Background

Mentalizing is defined as the capacity to understand the intentional or inner mental states of other persons, taking into account one's own intentional states (e.g., beliefs, thoughts, feelings, desires or goals) (Fonagy, 1991; Fonagy & Bateman, 2008). In recent years, a new concept, Mentalization-Based Treatment (MBT), in psychodynamic psychotherapy and research has become increasingly popular (Bateman & Fonagy, 1999, 2009, 2016; 2019; Fonagy, 1998). The objective of MBT is to improve the mentalization capacity of patients with severe personality disorders (Bateman & Fonagy, 2019).

The mentalizing approach was developed by Fonagy, Steele, Morgan, Steele, and Higgitt (1991) within the London-Parent-Child project. The authors observed that a secure attachment of the child is not only dependent on the mother's attachment security (Fonagy, Steele, & Steele, 1991) but also results from the mother's insight to acquire a psychological understanding of her own early childhood relationship with her parents (Fonagy, Steele, Morgan, et al., 1991). With regard to their predominantly positive attachment experiences and the associated secure attachment representations, 79% of these women also had securely attached children compared to 28% of mothers with an insecure attachment style. In particular, it could be confirmed that mothers who reported more negative childhood experiences during interviews were nevertheless able to provide coherent statements and coping strategies (Fonagy, Steele, Morgan, et al., 1991).

To measure mentalization, Fonagy, Target, Steele, and Steele (1998) developed the Reflective Functioning Scale (RFS) based on the Adult Attachment Interview (AAI, George, Kaplan, & Main, 2012) and examined the relationship between attachment and interaction behaviour between mothers and their infants (Bateman & Fonagy, 2019; Fonagy, 1991). The studies demonstrated that women who were more likely to be securely attached achieved higher values on the RFS than mothers with an avoiding attachment behaviour (Fonagy, Steele, & Steele, 1991). The empirical findings of their study made Fonagy, Steele, and Steele (1991); Fonagy, Steele, Morgan, et al. (1991) suggest that there must be a connection between early attachment experiences, self-representation and affect regulation even in patients with severe personality disorders (Bateman & Fonagy, 2019). Empirical studies have demonstrated that patients with severe personality disorders have lower mentalizing capacities than individuals without psychopathology (Fischer-Kern et al., 2010; Fonagy, 2004; Fonagy & Target, 2000; Skarderud, 2007a, 2007b, 2007c). This has also been shown for patients with other mental disorders such as depression (Lemma, Target, & Fonagy, 2011; Markowitz & Meehan, 2009; Staun, Kessler, Buchheim, Kachele, & Taubner, 2010; Taubner, Kessler, Buchheim, Kachele, & Staun, 2011) or eating disorders (Robinson et al., 2016). Furthermore, contemporary research has revealed associations between impairments in mentalization and other psychological phenomena, such as difficulties in social interactions (Fonagy, Luyten, & Bateman, 2015; Hayden, Müllauer, Gaugeler, Senft, & Andreas, 2018), or the use of maladaptive defence patterns (Hayden et al., 2021).

Since the development of the concept of mentalization, various instruments have been developed for the structured assessment (Luyten, Malcorps, Fonagy, & Ensink, 2019). It is possible to measure mentalization through self-report, expert-/observer-rated assessment, interviews, questionnaires or experimental tasks. Currently, one of the most widespread and commonly used instruments for measuring mentalization is the RFS based on the AAI (Fonagy, Steele, & Steele, 1991; Luyten et al., 2019). However, the primary objective of the AAI (George et al., 2012) is to identify the attachment style of a person, which is why not all questions of the AAI are relevant for the assessment of the interviewee's capacity to mentalize (Taubner et al., 2013). Taubner et al. (2013) examined the internal structure and reliability of the RFS in N = 196 subjects. They were able to show that those questions that were supposed to provoke a mentalization in the AAI (George et al., 2012) (demand questions) were included in the calculation of the overall score. However, they were able to show that raters also use the information from those questions that do not prompt mentalizing behaviour (permit questions) (George et al., 2012) to form the overall score. In this respect, focussing only on the demand questions may lead to a systematic underestimation of the reflective functioning (RF) score (Taubner et al., 2013).

Since the AAI (George et al., 2012) has an average interview duration of 1–2 hrs plus approximately 8 hrs of transcription time, Rudden, Milrod, Target, Ackerman, and Graf (2006) developed a more specific interview, the Brief Reflective Functioning Interview (BRFI). The questions in the BRFI are intended to provoke mentalization on attachment experiences and were developed on the basis of the AAI. The BRFI consists of ten questions (e.g., ‘Tell me something about one of your parents – how is your father or mother?’ or ‘What do you think, why did your father/mother become what he/she is now?’). Rudden et al. (2006) used seven of the eight demand questions from the AAI and added two more questions on current relationships. The assessment of mentalization is based on the same principle as for the AAI (see RFS, Fonagy et al., 1998), but in contrast to the RFS based on the AAI, each question is rated as a demand question. Rutiman and Meehan (2012) examined the validity of the BRFI in comparison to the AAI in N = 27 students. They found an average shorter interview duration of M = 24 min for the BRFI in contrast to M = 44 min for the AAI. Furthermore, the authors reported a significant positive correlation (r = .71) between the RFS scores of the AAI and the BRFI. The raters achieved a good interrater reliability of ICC = .79.

Interviews on the assessment of mentalization can be supplemented by self-assessment questionnaires. Hausberg et al. (2012) developed the Mentalization Questionnaire (MZQ) for the economical assessment of mentalization on the basis of people with different mental disorders. The MZQ is primarily used for screening measurements and follow-up assessments of patients undergoing psychotherapeutic treatment. It measures mentalization with 15 items and is presented as a self-evaluation tool. An example item is ‘Talking about feelings would mean that they become more and more powerful’. Self-reports and expert-rated instruments such as the AAI (Fonagy et al., 1998) have different advantages and disadvantages (Andreas et al., 2007). While self-reports focus on the perspective of the person, interview procedures focus more on an observer's perspective. There are also differences in terms of feasibility. The MZQ (Hausberg et al., 2012), for example, can be completed in 10 min, whereas the AAI including transcription and rating takes about 10 hrs. Self-report and expert-rated interviews for measuring reflective functioning always take on complementary, mutually reinforcing functions (Andreas et al., 2007). There are several studies that evaluated the psychometric properties of the MZQ (Hausberg et al., 2012; Innamorati et al., 2017; Ponti, Stefanini, Gori, & Smorti, 2019). However, no study has investigated the relationship between expert-rated RFS scores of the AAI or the BRFI and self-reported mentalization.

The aim of the present study was to examine the psychometric properties of the German version of the BRFI. In addition, reliability and validity of the BRFI in a student sample (Study 1) as well as a mixed student and clinical sample (Study 2) should also be investigated.

Study 1: Comparison Of The Rfs Scores Measured By The Aai And Those Measured By The Brfi – Psychometric Properties Of The Rfs Of The Brfi

Study 1 aimed to compare the psychometric characteristics of the RFS of the BRFI with those of the AAI. We investigated the distribution characteristics of the RFS items of the BRFI in comparison with those of the AAI. In addition, we aimed to determine the feasibility of the administration of the BRFI in comparison with the AAI. In light of the results of Rutimann’s and Meehan's study (2012), we assume that the psychometric properties of the BRFI are comparable to the psychometric properties of the AAI.

Method Translation process of the BRFI

The BRFI was translated into German by our research group (MH, PM and SyAn), as authorized by the test authors (MR) in 2014. The translation and re-translation followed steps 1 to 6 of the translation guidelines of the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) (Wild et al., 2005). In detail, the 10 questions of the BRFI were first translated from English into German by the members of our group (MH, PM, and SyAn). The translation was discussed in our group and modifications were made. A consensus version was then translated back into English by an independent native speaker and presented to the test author (MR) for further review. The test author agreed with the English consensus version, which was then translated back into German.

Design

The design of Study 1 is illustrated in Figure 1. To obtain an appropriate dispersion for determining reflective functioning during the use of the interviews, the MZQ (Hausberg et al., 2012) was used as a screening method for the selection of participants from the total of 187 students who had agreed to participate in the study. Based on the distribution of the MZQ scores, 60 of the 187 students were selected for interviews. With N = 60 bachelor’s students in psychology and economics at the University of Klagenfurt, Austria, one AAI and one BRFI each were conducted at intervals of approximately one week. To avoid interviewer sequence effects, the sequence between the AAI and the BRFI was varied, that is, in half of the sample, the AAI was conducted first and then the BRFI. To recruit the study participants, those who were initially considered for the survey were contacted by e-mail either directly in lectures or indirectly via university mailing lists. They were provided with all relevant information on the objectives and the design of the study. Two raters (PP and KR), who had previously been trained by coaches from the Anna Freud Centre in London and achieved good interrater reliabilities (ICC > .75) according to Landis and Koch (1977), evaluated the AAIs and the BRFIs.

Screening MZQ: low – middle – high scores in self-reported reflective functioning. Note: AAI = Adult Attachment Interview; BRFI = Brief Reflective Functioning Interview; MZQ = Mentalization Questionnaire.

Sample

As mentioned above, to achieve an appropriate distribution of mentalization capacities in the overall sample, the MZQ (Hausberg et al., 2012) was distributed in advance to those students who had agreed to participate in the study. Based on the variance of the total score of the MZQ, the students were assigned to three groups according to the RFS manual by Fonagy et al. (Fonagy et al., 1998): (1) rather low RF (M ≤ 3.46–3.48), (2) average RF and (3) above-average RF (M ≥ 3.9–4.0).

The mean RF of N = 91 surveyed psychology students was M = 3.6 (SD = 0.51). The mean RF in the surveyed sample of economics students (N = 96) was comparable to that of psychology students M = 3.58 (SD = 0.5). To obtain a widely representative sample for the psychometric examination of the BRFI, N = 60 students were randomly selected from this population. In each of the three RFS groups, there were n = 20 students, of which 25% were economics students and 75% were psychology students. The mean age of the students was 23 years (SD = 4.6 years); 48% of the students were female and 52% were male. Nationality was also divided equally between German and Austrian students. Only 2% of the student sample was married, 73% single and 25% with a partner.

Instruments The Adult Attachment Interview (AAI)

The total of 20 questions of the AAI with their demand format allow the interviewer a semi-structured, clinical process (George et al., 2012). The interview contains questions on the interviewee’s life situation in childhood, on the description of significant others, on experiences of loss and how parents reacted when the participant was ill or injured as a child. The aim of the AAI was to provide the attachment style of a person in adulthood and the associated inner working models according to Bowlby (Holmes, 2014). The entire interview is audiotaped for later evaluation purposes to determine the attachment style, the cohesion in the narrative or mentalization.

Based on the questions of the AAI, a scale was developed to measure the RFS (Fonagy et al., 1998). It is the operationalization of the mentalization capacity in the form of a measurement of the subject’s capability to reflect about their own mental states and those of others. In particular, the assessment includes the following aspects: The types and frequency of the acknowledgement of the opacity of mental states, empathy and the ability to reflect on certain situations. Furthermore, efforts to combine observable behaviour with inner states of mind are observed and the ability to consider changes in inner states of mind, and the resulting changes in behaviour are assessed. These categories are assessed on an 11-point scale ranging from −1 (anti-reflective) to 9 (exceptionally reflective). Only the odd scores of RFS are provided with anchor examples, where −1 means anti-reflexive, one means missing RF, three means suspicious RF, five means average RF and seven means above-average RF. However, from a score of four, first reflexive functions are rated. Training is required to administer the RFS.

The Brief Reflective Functioning Interview (BRFI)

The Brief Reflective Functioning Interview (BRFI) (Rudden et al., 2006) is a short, semi-standardized interview. Although it is based on the questions of the AAI, the interviewees are asked only about one parent (Rutimann & Meehan, 2012). Similar to the AAI, the questions of the BRFI are intended to provoke mentalization on attachment experiences. In contrast to the AAI, however, there are fewer questions of episodic memories of attachment experiences. The interviewees are also asked to describe a person of importance in his or her lives to assess the current attachment representation. Rutimann and Meehan (2012) believe that the BRFI captures a general capacity for reflection in relationships rather than a specific attachment experience as in the AAI. As described earlier, the BRFI consists of ten questions. The assessment of mentalization follows the manual for coding mentalization (Fonagy et al., 1998). However, it is not divided into demand and permit questions, but an average score is calculated over all ten questions.

The Mentalization Questionnaire (MZQ)

The MZQ by Hausberg et al. (2012) is a self-assessment instrument for the determination of mentalization in mentally ill persons. The instrument is primarily used for screening examinations, assessments of the course of mental illnesses, and adequate measurements of MBT success. In total, the questionnaire contains 15 different items that can be assigned to the following four subscales and follow a 5-point answer scale (from 1 ‘do not agree at all’ to 5 ‘fully agree at all’). The reliability of the MZQ has proven to be satisfactory in previous studies, with an internal consistency of α = .57 to .81. In addition, satisfactory values for test-retest reliability (r = .54 to .76) were found for both the individual subscales and the total score (Hausberg et al., 2012). In particular, the convergent validity for all external criteria defined in the study (e.g., attachment styles, suicidal tendencies, etc.) showed significant group differences from medium to large effect sizes. Significant correlations between mentalization ability and symptom severity were also determined. Based on the psychometric evaluation and the results presented on the quality criteria, the MZQ can be used as a practical, reliable and valid method for assessments of mentalization from a patient perspective (Hausberg et al., 2012). The instrument has been translated into several languages, including Korean, Indian, Italian, Finnish, and Danish (Juul et al., 2019; Anupama, Bhola, Thirthalli, & Mehta, 2018; Belvederi Murri et al., 2017; Eloranta, Kaltiala, Lindberg, Kaivosoja, & Peltonen, 2020; Ponti et al., 2019; Song & Choi, 2017).

Statistical analyses

Statistical analyses were performed using SPSS version 25. The power calculations were carried out with the programme G*Power 3.1. With a given sample size of Nmin = 50 persons, an approximate mean effect size according to Cohen (1988) for correlative relationships of r = .35 with an alpha error probability of p < .05 can be inferentially statistically validated with a sufficiently high power of at least w = .80.

The psychometric characteristics for the feasibility and reliability of the RFS of the BRFI were assessed using frequencies of the items of the RF scale and item analyses. We calculated internal consistencies of the RFS of the BRFI in comparison with those of the AAI. Furthermore, we examined the administration time between both ratings. Moreover, we used a t test to investigate the effects of the sequential administration of both interviews. To determine the relationship between the RF scale of the BRFI and the AAI, we performed a Pearson correlation.

Results Descriptives

The frequencies of the RF BRFI and AAI ratings are shown in Table 1. In both interviews, there were no ratings of −1 (anti-reflective) in the student sample. The students mainly revealed average scores of reflectivity (RF = 3–5). However, we found a score of 9 (extraordinary reflectivity) only for the AAI (n = 1).

Table 1. Frequencies of the Reflective Functioning (RF) of Brief Reflective Functioning Interview (BRFI) and Adult Attachment Interview (AAI) Frequencies BRFI AAI RF = −1 0 0 RF = 0 1 0 RF = 1 3 3 RF = 2 4 5 RF = 3 12 9 RF = 4 11 8 RF = 5 13 17 RF = 6 9 7 RF = 7 6 9 RF = 8 1 1 RF = 9 0 1 Note AAI = Adult Attachment Interview; BRFI = Brief Reflective Functioning Interview; RF = Reflective Functioning. Distribution characteristics and internal consistency

As presented in Table 2, the means of the RF score of the BRFI were higher for some questions than for others. The highest mean score was revealed for the question ‘current relationship/behaviour’, followed by ‘changes in the relationship’. Other questions such as ‘specific memory’ or ‘reason for chosen the parent’ were more likely to result in lower mean RF scores. The scores of skewness and kurtosis ranged between nearly –1 and 1.

Table 2. Descriptive statistics of the Reflective Functioning scale (RFS) of the Brief Reflective Functioning Interview (BRFI) Mean SD Skewness Kurtosis Characteristics of a parent 3.70 1.7 −.11 −.63 Reasons behind the characteristics 3.88 1.93 −.03 −1.18 Relationship to parent 3.87 1.82 .01 −.85 Reasons for current relationship 4.18 1.96 −.17 −.68 Specific memory 3.02 1.85 .39 −1.39 Changes over time 4.15 1.84 .04 −.41 Parent’s influence/overall experience 3.70 1.88 .63 −.69 Reasons for the choosing parent 3.07 1.50 .30 −.95 Another important person 3.33 1.73 .37 −.91 Relationship to that person 3.63 1.57 .07 −.74 Note BRFI = Brief Reflective Functioning Interview; RFS = Reflective Functioning scale; SD = standard deviation.

The internal consistency of the questions of the BRFI was excellent, with a Cronbach’s α = .92. The Cronbach’s α for the questions of the AAI was .91.

Feasibility of the BRFI

We calculated the administration time to examine the feasibility of the BRFI in comparison with the AAI. On average, the BRFI lasted 26 min (SD = 9.5 min), in comparison with the AAI, which lasted 78 min (SD = 20 min).

Sequential effects in the interview sequence

We investigated whether there was any difference in the sequence of interviews. In half of the subjects, the AAI was conducted first, and the BRFI was conducted second. In the other half of the subjects, the BRFI was conducted first, and the AAI was conducted second. As depicted in Figure 2, there were no significant differences on the RF scale in terms of the sequential order of the interviews. Furthermore, there was also no significant difference in terms of the length of time between the interviews (3–8 days and 9–14 days, respectively) (see Figure 2).

Sequential effect of the interview order. Note: 1st = given first; AAI = Adult Attachment Interview; BRFI = Brief Reflective Functioning Interview; RF scale = Reflective Functioning Scale.

Comparison between the RF scale of the BRFI and AAI

We found a significant correlation with a large effect size according to Cohen (1988) between the RF scale of the BRFI and that of the AAI (r = .877, p < .001, n = 60).

Study 2: Interrater Reliability And Validity Of The Rf Scale Of The Brfi

Study 2 aimed to examine the interrater reliability of the RF scale of the BRFI in the clinical sample and the validity of the RF scale of the BRFI. We investigated the relationship between the expert-rated RF scale and the self-reported scale of the MZQ (Hausberg et al., 2012). To ensure variance in the MZQ responses, we used a mixed sample including the student sample from Study 1 and a clinical sample. Furthermore, we examined age and gender differences in the RF scale of the BRFI between the clinical and the non-clinical sample. We also investigated differences in the RF scores between the clinical and non-clinical sample. We assumed that the reliability should be excellent. We assumed that people who show high reflective functioning in the BRFI will also show high self-reported mentalization in the MZQ. Relating to previous research (Hausberg et al., 2012), we assumed higher scores of mentalization for the non-clinical sample than for the clinical sample. In addition, since persons may differ in age and gender, we explored the differences of the RF scale of the BRFI among the clinical and non-clinical sample.

Method Design

The design of the study is described elsewhere (Hayden, Müllauer, Gaugeler, Senft, & Andreas, 2019). Clinical data were assessed in two different medical centres that use different therapeutic approaches. All patients who met the inclusion criteria were asked to participate at the beginning of treatment. Participation was voluntary, and there were no negative consequences associated with refusal. Regarding inclusion criteria, all participants had to be between 18 and 65 years of age (Hayden et al., 2019). Reasons for exclusion were insufficient ability to understand and/or speak German, early childhood traumatization, acute manic or psychotic episode, dementia, and/or cognitive impairment. All participants provided informed consent that included the purpose and circumstances of the study, information about data privacy, and the right to refuse participation without any consequences, as well as information about the right to withdraw from the study during and after data collection.

Sample

In the clinical sample, N = 89 patients were willing to take part in the study (Hayden et al., 2019). The sample description is displayed in Table 3. Most of the participants had diagnoses of the F3 spectrum (mood disorders) and the F4 spectrum (neurotic, stress-related and somatoform disorders) according to ICD-10 (WHO, 1992) as their main diagnosis.

Table 3. Sample description of the clinical sample Clinic I Clinic II n 57 32 Age M 44.2 (SD 9.92) M 43.7 (SD 9.70) Sex Male 29 13 Female 28 19 Education Elementary 0 1 Main school 12 13 Professional school 8 7 High school 13 4 University 11 3 Other 13 4 Civil status Single 11 15 Partnership 11 5 Married 22 6 Divorced 13 5 Widowed 0 1 Children Yes 36 20 No 21 12 Note M = Mean; SD = standard deviation. Instruments The BRFI, AAI and MZQ

The BRFI (Rutimann & Meehan, 2012), the AAI (George et al., 2012) and the MZQ (Hausberg et al., 2012) were already described in Study 1.

Statistics

Statistical analyses were performed using SPSS version 25. To assess the reliability of the RF score of the BRFI in the clinical sample, Intraclass Correlation Coefficients (ICCs) between two trained raters were calculated. Fleiss and Cohen (1973) showed that the ICC is equivalent to weighted kappa for measures of reliability, and Landis and Koch (1977) provided ‘rules of thumb’ for the interpretation of kappa coefficients. According to these rules, kappa values between 0.21 and 0.40 are ‘fair’, those between 0.41 and 0.60 are ‘moderate’, those between 0.61 and 0.80 are ‘substantial’, and those between 0.81 and 1.00 are ‘almost perfect’.

We calculated t tests to investigate the relationship between self-reported mentalization capacity and expert-rated mentalization in a mixed sample (student and clinical sample).

Results Interrater reliability of the BRFI

The raters achieved excellent interrater reliability of ICC = .85 for the RFS of the clinical sample.

Validity of the BRFI

There were significant differences in the RFS of the BRFI between the clinical and the non-clinical sample (p < .001) (see Table 4). There were also significant gender differences, since higher RF scores of the BRFI were found for women (mean rank – clinical = 42.22; mean rank – non-clinical = 72.47) than for men in both samples (mean rank – clinical = 29.69; mean rank – non-clinical = 50.44) (p < .01).