A confirmatory factor analysis of the Mandarin-Chinese version of the Barrett-Lennard Relationship Inventory

ABSTRACT The aim of this study was to translate and provide an initial validation for a full Mandarin-Chinese version of the Barrett-Lennard Relationship Inventory (B-L RI:MC) to include forms Other toward Self-64 (OS-64) and Other toward Self-40 (OS-40) for use in the Mandarin-Chinese research and clinical contexts. B-L RI:MC OS-64 was translated by a bilingual panel and subsequently administered to 658 Mandarin-speaking Taiwanese respondents online using an age-stratified random sampling strategy. Through both the factor analytic strategy of principle component analysis (PCA) and confirmatory factor analysis (CFA), the reliability and construct validity were investigated. The final results support the original four subscale dimensionality of the inventory. B-L RI:MC OS-64 showed Cronbach’s alpha was .96 and KMO = .97. PCA using Varimax rotation yielded a four-factor model supporting the sub-scales: level of regard, empathic understanding, unconditionality of regard and congruence, which explained 49.911% squared loading of the total variance. B-L RI:MC OS-64 and OS-40 were supported by the structures in CFA, which displayed NFI = .95 and .95, CFI = .97 and .96, IFI = .97 and .96, and RMSEA = .092 and .091, indicating a promising construct validity. In conclusion B-L RI:MC OS-64 and OS-40 versions can be considered appropriate for measuring the Rogerian therapeutic relationship conditions within a Mandarin speaking community.


Introduction
The Barrett-Lennard Relationship Inventory (B-L RI) is the most well-known questionnaire developed specifically for evaluating the interpersonal therapeutic relationship as defined by Rogers (1957). The scale was originally developed by Barrett-Lennard when working in the University of Wisconsin where Carl Rogers and his colleagues carried out studies into psychotherapy with people with a diagnosis of schizophrenia (Thorne & Sanders, 2013, p. 112). As a pioneer of contemporary psychological research, Rogers hypothesized there to be 4 conditions in therapeutic relationships: empathic understanding, positive regard, congruence and unconditionality of regard (Barrett-Lennard, 1959a, 1959b, 1962Rogers, 1957). Acknowledging the positive impact of Carl Rogers' theory, the Barrett-Lennard Relationship Inventory has been gradually adapted into different forms, such as the full 64-item form and the basic 40-item form (Barrett-Lennard, 1978, 2015Gurman, 1977). It has been applied worldwide in evaluating different kinds of relationship, such as therapist-client relationship, immediate family, close peer friend relationship and teacher-student relationship (Barrett-Lennard, 2015, pp. 27-31;Berzon, 1964;Hollenbeck, 1961Hollenbeck, , 1965Snelbecker, 1967;Walder & Little, 1969).
There is a significant amount of evidence that points towards the positive relation between the therapeutic relationship conditions measured by the Barrett-Lennard Relationship Inventory with successful psychotherapy outcomes. This has been shown across a wide range of patients experiencing various forms of psychological distress, including adult out-patient services for depression in clinical trials (Ablon & Jones, 1999;Blatt & Zuroff, 2005;Zuroff & Blatt, 2006), treatment studies for depression (Watson & Geller, 2005;Watson, Gordon, Stermac, Kalogerakos, & Steckley, 2003), youth and family therapy (Karver, Handelsman, Fields, & Bickman, 2006), severe psychosis (Hewitt & Coffey, 2005;Rogers, Gendlin, Kiesler, & Truax, 1967), and within general counselling (Archer, Forbes, Metcalfe, & Winter, 2000). For this reason, there is ample justification to consider the effects of the therapeutic relationship conditions as set out by Rogers in Asian culture. However, before this can be done effectively the scale for measuring the therapeutic relationship conditions needs to be translated and validated.
The Barrett-Lennard Relationship Inventory has been either fully or partially translated into 20 languages: American Sign, Arabic, Mandarin Chinese (partially), Czech Republic, Dutch, French, German, Greek, Hebrew, Iranian, Italian, Japanese, Korean, Malaysian, Polish, Portuguese, Slovak, Spanish, Swedish, andTurkish since 1964 (Barrett-Lennard, 2015). The previous Mandarin-Chinese version was only partially translated encompassing just 14 items of the empathic understanding sub-scale of the full Barrett-Lennard Relationship Inventory 64-item version. These 14 items were initially translated in 2006 by Chu and Tseng (2013) to develop an inventory for evaluating the quality of relationships in medical care aiming to help improve relationships within public health work in Taiwan. Their study showed that empathy as measured by the Barrett-Lennard Relationship Inventory was a relevant factor in considering the health literacy and understanding information; with higher levels of physician empathy being related to higher health literacy and understanding information in patients. This study provides an important link between the therapeutic relationship condition empathic understanding with improved health outcomes in an Asian context.
There is a growing need for effective, culturally sensitive, psychotherapies as the reported level of mental health problems is increasing worldwide. It has been reported that 1 in 4 people in Taiwan, a country with a 23 million population (Table 1) located in the Asia Pacific region, are suffering from common mental health problems, such as depression and anxiety disorder (Department of Census, Directorate General of Budget Accounting & Statistics [DGBAS], 2016a, DGBAS, 2016b, DGBAS, 2016c, DGBAS, 2016dDepartment of Household Registration Affairs, 2016;Fu et al., 2013;Ministry of Health and Welfare Taiwan, 2015). The World Health Organization (WHO) reported that the number of registered psychiatrists in South East Asia and Africa has increased 25% more than the number in 2011 according to the report of the Mental Health Atlas in 2014 (World Health Organisation [WHO], 2014). The population of other mental health professionals, such as psychiatric nurses, has grown by 37% (WHO, 2014, p. 53). In the Mandarin speaking world there is currently a surge in the development of psychological services to support people's mental wellbeing (Ministry of Health and Welfare Taiwan, 2015). These reports have led to a growing understanding that the proportion of the population experiencing psychological distress is increasing in developing countries. To meet the needs of the population those staff providing services need to be trained and equipped with evidence based approaches. Therefore, translating and validating the most widely used therapeutic relationship inventories can contribute to the development of the quality of mental health care available (Lee, Li, Arai, & Puntillo, 2009;Murphy, Cramer, & Joseph, 2012;National Institute for Mental Health and Royal College of Psychiatrists, 2005;Priebe & Gruyters, 1993;Priebe, Richardson, Cooney, Adedeji, & McCabe, 2011;Rogers, 2004;Slovák et al., 2015). Access to a translated version of the Barrett-Lennard Relationship Inventory would also provide a scope for researchers and clinicians to make meaningful comparisons across cultural divides Pescosolido, Medina, Martin, & Long, 2013;Rogers & Murphy, 2017;WHO, 1988).
The Chinese speaking population is approximately 14.4% of the world's population. The language Mandarin-Chinese, otherwise known as Standard Chinese, shares the similar characteristics with other Chinese language groups, such as Wu, Min, Yue, Jin, Xiang, Hakka, Gan, Huizhou, and Pinghua (Lewis, Simons, & Fennig, 2015). It is considered as the official language of China and Taiwan, as well as one of the four official dialects in Singapore. Mandarin is also widely used in Malaysia and Indonesia (Kurpaska, 2010). The language Mandarin-Chinese often requires professional translation services within industry. However, translating one language into another is playing the role of an ambassador for languages which are carrying specific cultural images of nations, ethnic groups and individuals to introduce the cultural uniqueness to foreigners. Regardless of the variety of Chinese sub-languages, there are two written systems: Simplified and Traditional Chinese characters used in translating services, yet they represent the mutual meanings in Chinese contexts (Li, Ran, & Xia, 2010;Liu, 2014).
Some studies have argued that certain diversities in character or content of language and culture might cause a difficulty to address the cultural image in translating the work, such as, the religious belief and philosophy of life which could be distinguishable in Chinese society and English culture (Liu, 2014). In contrast, some researchers have found that brain areas, such as the ventral occipitotemporal regions and Cerebellum, are involved in reading in whichever language (Herbster, Mintun, Nebes, & Becker, 1997;Petersen, Fox, Posner, Mintun, & Raichle, 1988;Petersen, Fox, Snyder, & Raichle, 1990;Pugh et al., 1997;Rumsey et al., 1997). The comparison of brain images whilst reading Chinese orthographic characters and English alphabetic words through fMRI (the functional magnetic resonance imaging) experiments demonstrated that the left inferior prefrontal cortex was active and involved in processing both Chinese characters and English word recognition (Tan et al., 2001). Regardless of simplified and translational Chinese writing systems, a study in cognitive science displayed a high similarity between two written characters in Chinese reading and recognition (Liu, Chuk, Yeh, & Hsiao, 2016). Although readers of simplified Chinese might encounter difficulties when writing traditional characters, the data indicated their competence in learning to read and write using simplified characters was transferred to processing traditional characters conceptually and comprehensively (Liu et al., 2016). Thus, translating conceptually and comprehensively a complete Mandarin-Chinese version of Barrett-Lennard Relationship Inventory (B-L RI:MC) which crosses over two languages and two written systems is possible.
As the person-centred approach becomes more well established in the Asia-Pacific region and Eastern cultures it will be useful to have access to translated measures that can assess the theoretical constructs as originally intended (Motoyama & Murphy, 2017). This will enable and inform the cultural relevance for their clinical and research application. It will also be advantageous to have the Barrett-Lennard Relationship Inventory translated to Mandarin as person-centred approaches to mental health care and psychotherapy need to be evaluated using theoretically consistent measures. In addition, the American Psychological Association and the Chinese Psychological Society have been working towards greater integration of these two systems (American Psychological Society [APA], 2016) suggesting the likelihood of future cross over in professional activity between USA and China will continue to grow. Having access to psychological measures of the therapeutic relationship available in both English and Mandarin will support the development of intercultural collaborations in research and practice.
Hence, the purpose of this study was to translate the complete 64 items and construct a Mandarin-Chinese version of Barrett-Lennard Relationship Inventory from the original English Barrett-Lennard Relationship Inventory. The aim is to use the form Other toward Self-64 (OS-64), and then validate both Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 and OS-40 to provide a contextually comprehensive measurement to evaluate relationships in the Mandarin-Chinese speaking community. Confirmatory factor analysis (CFA) is one of the methods used to investigate construct validity of psychological measures (Fournier-Vicente, Larigauderie, & Gaonac'h, 2008). Instead of constructing an inductive theory like exploratory factor analysis (EFA), Confirmatory factor analysis is an instrument which extracts latent factors from the overall observed variables and specifies a model based upon hypotheses (McArdle, 1996). It is a procedure of theory deduction through the test of construct validity of hypothesis-based questionnaires (Atkinson et al., 2011). Therefore, we examined the collected data set to evaluate the construct validity of the form OS-64 and OS-40 of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version using confirmatory factor analysis. We hypothesized that the fourfactor model (i.e. level of regard, empathic understanding, congruence and unconditionality of regard) would be replicated in the analysis of the Barrett-Lennard

Method
This study was carried out in three stages. The first stage involved the linguistic translation of the English language scale to the Barrett-Lennard Relationship Inventory Mandarin-Chinese version. The second stage consisted of testing the measurement properties of the 64 items in the Barrett-Lennard Relationship Inventory Mandarin-Chinese version, which included the tests for reliability and construct validity of the items using principle component analysis (PCA). Subsequently, conducting an investigation on the fitness of models of Form OS-64 and OS-40 with the use of confirmatory factor analysis in the final stage. The research received ethical approval from the University Research Ethics Committee.

Cross-cultural translation
Cross-cultural translation of the original Barrett-Lennard Relationship Inventory English version to the Mandarin-Chinese version was completed in three stages.
Three bilingual translators, who each spoke Mandarin-Chinese as their first language and specialized in the person-centred approach to counselling and psychotherapy, translated the items of the Barrett-Lennard Relationship Inventory into Mandarin-Chinese. In order to retain the original meaning of the scale items accurately, and to generate an optimal comprehensive translation, it was considered essential for each translator to have inside knowledge of the theoretical constructs within the scale to also achieve conceptual and semantic equivalence (Flaherty et al., 1988;Wang, Lee, & Fetzer, 2006;Lee et al., 2009;Barrett-Lennard, 2015, p. 158). Two of the translators were person-centred counsellors and the third translator was completing doctoral research in to the person-centred approach (lead author of this article).
To begin, the forward translation approach was performed. Each translator was assigned a set of items across each of the different dimensions of the Barrett-Lennard Relationship Inventory and translated them individually. The translators translated the inventory in accordance with the knowledge and understanding of Carl Rogers's person-centred theory and the principle of maintaining the content and semantic equivalences in the translation. Secondly, the expert review panel was established. Each of the translators then reviewed the translations made by each of the other translators in a 'round-robin' to identify and modify any of the inaccurate expressions of concepts in the translation of each dimension. Lastly, a process of back translation was carried out by a language specialist. The back translator was a linguist, who did not have any prior knowledge or understanding of person-centred counselling and psychotherapy. In the back-translation process, the suitability of the amended Mandarin-Chinese version in the second stage was examined through reverse translation and comparison with the original English version, Barrett-Lennard Relationship Inventory.
The final stage of the translation process involved the pilot test of the penultimate Barrett-Lennard Relationship Inventory Mandarin-Chinese version. The scale was completed by three Taiwanese people who were not in the field of person-centred counselling and psychotherapy. Each respondent completed the Barrett-Lennard Relationship Inventory Mandarin-Chinese version and subsequently they were interviewed about any obstacles in completing the questionnaire and asked about their understanding of each item. All the suggestions and findings were considered to modify the final Barrett-Lennard Relationship Inventory Mandarin-Chinese version before going forward to further validation.

Participants
The target sample was to recruit approximately 640 Taiwanese potential respondents, who were 18 years old or over and spoke Mandarin-Chinese as their first language. According to Tabachnick (2007) when estimating the sample size of prospective respondents to a distributed questionnaire, theoretically, there should be at least 10 individuals multiplied by the total number of scale items in the questionnaire. It is important to ensure accuracy in the results of the validation in any study and proper determination of the number of respondents can help reduce research error and thus strengthen the impact of results (Martínez-Mesa & Bastos, 2014).
The stratified random sampling method was performed to ensure that at least one observation was picked from each of the strata and is a suitable method to recruit the samples into stratum (Carl-Erik, Swensson, & Wretman, 2003). By stratifying the target population, the measurement can be placed into manageable groups and the representativeness of each group can be estimated. Stratification also provides a smaller error in estimation when there is a lower standard deviation in the measurement. In this study, the respondents were recruited by age grouping: 18-25, 26-35, 36-45, 46-55, 56-65, and over 66 years (see Table 1). In the final analysis, there were 658 Taiwanese respondents that completed the Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 using an online survey, by following Watts' success in 1989(Barrett-Lennard, 2015Watts, 1989), over a two-month period. Information concerning the demography of the respondents was also collected, such as gender, occupation, and the target relationship evaluated when completing the scale (Table 1).

Instruments
The original English version of the Barrett-Lennard Relationship Inventory was developed from the core concepts of Rogers's theory of the necessary and sufficient conditions for personality change in the therapeutic relationship (Barrett-Lennard, 1964). The Barrett-Lennard Relationship Inventory was developed as a Likert-type measurement for assessing relationships. The scale is constructed to enable obtaining an equal number of positively and negatively worded items for each sub-scale. Each item is rated on differing strengths of No or Yes in the range −3 to +3 ( Barrett-Lennard, 2015, pp. 26-34, 40-41). The internal reliability of the original 64-item Barrett-Lennard Relationship Inventory exceeded .80 completed with a data set that consisted of 82 people, including 42 psychotherapy clients and 40 therapists (Barrett-Lennard, 2015, p. 43). An early review of the 64-item Barrett-Lennard Relationship Inventory indicated the internal reliability coefficients of four sub-scales: level of regard = .91, empathic understanding = .84, unconditionality of regard = .74 and congruence = .88 (Gurman, 1977).
The Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 was translated in accordance with the original English version which contains the four dimensions: level of regard, empathic understanding, congruence and unconditionality of regard. Each sub-scale dimension contains 16-items, including 8 positive items and 8 negatively worded items (Barrett-Lennard, 2015, pp. 26-34).
First, the level of regard refers to the regardfulness of one person's response to another, and it might embed positive or negative feelings (Barrett-Lennard, 2015, p. 11). Secondly, the concept of empathic understanding is defined as the degree to which one person truly recognizes the felt awareness and meaning of another (Barrett-Lennard, 2015, p. 10). Thirdly, the definition of unconditionality of regard is given as the nonjudging affective response of one person towards another (Barrett-Lennard, 2015, p. 11). Finally, the concept of congruence is centred on the consistency between whole present experience and underlying awareness, for example, a congruent person can be honest, sincere and direct to another without hesitation or feeling compelled during the communication (Barrett-Lennard, 2015, p. 11). All the items in each sub-scale were arranged in the same order in the original Barrett-Lennard Relationship Inventory Form OS-64: one item of level of regard followed by an item for empathic understanding then unconditionality of regard and finally an item for congruence. This repeating pattern continues throughout the entire scale.
In this study, the Barrett-Lennard Relationship Inventory Mandarin-Chinese version: From OS-64 was administered to the respondents as a 64-item self-report measure of the relationship with a specific friend. The responses to items was recorded using a six-point Likert scale with scoring options: −3 = No, I strongly feel that it is not true, −2 = No, I feel it is not true, −1 = No, I feel that it is probably untrue, or more untrue than true, +1 = Yes, I feel that it is probably true, or more true than untrue, +2 = Yes, I feel it is true, and +3 = Yes, I strongly feel that it is true, which was located from right to left.
All the data were collected using an online survey advertised through social network sites. To prevent missing items, the online survey had a pre-setting to ensure the completion of each item in the questionnaire. Therefore, there were no missing data in this survey.

Data analysis
This study aimed to validate the Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 and Form OS-40, with confirmatory factor analysis (CFA). The data were analysed using SPSS Version 23.0 (SPSS Inc., 2015) and LISREL Version 8.7 (Jöreskog & Sörbom, 2001) for Windows. Initially, the internal reliability was analysed by calculating Cronbach's alpha indicating the degree of relatedness among the 64 items in the entire inventory. Secondly, Kaiser-Meyer-Olkin Measure of Sampling Adequacy (KMO) and Bartlett's Test of Sphericity were calculated using principal component analysis (PCA) with Varimax rotation, which indicated the adequacy for running the factor analysis (Field, 2005, pp. 619-666). Thirdly, a parallel analysis between the Monte Carlo simulation (MC simulation) and the PCA was performed. The aim of this is to reduce the 'noise' within the factor structure and determine those significant components within the overall group of components in the model of Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 (Conedera et al., 2011;Inoue, Hukushima, & Okada, 2006;Nasser & Wisenbaker, 2003;Sarıyar, Perk, Akman, & Hortaçsu, 2006). In statistics, a Monte Carlo simulation is one approach used to determine the properties of some phenomenon with a large number of random sampling, yet it does not always request truly random numbers. Monte Carlo simulation provides an intuitive understanding of the estimated components in a composition (Inoue et al., 2006). Hence, 658 subjects, 64 variables, 100 sets of the desired number of parallel data, and desired percentile 95.5, the components in the PCA which have a lower eigenvalue than those in the Monte Carlo simulation can be excluded from the factor structures (Inoue et al., 2006;Nasser & Wisenbaker, 2003). Lastly, confirmatory factor analysis (CFA) was used to investigate the fitness of models of the data for the Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 and Form OS-40.

Reliability analysis
To evaluate homogeneity, the overall consistency of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version was analysed. The Barrett-Lennard Relationship Inventory Mandarin-Chinese version was found to have high internal consistency with Cronbach's alpha = .96 in the entire sample (n = 658), .95 in the male group (n = 162), and .96 in the female group (n = 495), where a high alpha indicates a strong internal correlation of each item (Table 2). A Cronbach's alpha between 0.70 and 0.95 would be considered excellent (Terwee et al., 2007).
Separate reliabilities were calculated for all four of sub-scale variables: level of regard, empathic understanding, unconditionality of regard, and congruence, all of which exceeded the minimum Cronbach's alpha and were .94, .84, .75, and .89 respectively; which represented a high internal consistency across each of the four sub-scales in the Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 (Table 2).

Principal components analysis
Principal Components Analysis (PCA) with the Varimax rotation method was performed. PCA results showed a Kaiser-Meyer-Olkin (KMO) coefficient of .97 in the entire data set (n = 658), .89 in the male group (n = 162) and .97 in the female group (n = 495) and were well above the recommended .70. The P-value of Bartlett's Test of Sphericity of .000 (approx. chi-square 24039.755, df 2016) also suggested satisfactory sampling adequacy ( Table 2).
The 64-item inter-correlation matrix was analysed and 11 components were extracted (eigenvalue >1) in the initial model of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version. Using this data, the first five components in the Barrett-Lennard Relationship Inventory Mandarin-Chinese version were extracted that showed an eigenvalue = 22.359 (>1.721 in Monte Carlo simulation), 3.794 (>1.656 in Monte Carlo simulation), 2.150 (>1.610 in MC simulation), 1.894 (>1.574 in Monte Carlo simulation) and 1.746 (>1.533 in Monte Carlo simulation), which were able to account for 34.936%, 5.928%, 3.360%, 2.959% and 2.727% of the total explanatory variance of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version respectively (Table 3). Furthermore, by comparing the eigenvalues with the suggested eigenvalue generated in the parallel analysis using the Monte Carlo simulation, the eigenvalue of the sixth and seventh components in the Barrett-Lennard Relationship Inventory Mandarin-Chinese version were 1.447 and 1.321. These were less than the suggested eigenvaules 1.503 and 1.474 in the Monte Carlo simulation. Thus, the sixth and seventh components were explained as the noises of the factor structure (Table 3).
This finding remains consistent with the Scree Plot that also indicated that the first five components could be extracted from the Barrett-Lennard Relationship Inventory Mandarin-Chinese version (Figure 1).
If the criterion for a fixed number of components was set at 5 and for high loading the level is set at equal to or greater than ±0.49, there were only 2 items (C12 and R53) correlated highly with more than one component. The loadings on five components are presented as follows (see Table 4): the first and largest component accounted for 34.936% of the variance. Since 11 of the 15 items were from the level of regard subscale, this component was best identified as reflecting level of regard. The remaining items were two from the empathic understanding, one from the unconditionality of regard, and another one from the congruence scale.
The second largest loading component accounted for 5.928% of the variance. Since6 of these 14 items were from the congruence sub-scale and two were from the level of regard (R49 and R53), three from empathic understanding (Em22, Em50, Em58), and three from the unconditionality of regard (U55, U27, U19), the second component could be classified as partially but predominantly representing the congruence subscale.
The third component accounted for 3.360% of the variances. Since all 8 of these items were from the congruence, this component was best interpreted as characterizing the congruence sub-scale. Furthermore, the fourth component accounted for 2.959% of the variances. As all six of these items came from the empathic understanding, this component was best labelled as representing the empathic understanding sub-scale. Finally, the fifth component accounted for 2.727% of the variances. All items came from the unconditionality of regard, and therefore were reflective of the sub-scale of unconditionality of regard.

Confirmatory factor analysis
The confirmatory factor analysis of Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 was completed using the same sample of individuals, who had reflected on one specific example of their relationship with a friend. The present data set satisfied the confirmatory factor analysis requirement of comparative fit (Schreiber, Nora, Stage, Barlow, & King, 2006). The correlation matrix of four latent factors: level of regard (R), empathic understanding (Em), unconditionality of regard (U), and congruence (C), is displayed in Table 5 where all the factors showed logical interrelationships. R correlated highly with Em (r = 0.89, p < 0.001), U (r = −0.82, p < 0.001) and C (r = 0.89, p < 0.001). Em correlated highly with U (r = −0.85, p < 0.001) and C (r = 0.88, p < 0.001), and lastly, C was also correlated significantly with U (r = −0.84, p < 0.001).

Discussion
The Barrett-Lennard Relationship Inventory Mandarin-Chinese version aims to provide a Mandarin-Chinese version of relationship measurement to the Mandarin-Chinese community. Historically, the previous studies on the development and validation of the Barrett-Lennard Relationship Inventory had reinforced the need for evaluating relationships using psychometric assessment instruments, such as for the therapeutic relationship, friendship, and teacher-student relationship, etc. Despite previous studies with the Barrett-Lennard Relationship Inventory using exploratory factor analysis supporting its validity, to our knowledge no other investigation has been conducted using confirmatory factor analysis. This means that for the first time in over half century there are data available to confirm the theoretical model underpinning the therapeutic relationship questionnaire proposed by Barrett-Lennard (1962). This study has also reported the CFA validation of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 and Form OS-40 in a sample of people (n = 658), whose first language is Mandarin-Chinese and are 18 years old and above evaluating their perceptions of a relationship with a friend. The findings of this study suggest that the translated 64-items in the Barrett-Lennard Relationship Inventory Mandarin-Chinese version can reliably assess the effect of level of regard, empathic understanding, unconditionality of regard and congruence in relationships. The Cronbach's alpha of the sample set of 658 variables (162 males, 495 females, and one other) exceeds .96 which was higher than the original English version (.80) and indicates a high internal consistency across individual items (Barrett-Lennard, 2015, pp. 17-18, 43). The Cronbach's alpha of both male and female sample groups also shows strong internal correlation between each item. Thus, the reliability of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version is as significant as the original version and it can be applied towards all genders. Furthermore, the separate Cronbach's alpha of each sub-scale is .94 for level of regard, .84 for empathic understanding, .75 for unconditionality of regard and .89 for congruence scale, which is considered an excellent level of reliability across 16 items in each sub-scale in the Barrett-Lennard Relationship Inventory Mandarin-Chinese version. The results of Gurman's validation study reported that the mean coefficient for the sub-scales were .91 for level of regard, .84 for empathic understanding, .74 for unconditionality of regard and .88 for congruence scale (Barrett-Lennard, 2015, p. 43;Gurman, 1977). It implies that the Barrett-Lennard Relationship Inventory Mandarin-Chinese version also provided strong internal consistency in the entire scale and individual sub-scales as did the Barrett-Lennard Relationship Inventory.
The Barrett-Lennard Relationship Inventory Mandarin-Chinese version has been confirmed theoretically. It has found that the first five components in the Barrett-Lennard Relationship Inventory Mandarin-Chinese version could be extracted and identified as representative of the four latent factors in the original Barrett-Lennard Relationship Inventory through PCA. In the CFA conducted within this study, the results of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version: Form OS-64 and Form OS-40 further support the original hypothesized four-factor structure which were designed to measure therapeutic relationships. Both the OS-64 and OS-40 models exhibit a good fit for the data and present a consistent correlation between each of the four factors. The result of our analysis of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version has revealed the scale translation can be used to measure relationships using the four factors of the Rogerian therapeutic relationship in a Mandarin-Chinese context. Due to the data was collected from people in general with different occupations evaluating their friendships, it further implies that Taiwanese people are capable of differentiating between the different relationship factors important to therapeutic relationships and based on Rogers's person-centred theory. Regarding the friendship length, 63.7% respondents associated the term 'friendship' intuitively with a person, excluding family relations or romance, with whom they have shared a long-term relationship. For example, a more-than-5year friendship in this study. This phenomenon might evoke a thought whether a therapeutic relationship in clinical settings can be developed in three to five sessions between the Mandarin-Chinese speaking therapists and clients? Barrett-Lennard (1962) has suggested that at least five sessions of psychotherapy are required before an accurate rating could be gained, yet Murphy and Cramer's study in 2014 revealed that after three sessions people can predict the outcome using the scores of the Barrett-Lennard Relationship Inventory and similarly other studies have shown early ratings predict later improvement (Barrett-Lennard, 2015;Murphy & Cramer, 2014;Murphy et al., 2012). However, the existence of a high percentage of the respondents evaluating the long-term friendships in this study did not countervail the reliability of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version but might have exceeded it instead, and it might also have reinforced the validity of the translated inventory.
In terms of those few low loading single items which exist in each sub-scale. For instance, there is one in level of regard, three in empathic understanding, nine in unconditionality of regard and three in congruence sub-scale in the OS-64. The Unconditionality of Regard items involve quite subtle distinctions, difficult to render clearly even in English and for respondents with differing attitudes. Even with the care taken in the translation process in the present study, it is possible the items were not quite as well translated and understood by respondents in Mandarin as they might in English wherein they were originally constructed. Nevertheless, the analytical result provided strong internal consistency, reliability and construct validity, thus the Barrett-Lennard Relationship Inventory Mandarin-Chinese version could be considered as a well-translated version that manages to evaluate relationships in Chinese culture.

Implication
The study has reported very promising results regarding the reliability and construct validity of the Barrett-Lennard Relationship Inventory Mandarin-Chinese version for the measurement of relationships. A survey on the Chinese counselling approach conducted in 1988 revealed that behaviourism, psychoanalysis and cognitive therapy have been the mainstream theoretical orientations of Chinese counsellors (Chang, Tong, Shi, & Zeng, 2005). Although there are few studies on the practical approaches of person-centred counselling in the development of counselling and psychotherapy in China, the result of validating the Barrett-Lennard Relationship Inventory Mandarin-Chinese version has provided evidence that the translation imparts and establishes the sensitivity of the original Barrett-Lennard Relationship Inventory. It also indicates that the therapeutic conditions of Rogers's person-centred therapy can be identified by the people in Taiwan and also applied in the Mandarin-Chinese community. Thus, transcending Carl Rogers's person-centred theory into a variety of fields in Chinese Society, such as the therapeutic relationship within social services and clinical settings, teacher-student relationship in education, family relationship, relationship between organizations, relationship in business, etc., to grow a mutual understanding of humanistic interaction.
Regarding the mental health professionals' training, Barrett-Lennard has revealed one of his studies on the experiences of helping mental health professionals' experiential learning in three 2-week workshops where Barrett-Lennard and his colleagues had indicated that the Barrett-Lennard Relationship Inventory was used as a rating instrument to evaluate the outcome effects of the learning ( Barrett-Lennard, 2017, pp. 331-338). Therefore, the Barrett-Lennard Relationship Inventory Mandarin-Chinese version could be used in educational settings in the Mandarin-Chinese speaking world. Taking China as an example, the counselling and psychotherapy services in medical settings like special counselling and mental health centres have been developed rapidly since the 1980s because of the increasing psychological problems, such as depression and suicide (Chang et al., 2005;Higgins et al., 2008;Qian et al., 2012). Using adapted forms of Barrett-Lennard Relationship Inventory Mandarin-Chinese version, such as the Other toward Self (OS) and Me toward Other (MO) forms, to evaluate the effectiveness of training programs from therapists' and clients' perspective (Barrett-Lennard, 2017, pp. 331-338).
Last but not the least, the Barrett-Lennard Relationship Inventory Mandarin-Chinese version would be one of the vehicles which give the current Chinese counselling and psychotherapy services the momentum to go beyond the medical settings for their communities. The theoretical structure in the Barrett-Lennard Relationship Inventory, which Carl Rogers's person-centred approach emphasizes an interpersonal relationship with unconditional positive regard would carrying out psychological changes for clients, would meet Chinese culture where Mandarin speaking clients' adjustment often relates to their relationship with others, such as family and friends, and the interaction with their social circles, such as neighbours, teachers at school, colleagues at work, etc. The full 64-item and 40-item Form of Barrett-Lennard Relationship Inventory has been gradually adapted into many versions, for instance 'other toward self' (OS), 'myself toward other' (MO-40), 'observer' (Obs-40), 'teachers toward students' (MO-40:TS), 'students toward teachers' (OS-40: T-S), 'other toward young children' (OS-40CH), 'other in close relationship toward self' (OS-LR-40) and relationships between 'groups or organizations' (GS-40) version, which the researchers use to measure relationships from different points of view (Barrett-Lennard, 1978, 2015, 2017Gurman, 1977). Through the previous contributions of the Barrett-Lennard Relationship Inventory, the validated 64-item and 40-item Mandarin-Chinese versions can easily be transformed into the specific versions when it is needed. Hence, this Barrett-Lennard Relationship Inventory Mandarin-Chinese version would contribute to the international research community and extend the borders of the community of counselling and psychotherapy to those in the Mandarin-Chinese speaking world.

Limitations
There are several limitations to the present study. Firstly, the instrument of data collection could be further refined. Among the 658 respondents, only 2 people appeared in the age group of 66 and over. This might be related to varying levels of familiarity of computer usage and willingness to access and complete online surveys. Thus, distributing not only an online survey but also a paper and pen version of a questionnaire could multiply and better represent the sample size of older age groups.
Secondly, understanding of the scale item wording was varied as suggested by the feedback in an open dialogue box included at the end of the survey. For example, there might be scope for more intuitive translation that could be provided. It took approximately 30 minutes to complete the scale, therefore tiredness might have affected some online respondents. However, it should be noted that the item wording in the English version of the scale is reported to be quite complex and the full 64-item scale is time consuming to complete. In this sense, there is modest difference between the two versions of the scale. Items that require some careful consideration before responding can make participants really work to understand what they are being asked. Doing this, to the end of the task, may have been too much for some. Face to face administration could yield higher or more even validity.
Lastly, this study has gone some way to being the first study to translate the original version of the Barrett-Lennard Relationship Inventory into the Mandarin-Chinese version and examine the construct validity using confirmatory factor analysis. Whilst this was reported with satisfactory results in this study, both the OS-64 and the OS-40 scales are long and time-consuming. The test-retest reliability could be conducted if a shorter version of the scale for future use could be developed. Therefore, it would be useful to pilot in a clinical sample as the next step which is using the shorter time frame for relationships and also test-retest statistics to be calculated.

Disclosure statement
No potential conflict of interest was reported by the authors. Godfrey Barrett-Lennard studied with Carl Rogers at the University of Chicago, graduating with a Ph.D. from there. He is Adjunct Professor in the School of Psychology and Exercise Science and in the School of Health Professions at Murdoch University in Perth, Australia, and he is also Fellow of APA and its Australian equivalent. His areas of interest and expertise involves exploring the theory and measurement of helping other relationships; self-concept theory and study. He has highly commended research in the areas of inquiry into experiential group processes, adult life events and change, community and identity. His career in Australia and North America has balanced scholarly contributions with teaching and practice. His later writings included journal articles, chapters and research questionnaires, and further adaptations of his Relationship Inventory. His latest book, Experiential Learning for Professional Helpers: A Residential Workshop Innovation (Springer, 2017), which describes a series of ground-breaking residential workshops in therapeutic counselling in the 1960s and makes a compelling argument that such an intensive participatory process is as powerful today as it was, is a great resource for students and scholars in applied psychology and psychotherapy, as well as for practicing therapists and trainees committed to meaningful work with their client groups.