Social interactions have gained increasing importance, both as an outcome and as a possible mediator in psychotherapy research. Still, there is a lack of adequate measures capturing relational aspects in multi-person settings. We present a new measure to assess relevant dimensions of quality of relationships and collective efficacy regarding interpersonal interactions in diverse personal and professional social systems including couple partnerships, families, and working teams: the EVOS. Theoretical dimensions were derived from theories of systemic family therapy and organizational psychology. The study was divided in three parts: In Study 1 (N = 537), a short 9-item scale with two interrelated factors was constructed on the basis of exploratory factor analysis. Quality of relationship and collective efficacy emerged as the most relevant dimensions for the quality of social systems. Study 2 (N = 558) confirmed the measurement model using confirmatory factor analysis and established validity with measures of family functioning, life satisfaction, and working team efficacy. Measurement invariance was assessed to ensure that EVOS captures the same latent construct in all social contexts. In Study 3 (N = 317), an English language adaptation was developed, which again confirmed the original measurement model. The EVOS is a theory-based, economic, reliable, and valid measure that covers important aspects of social relationships, applicable for different social systems. It is the first instrument of its kind and an important addition to existing measures of social relationships and related outcome measures in therapeutic and other counseling settings involving multiple persons.
Citation: Aguilar-Raab C, Grevenstein D, Schweitzer J (2015) Measuring Social Relationships in Different Social Systems: The Construction and Validation of the Evaluation of Social Systems (EVOS) Scale. PLoS ONE 10(7): e0133442. doi:10.1371/journal.pone.0133442
Editor: Ye Wu, Beijing University of Posts and Telecommunications, CHINA
Received: March 4, 2015; Accepted: June 26, 2015; Published: July 22, 2015
Copyright: © 2015 Aguilar-Raab et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: Funding was provided by the Heidehof Foundation (http://www.heidehof-stiftung.de/) to CAR and JS. We acknowledge the financial support of the Deutsche Forschungsgemeinschaft and Ruprecht-Karls-Universität Heidelberg within the funding programme Open Access Publishing. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Social relationships are an important predictor of health, well-being, and efficacy [1–4]. It follows that human behavior and experiences can be better understood through the interplay and interaction between individuals and contextual factors. Therefore, the evaluation of psychosocial or systemic interventions and their intended process of change raise the question of which measurements and methods should adequately be applied considering the complexity and outcome factors of multiple person therapy and counseling. Existing outcome measures that are generally applied in psychosocial intervention studies focusing on social systems have major disadvantages: Most measures such as the Family Adaptability and Cohesion Evaluation Scale—FACES I-IV , the Family Assessment Device—FAD  or the Systemic Clinical Outcome and Routine Evaluation—SCORE , are customized to specific types of systems such as couples or families. Different or unusual types of social systems such as broader family constellations, cross generational households or arranged teams for short-term trainings and interventions cannot be investigated. Comparisons between different social systems are additionally obstructed by the multitude of constructs, such as family climate, family functioning etc., that are measured. Most of the scales are long and time-consuming in application such as the Systemic Therapy Inventory of Change (STIC) , which limits a scale’s utility for both clients and counselors. Some measures focus on more or less stable system characteristics such as family culture. Even though many measures are related to multi-person settings, most only score the individual level , based on items addressing the first person singular (“I…”), rather than the perception of the system as a whole (“We…”).
Theoretical approaches for the understanding of social systems’ interactions and functioning
A social system can be defined as a group of elements, such as individual persons, who are connected to each other by virtue of their relations . Members of social systems interact and communicate in certain ways in order to maintain their dynamic balance and functional abilities [11, 12]. As a self-regulating and autopoietic system they use feedback-loops about their performances based on circular causality . Negative or positive feedback leads either to reducing or enhancing change. Repetitive interactions over time create, in turn, relatively stable structures [14, 15]. The interactive process of informational shift implies that communication always occurs [16–18]. In line with that, interaction and the quality of relational aspects contain more than verbal communication.
The joint interactional reality is ruled by subjective experiences, beliefs, and categories, which may differ strongly between members. Individuals perceive and interpret any kind of stimuli, which in turn gives rise to a certain kind of behavior towards each other [19, 20].
In addition, Bandura’s extended model of perceived collective efficacy depicts collective agency and functioning as based on the individuals’ belief in the capability of the group to function as a whole . Unsurprisingly, family efficacy was found to exhibit a substantial influence on the quality of family functioning and satisfaction . The same holds true for organizational contexts .
Systemic intervention techniques and systemic outcome ideals
Despite the heterogeneity of systemic (family) therapy interventions, they all focus on the current problem-maintaining patterns rather than on the etiopathogenetic origins of interactional problems. The target of treatment is not a problematic and symptom-laden individual, but the interactional dynamic between members of a social system. A symptom is always seen as an attempt to a solution and stimulation to further interaction at the same time. Thus, interventions shift from the question of whether any attitude or behavior may be “right” or “wrong”, “normal” or “abnormal”, to whether it has a purpose, beneficial implications, or side effects within the social system. Using techniques such as circular and reflexive questioning [23, 24], the main aims of a therapy are to increase client involvement and participation, and to foster clients’ own unique insight . Based on the guidelines of not-knowing what clients should do [26, 27] and curiosity about why they do what they do , clients are encouraged to find their own useful, unconventional, and uniquely creative way of dealing with problems. The same applies to organizational contexts, where similar systemic techniques are used. Based on this systemic framework, the current study proposes that outcomes should be measured based on the system members’ evaluation of how they get along, rather than their satisfaction with specific outcomes.
Development of a new scale
Our goal was to create an assessment tool that captures dimensions of the quality of a relationship relevant for systemic interventions in interdepended social systems. So far, no measure is applicable to different social systems, such as couple relationships, families, or working teams, and refers to a system in terms of a collective “we”–oneself as part of a certain social system. EVOS tries to fill this gap. The scale was designed to be highly economic to enable easy measurement in therapeutic and counseling settings. As an outcome measure, EVOS was theoretically created to be sensitive to change in order to be used as an evaluation tool for psychosocial or interactional interventions. As outlined above, its development is based on systemic theory and thinking. Multi-perspectivity was taken into account to reflect the systemic-constructivist idea that every person has her own valuable perspective. The aim was to develop a measure without normative presupposition of what constitutes a “good” relationship. The specific content of the items was taken from several models and theories within the field of family research and therapy, as well as from organizational psychology.
EVOS contains two subscales covering affective aspects on one hand with the quality of relationship subscale, and cognitive aspects with the collective efficacy subscale on the other hand. Items have been derived from different, established models of family therapy as well as models of organizational consulting psychology (c.f. Table 1). Additionally, the EVOS scale includes a consensus item measuring the perceived consensus about the quality of relationships and collective efficacy within the social system. Although the item is not part of the scale, it should enable additional evaluation in terms of the quality of the social systems.
Major constructs were established in three influential theories and constitute the theoretical basis of EVOS: In the Beavers Systems Model , family competence comprises the quality of organizing and managing performances. Family style emphasizes the ability of competent families to modify their style to upcoming needs. Important dimensions selected are structure of the family, mythology describing how the family functions as a group, goal-directed negotiation, and family affect etc. In the Circumplex Model of Marital and Family Systems, Olson and colleagues discussed three main dimensions: cohesion, as a feeling of togetherness towards each other, flexibility as a functional ability to change, and communication as a facilitating dimension [30, 31]. Lastly, in the McMaster Model of Family Functioning, Epstein and colleagues propose that family functioning is determined by emotional and physical health and problems of family members [32, 33]. The model contains six dimensions: problem solving, communication, roles, affective responsiveness, affective involvement and behavior control.
In the field of organizational psychology, counseling and team development procedures are of great importance in order to optimize team efficacy. Therefore, team diagnostics are applied assessing the team status to identify what needs to be operationalized and what may be improved . In a coordination network (member-task-tool relation) different component networks are established and investigated, e.g. member-member relationships, or member-task relations . The performance of interacting individuals, groups and superordinate bodies should lay the foundation for innovations. Interventions aim at enhancing effectiveness and motivation of task and work processes . The Four Factor Model of facet-specific climate for innovations has been a main focus and functions as a theoretical base for the Team Climate Inventory [37, 38]. It measures four climatic factors: participatory safety, support for innovation, vision and task orientation, which can be associated to various stages of the process of group innovations.
We will describe the development of the EVOS in several steps. Based on a pre-test, Study 1 describes the initial construction of the scale. Study 2 aims to confirm the established measurement model, assesses measurement invariance of the scale in various contexts, and demonstrates construct validity. Finally, an English language translation of the EVOS is evaluated in Study 3.
The first step in the development of the scale included the formulation of items based on the theoretical background presented in the introduction. Twenty-four experts working more than 10 years in the field of systemic research and therapy participated in the process. This resulted in a pool of 30 items, which was tested in a small sample of N = 31 participants, both in private (n = 16) as well as in organizational (n = 15) settings. Participants were asked to comment on various aspects of the questionnaire, such as comprehensibility and clarity of the items. Two different response scale formats were considered, a visual analog scale marked from 0% to 100% and a 4-point rating scale. Participants’ feedback was overall positive and indicated a preference for the 4-point rating scale format ranging from “very poor” (0) to “very good” (3). To get a first impression of test-retest reliability, participants completed the same questionnaire after an interval of three weeks. Test-retest correlations amounted to r = .66 for quality of the relationship and r = .56 for the collective efficacy subscale. This indicated a moderate stability. Following the pre-test, for each item representing one dimension at least one additional alternative item wording were generated without altering the related theoretical dimensions. This resulted in a pool of 80-items that was subsequently tested in a large sample.
Study 1: Construction of the EVOS
Participants and Procedure.
Participants for the first study were recruited online and by means of a paper-pencil-survey, at the University of Heidelberg and in different local training institutions, via local newspapers as well as private and public networks. There was a clarification of objectives of the study, voluntary participation and the possibility of cancellation at any time. Participants did not provide written formal consent as this study was in part conducted online. They were informed that by returning the questionnaires they would consent to participation in the study and their data being analyzed. The study including the procedure was approved by the ethics committee of the Heidelberg Medical Faculty (S-508/2012). Each participant was free to choose which social context (partnership, family, working team) to evaluate. A total of 546 individuals took part in the study. Nine participants were excluded from the data analysis due a large amount of missing data (>10%), resulting in N = 537 participants (41% pen-paper-version, 59% online; n = 402 (74.9%) female and n = 131 (24.6%) male). The average age of the participants was 32 (M = 32.00, SD = 13.04). The majority of the participants, n = 253 (47.5%), were school and university students or apprentices, n = 244 (45.8%) were employed persons and n = 36 (6.7%) stated to be unemployed or “other”. Concerning the possible target contexts of the EVOS, n = 211 (39.3%) evaluated their couple relationship, n = 177 (33.0%) their family relationship (ranging from two to ten family members), and n = 149 (27.7%) rated their working teams (ranging from two to sixty team members). A full information maximum likelihood estimator was used to handle missing data during parameter estimation (a total of 0.4% of all cells).
SPSS 21  was used for descriptive and exploratory factor analyses (EFA), and Mplus 7.11  was used for confirmatory factor analyses (CFA). For all CFAs, a maximum likelihood estimator with robust standard errors (MLR) was used. Mplus provides MLR for maximum likelihood with robust ‘Huber-White’ standard errors and a scaled test statistic asymptotically equivalent to the Yuan–Bentler T2* statistic [41, 42]. Robust procedures were required due to the 4-point rating scale and because Small’s omnibus test indicated that multivariate normality did not hold in our sample, χ2 (22) = 200.74, p < .0001. Model fit was evaluated by 1. the—ideally non-significant—χ2 test  and as low as possible a χ2/df ratio, ideally as low as 2 ; 2. the comparative fit index (CFI) with values of .90/.95 and above indicating appropriate/good model fit [45, 46]; 3. the root mean square error of approximation (RMSEA) with values of .05/.08 and less indicating good/reasonable model fit ; and 4. the standardized root mean square residual (SRMR) with values less than .08 considered to reflect good fit . For comparisons of nested models later on, the Bayesian Information Criterion (BIC) was used with smaller values indicating better model fit .
Besides computing item difficulties, corrected item-to-total correlations and internal consistencies for the two sub-scales and the total scale, we conducted an exploratory factor analysis (EFA; principal axis factoring) as shown in Table 1.
In order to select items from the large 80 items pool and to extract the minimum amount of factors we used Promax rotation to allow for correlated factors. To test the sufficiency of the sample and data quality , we calculated the Kaiser-Meyer-Olkin criterion (.90) as well as Bartlett’s test of sphericity, χ2 (36) = 1956,154, p < .001. Both tests indicated sample adequacy and the necessary substantial correlations to conduct a factor analysis. After a successive process, each time deleting items with loadings lower than .40, lastly the Kaiser-Criterion (eigenvalue EV > 1) resulted in a two-factor solution with nine items in line with the theoretical framework: The first factor accounted for 50.5% of the variance and the second factor for 11.3%. Inter-item correlations can be seen in Table 2.
As expected, the difficulty indices were mediocre, but generally >.50. All items were "easy" to rate, so that both the relationship quality and effectiveness were mainly judged as positive–as would be expected in a non-clinical sample. The corrected item-to-total correlations with no item lower than .50 were consistently high, ensuring a high level of representativeness of the subscales. The factor analysis yielded two factors (subscales) on which the items loaded at least .40 with cross-loadings lower than .25. The internal consistencies can be judged as moderate, which is due to the short scale-length and weak to moderate inter-item-correlations (as an indicator for homogeneity).
To evaluate the accuracy of the two-factor structure of the 9 items obtained by EFA (the consensus item is not an integral part of the scale), a CFA was conducted separately for each social system (couple, family and working team) as well as overall across all social contexts. This was done for the reader’s convenience only and will be reevaluated in Study 2. Model fit was acceptable for all CFAs, while the scale showed best fit in a working team context and worst fit in the couple context Table 1. Additionally, the CFA enabled us to overcome the Alpha coefficients inadequacy to estimate reliability of a scale that lacks true tau-equivalence and strict uni-dimensionality. SEM based estimates can be used instead [50, 51]. A measure is considered to be reliable if at least half the variance can be extracted by the construct, AVE (average variance explained) > .50, and if the construct reliability (CR or Ωw) passes a threshold of .60. The EVOS met both criteria.
Study 2: Validation
With Study 2, we aimed to replicate the previously established measurement model for the EVOS. Additionally, we investigated measurement invariance between different contexts to demonstrate that EVOS assesses the same construct of social relationship, even in noticeably dissimilar situations. In a last step, we investigated correlations with other relevant measures to validate the EVOS.
Participants and Procedure.
Sample 2 included N = 564 participants, of whom six were excluded from the data analysis due to >10% of their data missing. This resulted in N = 558 participants (45.5% pen-and-paper-version, 54.5% online; n = 426 females, n = 132 males). Again, participants where free to choose which social context to evaluate, after being informed about the objectives of the study, the voluntariness and cancellation option. Participants were informed that by returning the questionnaires they would consent to participation in the study and their data being analyzed. Participants of this study did not provide written formal consent as this study was in part conducted online. The study including the procedure was approved by the ethics committee of the Heidelberg Medical Faculty (S-508/2012). Only participants in the pen-paper condition received additional questionnaires due to copyright reasons. The average age was 33 (M = 33.00; SD = 12.31). Sample composition was comparable to Study 1 with the majority of the participants being students (n = 266; 40.6%), followed by employed persons (n = 142; 25.5%) and unemployed or “else” (n = 189; 33.9%). Evaluation targets were balanced, with n = 182 (32.6%) participants evaluating their couple relationship, n = 188 (33.7%) evaluating their families and the same number of n = 188 rating their working teams. A total of 1.14% of all data were missing at random (MCAR) and subsequently handled using a full information maximum likelihood estimator. Again, robust procedures were required for the CFA, because Small’s omnibus test indicated that multivariate normality did not, χ2 (22) = 210.06, p < .001.
Measurement invariance (MI) refers to the measurement model being invariant across groups, settings, or times of measurement [52, 53]. In the case of EVOS, one crucial aspect is the applicability of the scale in different contexts. Testing for MI can provide a statistical assessment of EVOS’ ability to capture social relationships across a broad range of contexts. Testing for MI involves a series of tests using consecutive multi-group CFAs (MGCFA) with increasingly restrictive nested models. Model parameters are set equal across groups and model fit is evaluated to examine whether a more restrictive model can still similarly represent the empirical data. If invariance holds at a given level, fixing the relevant parameters to be equal across groups should not harm model fit.
Testing for MI includes several hierarchically ordered levels. At the first level of configurable invariance, a similar but not necessarily the same construct is measured in both groups. If the model holds, all items are associated with the same latent variable(s), though the factor loadings can differ across groups. A failure to replicate the two-factor model of the EVOS would indicate a severe difference across contexts. This step differs from prior analyses by using MGCFA, rather than single group CFA. At the level of metric invariance the factor loadings are set equal across groups. This demonstrates that respondents attribute the same meaning to the items. Therefore, the underlying latent construct is assumed to be identical. If metric invariance holds, relationships to other measures, i.e. correlations, can be meaningfully interpreted. This is a crucial test, as it could very well be possible that respondents attribute different meaning to the items in different contexts. For example, the item “For me, the way we talk with each other, is …” could be construed differently in a working team context (i.e., being polite and friendly) than in a couple relationship context (i.e., telling your spouse that you love him/her). The next level of scalar invariance means that items are calibrated equally in different contexts. Statistically, item intercepts are held equal across groups. Only then can we assume that scores not only have the same unit of measurement, but also the same origin in the regression equations underlying the MGCFA. If scalar invariance holds, different social contexts can be compared regarding their scores on the latent variable. Lack of scalar invariance could be due to some items being easier or more difficult to endorse in one context than another. With regard to the theoretical conception of the EVOS, it is highly unlikely that this level of invariance could hold. Based on this, any further steps in testing for MI are not expected to succeed, so we will not go into any more details.
When testing for MI, subsequent models can at each step be compared using an ideally non-significant χ² difference test. This test, however, tends to be too strict as sample size increases, so model fit indices can be used alternatively to assess model fit. Conventionally, a CFI drop is accepted if smaller than .01 and a RMSEA increase would still be acceptable if smaller than .015 [54, 55]. Additionally, BIC indicates at each step whether the model remains as accurate as before, representing a tradeoff between accuracy and parsimony.
Even though EVOS is applicable to different contexts, validation measures are more often than not specific. Thus, participants in the pen-paper-condition were presented different measures depending on the context they chose to evaluate.
Satisfaction with life scale (FLZ): The FLZ (“Fragebogen zur Lebenszufriedenheit”; ) is a commonly used German language measure of satisfaction with several aspects of life. The measure includes ten sub-scales with seven items each referring to 1. Health, 2. Work and job, 3. Financial situation, 4. Leisure, 5. Marital and couple relationship, 6. Relationship to one’s own children, 7. Own person, 8. Sexuality, 9. Friends and relatives, and 10. Living situation. Answers are given on 7-point scales marked from 1 (very dissatisfied) to 7 (very satisfied). Cronbach’s Alpha for the subscales ranged from .73 to .89 (FLZ_tot α = .95) in our sample. Sum scores can be computed for individual scales. Additionally, a sum score for global life satisfaction includes the subscales health, financial situation, leisure, own person, sexuality, friends and relatives, and living situation. To validate the EVOS, only those subscales which were applicable in a context were used.
Family scales (FB): The family scales ("Familienbögen"  are a self-evaluation measure to assess family functioning and resources. It is based on a general process model of family functioning  and includes seven sub-dimensions: task fulfillment, role behavior, communication, emotionality, affective responsiveness, control, and values/rules. Different versions of the FB are tailored towards specific contexts. The FB-A (or FB-fam) targets whole family systems, whereas the FB-Z (or FB-two) focuses on dyadic relationships.
The FB-two consists of 28 items in seven sub-scales. The FB-fam goes beyond the FB-two by additionally including 12 more items measuring social desirability and defensive mechanisms. Answers are given on four-point scales marked from 1 (completely true) to 4 (not true at all). All 28 items of the seven core scales can be summed up to global score of functioning. Cronbach’s Alpha amounted to α = .87 for FB-two and α = .74 for FB-fam. The FB scales have been shown to be responsive to therapeutic interventions  and resemble a global assessment of family relationships, as it is covered by the EVOS.
Work in a team scale (FAT): The work in a team scale ("Fragebogen zur Arbeit im Team"; ) is based on two models of team development, the SGRPI-Modell  and the theory of team reflexivity [62, 63]. The FAT includes 24 double-items. Answers are given on 6-point scales with one end marked with a positive sentence (e.g., “goals in our team are clear to us”) and the other end marked with a corresponding negative statement (e.g., “goals in our team are not clear to us”). Items are grouped in four factors: goal orientation, task achievement, coherence, and responsibility. A sum score can be computed over all items with lower scores proposing to indicate a need for a team development intervention. In our sample Cronbach’s Alpha reached α = .91 for the scale sum score.
Confirming the measurement model for the EVOS.
Inter-item correlations Table 2 were slightly higher than in the first study, varying in a medium range. The measurement model that was established previously was now subject to confirmation. Table 3 shows the results for the psychometric evaluation of the EVOS. Similar to Study 1, the item difficulty coefficients can be judged as mild, still all >.50, indicating a positive evaluation of the respective social contexts, overall. The high corrected item-to-total correlations (all >.50) demonstrated good representativeness for the subscales. Internal consistencies were satisfactory to very good. A model with two interrelated factors fitted the empirical data well across all contexts. Again, model fit was slightly better in the working team context than in the significant other context.
Results of the MI tests can be seen in Table 4. In agreement with our expectations, metric invariance holds across all contexts. Further restrictions, required by higher levels of invariance, severely diminished the model fit and could not be accepted. Our results confirm the structural validity of the EVOS. The previously developed measurement model could be replicated. Tests of measurement invariance indicated that EVOS indeed measures the same latent construct in all contexts.
Descriptives as well as correlations between the EVOS and the validation measures can be seen in Table 5. In the couple context, EVOS showed the expected high and positive correlations with scales of life satisfaction, especially with the subscale of marriage/partnership, whereas lower correlations with the subscale sexuality are plausible due to limited content overlap. Based on a negative or problem-oriented approach within the family scales–dyadic relationships version–substantial negative correlations have been identified, confirming our hypothesis. For the evaluation of the family context similar moderate to very strong results indicated high construct validity. In the working team context slightly lower correlations were found. Still, all correlations pointed in the expected direction. Overall, the highest construct overlap could be found for private social systems, especially with regard to the total score of the life satisfaction scale, which does not include all subscales of the FLZ. To summarize, EVOS seems to measure different aspects of social relationships, but still showed an adequate overlap with existing measures. Thus, EVOS demonstrated convergent, as well as discriminant validity.
Study 3: The English Version of the EVOS
Participants and Procedure.
The EVOS was translated into English and back translated with the support of two native speakers. Afterwards, the recruitment took place in England in three centers of KIDS Company in London (http://www.kidsco.org.uk/), an institution that works with children and adolescents applying systemic interventions . All participants were informed about the study objectives, the voluntariness and the possibility to drop out at any time. The recruitment took place in a staff member meeting. After formal written consent N = 330 employees of KIDS Company participated. Data from n = 13 were excluded as a result of more than 10% of their data missing. The validation study included n = 317 participants (239 female, 75.4%; 77 male, 24.3%; one undetermined, 0.3%) with a mean age of 36.07 years (SD = 10.56). Participants were free to choose whether they wanted to evaluate their couple relationship (n = 90), family (n = 112), or working team (n = 115). A total of 1.2% of all data were missing (at random) and subsequently handled using full information ML in the CFA.
Descriptive Data Analysis.
Comparable to the German participants in the earlier studies, the UK participants also rated their couple relationships more favorably than their family relationships and working team relationships, respectively, with regard to the EVOS means, F(2, 314) = 8.83, p < .001, η2 = .05, Ms = 2.31, 2.06, 1.98, SDs = 0.06, 0.05, 0.05. Both subscales of the EVOS reflected this tendency; quality of the relationship: Ms = 2.47, 2.12, 2.12, SDs = 0.06, 0.06, 0.06; collective efficacy: Ms = 2.19, 1.95, 1.88, SDs = 0.07, 0.06, 0.06. This also implies that participants’ ratings, across all contexts, were higher for quality of the relationship than for collective efficacy, t = 10.48, df = 316, p < .001, Cohen’s d = 1.18. Cronbach’s Alpha for the scale amounted to α = .93. Item to total correlations ranged from ritc = .70 to .78.
Confirmatory Factor Analysis.
We aimed to replicate the initially presented two-factor model for the English language translation of the EVOS. The model fitted the data well, χ2(26) = 54.36, χ2/df = 2.09, p < .001, RMSEA = .063, CFI = .977, SRMR = .030. Fig 1 shows standardized loadings for this model. The construct reliability yielded Ωw = .95 while the AVE amounted to .66. To summarize, the English translation of the EVOS was successfully tested. The measurement model that was developed for the original German version also held for the English adaptation.
The aim of the present research was to construct and validate a short scale to measure quality of relationships and the collective efficacy in different social systems as systemic interventions are similarly applied in different social contexts. Theories of functional social systems with regard to the context of couple/family therapy and organizational psychology, as well as systemic interventional guidelines provided the theoretical framework based on systemic, non-normative and constructivist approaches and referring to important theories of family functioning, family style as well as organizational theories of working teams. Study 1 presented the initial construction of the scale, resulting in a 9 plus 1 item two-factor solution. Study 2 confirmed the good psychometric properties of the EVOS and the established measurement model. Moreover, we were able to confirm metric measurement invariance across different social contexts, confirming the applicability in many different contexts. Construct validity of the scale was demonstrated by correlations with measures of life satisfaction, family functioning, and team assessment. In Study 3 we validated an English translation of the EVOS, which also demonstrated good psychometric qualities and confirmed the original measurement model.
With EVOS, we present a short, economic, reliable, and valid measure answering a need for a measure taking into account affective and cognitive aspects of social systems relationships. Despite high correlations with corresponding measures that assess specific aspects of functioning or pathology, EVOS covers theoretically and directly the core concept of change behind systemic interventions, focusing on relational and dynamic aspects that are important in different social systems.
Within the theoretical framework of systemic counseling, the diagnosed ‘problem’ cannot be understood as an objectified ‘true psychopathology’. The primary concerns of systemic interventions are to address the requests of all members of a social system, to ensure task clarification, and to critically question the functionality of symptoms, in order to contextualize them within the interactional, reciprocal framework. This should ultimately initiate a change with no predictable linear outcome. The evaluation of such interventions becomes even more difficult when the intended goal is highly individual or when the fluctuation of the ‘problem definition’ is taken into account. For example, the problem of a non-eating child at the beginning of therapy might be replaced by the problem of parents not talking to each other anymore during the progress of the therapy process. A measure, which directly addresses crucial aspects of social relationships, could reveal such structures important for the counseling process. Still, the changeable nature of initial problems contains several problems in evaluating interventions that can hardly be overcome by quantitative measures.
Directions for Future Research and Limitations
Our long-term goal was to develop a scale capable of measuring the change following interventions addressing relational aspects. As we conducted primarily cross-sectional studies, we are unable to evaluate EVOS ability to measure change. This highly important issue will be dealt with in future longitudinal studies. Moreover, in almost all of the studies presented here, we got an unintended unbalanced composition of the samples with respect to gender. Future studies will need to focus on verification of invariance across genders in order to take into account that relationship perception and satisfaction might vary across genders [65–67]. Further on, all participants answered the measures with regard to only one social context due to practical reasons. Hence, the subsamples evaluating different contexts (partnership, family, working team) include different individuals and therefore might differ from each other.
Unsurprisingly, the sole use of self-report measures has essential limitations . In order to broaden the perspective of the evaluation of one particular social system and to investigate the fit between different perspectives referring to second order cybernetics, a parallel external rating version for counselors and therapists (EVOS-E) was also created, but is still subject to evaluation. Furthermore, it is yet to be established whether EVOS can be applied to children between the age of 12 and 18, as they are important family members giving genuine insights of a family’s system quality. Moreover, the application in a clinical sample is necessary to identify its potential to detect the correspondence between pathology and negative evaluation of the quality of social relationships.
Simple questionnaires can never fully meet the requirements concerning the complexity of social systems. Nonetheless, they can still form the basis for many complex methodologies such as time series panel analysis  or multilevel growth modeling . To get a more comprehensive view of social systems and their characteristic hierarchical data sets, we need to investigate different approaches that go beyond sum scores, means, and difference values . The correlations between members of a social system can be of special insight for the intervention process. Therefore, the EVOS scale includes one additional item assessing the perceived consensus in a system. The differences between actual and expected consensus of the evaluation by each member could potentially illuminate essential aspects of a social system’s dysfunctional or problematic relationship patterns [72–74]. This might provide valuable new insights for therapy processes and outcome research.
EVOS is a reliable, valid and economic tool measuring the perceived quality of relationship and collective efficacy in personal and organizational social systems. We see it as an important addition to fill the gap in current diagnostic research particularly with regard to systemic and other relational psychosocial interventions.
S1 File. EVOS Scale English Version.
S2 File. EVOS Scale German Version.
The authors thank Lisa Mühlhan, Christina Hunger, Julian Geigges and all lab members for help with the data collection.
Conceived and designed the experiments: CAR DG JS. Performed the experiments: CAR DG JS. Analyzed the data: CAR DG JS. Contributed reagents/materials/analysis tools: CAR DG JS. Wrote the paper: CAR DG JS.
- 1. Bandura A. Exercise of human agency through collective efficacy. Cur Dir Psychol Sci. 2000;9(3):75–8.
- 2. Cohen GL, Sherman DK. The psychology of change: Self-affirmation and social psychological intervention. Annu Rev Psychol. 2014;65(1):333–71.
- 3. Miller G, Chen E, Cole SW. Health Psychology: Developing biologically plausible models linking the social world and physical health. Annu Rev Psychol. 2009;60(1):501–24.
- 4. Rusbult CE, Van Lange PAM. Interdependence, interaction, and relationships. Annu Rev Psychol. 2003;54(1):351–75.
- 5. Olson DH. Circumplex Model VII: Validation studies and FACES III. Fam Process. 1986;25(3):337–51. pmid:3758310
- 6. Epstein NB, Baldwin L.M. & Bishop D.S. The McMaster family assessment device. J Marital Fam Ther. 1983;9(2):171–80.
- 7. Stratton P, Bland J, Janes E, Lask J. Developing an indicator of family function and a practicable outcome measure for systemic family and couple therapy: the SCORE. J Fam Ther. 2010;32(3):232–58.
- 8. Pinsof WM, Zinbarg RE, Lebow JL, Knobloch-Fedders LM, Durbin E, Chambers A, et al. Laying the foundation for progress research in family, couple, and individual therapy: The development and psychometric features of the initial systemic therapy inventory of change. Psychother Res. 2009;19(2):143–56. doi: 10.1080/10503300802669973. pmid:19235092
- 9. Sanderson J, Kosutic I, Garcia M, Melendez T, Donoghue J, Perumbilly S, et al. The measurement of outcome variables in couple and family therapy research. Am J Fam Ther. 2009;37(3):239–57.
- 10. Willke H. Systemtheorie. 4., überarb. Aufl. ed. Stuttgart: G. Fischer; 1993.
- 11. Bateson G. Steps to an ecology of mind. Collected essays in anthropology, psychiatry, evolution, and epistemology. San Francisco, CA: Chandler Pub. Co.; 1972.
- 12. Bateson G. Mind and nature: A necessary unity. New York, NY: Dutton; 1979.
- 13. Maturana HR, Varela FJ. The tree of knowledge: The biological roots of human understanding. Boston, NY: Shambhala; 1992.
- 14. Haken H. Synergetics: Introduction and advanced topics. Berlin: Springer; 2004.
- 15. Tschacher W, Schiepek G, Brunner EJ. Self-organization and clinical psychology: Empirical approaches to synergetics in psychology. Berlin: Springer; 1992.
- 16. De Mol J, Buysse A, Cook WL. A family assessment based on the social relations model. J Fam Ther. 2010;32(3):259–79.
- 17. Watzlawick P, Beavin J. Some formal aspects of communication. Am Behav Sci. 1967;10(8):4–8.
- 18. Watzlawick P, Bavelas JB, Jackson DD. Pragmatics of human communication: A study of interactional patterns, pathologies, and paradoxes. New York: Norton; 1967.
- 19. von Foerster H. Understanding understanding: Essays on cybernetics and cognition. New York: Springer; 2003.
- 20. Watzlawick P. The Invented reality: How do we know what we believe we know? Contributions to constructivism. New York, NY: Norton; 1984.
- 21. Bandura A, Caprara GV, Barbaranelli C, Regalia C, Scabini E. Impact of family efficacy beliefs on quality of family functioning and satisfaction with family life. Applied Psychology. 2011;60(3):421–48.
- 22. Goncalo JA, Polman E, Maslach C. Can confidence come too soon? Collective efficacy, conflict and group performance over time. Organ Behav Hum Decis Process. 2010;113(1):13–24.
- 23. Simon FB, Rech-Simon C. Zirkuläres Fragen: Systemische Therapie in Fallbeispielen. Ein Lehrbuch. 10., unveränd. Aufl. ed. Heidelberg: Carl-Auer; 2013.
- 24. von Schlippe A, Schweitzer J. Lehrbuch der systemischen Therapie und Beratung I: Das Grundlagenwissen. Göttingen: Vandenhoeck & Ruprecht; 2012.
- 25. Sprenkle DH. Systemic assessment. In: Cierpka M, Sprenkle DH, Thomas V, editors. Family assessment: Integrating multiple perspectives. Cambridge, MA: Hogrefe; 2005. p. 211–30.
- 26. Anderson H, Goolishian H. The client is the expert: A not-knowing approach to therapy. In: McNamee S, Gergen KJ, editors. Therapy as social construction. Inquiries in social construction. Thousand Oaks, CA: Sage; 1992. p. 25–39.
- 27. Goolishian HA, Anderson H. Strategy and intervention versus nonintervention: A matter of theory? J Marital Fam Ther. 1992;18(1):5–15.
- 28. Cecchin G, Lane G, Ray WA. Influence, effect, and emerging systems. J Syst Ther. 1994;13(4):13–21.
- 29. Beavers WR. A systems model of family for family therapy. J Marital Fam Ther. 1981;7(3):299–307.
- 30. Olson DH, Portner J, Lavee Y. FACES III—Family Adaptability and Cohesion Evaluation Scales. St. Paul, MN: University of Minnesota; 1985.
- 31. Olson DH, Sprenkle DH, Russell CS. Circumplex model of marital and family systems: I. Cohesion and adaptability dimensions, family types and clinical applications. Fam Process. 1979;18(1):3–28. pmid:437067
- 32. Epstein NB, Bishop DS, Levin S. The McMaster model of family functioning. J Marital Fam Ther. 1978;4:19–31.
- 33. Epstein NB, Bishop D, Ryan C, Miller IW, Keitner GI. The McMaster model view of healthy family functioning. In: Walsh F, editor. Normal family processes New York: Guilford Press; 1993.
- 34. Kauffeld S. Teamdiagnose. Göttingen: Hogrefe; 2001.
- 35. McGrath JE, Arrow H, Berdahl JL. The study of groups: Past, present, and future. Pers Soc Psychol Rev. 2000;4(1):95–105.
- 36. Faulstich P. Strategien der betrieblichen Weiterbildung: Kompetenz und Organisation. München: Vahlen; 1998.
- 37. Agrell A, Gustafson R. The Team Climate Inventory (TCI) and group innovation: A psychometric test on a Swedish sample of work groups. J Occup Organ Psychol. 1994;67(2):143–51.
- 38. Anderson N, West MA. The Team Climate Inventory: Development of the TCI and its applications in teambuilding for innovativeness. Eur J Work Organ Psy. 1996;5(1):53–66.
- 39. IBM. IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp 2012.
- 40. Muthén L, Muthén B. Mplus User's Guide. Los Angeles, CA: Muthén & Muthén; 1998–2012.
- 41. Yuan KH, Bentler PM. Normal theory based test statistics in structural equation modelling. Br J Math Stat Psychol. 1998;51(2):289–309.
- 42. Yuan KH, Bentler PM. Three likelihood-based methods for mean and covariance structure analysis with nonnormal missing data. Sociol Methodol. 2000;30(1):165–200.
- 43. Bentler PM, Bonett DG. Significance tests and goodness of fit in the analysis of covariance structures. Psychol Bull. 1980;88(3):588.
- 44. Tabachnick BG, Fidell LS. Using Multivariate Statistics: Pearson; 2007.
- 45. Bentler PM. Comparative fit indexes in structural models. Psychol Bull. 1990;107(2):238–46. pmid:2320703
- 46. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equ Modeling. 1999;6(1):1–55.
- 47. Browne MW, Cudeck R. Alternative ways of assessing model fit. In: Bollen KA, Long JS, editors. Testing Structural Equation Models. Beverly Hills, CA: Sage; 1993.
- 48. Schwarz G. Estimating the dimension of a model. The annals of statistics. 1978:461–4.
- 49. Gorsuch RL. Factor analysis. Hillsdale, NJ: L. Erlbaum; 1983.
- 50. Bacon DR, Sauer PL, Young M. Composite reliability in structural equations modeling. Educ Psychol Meas. 1995;55(3):394–406.
- 51. Fornell C, Larcker DF. Evaluating structural equation models with unobservable variables and measurement error. J Mark Res. 1981;18(1):39–50.
- 52. Meredith W. Measurement invariance, factor analysis and factorial invariance. Psychometrika. 1993;58(4):525–43.
- 53. Vandenberg RJ, Lance CE. A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organ Res Meth. 2000;3(1):4–70.
- 54. Chen FF. Sensitivity of goodness of fit indexes to lack of measurement invariance. Struct Equ Modeling. 2007;14(3):464–504.
- 55. Cheung GW, Rensvold RB. Evaluating goodness-of-fit indexes for testing measurement invariance. Struct Equ Modeling. 2002;9(2):233–55.
- 56. Fahrenberg J, Myrtek M, Schumacher J, Brähler E. Fragebogen zur Lebenszufriedenheit (FLZ). Handanweisung. Göttingen: Hogrefe; 2000.
- 57. Cierpka M, Frevert G. Die Familienbögen [FB]: ein Inventar zur Einschätzung von Familienfunktionen. Göttingen: Hogrefe; 1994.
- 58. Steinhauer PD, Santa-Barbara J, Skinner H. The process model of family functioning. Can J Psychiat. 1984;29(2):77–88.
- 59. Kreppner K, Ullrich M. Ablöseprozesse in Trennungs- und Nicht-Trennungsfamilien. In: Walper S, Schwarz B, editors. Was wird aus den Kindern? Chancen und Risiken für die Entwicklung von Kindern aus Trennungs- und Stieffamilien. Weinheim: Juventa; 1999. p. 91–120.
- 60. Kauffeld S. Fragebogen zur Arbeit im Team: FAT; Manual. Göttingen: Hogrefe; 2004.
- 61. Beckhard R. Optimizing team-building efforts. J Contemp Bus. 1972;1(3):22–32.
- 62. West MA. Effective teamwork. Exeter: BPC Wheatons Ltd; 1994.
- 63. West MA. Reflexivity and work group effectiveness: A conceptual integration. In: West MA, editor. Handbook of work group psychology Chichester: Wiley; 1996.
- 64. Hunger C, Bornhäuser A, Link L, Geigges J, Voss A, Weinhold J, et al. Experience in Personal Social Systems: Development and Psychometric Properties of the German and English Experience in Personal Social Systems Questionnaire. Unpublished document.
- 65. Kenny DA, Acitelli LK. Accuracy and bias in the perception of the partner in a close relationship. J Pers Soc Psychol. 2001;80(3):439. pmid:11300577
- 66. Santtila P, Wager I, Witting K, Harlaar N, Jern P, Johansson A, et al. Discrepancies between sexual desire and sexual activity: Gender differences and associations with relationship satisfaction. J Sex Marital Ther. 2007;34(1):31–44.
- 67. Shulman S, Scharf M. Adolescent romantic behaviors and perceptions: Age- and gender-related differences, and links with family and peer relationships. J Res Adolesc. 2000;10(1):99–118.
- 68. Podsakoff PM, MacKenzie SB, Lee J-Y, Podsakoff NP. Common method biases in behavioral research: a critical review of the literature and recommended remedies. J Appl Psychol. 2003;88(5):879. pmid:14516251
- 69. Tschacher W, Ramseyer F. Modeling psychotherapy process by time-series panel analysis (TSPA). Psychother Res. 2009;19(4–5):469–81. doi: 10.1080/10503300802654496. pmid:19585371
- 70. Keiley MK, Martin NC, Liu T, Dolbin-MacNab M. Multilevel growth modeling in the context of family research. In: Sprenkle DH, Piercy FP, editors. Research methods in family therapy. New York, NY: Guilford Press; 2005. p. 405–30.
- 71. Fisher L, Kokes RF, Ransom DC, Phillips S, Rudd P. Alternative strategies for creating "relational" family data. Fam Process. 1985;24(2):213–24. pmid:4018242
- 72. Escudero V, Friedlander ML, Varela N, Abascal A. Observing the therapeutic alliance in family therapy: associations with participants' perceptions and therapeutic outcomes. J Fam Ther. 2008;30(2):194–214.
- 73. Flemons DG. Konsens—Dissens: Eine relationale Alternative zur Macht-Metapher. Zeitschrift für systemische Therapie. 1989;7(3):159–68.
- 74. Marcus DK, Holahan W. Interpersonal perception in group therapy: A social relations analysis. J Consult Clin Psychol. 1994;62(4):776. pmid:7962881