The Client Attachment to Therapist Scale ( CATS ) in a Greek Clinical Population : A Validation Study

Client attachment to therapist has been long considered a prominent process variable in psychotherapy research. The purpose of this study was to translate in Greek and evaluate the structure, reliability, convergent and discriminant validity of the well-known and widely administered Client Attachment to Therapist Scale (CATS) with exploratory and confirmatory factor analysis in a clinical sample of 153 clients of different treatment modalities and at different agencies. The results indicated that the EFA resulted in a model almost identical to the Mallinckrodt, Gantt and Coble’s [1] original model after the deletion of six items. Moreover, the findings indicated adequate internal consistency, test-retest reliability, convergent and discriminant validity of the three factors. Normalized scores were provided to help mental health professionals interpret the scale scores. In conclusion, CATS is a valid and reliable tool to measure client attachment to therapist, whose use can promote psychotherapeutic process and outcome. Thus, more validation studies should be conducted in several cultural contexts to better understand its structure and psychometric characteristics.


Introduction
First introduced in Bowlby's attachment theory [2] [3], the "secure base" has become the backbone of the burgeoning attachment-informed psychotherapy research.As it was initially highlighted [4], the role of attachment security in the therapeutic relationship is a key process variable in psychotherapy, since it faci-found to play a major role in the client-therapist relationship [21] [22].Particularly, when clients employ a deactivating strategy, they tend to adopt downgrading behaviors of each of the five aforementioned secure attachment features, while clients who hyperactivate in the relationship with the therapist tend to exaggerate in expressing the five security attachment core elements [13].
Guided by the tripartite typology of attachment in Ainsworth's work, the self-report Client Attachment to Therapist Scale (CATS) was developed by Mallinckrodt, Gantt and Coble [1] to conceptualize the therapeutic relationship from an attachment perspective and to assess the quality of the client attachment to therapist.By using a rational-empirical approach to construct the CATS [23], a panel of nine therapists generated theoretically-driven items, which were then tested to 138 clients having completed at least 5 sessions of counselling.The exploratory factor analytic strategy indicated three CATS subscales, the Secure (14 items), the Preoccupied-Merger (10 items) and the Avoidant-Fearful (12 items), as three distinct patterns in psychotherapy attachment.What is perhaps striking is that the CATS attachment facets well-reflect the later developed two-dimensional models of adult insecure attachment involving anxiety and avoidance [20], but also the hyperactivating and deactivating attachment tendencies [12].
Hence, clients with a Secure attachment perceive their therapists as emotionally responsive and reliable as well as "a secure base" from which to explore threatening psychological territory (e.g., "My counselor is sensitive to my needs", "My counselor helps me look closely at the frightening or troubling things that have happened to me").In contrast, clients with an Avoidant-Fearful attachment pattern express mistrust, fear of rejection, avoidance of self-disclosure, and the lowest levels of therapeutic alliance (e.g., "It's hard for me to trust my counselor", "Talking over my problems with my counselor makes me feel ashamed or foolish").Clients with a Preoccupied-Merger attachment wish to blur the boun- daries of therapy, long for a close proximity with the therapist (e.g., "I wish my counselor could be with me on a daily basis", "I yearn to be 'at one' with my counselor"), and although they easily develop therapeutic alliance in terms of bonding, this is not the case with regard to the goals and the tasks of therapy.
Cluster analysis of the CATS identified one further pattern of attachment in psychotherapy, which was labelled as Reluctant.Clients with a reluctant attachment show unwillingness to self-exploration, albeit their good therapeutic alliance and the positive regard of the therapist [1].Moreover, current research demonstrated two subtypes of clients based on a combination of CATS subscales, including "pseudosecure" clients having high CATS Secure and high CATS Preoccupied scores, and "individuated-secure" clients having high CATS Secure and low Preoccupied scores [24].
Regarding the psychometric properties of the original CATS, the internal consistency and the retest reliability coefficients in an average of 3.24 weeks after the first measurement were found to be greater than 0.63 for all of the subscales.As regards subscales' intercorrelations, negative associations were found between Open Journal of Medical Psychology the Secure and the Avoidant-Fearful subscales, while positive ones were found between the Secure and the Preoccupied-Merger subscales.Additionally, concurrent validity was indicated through significant positive associations between the secure subscale and social competencies (i.e.free of object-relations deficits) as well as working alliance.In contrast, the Avoidant-Fearful and the Preoccupied-Merger subscales were negatively correlated with object-relations, working-alliance and self-efficacy constructs [1].Although the CATS has prevailed mainly as a continuous measure of psychotherapy attachment behavior, its use as a categorization tool to typologize clients across dimensions of attachment patterns in therapy was highlighted as a need for further research [23].
The CATS has been so far the most well-known and widely administered self-report measure of client attachment.According to a recent meta-analysis [19], the original article introducing the CATS [1] has been cited more than 100 times to date.Other relevant instruments include the 45-item self-report Components of Attachment Questionnaire (CAQ) [25] with nine theoretically derived components of psychotherapy attachment, which however was only applied in a psychoanalytic study to assess whether therapists function as attachment figures [11].On the other hand, the recently developed Patient Attachment to Therapist Rating Scale (PAT-RS) [26] and the Patient Attachment Coding System (PACS) [27] are both promising measures for assessing the quality of attachment to therapist, yet the first one is an observer-rated interview-based instrument and the second one is a transcript-based instrument that is proposed as an alternative to the AAI in psychotherapy research.Finally, the Attachment Avoidance in Therapy Scale (AATS) [28] that measures patient-therapist attachment avoidance portrayed reliability in its all six scales, albeit therapeutic attachment anxiety showed no adequate validity scores.
Findings based on CATS suggest that the client attachment is a prominent variable in treatment [29] since whether a client would be securely, or insecurely, attached to the therapist raises many challenges in the clinical work and influences client change [13].In the meta-analysis of Mallinckrodt and Jeong [13], the CATS was found to be significantly associated with working alliance and pretherapy adult attachment.Several other studies showed that clients' attachments influence in-session exploration [30] [31] [32], self-disclosure [33], transference [21] [34], the working alliance [35] [36] [37] [38] and therapeutic outcome [39] [40] [41].Additionally, evidence on the attachment relationship and the real relationship indicated that clients' secure attachment positively correlated with the client-rated real therapeutic relationship [42] [43] as well as with the clients' perceptions of the therapists' empathy [42].
Given the importance of the client attachment to therapist as measured by the CATS as well as the contention that there may be a strong cultural basis on how secure attachment is perceived [44], the aim of this study was to validate the CATS in the Greek context and explore its psychometric properties among a clinical Greek population.More specifically, the objectives of this study were the

Materials
Client Attachment to Therapist Scale (CATS) [1] is a 36-item self-report measure developed to assess clients' perceptions of the client-therapist relationship from the perspective of attachment theory.Participants respond on a 6-point scale ranging from strongly agree (1) to strongly disagree (6).The scale consisted of three dimensions: 1) secure attachment (clients' feeling encouraged to explore frightening or troubling material in therapy and experiencing the therapist as responsive, sensitive, emotionally available, and a comforting pres-Open Journal of Medical Psychology ence), 2) avoidant-fearful attachment (suspicion that the therapist is disapproving and likely to be rejecting if displeased, reluctance to make personal disclosures, and feeling threatened or humiliated in the sessions), and 3) preoccupied-merger attachment (longing for more contact and to be "at one" with the therapist, wishing to expand the relationship beyond the bounds of therapy, and preoccupation with the therapist and the therapist's other clients).In our sample, the 3 subscales as well as the total score of the scale demonstrated satisfied psychometric properties.Experiences in Close Relationships-Revised.The Greek Version of the Revised Experiences in Close Relationships (ECR-R) [45] [46] is a 36-item, self-report instrument of adult romantic attachment.The instrument comprises two subscales, 18 items each, which assess attachment anxiety and avoidance.Sample items are "I'm afraid that I will lose my partner's love" (anxiety subscale) and "I prefer not to show a partner how I feel deep down" (avoidance subscale).
Participants indicate their level of agreement to each item on a 7-point Likert-type scale, where 1 = strongly disagree and 7 = strongly agree.Instructions for completing the scale read: "Please take a moment to think about your overall experiences in romantic relationships including both your previous and current romantic experiences and answer the following questions".
Working Alliance Inventory.The Working Alliance Inventory (WAI) [47] [48] is a self-report questionnaire with 36 items which measures the strength of the therapeutic alliance during the therapeutic procedure.Each item is rated on a 7-point Likert scale (1 = Does not correspond at all; 7 = Corresponds exactly).This scale is available in three forms: one for the therapist, one for the client and one for the independent judge.In the present study, the scale for the client was used.In our sample, the scale demonstrated good internal consistency (α = 0.96).
Therapeutic Reactance Scale.The Therapeutic Reactance Scale (TRS) [49] is a 28-item self-report scale and it is used as the measure of psychological reactance, as it is defined by Brehm [50].This instrument consists of a total score as well as two subscale scores, which are labelled verbal and behavioural reactance.Items on the Therapeutic Reactance Scale tend to include statements regarding verbal and behavioural oppositional behaviour (i.e., "If I am told what to do, I often do the opposite"; "I am relatively opinionated"; and "I usually go along with others' advice").These are rated on a 4-point Likert Type scale anchored from strongly agree to strongly disagree.
Hope Scale.The Greek version of the Hope Scale (HS) [51] [52] was used to measure individuals' sense of successful goal-directed determination and planning of ways to meet goals using eight items rated on a 4-point Likert scale ranging from "Definitely False" to "Definitely True".In our sample, the scale demonstrated good internal consistency (α = 0.89).

Procedure Open Journal of Medical Psychology
the inclusion criteria of the study (i.e.older than 18 years, with no major psychopathology, and at least 5 completed sessions) were informed by the researchers about the purpose of the study, as being a validation assessment of a self-administered questionnaire that describes "aspects of the therapeutic relationship".Data were collected by the researchers with the confirmation that participation in the study was voluntary and had no impact on the treatment services received.Participation in the study did not include any incentives.All clients signed a written consent prior to their participation and anonymity was ensured.A subsample of 52 consecutive clients from the community mental health centre provided retest data, labelled with anonymous identifiers, by completing the CATS a month and two months after the initial survey.To ensure homogeneity in the retest process, all clients should have completed 5 -8 sessions when data collection took place, while in the rest of the sample clients were allowed to participate no matter how many sessions they had completed beyond the fifth session so as to increase variability [1].With regard to the clients from the training psychotherapy centre, all trainees participated in the study.Clients from the private practice were recruited by post-graduate psychology students who administered the questionnaire to adults of their social environment who underwent psychotherapy.In all cases, the time needed to complete the CATS was estimated to be approximately 15 minutes.

Statistical Analysis
Translation and adaptation into the Greek cultural context The forward-backward translation was adopted for the Greek version of the CATS, as it is the most common translation procedure [53].Specifically, two bilingual independent translators translated the English items to the Greek language (forward translation) and then two other translators translated back the Greek items into the original language (backward translation).Inaccuracies between the two versions were retranslated until full agreement was achieved between the authors and the independent translators (Table 1 In order to assess model fit, several fit indices were evaluated on the basis of the suggestions of Hu and Bentler [58]: x 2 ratio (x 2 /degrees of freedom less than 3), RMSEA [59] and SRMR [58] less than 0.08, CFI [60] and TLI [61] higher than 0.90.
Exploratory factor analysis.In order to further examine the factorial structure of the CATS, Exploratory Factor Analysis (Principal Axis Factoring) was conducted using the IBM SPSS, version 21, after the testing of the data suitability (sample size higher than 3:1 N:p ratio, KMO index higher than 0.50, and statistically significant Bartlett's Test of Sphericity 0) [56].Based on the factorial structure of the original version of the test we expected factor loadings on three factors based on the Kaiser's criterion of eigenvalues higher than 1, scree plot and parallel analysis using Monte Carlo PCA for Parallel Analysis.
Internal consistency reliability.Convergent and discriminant validity.Norms.Sten scores for the CATS factors were computed using Stanscore 4.

Normality testing
The distributional indices for the 36 CATS' were computed (see Table 2).
Item means ranged from 1. 35

Inter-item correlations
The correlations amongst the 36 CATS' items range from 0.00 to 0.63 (see Table 3).The results indicate that the correlation matrix is factorable, since many correlations are higher than 0.30, however, a lot of CATS items do not correlate significantly, which shows that possibly some items will be excluded from the factor analysis, since they will not load on any factor.

Confirmatory factor analysis
Firstly, the three-factor model that was identifiable using AMOS was created.
Then the normality of the items was assessed again, and the 8 items mentioned above showed unacceptable values to perform a CFA, thus, they were deleted one at a time, after testing the values again.Also, items 18 and 31 showed unacceptable factor loadings (less than 0.30) and they were deleted.Then, the multivariate normality of the CATS items was examined using Mardia's test, whose value was 30.066, less than 1368, which means that multinormality was achieved.
At that point, the model fit was tested; the extraction method was the maximum likelihood.The evaluation of the model fit indices showed that x 2 ratio value was acceptable (2.35 < 3), however, CFI and TLI values were 0.74 and 0.70 respectively, which is lower than the cut-off value of 0.90.The same applied for the RMSEA and SRMR values, which were 0.09 and 0.13 (>0.08).These values are indicative of an unacceptable model fit.

Exploratory factor analysis
In order to further examine the factorial structure of the CATS, an Exploratory Factor Analysis was performed.First of all, data suitability was tested; Sample size was higher than 3:1 N:p ratio, KMO index was 0.80 (higher than 0.50), and Bartlett's Test of Sphericity was statistically significant (p = 0.00).Then the number of factors was tested.Based on Kaiser's criterion of eigenvalues higher than 1, eight factors showed eigenvalues from 1.027 to 6.165.Based on the parallel analysis results, three factors had higher eigenvalues than the random eigenvalues and the same results were visible on the screen test.Also, the same number of factors has been identified by Mallinckrodt and his colleagues [1].
The factor correlation matrix was then examined in order to decide which rotation method to use.According to the results, the three factors showed low to moderate correlations to each other (r 1,2 = 0.21, r 2,3 = 0.16, r 1,3 = −0.40),thus an Oblique Rotation Method was used.The initial EFA results showed that items 11, 17, 24, 27, 30, and 36 did not load sufficiently on any factor (loadings lower than 0.30), thus they were deleted.
The results of the Principal Axis Factoring indicate that the three latent variables consisted of 30 items explain 42.43% of the total variance, which is adequate and shows how much of the variability in the data has been modelled by the extracted factors.More specifically, the first factor consists of eleven items (as depicted by the Secure Attachment to Therapist subscale of the original model), the second of nine items (as depicted by the Preoccupied-Merger Attachment to Therapist subscale of the original model), and the third of ten items (as depicted by the Avoidant-Fearful Attachment to Therapist subscale of the original model).The only divergence in comparison with the original model [1] was that item 31 ("I often wonder about my counselor's other clients.")loads on the Avoidant-Fearful instead of the Preoccupied-Merger attachment style (see Table 4).

Internal consistency reliability
The Cronbach alpha and Spearman-Brown coefficients were assessed in order to examine the internal consistency of the CATS's subscales.The secure, avoidant-fearful and preoccupied-merger subscales demonstrated good internal consistency.More specifically, their Cronbach alpha values were 0.81, 0.77, and 0.81, while their Spearman-Brown coefficient values were 0.82, 0.70, and 0.84, respectively.

Test-retest reliability
The test-retest reliability assessment showed satisfactory results for the three CATS subscales between the three time points.In particular, the test-retest values among time 1 and times 2, and 3 for the secure attachment subscale were 0.46 and 0.59 (p < 0.001), for the avoidant-fearful attachment subscale were 0.67 and 0.75 (p < 0.000), and for the preoccupied-merger attachment subscale were 0.81 and 0.67, respectively.Taking into consideration the existence of possible therapeutic change, possible psychopathological symptoms and the content of the "attachment to therapist style" construct, the test-retest results are adequately good.

Convergent and discriminant validity
To test convergent validity, the correlations between the CATS factors and similar constructs were examined.The results showed that 1) secure attachment subscale positively correlated to working alliance, 2) avoidant-fearful attachment Open Journal of Medical Psychology subscale was positively correlated to anxiety and avoidance in close relationships, and 3) preoccupied-merger attachment subscale positively correlated to anxiety in close relationships.Furthermore, the results of the discriminant validity testing showed that 1) secure attachment subscale negatively correlated to avoidant-fearful attachment subscale, 2) avoidant-fearful attachment subscale was negatively correlated to working alliance, 3) preoccupied-merger attachment subscale was not correlated with other attachment styles and avoidance in close relationships, and 4) all three subscales were not correlated to therapeutic reactance and hope levels (see Table 5).

Mean differences
The mean differences among the subgroups created by gender and age were examined and the results indicated that there were no statistically significant differences across the different groups (see Table 6).

Norms
To help psychotherapists and counsellors interpret the scores of the three subscales, the raw scores were converted to normalized scores (sten scores; ranging from 1 to 10; see Table 7).

Discussion
The main aim of this study was to examine the psychological properties of the CATS, which is a measure that assesses the client-therapist relationship from the perspective of attachment theory, in a clinical Greek population.More specifically, the objectives of this study were to establish the construct validity of the Client Attachment to Therapist, to study the internal consistency reliability, and construct reliability, to examine mean differences of CATS factors across gender and age and to evaluate the convergent and discriminant validity of the CATS with other constructs.
Firstly, forward-backward translation and bilingual participants verified, Note.S = secure attachment to therapist subscale, A-F = avoidant-fearful attachment to therapist subscale, P-M = preoccupied-merger attachment to therapist subscale.There are no significant differences.t = t-statistic in independent samples t-test, F = F-statistic in one-way analysis of variance (ANOVA), df = degrees of freedom.To further examine construct validity, the correlations between scale's factor scores and other measures were studied.The results indicated that there is a positive correlation between Secure Attachment and Working Alliance which is in accordance with a previous study [19].In addition, there is a positive relationship between Avoidant-Fearful Attachment and Avoidance as well as a positive relationship between Avoidant-Fearful Attachment and Anxiety in Close Relationships' Attachment.Finally, yet importantly, a positive correlation between Preoccupied-Merger Attachment and Anxiety in Close Relationships' Attachment was found.A number of recent studies have linked attachment avoidance and anxiety to distinct profiles of psychological distress and maladaptive interpersonal functioning [67], perfectionism [68] and social competencies [69].
Furthermore, the results showed that secure attachment subscale negatively correlated to avoidant-fearful attachment subscale and avoidant-fearful attachment subscale was negatively correlated to working alliance.These results are in accordance with other studies [19] [42] [43].Preoccupied-merger attachment subscale was not correlated with other attachment styles and avoidance in close relationships.Finally, all three subscales were not correlated to therapeutic reactance and hope levels.Some studies have indicated that many attachment associated outcomes are linked to the hope construct [70] [71] [72] but further research should be conducted in order to examine this relationship.
The strength of this paper rests on the lack of studies in Greece regarding client-therapist relationship because of the absence of a valid and reliable instrument as well to the statistical analysis that were used.As with all research, the present paper has a few limitations worth considering.Limitations of this study include those inherent to self-report measures as well as the relatively small samples used for analyses.Research with larger and more diverse samples should further identify the strengths and weakness of the CATS, as well as its predictive ability of outcome.

Conclusion
The study provides useful information regarding the use of the CATS in future studies in Greek speaking populations and it could expand psychology research.
Although more research is needed, the results of this study show that CATS is a valid and reliable instrument for measuring client attachment in the Greek cultural context, with good psychometric strength.Future research regarding the validation of the CATS in the Greek population could focus on more specific psychology measures while also exploring the cultural differences between different populations regarding clients' therapeutic attachment to the therapist.
Cronbach's alpha and Spearman-Brown prophecy coefficients were evaluated for the CATS factors, where values higher than 0.70 indicate good internal consistency [62] [63].Test-retest reliability.The consistency and stability of the client's responds were evaluated in three time points, at time 1 after having completed 5 -8 sessions, a month later (time 2) and then another month later (time 3).Values amongst 0.70 and 0.90 were considered as acceptable[63] [64].However, in the interpretation of the test-retest reliability values it should be kept in mind the time between tests, the context, the type of the sample and the content of the measured construct[65].
Note. S = secure attachment to therapist subscale, A-F = avoidant-fearful attachment to therapist subscale, P-M = preoccupied-merger attachment to therapist subscale.through test-retest (and high correlations), that the translated version was equivalent to the original one.Secondly, even though the confirmatory factor analysis does not confirm the theoretical model of the three factors suggested by Open Journal of Medical Psychology Mallinckrodt, Gantt and Coble[1], the exploratory factor analysis results in the same model after the deletion of six items that do not load on any factor (q11, q24, q30, q17, q36, q27).Also, one item (q31) loads on factor Avoidant-Fearful instead of Preoccupied-Merger type, as suggested by the authors of the original version.Test-retest reliability of the three factors is satisfactory and confirms the consistency of the Greek version and construct validity and item-total correlations supported the strength of the Greek scale and alignment to the original English version.Internal consistency showed high alpha coefficients, which indicate good internal consistency[62] [63].

Further research is needed
to test the psychometric properties of the CATS with other samples of population.Having established the CATS as a valid and reliable measure in the Greek cultural context, a number of interesting directions for further research would be opened.In particular, the CATS could be administered to assess the therapeutic relationship in specific populations in relation to their needs and attitudes towards counselling and psychotherapy services.For Open Journal of Medical Psychology example, relevant studies in the Greek cultural context have shown that Greek university students held a negative attitude towards visiting a university psychotherapy centre[73], despite the beneficial effects of psychological services in students' anxiety and depression levels as well as their self-esteem[74].Moreover, the existing needs of the Greek population for therapeutic services may have differentiated due to the recent global economic crisis[75] [76].Such an issue requires further investigation in relation to the role of the therapeutic relationship, so as to tailor the mental health services to adequately address the emerging needs.Overall, the use of the CATS is expected to shed light on major dimensions of the therapeutic relationship and the therapeutic process generally, an under-examined research subject so far in the Greek context.Finally, cross cultural studies would add strength to the research area of clinical psychology and psychotherapy.
To further examine construct validity, the correlations between the scale's factor scores and other measures were examined.Positive correlations with similar constructs would indicate good convergent validity and negative or non-correlations with totally different constructs would indicate good discriminant validity of scale's factors.
Note.SD = standard deviation, SE = standard error of mean, Var = variance, R = item should be reverse keyed.

Table 4 .
CATS item loadings per factor.

Table 5 .
Convergent and discriminant validity: Average correlations of CATS factors with other constructs.

Table 6 .
Mean differences (means and standard deviations) of CATS factors by demographics.

Table 7 .
Norms for the CATS factors