8 items to improve the subscales’ internal consistency or because they presented double loadings. Consequently, the version of the scale used in this study is composed of 22 items aggregated in the three original dimensions: clarity (9 items, e.g., “I am rarely confused about how I feel”, α=.82); attention (8 items, e.g., “I do not pay much attention to my feelings”, α=.78) and repair (5 items, e.g., “I try to think good thoughts no matter how badly I feel”, α=.77).

Secure base

The SBQ (Carvalho & Matos, 2021) is a 17-item self-report questionnaire developed for assessing therapists’ representations as secure base figures for their clients. The questionnaire’s content validity was assured through input and revision by experienced senior researchers, and it included the establishment of the instruments’ facial validity with the participation of ten psychotherapists. The final structure of the scale was achieved through principal component analysis (with varimax rotation), which resulted in a reliable and theoretically grounded 4-factor structure, explaining more than 51% of the variance: i) sensitivity [5 items, e.g., “I feel I am able to help my clients feel hope and security”, α=.69, mean inter-item correlation (MIC)=.31]; ii) compulsive caregiving (5 items, e.g., “I often feel overwhelmed by my client’s problems and difficulties”, α=.68, MIC=.30); iii) avoidance of uncertainty (4 items, e.g., “I feel uncomfortable when I face uncertainty in the therapeutic process”, α=.69, MIC=.36); iv) encouragement of exploration (3 items, e.g., “I usually encourage the client to reflect on his/her relational patterns”, α=.65, MIC=.39). Participants respond to a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). Given the reduced number of items per subscale and the relation between the number of items and Cronbach’s α, both Cronbach’s α (along with the number of items in each subscale) and the MIC should be used to assess internal consistency (Field, 2018; Lance et al., 2006). As Cronbach’s α scores ranged from .65 to .69 and the lower MIC was .30, we conclude that all factors can be considered to present a good level of homogeneity.

Results

Preliminary and descriptive analysis

All scales used went through a confirmatory factor analysis process. When necessary, adjustments to the scales were made. Priority was given to the (re)validation of the original factor structure achieved for the three used measures, as detailed in the measure’s subsection.

The preliminary analysis included checking the percentage of missing values for each variable as well as the extent to which values of kurtosis and skewness were within reasonable limits. The percentage of missing data for each variable was considerably low, ranging from 0% to a maximum of 1.8%. As Newman (2014) recommended, maximum likelihood estimation (with robust standard errors) was used. All but one variable were within the common threshold of skewness below 3 and kurtosis above 10 (Kline, 2016; Weston & Gore, 2006).

Specifically, one item (TMMS_3_inv) presented a marginally higher value regarding skewness (-3.1) but a kurtosis value of 13.7. Given this item’s importance to the maintenance of the original factor structure, we opted to conduct a maximum likelihood estimation with robust (Huber-White) standard errors and the scaled chi-square statistic, therefore adjusting for the impact of non-normality on the results (Satorra & Bentler, 1988). Structural equation modeling (SEM) analysis, including confirmatory factor analyses and bootstrapping procedures, was conducted with R’s Lavaan package. The mediating bias-corrected bootstrap CI were generated with 1000 bootstrapping resamples and a 95% CI to test significance. If the 95% CI for the average estimates of these 1000 indirect effects did not include 0, using the method of Shrout and Bolger (2002) we concluded that the indirect effect is statistically significant at the .05 level.

M, SD, and 0-order correlations for all the 11 latent variables used in the SEM models are shown in Table 1.

Analytical plan

As a first step, we run a full model estimating all parameters between the latent variables (referred to below as the Initial Model and presented in Figure 1), allowing for the identification of which independent variables (either directly or through mediation) carried any explanatory power over the dependent variables. Secondly, after carefully analyzing these relations, variables that did not directly or through mediation present statistically significant relations with the dependent variables were removed. This resulted in the final model presented in Figure 2. For readability ease, neither Figure 1 nor 2 include the measurement model. More information regarding each measure is presented above under the Measures section. Nevertheless, it is important to make clear that no parameters were estimated directly between items. In other words, no item errors (or unique variances) were correlated.

Initial model

The adequacy of the models’ fit was based on the selection of the following indexes (DiStefano & Hesse, 2005; Hu & Bentler, 1999; Kline, 2016) and cut-off values: the ratio between (the scaled) χ2 statistic and degrees of freedom, with values below 3 considered acceptable and good below 2 (Schermelleh-Engel, et al., 2003); CFI, with values above .90 considered acceptable and good above .95 (Lance, et al., 2006; Schermelleh-Engel, et al., 2003); Tucker Lewis index (TLI), with values above .90 considered acceptable and good above .95 (Hu & Bentler, 1999); RMSEA, with values below .08 considered acceptable and below .05 good (Hu & Bentler, 1999); and standardized root mean square residual (SRMR), with values below .08 to be considered acceptable (Hu & Bentler, 1999).

The (exploratory) initial model depicted in Figure 1 presented overall satisfactory adjustment indices. Specifically, the chi-square test statistic was 1381 for 890 df (χ2/df=1.55); CFI=.89; TLI=.88; RMSEA=.040 (.036-.044); and SRMR=.057. As depicted in Figure 1, several regression paths were deemed non-significant (grey lines).

As shown in Figure 1, RAQ’s avoidance does not directly or indirectly yield explanatory power over any outcome variables and was therefore removed from the final model. Contrarily, both RAQ’s dimensions of ambivalence and dependence presented a statistically significant direct path with SBQ’s outcomes and are therefore maintained in the final model. Furthermore, although RAQ’s trust factor was not directly associated with any of the outcomes, it showed statistically significant relations with SBQ’s sensitivity through the TMMS’s dimension of clarity. For this reason, both RAQ’s trust and the mediator’s clarity were maintained in the final model.

As TMMS’s dimension repair did not yield statistically significant associations with both RAQ’s and SBQ’s dimensions, they were not included in the final model. Hence, the final model to be tested is composed of three RAQ’s dimensions (trust, ambivalence, and dependence), clarity as a mediator, and the four SBQ outcomes (sensitivity, compulsive caregiving, avoidance of uncertainty, and encouragement of exploration).

Final model

The resulting final model, with the above-specified variables, is depicted in Figure 2. Adjustment indexes confirm an overall good fit of the model to the data, according to the above-specified thresholds. Specifically: the chi-square (scaled robust) statistic was 757 for 473 df (χ2/df=1.60); CFI=.92; TLI=.91; RMSEA=.041 (.036-.047); and SRMR=.058.

Figure 1.

Regression coefficients for mediation model testing the indirect/direct effects of attachment (trust, ambivalence, dependence and avoidance) on secure base behaviors toward clients (sensitivity, compulsive caregiving, avoidance of uncertainty and encouragement of exploration) through emotion regulation processes (clarity, attention and repair). Values are in standardized units. Grey lines represent non-significant paths. Solid lines depict direct effects. Dotted lines depict indirect effects. C. caregiving, compulsive caregiving; A. uncertainty, avoidance of uncertainty; E. exploration, encouragement of exploration.

Table 1.

Descriptives and correlations of the latent variables.

Variable	1	2	3	4	5	6	7	8	9	10	11
1. RAQ Trust	-
2. RAQ dependence	.09	-
3. RAQ avoidance	-.47**	-.24**	-
4. RAQ ambivalence	-.68**	.04	.45**	-
5. TMMS clarity	.17**	-.14**	-.08	-.14**	-
6. TMMS attention	.07	-.20**	-.19**	-.15**	.31**	-
7. TMMS repair	.20**	-.18**	-.03	-.17**	.38**	.11*	-
8. SBQ sensitivity	.15**	-.02	-.11*	-.09	.43**	.14**	.30**	-
9. SBQ e. exploration	.12*	-.01	-.13*	-.10*	.29**	.12*	.24**	.54**	-
10. SBQ c. caregiving	-.11*	.21**	.06	.24**	-.26**	-.14**	-.14**	-.17**	-.14**	-
11. SBQ a. uncertainty	-.08	.21**	.04	.20**	-.17**	-.09	-.14**	-.28**	-.21**	.38**	-
Possible range	1-6	1-6	1-6	1-6	1-6	1-6	1-6	1-7	1-7	1-7	1-7
M	5.10	2.76	1.91	2.16	4.56	5.46	4.29	5.50	5.73	2.95	3.40
SD	.80	.98	.82	1.04	.76	.61	.80	.65	.69	.98	1.02

RAQ, Romantic Attachment Questionnaire; TMMS, Trait-Meta Mood Scale; SBQ, Secure Base Questionnaire; M, mean; SD, standard deviation; **statistically significant.

As observable in Figure 2, RAQ’s trust dimension has been shown to be directly and positively associated with SBQ’s compulsive caregiving. Ambivalence has yielded positive relations with both SBQ’s compulsive caregiving and avoidance of uncertainty, but not with exploration or sensitivity. Lastly, RAQ’s dependence has proved to relate to both compulsive caregiving and avoidance of uncertainty (but not with encouragement of exploration or sensitivity).

Mediation analysis

Clarity has been shown to mediate the relation between RAQ’s trust dimension and three of the four outcome variables, with none of the three 95% bootstrap CI including 0. In more detail, clarity has proven to fully mediate the relation between trust, on the one hand, and sensitivity and encouragement of exploration, on the other. Additionally, clarity partly mediates the association between RAQ’s trust and SBQ’s compulsive caregiving. Lastly, the indirect effects between trust and avoidance of uncertainty through clarity did not reach statistical significance. Indirect effects estimates are presented in Table 2.

Discussion

The implications of psychotherapists’ personal characteristics, namely attachment dimensions, on psychotherapy have been consistently addressed by research (Bennett, 2008; Dozier et al., 1994; Fuertes et al., 2019; Lingiardi et al., 2018; Ryan et al., 2023; Tyrrell et al., 1999). Nevertheless, scarce knowledge has been produced regarding the underlying mechanisms of these processes (Carvalho & Matos, 2021; Mikulincer & Shaver, 2007; Talia et al., 2017; Talia et al., 2020). By framing therapeutic relationships through the complementarity between attachment and caregiving systems, this study offers important contributions for gaining a deeper knowledge of the processes implicated in the therapeutic encounter, namely the ones deeply rooted in relational dimensions such as attachment and emotion regulation.

One contribution relates to the association between attachment dimensions and the provision of a secure base in clinical settings. This result brings further evidence to research that addresses the relational dimensions of therapy and the crucial role that personal characteristics play in the way therapists represent themselves as secure base figures in psychotherapy (Daniel, 2006; Degnan et al., 2016; Steel et al., 2018). Additionally, this result confirms previous studies that show psychotherapy is significantly influenced by the complex dynamics underlying therapists’ history of receiving care (Slade & Holmes, 2019) and therapists’ predispositions to form specific relationships with others (Steel et al., 2018). In this context, SBQ (Carvalho & Matos, 2021) offers an essential tool for capturing different dimensions associated with a central and subtle concept of attachment relationships in adulthood – the secure base one (Crowell et al., 2002; Waters & Cummings, 2000).

The second contribution relies on the possibility offered by this study to develop a closer look at the processes underlying the link between attachment and relational dimensions in the psychotherapeutic context. Distinct interactions were found between attachment insecurity/security dimensions and the psychotherapists’ representation as a secure base figure for clients. Direct effects have shown that ambivalence and dependence attachment dimensions are primarily and positively associated with psychotherapists’ compulsive caregiving and avoidance of uncertainty (Carvalho & Matos, 2021). Both compulsive caregiving and the avoidance of uncertainty could work as important emotional triggers for psychotherapists with insecure internal working models. Previous research has found that these components of caregiving in adult relationships could be primarily found in individuals who are more prone to experiencing negative models of themselves, namely anxious and fearful attachment patterns (Clark et al., 2020; Kunce & Shaver, 1994; Mikulincer & Shaver, 2013; Shaver et al., 2005).

In fact, fusional and preoccupied care behaviors can also be characteristics of close and intimate relationships (Bowlby, 1977).

Figure 2.

Regression coefficients for mediation model testing the indirect/direct effects of attachment (trust, ambivalence, dependence) on secure base behaviors toward clients (sensitivity, compulsive caregiving, avoidance of uncertainty and encouragement of exploration) through emotion regulation processes (clarity). Values are in standardized units. Grey lines represent non-significant paths. Solid lines depict direct effects. Dotted lines depict indirect effects. C. caregiving, compulsive caregiving; A. uncertainty, avoidance of uncertainty; E. exploration, encouragement of exploration.

Table 2.

Bootstrap analysis of the magnitude and statistical significance of the indirect effects.

Paths for the indirect effects	β	95% CI	p
Trust → clarity → sensitivity	.22 X .38 = .074	[.013, .146]	.025
Trust → clarity → compulsive caregiving	.22 X - .31 = -.068	[-.142, -.012]	.040
Trust → clarity → avoidance of uncertainty	.22 X - .19 = -.041	[-.111, .005]	.176 (n.s.)
Trust → clarity → encouragement of exploration	.22 X .22 = .048	[.008, .106]	.050

CI, confidence interval; β, paths coefficients and product; n.s., not significant.

This is particularly relevant in the therapeutic process since most clients seeking professional help experience emotionally vulnerable states and are eager for genuine and interested care (Farber & Metzger, 2009). Sauer et al. (2003) provided additional information regarding the increased ability of anxious attached psychotherapists to develop stronger working alliances at the beginning of the therapy; nevertheless, from a processual view, this bond could be more fragile and less resistant to therapeutic tensions or challenges. Compulsive care could be particularly detrimental during the intermediate phase of the therapy when therapists and clients have a lower chance of accordance (Kanninen et al., 2000). As in other adult relational dynamics, exerting a more controlled and fusional relationship may respond to emotionally perceived threats of abandonment or being inadequate (Bowlby, 1977; Kunce & Shaver, 1994).

In the same way, avoidance of uncertainty could be significant for engaging in reassurance-seeking processes and dealing with uncertainty processes in therapy (Clark et al., 2020; Shaver et al., 2005). Less positive models of self-constraint, a more open attitude and flexibility toward new information, and ambivalence can ultimately prevent the embracement of uncertainty in the therapeutic process (Mikulincer, 1997; Mikulincer & Shaver, 2013). These results align with previous research that shows that therapist attachment threats could be activated in more stressful moments in therapy or when relating to more challenging clients (Sauer, 2003; Schauenburg et al., 2010).

Another significant result closely connected to this last contribution regards the absence of emotion regulation as a mediator between ambivalence and dependence and the secure base dimensions in clinical settings. It is possible that hyperactivation attachment strategies (primarily used in individuals with higher scores on dependence and ambivalence attachment dimensions) or higher vulnerability to activation of own attachment-related worries during psychotherapy (Mikulincer & Shaver, 2013, 2019) could undermine the therapist’s ability to consciously acknowledge these processes in the therapeutic relationship. Paralleling other regulation processes in psychotherapy, such as mentalization, hyperactivation may hinder the capacity to integrate uncertainty and develop a curiosity attitude towards the therapeutic process. These processes could ultimately interfere with the psychotherapist’s ability to be responsive to the client’s underlying attachment needs (Bowlby, 1988) or even heighten the patient’s defenses in therapy (Slade & Holmes, 2019) since the therapist’s own attachment fears overlap with responsive caregiving centered on the client’s attachment needs.

A different result has been shown by the secure attachment dimension represented by trust. Two distinct associations were found between trust and compulsive caregiving through a mediated and a direct path. The mediating effect of clarity had two major implications for the association between psychotherapist attachment and secure base representations. First, we observed a negative association between these two dimensions through clarity. Second, there was a positive association between trust and representations of compulsive care provision in therapy. Note that the correlation analysis shows a negative association between trust and compulsive caregiving, indicating that even when inspecting the direct effects, emotion regulation processes interfere with this direct association. These results stress the complex nature of the dynamics involved in providing secure base processes in psychotherapy, requiring an integrated analysis of this effect. In this sense, there are facets of compulsive caregiving that could be shared by secure and some insecure attachment dimensions. The distinctive processes of the secure and insecure dimensions concern the role of the emotional regulation process. The ability to acknowledge one’s own emotions affects experiences and states involved in the emotion regulation process of clarity, which could, as in other relational dimensions of the therapeutic encounter, facilitate empathic processes and re-centered the psychotherapist’s focus on the client’s attachment needs (Ruiz-Aranda et al., 2021; Ryan et al., 2023). This mechanism could be fundamental in situations where the therapist is most reflexive or more uncertain about the therapeutic relationship and process. An interesting finding of this study is that higher levels of trust do not seem to prevent therapist caregiving representations of compulsive caregiving behaviors in psychotherapy. Nevertheless, this association is challenged by emotional self-awareness processes induced by clarity. Emotion regulation processes allow therapists to attend and gain a deeper understanding of their own emotions and feelings in this process. Clarity was also an essential mediator between trust and the encouragement of exploration in clinical settings. Note that although the magnitudes of effects were smaller when addressing indirect paths (Table 2), this study offers important ground by acknowledging the role of emotion regulation as one of the mechanisms that could interfere with relational processes in psychotherapy. This finding lends support to earlier research that addresses the critical role of the therapists’ emotion regulation process on their ability to develop responsive therapeutic relationships (Ruiz-Aranda et al., 2021).

Additionally, this study offers essential ground for disentangling challenges that emerge from psychotherapeutic processes, such as questioning the proximity and investment in therapy and dealing with the uncertainty/erratic nature of the therapeutic process and the mechanisms underlying responses in therapists that exhibited distinct levels of secure and insecure attachment dimensions. The challenge is to understand if these mechanisms, which may function as potentially protective mediator factors, can be (and in what ways) improved in therapeutic training and supervision. Given the “unconscious predispositions of therapists to form certain styles of relationships with themselves and others” (Steel et al., 2018, p. 34), it would also be interesting to understand how clients’ distinct attachment dimensions elicit distinct responses as secure base figures, as well as the role of emotion regulation processes in this association.

Several limitations and constraints should be acknowledged in this study. The first one concerns the small ratio between sample size and the number of parameters to be estimated, which may warrant some caution in the interpretation of the results (Bentler & Chou, 1987; Kline, 2016; Mueller, 1997). Nevertheless, the relatively large sample size of therapists (n=384), which is significantly above the commonly suggested 200 thresholds (Jackson, 2003), the reliability of the indicators used, and the adequate number of indicators per latent variable important criteria mentioned by different authors (Bentler & Chou, 1987; Jackson, 2001, 2003; Kline, 2015; Mueller, 1997), give robustness to our findings.

The second one concerns the use of exclusive psychotherapist self-report assessments for evaluating relational dynamics in psychotherapy. Research has stressed the differentiating role that client attachment could elicit in therapist responses, namely in the most insecure ones (for a review, see Daniel, 2006; Romano et al., 2008). As previously stated, future studies should acknowledge these interactional processes that collect data on both clients’ and therapists’ personal and relational characteristics. Considering potential limitations to dyadic data collection in real settings, using performance-based tasks and observational studies could be an important alternative for addressing therapists’ contributions, such as attachment, to therapeutic processes and outcomes (Anderson et al., 2020) and to control potential self-appraisal bias. Secondly, considering the bidirectional influences between psychotherapeutic processes, outcomes, and psychotherapist attachment (Degnan et al., 2016), future studies should include outcome variables and dyadic and longitudinal designs. In fact, the cross-sectional nature of the current study represents an explicit limitation, namely when addressing such dynamic dimensions of the therapeutic encounter. Multiformat and repeated measures design could help us gain a more comprehensive insight into the stability of secure base constructs throughout distinctive therapeutic dyads and across therapeutic moments.

Contributions to psychotherapeutic training, supervision and practices

It appears that the change process among this small and unique sample of therapists closely parallels how clients change during therapy: Within a supportive interpersonal context, the person becomes aware of things in one s life that are remnants of the past and do not necessarily work in the current situation (…) This is how our clients change. As Sullivan once suggested, people are more human than otherwise (Goldfried, 2001, p. 326)

While there is a consistent body of research addressing the impactful nature of therapists’ relational characteristics in psychotherapeutic processes and outcomes, there are still scarcely any specific guidelines for helping psychotherapists gain a deeper understanding of their own relational stories. Attachment history could have complex, unconscious, and dynamic implications for therapeutic interactions that could inspire or inhibit change. Nevertheless, offering space-time opportunities for improving self-knowledge of emotional triggers and working on emotion regulation processes could enhance therapists’ understanding of both their own and client’s mental states (Ryan et al., 2023).

Supervision and intervision practices should offer secure base contexts for improving therapists’ emotional awareness by exploring distinct emotional experiences elicited by clients’ distinct relational patterns, distressful events, and working alliance challenges. In this sense, SBQ could be a useful tool for monitoring processes and ultimately enhancing mentalizing abilities as an important mechanism for compensating insecure attachment patterns of oneself and others (Cologon et al., 2017). Also, considering the close link between attachment and secure base behaviors in psychotherapy, psychotherapists’ training could benefit from intervention strategies used in attachment-based interventions for improving sensitivity and attunement.

Psychotherapists’ attachment-related issues should be further recognized and explored using distinct strategies that go from video feedback, discussion of research on therapist and client effects on therapeutic sessions and offering individual psychological support for helping therapists improve mentalization abilities, integrating trauma, or the most difficult emotional experiences and memories (Dozier et al., 2018; Talia et al., 2019; Woodhouse et al., 2018). The supervising/training relationship could offer an optimal context for framing the therapeutic act while considering the relational attachment framework, promoting improvements in sensitive and attunement abilities but also preventing conditions for feeling unhelpful and, ultimately, burnout (Hiebler-Ragger et al., 2021; Talia et al., 2019).

Conclusions

In summary, this study supports the role of emotional regulation processes, namely concerning the ability to attend, discriminate, and acknowledge emotions, as robust processes for grasping therapeutic outcomes (Najavits, 2000). Humanizing the therapeutic encounter also implies recognizing that psychotherapists’ actions are, as in their clients, inspired by their history of love and affection, their emotional comprehension of human nature, and what thrives motivation and behaviors. If there is a profound ethical compromise in becoming better professionals, providing better responses, and helping people thrive through “dark waters”, acknowledging and considering these profound human dimensions of care could also contribute to improving the psychotherapists’ sense of self-worth, competence, and security, which are necessary conditions for embracing uncertainty and complexity in psychotherapy (Stricker, 2002).

Funding Statement

Funding: this work has been funded by Foundation for Science and Technology (FCT), namely through grants SFRH/BD/22333/2005 (Helena Carvalho); UIDB000502020 and CEECIND/00646/2018 (Gil Nata).