Introduction

The prevalence and natural course of couple disorders are increasingly well delineated. Rates of couple distress in the general population range from 20 to 30 % (Whisman et al. 2008) and longitudinal studies show a general decline of dyadic adjustment in the long run (Kurdek 2008). Relationship help-seeking is, however, not a straightforward process and less than 20 % of distressed couples seek treatment, generally after problems have become chronic (Doss et al. 2009). In addition, whereas couple therapy has grown into an evidence-based practice (Shadish and Baldwin 2005; Snyder et al. 2006), early termination of treatment remains a widespread but understudied phenomenon. Yet, in intent-to-treat analyses, large early termination rates severely curb treatment response estimates (Tremblay et al. 2008). They also reduce organizational effectiveness and impede the delivery of couple therapy in community settings, increase overall therapy costs, and lower professionals’ morale.

An examination of two recent meta-analytic studies of couple therapy research (Shadish and Baldwin 2005; Wright et al. 2007) showed extensive heterogeneity in the likelihood of early termination. When reported (in 51 out of 68 studies), early termination rates varied from 0 to 44 % (M = 11.0 %, SD = 14.5 %) in studies of couple therapy for the alleviation of marital distress, and from 0 to 50 % (M = 15.6 %, SD = 12.7 %) in studies of couple treatment of marital distress and a comorbid condition such as substance abuse or depression, t (49) = 1.11, p > .05. These observations, derived from well-controlled outcome studies, most probably—as in individual psychotherapy practice (Barrett et al. 2008)—provide an underestimation of couple therapy early termination in actual practice. In fact, there are some preliminary indications that when conducted under clinically representative conditions and outside of a research context, the incidence of early termination in couple therapy can raise to nearly 50 % (Allgood et al. 1995; Tremblay et al. 2008; Ward and McCollum 2005). In that context, couples are not recruited, preselected, and fully assessed by paid assistants, using a multimethod, multivariable protocol. Furthermore, treatment is not free and there is no intensive monitoring and supervision of ongoing cases to optimize treatment observance and clinical expertise.

Most couples in efficacy/effectiveness studies show significant clinical distress (Shadish and Baldwin 2005; Wright et al. 2007) and researchers commonly assume that people seek couple therapy to reduce relationship distress and are committed to their union. In actual practice, that is generally, but not always true, nearly 30 % of couples (Tremblay et al. 2008) also consult to determine whether they should stay together. If such is the case and couples come to treatment for other reasons than to improve their relationship, therapists need to assess and address these ambivalence issues more directly. They may also need to adjust their outcome criteria, including the criteria used to define early termination, accordingly.

Most clinicians conceptualize therapeutic goal formation as a joint negotiation process where partners’ expectancies and short-term plans about the continuation of the relationship are taken into consideration (Gurman 2008). In addition, in a careful study of couples’ reasons for seeking therapy, Doss et al. (2004) have shown that one of the most frequently reported motives is divorce/separation preoccupations (e.g., discover if relationship is workable or deciding whether to divorce/separate). However, this is an area of concern not frequently discussed in the couple literature. Tremblay et al. (2008) introduced the notion of therapeutic mandates in couple therapy to better operationalize how therapists take into account the general orientations that partners want to give to the therapy process. A therapeutic mandate is defined as an agreement reached between the couple and the therapist as to the direction treatment will take. In traditional and integrative cognitive-behavioral couple therapy, the mandate would generally be negotiated around the fourth session. At this moment, the diagnostic assessment is completed and assessment feedback about the couple’s main difficulties, the severity of these problems and their potential causes has been discussed (Jacobson and Christensen 1996). The therapeutic mandate notion was developed to recognize that some partners, even after having completed the assessment process, cannot decide whether they want to invest in couple therapy to improve their relationship or if they need to further explore their ambivalence toward the relationship.

To more formally address these situations, three categories of mandates have been described by Poitras-Wright and St-Père (2004) and studied by Tremblay et al. (2008): alleviation of distress, ambivalence resolution, and separation intervention. Tremblay and colleagues demonstrated that theses mandates can be reliably classified, they are sometimes revised over the course of treatment to take into account the specific needs of couples and they are associated with pre-treatment relationship distress and therapeutic outcome. This diversity of couple therapy mandates may partially explain the elevated early termination rates in actual practice, ambivalence resolution and separation intervention potentially creating volatile treatment situations. To our knowledge, no study has examined the characteristics of couples seeking therapy to resolve relationship ambivalence, even though this represents a sizable proportion of the cases therapists treat. The vast majority of treatment procedures used by the therapists for couples pursuing an alleviation of distress mandate are well defined in recent clinical literature from CBCT, IBCT and EFT. Treatment guidelines for an ambivalence resolution mandate are described in Wright et al. (2008). Therapists working to resolve ambivalence help spouses explore feelings, cognitions and behavioral intentions concerning the past and future of their relationship and develop a new and more adaptive model of the nature and causes of their ambivalence. Constructive decision making is attempted around questions such as: Should we separate now? Should we try to improve our relationship? Do we believe in the relationship? Do we feel it is possible to be happy together? Are we ready to work on the relationship? Is it too late? Given we chose not to separate, should we try to develop strategies to handle the ambivalence more constructively? The goal of an ambivalence resolution mandate is to help spouses make a well-informed decision, whether it be to separate or to stay together, and separation can be considered a positive outcome (Christensen and Heavey 1999).

Thus, a first objective of the present study was to examine whether early termination is differentially associated with the nature of mandates in couple therapy. The additional contributions of demographic and clinical variables to the prediction of early termination of couple treatment also needed to be tested.

Past studies of the early termination phenomenon in couple therapy are sparse and they mostly bear on readily-available sociodemographic factors. Early termination has been associated with cohabitation (Davis and Dhillon 1989; Werner-Wilson and Winter 2010), a low number of children (Allgood and Crane 1991), fewer financial difficulties (Davis and Dhillon 1989), income (Tremblay et al., 2008; Werner-Wilson and Winter 2010), and an ethnic minority status (Boddington 1995). Comorbid mental disorders as a presenting problem (Allgood and Crane 1991) as well as elevated personal (Allgood and Crane 1991; Tambling and Johnson 2008), marital distress (Tambling and Johnson 2008; Tremblay et al. 2008) and lack of previous therapy experience (Werner-Wilson and Winter 2010) were the most significant psychosocial prognostic indicators of an elevated early termination rate. These results draw a preliminary, albeit sometimes contradictory (e.g., for financial resources), portrait of factors increasing the risk of early termination in couple treatment. However, there is a need for the investigation of other less-studied but theoretically-relevant, dynamic factors that might increase the risk of early termination from couple therapy. Romantic attachment is one of the potential predictors of early termination that needs to be examined.

In recent years, adult romantic attachment has widely been investigated as a correlate of couple relationship functioning (Milkulincer and Shaver 2007, for a review). Bartholomew and Horowitz (1991) conceptualised romantic attachment as a bidimensional phenomenon: anxiety over abandonment and proximity avoidance. Anxiety over abandonment reflects a working model of self as unlovable and a tendency to excessively worry about being rejected by a partner. Proximity avoidance reflects a working model of others as unresponsive in times of need and a tendency to avoid being too close or dependent in close relationships. Studies conducted on community samples of married couples have demonstrated a consistent association between attachment anxiety and avoidance and couple distress (e.g., Davila et al. 1998; Feeney 1994). Since attachment orientations describe how people tend to experience relationships, they are likely to influence the way clients perceive and behave in the therapeutic context. Indeed, attachment security (i.e., low anxiety and low avoidance) has been associated with a stronger alliance and a more positive outcome in individual therapy (Eames and Roth 2000; Reis and Grenyer 2004) and attachment avoidance has been related to early termination in individual cognitive-behavioral therapy (Tasca et al. 2006). In addition, the clinical relevance and empirical validity of romantic attachment orientations with couples in therapy was recently underlined by Parker et al. (2011) and some preliminary support as a predictor of treatment responsiveness in emotionally focused couple therapy (Johnson and Talitman 1997).

The association between romantic attachment and couple therapy termination has yet to be examined. However, attachment theory and clinical experience suggest that elevated abandonment anxiety and proximity avoidance might disrupt couple treatment processes. The initial experiences surrounding couple therapy require a tolerance for self- and other-disclosure of negative feelings and for uncertainty as to the future of the union. This might be an especially challenging task for highly anxious partners, exacerbating their already prominent fear of abandonment and feeling of being unlovable. These difficulties are probably increased when the therapeutic mandate consists of resolving ambivalence toward the relationship. In addition, successful treatment requires the formation of a positive therapeutic alliance with the therapist (Pinsof et al. 2008), which might be challenging for highly avoidant partners who tend to emphasize self-reliance and might thus be sceptical about the usefulness of therapy. Highly avoidant partners might also feel uncomfortable talking about their feelings and vulnerabilities, which might contribute to a desire to interrupt treatment early. These hypotheses call for empirical verification and gender differences or interaction effects between female and male romantic attachment should also be scrutinized.

Finally, there are heterogeneous definitions and criteria for assessing early termination: after the first phone interview, during the assessment period, following an arbitrary fixed number of sessions, without therapist’s agreement on goal attainment, etc. The various definitions of early termination suggest caution. They limit between-study comparability and probably represent diverse, partially overlapping, constructs with some specific predictors and outcomes that should all be researched (Barrett et al. 2008). Although most previous studies have defined early termination as attending less than an arbitrary fixed number of sessions—perhaps for methodological convenience—this definition has widely been criticized for its lack of theoretical relevance (Bischoff and Sprenkle 1993; Garfield 1994; Wierzbicki and Pekarik 1993). In the present study, early termination was defined, following Garfield’s (1994) recommendation, as cases where clients unilaterally decided to terminate treatment without therapist agreement, at a time where therapists perceived little improvement had been made. In addition, because of our specific interest about the profile of couples in which a partner had significant doubts about the relationship and did not accept to participate in a treatment exclusively designed to improve the relationship, only couples that had completed the assessment period and agreed on a specific therapeutic mandate were selected. Thus, our definition excluded couples who terminated during the assessment period (i.e., before the 4th session). Moreover, the decision to rely on therapists’ judgment rather than on partners’ perspective does not take into account the possibility that clients may stop treatment early because they reached their own objectives. As a consequence, early termination should not automatically be equated with poor outcome. This should be empirically tested. In fact, couple therapists working from a systemic perspective have suggested that the meaning of early termination should be examined mainly from the couples’ perspective (Shoham et al. 2008). There is also some recent evidence supporting the claim that some couples achieve significant gains over a very short interval of time (Doss et al. 2011). However, these sudden gains were generally observed around the fifth session and in treatments where couples continued to at least eight sessions. Even if sudden gains are less likely to be reported during the assessment period, it is important to keep in mind that our research design is not sensitive to this possibility.

To summarize, the main purpose of the present study was to determine the role of sociodemographic variables, therapeutic mandates (i.e., alleviation of couple distress or ambivalence resolution) and pre-treatment prognostic indicators (i.e., relationship distress and attachment) in the prediction of early termination of couple treatment. Based on the slowly growing data-base on early termination, we hypothesized that elevated initial couple distress would increase the likelihood of early termination, as assessed by therapists. We hypothesized that partners pursuing an ambivalence resolution mandate would present a higher rate of early termination as compared to couples seeking help for the alleviation of relationship distress. Finally, it was hypothesized that elevated attachment insecurity, expressed through either anxiety or avoidance, would increase the likelihood of couple therapy early termination. As previous studies have failed to report consistent findings as to the influence of sociodemographic variables on couple therapy completion, we did not expect significant associations, but included sociodemographic variables in the analyses as covariates.

Method

Participants

Between September 2005 and February 2009, 148 heterosexual couples who initiated therapy at a fee-for-service couple therapy clinic affiliated with the University of Montreal and located in the Montreal, Quebec, Canada area were recruited. Couples were self-referred or referred by a mental health professional. Heterosexual couples seeking therapy to alleviate relationship distress or to resolve ambivalence about relationship continuation and who were married or cohabiting were eligible for inclusion in the study. Men were on average 43.4 years old (SD = 8.0, range 25–66) and women were on average 40.6 years old (SD = 8.3, range 23–65). Employment rates were 95.7 % for men and 81.8 % for women, and 50.7 % of men and 56.9 % of women had obtained at least an undergraduate degree. Average income was in the CAN$70 000—79 999 range for men and in the CAN$40 000—49 999 range for women. Couples in the present sample had been together on average 13.2 years (SD = 8.5, range 1–42). Most participants were parents: 10.6 % of men and 8.5 % of women had no child, 19.1 % of men and 24.1 % of women had one child, and 76.4 % of men and 67.3 % of women had two children or more. Fifty-five percent of couples were legally married and 45 % were cohabiting, which is representative of couples in the province of Quebec (Le Bourdais and Lapierre-Adamcyk 2004).

Therapists

The therapists who participated in the study were 6 licensed psychologists (n = 129 cases treated), 2 clinical psychology doctoral interns (n = 5 cases) and 8 clinical psychology graduate trainees (n = 7 cases). The psychologists held a master’s (3) or doctoral (3) degree in clinical psychology and averaged 16.0 years experience in couple therapy at the beginning of the study (SD = 10.8, range 4–32). Six therapists reported that their primary theoretical allegiance was cognitive-behavioral, 1 emotion-focused and 1 psychodynamic. The graduate trainees were all under the supervision of the 3rd author and followed procedures consistent with Cognitive-behavioral couple therapy (CBCT; Epstein and Baucom 2002), Integrative behavioral couple therapy (IBCT; Jacobson and Christensen 1996) or Emotion-focused couple therapy (EFT; Johnson 2004).

Procedures

This study was conducted as part of a broad investigation of couple therapy outcome in a fee-for-service clinic that has been approved by the University of Montreal Ethical Review Board. Couples were referred to the research team by therapists who introduced the research project during the first therapy session. Couples who agreed to participate signed a consent form and partners independently filled in, before the second session, the pre-intervention self-report measures battery. This battery included a demographic questionnaire, the Dyadic Adjustment Scale (Spanier 1976) and the Experiences in Close Relationships Questionnaire (Brennan et al. 1998). Assessment data were analyzed and transmitted to the therapists, who used them as additional information in the assessment process. When the therapeutic mandate was set with couples at the 4th session, the therapists filled in the Classification of Therapeutic Mandates Questionnaire (Poitras-Wright and St-Père 2004). As this study was designed to be highly clinically representative (Wright et al. 2007), no treatment manual was followed by the therapists. They conducted treatment-as-usual, using a varied range of techniques and approaches in order to best respond to clients’ needs, and their interventions were not monitored. When the couples terminated treatment the therapists answered the Termination Form.

Measures

Demographic Information

The variables of interest for the current study were age, educational attainment, employment status, annual individual income, marital status (married or cohabiting), relationship length (in years), number of children, number of previous significant relationships, family structure (nuclear family or stepfamily), prior experience in couple therapy (yes/no), prior experience in individual therapy (yes/no), and duration of relationship problems (in years and/or months).

Couple Distress

The Dyadic Adjustment Scale (DAS; Spanier 1976) is a 32-item self-report questionnaire that is widely used to assess global couple distress. Global scores range from 0 to 151 with higher scores indicating better couple adjustment. Reliability coefficients range from .91 to .96 for the original version (Spanier 1976) and from .89 to .91 for the French translation (Sabourin et al. 1990). In the present sample, Cronbach’s alphas were .91 for men and .90 for women.

Romantic Attachment

The Experiences in Close Relationships Questionnaire (ECR; Brennan et al. 1998) is a 36-item self-report questionnaire widely used to assess attachment orientations. Separate scores are obtained for the anxiety over abandonment and proximity avoidance dimensions. Alpha coefficients of .94 for avoidance and .91 for anxiety have been reported for the original version (Brennan et al. 1998). Using the French-language version, Lafontaine and Lussier (2003) reported alpha coefficients ranging from .86 to .88 for both scales. In the present sample, alphas for the avoidance scale were .85 for men and .90 for women, and alphas for the anxiety scale were .87 and .89 respectively for men and women.

Therapeutic Mandates

The Classification of Therapeutic Mandates Questionnaire (Poitras-Wright and St-Père 2004) was completed by therapists at the fourth assessment session, using a coding manual based on three general categories: alleviation of distress, ambivalence resolution, and separation intervention. The alleviation of distress mandate aims at increasing relationship satisfaction and is the traditional mandate pursued in couple therapy. The ambivalence resolution mandate aims at helping partners resolve ambivalence and make an informed decision toward relationship continuation. The separation intervention mandate pursues the goal of helping the partners achieve a constructive separation and to minimize the negative effects of separation. The couples pursuing that last mandate were excluded from the present study. Because assessment sessions were not recorded, therapeutic mandates could not be coded by external judges and reliability estimates could not be computed. However, validity coefficients for the Classification of Therapeutic Mandate Questionnaire with an independent sample of clinical couples were reported by Tremblay et al. (2008): therapeutic mandates were related to pre-treatment couple distress and dysfunctional responsibility attributions.

Termination Form

When treatment ceased, in order to determine termination status, the attending therapist was interviewed by one of three research assistants (supervised by the first author). Two choices were given to classify the reasons for termination: (a) Treatment completers: cases of complete (treatment goals were achieved) or partial recovery (significant improvement had been made but more treatment would have been beneficial) were grouped together to take into account the fact that therapists tend to have higher goals for clients and to expect treatment to last longer than clients do (Barrett et al. 2008), (b) Early terminators : cases where clients unilaterally decided to terminate treatment without therapist agreement, at a time where therapists perceived little improvement had been made, and therapy is viewed as not completed by the therapist (Garfield 1994). Since the goal of an ambivalence resolution mandate is to achieve an informed decision on the future of the relationship, couples were considered completers if they had achieved that goal, regardless of the relationship outcome (whether the decision was remaining together or separating). As in most studies of early termination in actual couple therapy practices (e.g., Allgood et al. 1995; Ward and McCollum 2005), interobserver agreement on termination status could not be computed because treatment sessions were not audio- or videotaped.

Results

Average relationship satisfaction at pre-treatment was 92.51 (SD = 16.71) for men and 90.28 (SD = 16.10) for women, which is considered mildly distressed, and 61.0 % of men and 66.7 % of women scored in the distressed range (DAS ≤ 97). Couples estimated that when they contacted the clinic, they had experienced relationship problems for on average 5.24 years (SD = 5.54). A majority of couples sought treatment to alleviate relationship distress (n = 106) but a noteworthy portion of the sample consulted to resolve their ambivalence toward the relationship (n = 35). As this study was conducted in a fee-for-service clinic with various therapists, the therapist’s effect was examined in preliminary analyses as this variable was seen as a potential confounder. A logistic regression showed that early termination frequency was not significantly related to therapists (Wald = 4.5, df = 8, p = .81) and therefore data were collapsed across therapists for the remaining analyses.

Prediction of Early Termination from Couple Therapy

Therapists classified 106 couples as completers (alleviate relationship distress group = 86 (81.1 %); ambivalent group = 20 (57.1 %)) and 35 as early terminators (alleviate relationship distress group = 20 (18.9 %); ambivalent group = 15 (42.9 %). There were significantly more early terminators in ambivalent couples than in couples where the therapeutic mandate was to alleviate relationship distress, χ2 (1, N = 141) = 8.11, p = .004). Completers attended on average 11.4 therapy sessions (SD = 6.1) whereas early terminators remained in treatment for on average 6.5 sessions (SD = 3.7), t (97) = 5.66, p < .001. Treatment length also varied as a function of the therapeutic mandate: couples pursuing an alleviation of distress mandate remained in treatment for on average 11.1 sessions (SD = 6.0) whereas couples pursuing an ambivalence resolution mandate attended on average 7.4 sessions (SD = 4.9), t (71) = 3.60, p < .001. Analyses of variance were conducted as preliminary analyses to identify relevant predictors of early termination (see Table 1). Using relationship satisfaction and romantic attachment as dependent variables, gender was entered as a within-subject variable to account for the non-independence of couple data and termination status (completion or early termination) was treated as a between-subject variable. The relation between therapeutic mandates and early termination was assessed using a Chi square analysis. Demographic variables and romantic attachment were not related to termination status. However, elevated pre-treatment couple distress, F (1, 139) = 8.14, p < .01, η2 = .06, and a therapeutic mandate oriented toward ambivalence resolution, χ2 (1, N = 146) = 8.11, p < .001, proved to be significant prognostic indicators of early termination of couple therapy.

Table 1 Univariate analyses comparing completers and early terminators on relationship satisfaction, attachment avoidance, attachment anxiety, and therapeutic mandate
Full size table

To determine the variables related to termination status, a binary logistic regression analysis was conducted.Footnote 1 Variables significantly related to termination status in preliminary analyses (i.e., relationship satisfaction and therapeutic mandate) were entered in the model, and interactions between pre-treatment satisfaction, therapeutic mandate, attachment avoidance, and attachment anxiety were also examined. Results showed that women’s relationship satisfaction, therapeutic mandate and the interaction between women’s relationship satisfaction and men’s attachment anxiety predicted early termination. The model explained 22.7 % of the variance in couple therapy termination status, χ2 (5, N = 141) = 23.37, p < .001. Each incremental decrease in women’s relationship satisfaction was associated with an increased probability of early termination by a factor of .53 (95 % CI = .32–.88). Couples pursuing an ambivalence resolution mandate were 4.10 times more likely to discontinue treatment early than couples pursuing an alleviation of distress mandate (95 % CI = 1.57–10.74). A significant interaction between women’s relationship satisfaction and men’s attachment anxiety was found, OR = .54 (95 % CI = .34–.88). Further examination of the interaction effect showed that a decrease in women’s relationship satisfaction was significantly related to increased chances of early termination only when men scored above average on attachment anxiety, OR = .38 (95 % CI = .20–.69).

Discussion

Many significant findings emerged from the present study. First, the prevalence rate of early termination of couple treatment in a fee-for-service clinic reached 25 %. Ambivalent couples were more prone to drop out of couple treatment (42.9 %) than couples in which the main treatment goal was to alleviate relationship distress (18.9 %). Whereas our early termination prevalence estimate is lower than what has been reported in studies of actual couple therapy practices (Allgood et al. 1995; Tremblay et al. 2008; Ward and McCollum 2005), it is clearly higher than what has been observed in past couple therapy outcome research (Shadish and Baldwin 2005; Wright et al. 2007). If the observed trend showing increased occurrence of early termination in actual practice is replicated, future work will need to scrutinize the barriers specific to couple therapy continuation in these traditional clinical settings.

Second, our efforts to identify prognostic indicators of early termination in couple treatment proved fruitful on several fronts. As expected, demographic factors were not significantly related to treatment discontinuation, but theoretically-relevant predictors were identified. More specifically, relationship distress, therapeutic mandates, and attachment anxiety were, alone or in interaction, consistent predictors of early treatment discontinuation. As reported in past studies (Tambling and Johnson 2008; Tremblay et al. 2008), pre-treatment relationship distress significantly increased the risks of early treatment termination. In addition, past longitudinal studies have rigorously established that self-reports of couple distress or satisfaction predicts trajectories of relationship quality and stability (Kurdek 2008; Tilden et al. 2010). It may well be that the best prognosis indicators of future couple well-being will prove to be the most efficient variables to predict early termination in couple therapy. Future studies will need to determine if relationship distress influences treatment continuation through one of several mediator variables: learned helplessness, lower subjective barriers to couple dissolution, different expectations about therapy duration, diminished capacities to form a productive working alliance with their partner and the therapist, etc.

The present results also reveal that, for women, the role of pre-treatment relationship distress is contextualized by men’s attachment anxiety. More specifically, the association between women’s relationship distress and early termination was significant only when men evidenced high attachment anxiety. This result is difficult to interpret in the absence of data specifying how the decision to terminate treatment is reached for each couple. However, it may be hypothesized that attachment anxiety in men increases their likelihood to disengage from couple treatment because they are easily overwhelmed by their partner’s disclosure of dissatisfaction in major relationship areas. The gender-inconsistent pattern of elevated attachment anxiety in men (Brassard et al. 2007) may put these couples at greater risks of early termination: men confronted with attachment anxiety issues may react more strongly than women to expression of elevated relationship dissatisfaction in the presence of a third party and elect to quit treatment to avoid being confronted not only with their partner’s public disclosure of dissatisfaction but also with their gender-inconsistent (and potentially humiliating) pattern of attachment.

In the present study, the most robust predictor of early termination in couple treatment was an ambivalence resolution mandate. In contrast, couples seeking help for an alleviation of distress mandate were four times more likely, according to their therapist’s judgment, to complete couple therapy. This is an important finding for two related reasons. First, to our knowledge, there is a lack of empirical data delineating the frequency of specific therapeutic mandates and their implications for the outcome of couple therapy. Overall, the results revealed that 25 % of couples in the present sample sought help to resolve ambivalence toward their relationship. This represents a sizable proportion of the participating couples and future studies will be needed to determine with some accuracy the prevalence of couples consulting in a context where indecision about the outlook of the relationship is a significant clinical issue for one or both partners. Second, there are now two independent investigations supporting the empirical validity of the therapeutic mandate concept in couple therapy (Tremblay et al. 2008) and this construct is associated with early termination. These findings appear sufficiently promising to warrant a larger-scale examination of the meaning and consequences of early termination in couple therapy from the clients’ perspective, especially when one or both partners are ambivalent about their relationship. It would be important to understand more fully how these partners experience early termination. The decision of ambivalent couples to terminate treatment early may be associated with good outcomes related to sudden gains and rapid fulfillment of pre-treatment expectations (Doss et al. 2011) or to negative outcomes, i.e., increased couple distress through escalation in conflict or distance behaviors. The “dropout” literature in couple therapy is based on the largely untested assumption that early termination is an unambiguous sign of poor outcomes. Indeed, empirically-supported couple therapy manuals from cognitive-behavioral (Epstein and Baucom 2002), emotionally-focused (Johnson 2004) or integrative (Jacobson and Christensen 1996) perspectives provide treatment duration guidelines ranging from 12 to 20 sessions. These guidelines rest partly on a dose–effect model for which there is some evidence in patient-focused research (Leach and Lutz 2010). However, in systemic couple therapy, clients’ expectations about short treatment duration have led to flexible adaptations and crisis-oriented brief interventions. The respective value of these approaches to the meaning of early termination needs to be further examined.

Overall, the present results highlight the need for more research to better understand the specific needs of the ambivalent couples, the mechanisms leading to an elevated rate of early treatment termination, and, ultimately, to develop treatment approaches specifically targeted to effectively resolve ambivalence issues. Future studies should be based on larger samples of ambivalent couples to identify with more precision the specific predictors of early termination in this subgroup. In the present study, it was not possible to conduct such analyses because of statistical power problems. In light of Pinsof et al. (2008) finding that low within-couple agreement on alliance in early treatment sessions predicted lower retention and observed outcome by the 8th session, it may well be that ambivalent couples are at higher risks of disagreement on therapy goals in the first place, and that, in these cases, poor alliance is related to early termination. In fact, it is sometimes (but not always) the case that an ambivalence resolution mandate is set because only one of the partners is unsure about the relationship future. On the other hand, an alleviation of distress mandate is always chosen when both partners want to improve their relationship, thus minimizing the possibility of a major within-couple alliance problem. At the present time, it is difficult to ascertain whether ambivalent couples are systematically excluded (or exclude themselves) from randomized trials assessing the efficacy/effectiveness of couple therapy, a factor that can limit the clinical representativity of these studies and the transportability and dissemination of evidence-based couple treatment to actual practice settings (Wright et al. 2007).

There are some limits to the present study. First, therapeutic mandates and early termination were assessed strictly from the therapists’ perspective. There is a need to measure these two variables using independent observers and both partners’ viewpoints. This would provide a unique occasion to establish interjudge reliability and to examine multi-source convergence. Second, therapeutic mandates were coded at the fourth session, following a full diagnostic workup. Our decision to wait after this period to classify mandates was based on traditional assessment guidelines in cognitive-behavioral couple therapy (Epstein and Baucom 2002) and on the need to develop a comprehensive conceptualization of couple difficulties before negotiating a therapeutic mandate. It would be interesting to determine if a therapeutic mandate could be agreed upon and reliably coded earlier in the diagnostic process. This would allow the inclusion in the sample of a larger number of couples who terminate treatment before the fourth session. Finally, post-treatment measures of couple distress should be included in future studies to determine more rigorously whether early termination in ambivalent couples is objectively associated with poor outcomes.

Clinical Implications

The present findings indicate that thorough assessment of ambivalence toward the relationship should be conducted during the diagnosis period and used to develop a clinically meaningful therapeutic mandate adapted to the situation of each couple. The reliable measurement of the nature and strength of ambivalence can help therapists identify potential cases of early termination. Even if the present results did not allow the identification of risk factors of early termination in ambivalent couples, therapists might need to consider adjusting their treatment plan to include short-term and crisis-oriented interventions. Moreover, such an evaluation will help to clarify complex clinical issues where some partners may not want to discuss their own or their partner’s ambivalent feelings toward the relationship. Addressing ambivalence in couple therapy creates highly emotional situations, especially when only one partner is questioning the relationship future. Therapists need to be highly sensitive to these situations and to acknowledge how demanding therapy can be for partners. In cases where only one partner is ambivalent and partners have divergent therapy goals, therapists should first have both partners agree on a common mandate before pursuing therapy. To do so, therapists can explain to the non-ambivalent partner that ambivalence resolution is also in his/her best interest, and that rushing the ambivalent partner into an alleviation of distress mandate may prove counterproductive. However, there probably exists a wide range of attitudes in therapists about the usefulness of fully addressing ambivalence issues in couple treatment. More research is needed to examine the relevance of different pairings of partners’ and therapists’ attitudes toward a therapeutic mandate centered on ambivalence resolution.

The finding that lower pre-treatment relationship satisfaction has been associated with early termination suggests caution in relying on a significant level of relationship distress to provide strong motivational incentives to persist in couple treatment. Quite to the contrary, most couple therapists generally report that high levels of couple distress is associated with lower motivation to engage in couple treatment and studies have shown that couple treatment is more effective with mildly or moderately distressed couples than with severely distressed couples (Snyder et al. 2006). Therapists might need to specifically address treatment motivation and give hope to the most distressed couples. Furthermore, Tambling (2012) indicated that more studies are needed about the ways in which expectations influence short term outcomes, such as alliance formation or treatment persistence. The relationship between treatment expectations and short-term outcomes may help understand long-term psychotherapy outcomes.

Finally, the current results suggest that therapists may routinely assess and address attachment orientations in couple therapy, as attachment insecurities might interfere with treatment continuation. Although at first the hallmark of EFT, interventions focused on attachment issues are becoming increasingly integrated in various therapeutic models (Johnson and Whiffen 2003), and the current results support the usefulness of attachment theory in couple therapy.