Introduction

The working alliance has been one of the most intensely studied topics in psychotherapy and is a reliable predictor of successful outcomes in therapy (Horvath et al., 2011). The alliance in couple therapy is more complicated than individual therapy as the therapist must form multiple alliances simultaneously with both members of the couple, and pre-existing relationship dynamics can influence the formation of the alliance as well as the trajectory of therapy (Friedlander et al., 2018).

Both members in a couple often report on their perceptions of the same construct. For example, each partner may provide information on their perceptions of relationship satisfaction, communication problems, or conflict. Couple researchers commonly apply dyadic data analysis models such as the actor–partner interdependence model (APIM) when analyzing data containing reports from both partners (Kenny et al., 2006). Researchers in the developmental sciences often use an overlapping but complimentary perspective called cross-informant discrepancies (De Los Reyes, 2011; De Los Reyes et al., 2019). An informant discrepancy occurs when two raters (informants) provide different reports on the same construct, such as marital satisfaction. The difference, or discrepancy, between the informants is the subject of interest. In this study we adopt a cross-informant approach to examine how discrepancies in relationship satisfaction affect the development of the working alliance at the start of couple therapy.

Literature Review

The Working Alliance

The working alliance is the collaborative relationship between the client and therapist. Bordin’s (1979) pan-theoretical tripartite model of the working alliance, consisting of the emotional bond, agreement on goals, and agreement on tasks, is the most commonly used definitions of the working alliance. Bordin’s definition captures not only the emotional bond between the therapist and client but also the congruence between therapist and client on the direction of therapy (goals) and the means of achieving those goals (tasks). The working alliance is a common factor found in nearly every therapy model in individual, couple, and family therapy (Blow et al., 2007).

Several meta-analyses conducted measuring the overall effect of the working alliance in psychotherapy found a modest, yet significant impact on the relationship of the working alliance to clinical outcomes in individual therapy (Horvath et al., 2011; Martin et al., 2000; Tryon et al., 2007) as well as couple therapy (Friedlander et al., 2018). The majority of research on the alliance has focused on alliances in individual therapy and relatively fewer articles have focused on the alliance in couple or family therapy (Friedlander et al., 2011). Alliances in couple or family therapy are uniquely challenging to develop as they involve the formation of multiple and simultaneous alliances with different family members (Friedlander et al., 2018). Additionally, couples coming into therapy have previous and ongoing relationships with one another that may impact the formation of alliances (Glebova et al., 2011). Members of the couple may also disagree on the goals or tasks of therapy, as well as how much they like or trust the therapist. These different alliances interact systemically, with a difference between one person or another influencing the alliance. Nevertheless, understanding factors that underlie the formation of the working alliance is an important endeavor as better understanding the alliance may lead to enhancing outcomes in psychotherapy.

Predictors of the Alliance in Couple Therapy

Predicting what factors promote a positive working alliance in couple therapy is difficult, with unique challenges that are not present in research on individual therapy. The difficulty in predicting a couple’s working alliance can be attributed to its increased complexity involving the added relationship dynamic of the couple (Glebova et al., 2011). Additionally, couples must work to integrate the alliance they have with a therapist into their preexisting allegiance to one another (Symonds & Horvath, 2004). A more specific factor that has been discussed as a predicting factor in the alliance is marital adjustment. Despite being considered insignificant in the past, marital adjustment has been shown to predict couples’ working alliance (Mamodhoussen et al., 2005). Mamodhoussen et al. (2005) has also indicated that while psychiatric symptoms have been shown to predict alliance differently between men and women, they do not predict overall couple alliance outcomes.

Relationship satisfaction has also been observed to impact the therapeutic alliance. Overall, lower relationship satisfaction tends to predict a lower initial therapeutic alliance (Anderson & Johnson, 2010). Symonds and Horvath (2004) suggest there is an interplay between the couple’s relationship satisfaction, and the alliance each of them develops overtime with the therapist. More specifically, one partner’s thoughts about the other’s alliance with the therapist may positively or negatively impact their current relationship satisfaction. The connection between relationship satisfaction and the therapeutic alliance becomes increasingly complicated when factoring in which session the measure is taking place, and the differences in how individuals within a couple interpret their partner’s alliance with the therapist.

Differentiation has also been shown to be a factor in predicting the therapeutic alliance. As described by Bowen (1978), differentiation refers to one’s ability to deeply reflect on their own sense of self while also being able to connect meaningfully and intimately with others. In the context of couple therapy, there has been evidence pointing towards differing levels of differentiation playing a significant role in predicting the quality of the therapeutic alliance. According to Knerr and Bartle-Haring (2010), husbands’ and wives’ ability to differentiate predicts levels of relationship satisfaction, which then often predicts alliance outcomes. Knerr et al. (2011) supports the notion of differentiation playing a role in the therapeutic alliance through observing that with higher levels of differentiation, there were higher levels of collective bond.

Gender Differences in the Alliance

When discussing factors that predict the therapeutic alliance among couples, there are often gender differences found in the measured variables. Marital adjustment, relationship satisfaction, and differentiation have all been found to have an impact on the alliance in couple therapy. Within these variables, there have been additional observations of gender differences and how they correlate with the alliance. As for marital adjustment, Mamodhoussen et al. (2005) finds that men’s marital adjustment in heterosexual couples predicts the therapeutic alliance more that his female partner’s marital adjustment. Despite not being considered significant from a couple’s alliance standpoint, women’s psychiatric symptoms were seen to be more likely to predict the alliance they have with the therapist. There are also gender differences in how relationship satisfaction affects the therapeutic alliance and develops throughout proceeding sessions. Glebova et al. (2011) found that husbands often have a greater impact on the therapeutic alliance. When considering relationship satisfaction and the perceived alliance, the husband’s perceptions of these accounted for changes with both partners in the third session. Although there is debate as to whether these differences in outcomes of the therapeutic alliance are more impacted by gender rather than pre-existing relationship satisfaction, Glebova et al. (2011) conclude there is a reciprocal relationship between perceived relationship satisfaction and therapeutic alliance, in which it is more affected by the male partner. When looking closely at the gender differences in variables that predict the therapeutic alliance among couples, one can see the added complexity couples bring in establishing and maintaining a therapeutic alliance.

Split Alliance

In contrast to individual therapy, couples may disagree on the strength of the working alliance, leading to a split in the alliance (Pinsof & Catherall, 1986), which is associated with poorer treatment outcomes (Friedlander et al., 2018). The split alliance has been typically calculated by dichotomizing couples who have a difference score of more than one standard deviation (e.g. Mamodhoussen et al., 2005). The next section will describe limitations of using discrepancy scores, but dichotomizing a continuous variable is also statistically problematic. Bartle-Haring et al. (2012) describe the limitations of dichotomizing a variable including the loss of variance. However, using a simple difference score also ignores the overall level of the alliance. In the next section we describe a relatively unused analytic strategy that can overcome both of these limitations, the latent congruence model (LCM) (Cheung, 2009).

Discrepancies in Relationship Satisfaction

It is commonly believed that members of a couple may enter treatment with differing levels of relationship satisfaction, in fact a recent meta-analysis of gender differences in relationship satisfaction found that in clinical samples women report lower levels of relationship satisfaction relative to male partners (Jackson et al., 2014). Researchers who study couples recognize the value of incorporating each partner’s perception into analysis. Although couple therapists and researchers often use dyadic data analysis, developmental science adopts a complimentary but alternative approach. Developmental scientists also obtain information from multiple sources (informants) using the cross-informant discrepancies perspective (De Los Reyes et al., 2019).

Informant discrepancies occur when two or more individuals (informants) report on the same phenomenon or construct, such as a child’s behavior problem, or the parent–child relationship. One of the earliest studies on informant discrepancies came from Achenbach et al. (1987) a meta-analysis that found only a modest correlation between multiple informants (e.g. parents, teachers) on child behavior problems. Although early research on informant discrepancies primarily focused on differences across raters on child and adolescent behavior problems, a second generation of research uses an informant discrepancy as an independent variable (Rescorla, 2016). These studies examine how differing perceptions of a phenomena (e.g. parent–child relationship) are associated with various developmental outcomes such as adolescent substance abuse (Kliewer et al., 2018) delinquency (de Los Reyes et al., 2010; Ksinan & Vazsonyi, 2016) depression (Laird & De Los Reyes, 2013) or youth anxiety treatment (Goolsby et al., 2018).

Measuring Informant Discrepancies

Early research on informant discrepancies often relied on using observed difference scores which are calculated by simply subtracting the score of one partner (e.g. husband) from the other (e.g. wife). In addition to methodological and statistical problems, Laird (2020) notes a conceptual problem in using a difference score. Difference scores fail to take into account the direction of the association and ignore the effect of the overall level (i.e. high or low) of the variable. Laird (2020) illustrates this problem using the example of discrepancies in parental warmth. A large discrepancy between parents and adolescents in reports of parental warmth may be associated with negative developmental outcomes. However, low levels of warmth reported by both parent and adolescent may also be associated with poor outcomes, but it is impossible to examine the effect of the level and the discrepancy in reports by using a simple difference score. Bartle-Haring et al. (2012) note a similar concern when examining discrepancies in the alliance. A discrepancy in the alliance (i.e. split alliance) may not be problematic for couples with an otherwise relatively high alliance.

The Latent Congruence Model

The LCM was developed in organizational research to overcome many of the challenges of using simple difference scores (Cheung, 2009). The LCM consists of two indicators, one for each dyad. The two indicators load onto two latent factors, one which measures the average (level) and the other which represents the difference (congruence) between members of the dyad. The loadings for the partners are fixed, with the loadings for the level latent variable fixed at 1, and the loadings for the congruence latent variable fixed to .5 and − .5 (see Fig. 1). Using these parameter constraints is mathematically equivalent to creating two latent variables that measure the average score and difference score between partners in a dyad (Cheung, 2009; Ledermann & Kenny, 2017). These two latent variables can then be flexibly applied as either independent, dependent or even intermediary variables (e.g. mediator or moderator) to answer substantive research questions.

Fig. 1
figure 1

LCM of relationship satisfaction predicting LCM of alliance. Covariances between variables omitted for simplicity. Fully standardized parameter estimates. RDAS revised dyadic adjustment scale. *p < .01

Full size image

To our knowledge, researchers examining couple therapy or couple processes have rarely applied this perspective to analyze couple level data instead use other dyadic designs, such as the actor partner interdependence model (APIM). We argue informant discrepancies can be used as an alternative analysis strategy to traditional dyadic analyses for research with couples or in couple therapy. The LCM has several advantages to observed difference scores and may be better suited compared to other dyadic designs for answering certain research questions (Iida et al., 2018). The LCM estimates both the level (dyadic average) and the discrepancy (difference score between the partners in a couple) as latent variables, allowing researchers to simultaneously consider the effects of both the level and difference. The LCM specifically examines how the degree of congruence (discrepancy) in reports affects an outcome, while also controlling for the effect of the overall level (average). Research questions specifically focused on the degree of similarity or difference, where there may be theoretically salient reasons to suspect divergent reports may be associated with an outcome, can be answered using the LCM. For example, the LCM may be well suited to better measure how a split alliance affects treatment outcomes. The LCM operationalizes the discrepancy as a continuous variable, and also controls for the overall level of the alliance. Couples with a high discrepancy in relationship satisfaction may also report lower overall alliance scores, which could confound the association between a split alliance and worse outcomes.

The Present Study

Although it is commonly believed that couples often enter treatment with differing perceptions of relationship satisfaction, to our knowledge no one has examined the effect of that discrepancy on the development of the working alliance. Previous research indicates relationship satisfaction at the start of treatment significantly predicts the alliance for both partners. Partners who come into therapy with different perceptions of relationship satisfaction may also have different motivation levels for working in therapy, as one partner may not view therapy as necessary or may worry about being blamed. It may also indicate one partner may be more disengaged from the relationship, less aware of their partners concerns, or avoidant of discussing their own concerns. All of these factors could impede the ability of a therapist to effectively form an alliance with both partners. Likewise, a discrepancy in relationship satisfaction may also be associated with a larger discrepancy in reports of the alliance. However, the overall level of the relationship satisfaction may influence whether discrepancies in relationship satisfaction may influence the association between the discrepancy and the formation of the alliance. Since low relationship satisfaction is also associated with a lower alliance, it is important to control for the effect of the level of relationship satisfaction. Likewise, when examining discrepancies in the alliance, it is important to also consider the overall level of the alliance.

In this study, we investigated how discrepancies in relationship satisfaction during the early phase of therapy influence the development of the working alliance in couple therapy, using the following two research questions: Does a discrepancy in relationship satisfaction predict discrepancies in alliance scores? Does a discrepancy in relationship satisfaction, predict the alliance scores for either male or female partners? For research question one we predict that discrepancies in relationship satisfaction will predict discrepancies in the alliance report, even when controlling for the level in both variables. For research question two, we predict that discrepancies in relationship satisfaction will predict both the male and female partner’s alliance. To our knowledge, this is the first study to operationalize a split alliance (discrepancies in alliance) using the LCM which controls for the level of the alliance.

Methods

Participants

Data were collected across a span of 18 months and included 22 heterosexual couples seeking outpatient mental health services from a South-Central university-based marriage and family therapy university training clinic associated with a COAMFE accredited master’s degree MFT program. Inclusion criteria included both partners consent to participation in the study. 46.3% of the couples reported being in their first marriage, 26.8% reported they were in a relationship but not married (pre-marital counseling), 17.1% reported they were cohabiting but not married, 4.9% reported the current relationship was their second marriage, and 4.9% reported they were married but currently separated. Members of the couple were predominantly Caucasian (82.9%), 9.8% American Indian, 2.4% Hispanic/Latino, 2.4% African-American, and 2.4% multiracial/multiethnic. 50% reported of the couples reported achieving a bachelor’s degree or higher, 30% of the partners reported they did not graduate high school, 15.0% had at least a high school diploma, and 5% reported some college. A majority of partners (53.7%) reported a household income ranging from < 5000 to 15,000 annually. Partners’ ages ranged from 20 to 60 years, with an average of 29.6 years.

Therapists

Therapists (n = 12) included master’s level interns part of a COAMFTE-accredited MFT master’s program. Therapists were in various stages of their training; some therapists were in their second semester of training while other therapists were completing their third year. In the study, 83% of therapists were female. All therapists received supervision from doctoral level faculty within the program. Therapists practiced from an integrative framework using a combination of models taken from the classic schools of family therapy.

Procedures

Following the standard procedures of the clinic, clients phoned the clinic to request services and completed a brief intake questionnaire over the phone. During the initial phone call to the clinic, clients were informed of the training nature of the facility. Before starting their first session several measures were administered to all clients seeking therapy as part of the standard intake packet. During the first session, clients read and signed the counseling agreement outlining the use of data collection through clinic assessments for research purposes. At the end of the second session, clients completed the Working-Alliance Inventory short form revised (WAI-SR; Hatcher & Gillaspy, 2006) in the clinic waiting room. The working alliance is most commonly measured early in treatment, typically within the first three sessions. Previous research has found the alliance is relatively stable after the second session (and does not significantly change during the early stages of therapy Glebova et al., 2011; Knobloch-Fedders et al., 2007). The WAI-SR was administered by a trained administrative assistant (non-clinician) to ensure clients’ answers are not influenced by the presence of their therapists. At the end of the second session, the therapist working on the case was instructed to fill out the therapist version of the alliance for each client, as well as a rating of what stage of change the therapist believes the client is in. Although the therapist version was used, the results of it were not used in this study. Only couples who attended a minimum of two sessions and completed WAI-SR were included in the study.

Measures

The Working Alliance

The Working Alliance Inventory Short form revised (WAI-SR) (Hatcher & Gillaspy, 2006) is a 12 item self-report measure. The WAI-SR is based on Bordin’s (1979) tripartite model of the alliance and assesses (a) the emotional bond between client and therapist (b) the agreement on goals in treatment and (c) agreement on the relevant tasks for achieving those goals. Items on the measure are rated on a Likert scale from strongly disagree (1) to strongly agree (5). Higher scores indicate a stronger alliance. A global scale can be calculated by summing the 12 items. The WAI-SR correlated highly with the original Working Alliance Inventory, and several other measures of the alliance. In this study, the reliability was high with an overall α = .91. The working alliance was measured at the end of the second session.

Revised Dyadic Adjustment Scale

We used the Revised Dyadic Adjustment Scale (RDAS) to measure relationship satisfaction (Busby et al. 1995). The RDAS consists of 14 items that measure three constructs that measure consensus, cohesion and satisfaction, with a reported alpha of α = .90 for the summed total score. The RDAS was given to both partners prior to the first session of therapy.

Analytic Plan

To test our hypotheses about the association between discrepancies in reports of dyadic satisfaction between members of a couple and the working alliance, we used the LCM (Cheung, 2009; Ledermann & Kenny, 2017). All substantive analyses were run using Mplus version 8.1 (Muthen & Muthen, 2018) with maximum likelihood estimation. We ran two sets of analyses: In the first analysis, we fit two LCMs simultaneously, one for relationship satisfaction and one for the alliance. We tested whether the level and congruence of relationship satisfaction predicted the level and congruence of the working alliance. In the second analysis, we fit a LCM for relationship satisfaction to predict the individual alliance scores of husbands and wives.

Results

Descriptive Statistics

Table 1 displays the means, standard deviations and correlations among the study variables (male and female partners’ scores on the working alliance and relationship satisfaction). The correlation for the male and female reports of relationship satisfaction (r = .83 p < .01) and the alliance (r = .61 p < .01) were both large and statistically significant. However, no other correlations were statistically significant. The means for both the husbands’ (m = 46.7) and wifes’ (m = 46.6) relationship satisfaction were slightly below the cut-off of 47.31 proposed by Anderson et al. (2014), indicating relatively low levels of relational distress. Paired sample t-tests indicated couples did not report a significant difference in the alliance t = − .68 (21), p = .50 or relationship satisfaction t = − .09 (21), p = .93.

Table 1 Descriptive statistics for study variables
Full size table

Size of the Discrepancy in Relationship to Satisfaction and Gender

To better understand this discrepancy in relationship satisfaction, we examined the size of the discrepancy, the range of discrepant scores and how frequently the male or female partner reported higher relationship satisfaction. We examined the mean of the discrepancy latent variable which represents the average difference in husband-and-wife reports of relationship satisfaction. In this sample that value was .12. This indicates on average husband’s relationship satisfaction is .12 points higher than their wife’s report of relationship satisfaction. The absolute value of the range of discrepancy scores for relationship satisfaction varied from 0 to 12 points. Our sample was fairly evenly split on whether the husband or wife reported higher relationship satisfaction. The wife reported higher satisfaction in 11 (50%) of the twenty two cases. In the remaining cases, the husband reporting higher satisfaction in 10 (45.45%) of the cases, and in one case (4.54%) there was no discrepancy.

Does a Discrepancy in Relationship Satisfaction Predict a Discrepancy in the Working Alliance?

We fit two LCMs, one for relationship satisfaction and one for the working alliance. We regressed the level and discrepancy factors for relationship satisfaction onto the level and discrepancy factors of the working alliance. The LCM with manifest indicators is a saturated model (Cheung, 2009; Iida et al., 2018). A saturated model has zero degrees of freedom as the number of parameters estimated equals the number of variances and covariances. As such, the model fits the data perfectly and no fit indices are reported. Figure 1 displays a simplified path diagram for this model. The discrepancy in relationship satisfaction predicted the level of the alliance score with β = − 0.43, p = .02, but not the discrepancy in alliance reports β = − 0.05, p = .80. This indicates that greater discrepancies in relationship satisfaction predict a lower alliance at the dyadic level. The level of relationship satisfaction did not predict either the alliance level, or the discrepancy in the alliance.

Does the Discrepancy Relationship Satisfaction Predict the Individual Partners’ Working Alliance?

We fit a LCM for relationship satisfaction and then regressed both the congruence and level variables onto the individual alliance scores. The path diagram is displayed in Fig. 2. The path from the discrepancy latent variable to the husband’s alliance was the only significant path with a β = − .46, p < .01. This indicates that larger discrepancies in relationship satisfaction are associated with a smaller alliance report for husbands.

Fig. 2
figure 2

LCM of relationship satisfaction predicting male and female alliance. Covariances between variables omitted for simplicity. Fully standardized parameter estimates. RDAS revised dyadic adjustment scale. *p < .01

Full size image

To better understand this discrepancy in relationship satisfaction, we examined the mean for the relationship discrepancy latent variable. This value represents average difference in husband-and-wife reports of relationship satisfaction, in this sample the average discrepancy was .12. This indicates on average husband’s relationship satisfaction is .12 points higher than their wife’s report of relationship satisfaction.

Discussion

In this study we examined how a discrepancy in relationship satisfaction at the start of therapy impacts the early therapeutic alliance in couple therapy. We answered two research questions (1) Does a discrepancy in relationship satisfaction predict discrepancies in the alliance? And (2) Does a discrepancy in relationship satisfaction predict the individual alliance reports of male and female partners in a couple? We adopted the cross-informant discrepancy perspective (De Los Reyes et al., 2019) and used the LCM to answer these questions.

This study contributes to the literature in the several ways. We described a complementary and alternative approach to analyzing dyadic data, the cross-informant discrepancy. We demonstrated the utility and applicability of analyzing dyadic data using the LCM and compare it to a commonly used dyadic analysis, the APIM. We provided insight into the complexities of the therapeutic alliance in working with couples, providing potential support for the clinical intuition that discrepancies in relationship satisfaction affect the therapeutic alliance. In doing so, we added support to the broader literature on what factors predict the development of the working alliance in couple therapy.

Summary and Interpretation of Findings

We found differences in reports of relationship satisfaction in a couple predict a lower dyadic level (couple average) of the alliance, but not a greater discrepancy. This indicates that the larger the difference in reports of relationship satisfaction, the lower the overall level of the alliance for the couple. The standardized estimate for this effect was β = − .42 which is a medium effect size. This indicates that for every one-point increase in the difference between a couple’s rating on part of the measure of relationship satisfaction, that is an increase in their disagreement in relationship satisfaction, the overall average level of the alliance decreases by almost half a standard deviation. This did not provide support for our initial hypothesis that discrepancies in relationship satisfaction would predict discrepancies in the alliance, however the second analysis clarified this finding.

A lower alliance level could result from either the male, female, or both partners reporting lower alliances. The second analysis found that the discrepancy in relationship satisfaction predicted a lower alliance for male partners, but not female partners. This provided partial support for our second hypothesis that discrepancies would predict both partners’ alliances. This finding emerged even in a sample of relatively low-distress couples, who did not report having significantly different perceptions of their relationship satisfaction. In this sample, on average husband’s reported a higher relationship satisfaction of .12 points. This may provide evidence that discrepant reports of relationship satisfaction are clinically relevant, even when differences in relationship satisfaction are not especially large.

In the broader literature on the alliance in couple therapy, the present findings offer a unique perspective on factors contributing to the alliance, add empirical support for clinical intuition on attending to discrepancies in couple satisfaction, and offers a unique application of a statistical model to examining effects of discrepancies in dyadic relationships. These findings affirm the complexities in working with couples in therapy (Glebova et al., 2011) by displaying the association between couple relationship satisfaction and the alliance for therapeutic work. Furthermore, it has been previously demonstrated that relationship satisfaction has implications for the alliance (Anderson & Johnson, 2010). The present findings support this finding, and further add that not only the relationship satisfaction, but a schism in a couple’s experience of satisfaction predicts a lower report on the working alliance.

Cross-Informant Discrepancies and Dyadic Data Analysis

In this study we presented an application of a relatively novel analytic model, the LCM (Cheung, 2009). We adopted a perspective often applied in developmental sciences, the cross-informant discrepancy (De Los Reyes et al., 2019) that has not been used as frequently in research with couples. There is substantial overlap between dyadic analyses, including the APIM, and research that adopts the cross-informant discrepancy perspective. Iida et al. (2018) further describes the similarities and differences between the APIM and LCM, noting that since both models are saturated (just-identified) they are statistically equivalent. In fact, estimating the APIM using pooled regression (Kenny et al., 2006; Tambling et al., 2011) is quite similar to the LCM, with both models examining both within-dyadic processes (a difference score) and between dyad processes (a dyadic average). Although the APIM and LCM are statistically similar, they are conceptually distinct with differing emphases and research implications. The LCM provides a different analytic strategy for handling dyadic data, that can be adopted by future researchers to further examine how discrepant reports influence couple therapy.

Gender and the Alliance

Although research indicates that therapist gender does not influence treatment outcomes directly (Beutler et al., 2004; Blow et al., 2008; Okiishi et al., 2006), therapist’s perceptions of their clients do vary by gender (Bowers & Bieschke, 2005). Further, studies strongly suggest that client gender does not appear to have an effect on treatment outcomes (Griner & Smith, 2006), however, split expectations of couple clients may. Thus, attending to the broader domain of gender in couple and family therapy seems to be critically important in terms of shifting our understanding from therapist gender influences on therapy to client gender experiences of therapy with clinicians. Moreover, attending to therapeutic alliance and client gender and the role of the therapist in actively attending to gender differences moves an underlying processes that can impact the alliance into an explicit space for conversing in therapeutic interactions.

Clinical and Training Implications and Discussion

When a couple presents to therapy with a mixed agenda, one partner leaning into the relationship and the other one leaning out, effective work in couple therapy may be disrupted (Doherty, 2011). This disruption may be related to the discrepancies in commitment to repairing the relationship which could impede developing a therapeutic alliance with both partners. Our finding suggest additional factors may impact the working alliance in therapy. In our sample the discrepancy in relationship satisfaction predicted a lower alliance rating for the male partner. This suggests therapists should be attuned to discrepancies present in a couple’s satisfaction about the relationship. It also suggests that male partners in heterosexual couples who present to therapy are at greater risk for having a lower working alliance when there is a discrepancy in reports of relationship satisfaction.

In clinical work, these findings suggest couple therapy may be enhanced by explicitly discussing discrepancies in the relationship satisfaction, and by informing the therapist’s stance and attitudinal disposition in therapy. The present findings speak to the importance and utility of utilizing and attending to measures of relationship satisfaction and discrepancies between clients on those measures. Furthermore, these findings may indicate a need to overtly present clients’ responses to assessments, which may create a natural therapeutic moment of bringing couple systemic processes to the forefront of a session. Pulling the couple from the in-the moment conversation to a meta-processing and meta-communication level may make explicit the underlying processes that are constraining the effective development of a working alliance in therapy. Indeed, clinicians may find that in highlighting discrepancies in the relationship satisfaction, an opportunity to help clients describe and express the meaning behind their answers to a relationship measure will emerge. This client feedback may also provide an opportunity for clinicians to further explore and help clients express their underlying primary (Johnson, 2012) or accompanying soft (Christensen et al., 2020) emotions and create an experiential moment of healing and acceptance.

The present findings also highlight a potentially useful reframe clinicians can utilize to inform and promote their own helpful attitudinal disposition in their approach with a couple client. Therapists who notice a discrepancy in relationship satisfaction and experience discouragement when attempting to build an effective alliance with a couple may take reassurance that the discrepancy itself, separate from something the clinician is or is not doing, may be constraining the development of a stronger alliance. This may aid in buoying up a therapist’s own hope for the couple to experience their desired outcome, a factor found to be related to improved client outcomes (Coppock et al., 2010). Furthermore, a therapist could conceptualize the discrepancy in relationship satisfaction as a systemic factor affecting the formation of an alliance, rather than a “deficit” in one member of a couple, further reinforcing a systemic frame of reference through which a therapist can operate.

Those in the role of training therapists may consider assisting newer clinicians in utilizing assessments of relationship satisfaction along with conversing skills to re-frame conflict and discrepancies into systemic terms to help them bypass or overcome barriers to forming an effective alliance and thereby improve the overall therapeutic process. If a couple is presenting with discrepancies in relationship satisfaction, it may be that they have differing goals for therapy as well. As shared goals is a theoretical cornerstone of forming an effective therapeutic alliance (Bordin, 1979), clinicians may utilize a discrepancy in opening a dialogue about explicit goals and possible difference in goals or preferred therapeutic tasks between members of a couple in therapy. In some situations, this may also involve a therapist sitting in the confusion discrepancies create for clients. In a case involving satisfaction discrepancies, the first goal of the therapist should be to allow space for each partner to understand what the other is thinking and expecting for their relationship and therapy. Further consideration in training should be made for the utilization of clinical models in understanding the role of the therapist in shaping client experiences during different phases of therapy. Specifically, using theory to guide opportunities to challenge client assumptions about themselves (e.g., the therapist using self and the alliance in an EFT framework to deepen and assemble each partner’s affect across stages), and the therapist (e.g., strategically getting clients to work together even if against the therapist).

Limitations

Several important limitations to this study may have impacted the findings. The study had a small sample with 22 couples and is cross-sectional. Such a small sample size leads to a decrease in power which renders finding only the largest effects likely. As such, moderate or small effects may not have been detected. In addition to the small sample, the sample was relatively homogeneous, for example containing 82.9% Caucasian participants and consisted entirely of heterosexual couples. Additionally, participants consisted of couples in various stages of their relationship, including some who were receiving premarital counseling and others who were separated. The variance in couple relationship stage may impact the relationship between discrepancies in relationship satisfaction and the therapeutic alliance. For example, a couple in an earlier phase of a relationship who experience a discrepancy in their relationship satisfaction may experience a differential impact on the therapeutic alliance than a couple in a later stage of their relationship. Furthermore, while both male and female partners relationship satisfaction scores were below the clinical cut-off, they were still higher than the average clinical score reported by Anderson et al. (2014). The data also came from a COAMFTE accredited training clinic using student clinicians, which may reduce the generalizability of findings to more experienced clinicians, or more distressed samples. The measure of the working alliance we used was initially developed for individual therapy and is thus not explicitly systemic in its conceptualization of the alliance. Other measures of the alliance that incorporate a systemic conceptualization also features items that capture dyad or family level processes such as a shared sense of purpose in the System for Observing Family Therapy Alliances (SOFTA) (Friedlander et al., 2006) should be used to replicate these findings.

Directions for Future Research

In addition to relationship satisfaction, other factors that impact the alliance in couple therapy include marital adjustment (Mamodhoussen et al., 2005) and differentiation (Knerr et al., 2011). While the present study examined how discrepancies in relationship satisfaction impacts the alliance in couple therapy, greater clarity of what influences the formation of an effective alliance beyond relationship satisfaction discrepancies may improve clinicians’ ability to shape their therapeutic stance and optimize the likelihood of developing an effective alliance. Future dyadic discrepancy research can add to the literature on alliance development by examining how discrepancies in reports of marital adjustment and differentiation impact the development of the alliance. Moreover, research with larger samples could examine the gender effect in greater detail. For example, researchers could examine if the strength of the alliance is affected by whether the male or female partner reports lower relationship satisfaction.

Utilizing a more robust system of collecting alliance data from both therapist and clients across time as well may allow for a more nuanced understanding of the impact of alliance on couple processes. This could allow researchers to answer more complicated questions regarding discrepancies in therapist and couple reports, as well as discrepancies within the couple. Examining the alliance using an observational measure or with an instrument that captures systemic processes more explicitly, such as in the SOFTA (Friedlander et al., 2006) could further enhance confidence in the validity of the findings. The LCM could be used as a way of measuring a split alliance that estimates the effect of the discrepancy in alliance reports while controlling for the overall level (e.g. high or low) of the alliance. This could be used as an additional strategy for examining the association between split alliances and premature termination or poorer treatment outcomes. Longitudinal research that examines how discrepancies in relationship satisfaction may affect the trajectory of the alliance across different phases of treatment and in different therapy models are also warranted.

Conclusion

In this study we found that discrepancies in relationship satisfaction predict a lower overall participant rating of the strength of the couple alliance, and lower alliance scores for male partners in a couple. This study provided evidence that discordant reporting on relationship satisfaction influences the initial alliance in couple therapy. Therapists may benefit from assessing discrepancies in reports of relationship satisfaction and be cognizant of its potential impact on the working alliance.