The Role and Relevance of the Working Alliance in Forensic Treatment: A Meta-Analysis

Abstract

The working alliance has been extensively studied and consistently linked to improved treatment outcomes; however, existing reviews have almost exclusively focused on non-forensic samples, which differ significantly in client characteristics, treatment goals, and settings. Therefore, prior reviews may not generalize to forensic contexts. Seeking to fill this gap in the literature, a meta-analysis of the working alliance in forensic treatment contexts was conducted and its relationship to risk-relevant treatment outcomes was examined featuring k = 21 unique studies. All three treatment outcomes examined—treatment dropout (d = 0.390), institutional misconduct (d = 0.374), and community recidivism (d = 0.196)—were significantly and negatively associated with the working alliance. These findings demonstrate the working alliance as a meaningful predictor of treatment outcomes in forensic contexts. Implications for research and practice are discussed to promote more favorable treatment outcomes for forensic clients.

Plain Language Summary

Client-Service Provider Relationships in Forensic Treatment and Links to Rehabilitation and Reoffending Outcomes: A Quantitative Review

The bond between client and therapist, conventionally known as the working alliance, has been extensively studied and consistently linked to improved treatment outcomes; however, existing reviews have almost exclusively focused on samples of people who are not involved in the justice system. These groups differ significantly in client characteristics, treatment goals, and settings. To address this gap, we conducted a quantitative review of the working alliance literature and its relationship to important justice system outcomes, featuring 20 studies. We focused on three outcomes—treatment dropout, prison misconduct, and reoffending in the community post-release. Stronger working alliances were associated with increased treatment completion, decreased prison misconducts, and decreased reoffending in the community. These findings clearly establish the working alliance as a robust predictor of treatment outcomes in forensic contexts. Implications for further research and clinical practice are discussed to promote more favorable treatment outcomes for forensic clients.

Keywords

working alliance forensic therapy treatment dropout institutional misconduct community recidivism

The Role of the Working Alliance in Forensic Treatment: A Meta-Analysis

In psychotherapy research, the working alliance—defined as the collaborative bond between therapist and client (Bordin, 1979)—has been shown to be one of the most reliable predictors of treatment outcomes (Flückiger et al., 2018). Consequently, the working alliance has become the focus of extensive research exploring how to measure it, its correlates, methods for strengthening it, and more (Flückiger et al., 2018). Despite this large body of general research, there is little research on the working alliance and treatment outcomes in forensic settings. Existing reviews from traditional psychotherapeutic contexts include limited representation of forensic studies and these have not been examined using meta-analysis. Addressing this knowledge gap is critical because forensic contexts present distinct challenges and pursue alternative treatment goals than traditional psychotherapy; for instance, the pursuit of enhanced client self-esteem is considered a laudable therapeutic endeavor in traditional contexts, yet doing this with correctional populations can be iatrogenic, absent formal attention to dynamic risk factors linked to crime and violence (Bonta & Andrews, 2023). These considerations necessitate a quantitative review of the literature examining the working alliance specific to forensic populations, which is focus of the current study. We examine the strength of this relationship and identify moderating factors that may influence it to inform improvements in the delivery of correctional treatment and reductions in recidivism; ultimately, to promote greater public safety.

Current State of the Working Alliance

The modern definition of the working alliance originated with Bordin (1979)’s model of task, bond, and goal (Doran, 2016). In this model, Bordin defined the working alliance as a collaborative agreement between client and therapist, comprising three components: agreement on therapeutic goals (goals), focusing on relevant therapeutic tasks (tasks), and fostering a positive, warm relationship (bond). Bordin believed that this concept could be applied to all interpersonal change processes, including different therapeutic modalities (Hatcher, 2010). At the time, research in psychology was increasingly recognizing that various therapeutic modalities appeared to have similar levels of effectiveness, a phenomenon known as the “dodo bird effect” (Bordin, 1979). Accordingly, interest proliferated in how much of the outcome could be predicted by factors not specific to any particular therapeutic intervention, such as the working alliance (Horvath & Greenberg, 1989).

Consequently, a large literature developed linking the working alliance to treatment outcomes. According to a landmark meta-analysis by Flückiger et al. (2018), nearly 300 studies (k = 295) have assessed the relationship between the working alliance and treatment outcomes. Flückiger et al. found that stronger working alliances were correlated with better treatment outcomes, at a medium effect size of d = 0.579 (Cohen, 1988). This is a significant finding in the context of psychotherapy research, where effect sizes are typically small (d < 0.20; Cohen, 1988; Messer & Wampold, 2002). For perspective, specific treatments account for approximately 1% of treatment outcomes (Messer & Wampold, 2002), whereas the working alliance explains 7.7% (r² = .077), or approximately seven times greater (Flückiger et al., 2018). Despite the clinical significance of the working alliance and the extensive research dedicated to it, there has been limited focus on its role within correctional treatment settings, which differ substantially from traditional therapeutic environments.

Effective Correctional Treatment

To understand the differences between correctional treatment and general psychotherapy, it is important first to examine the foundational theories and evidence-based models specific to forensic contexts. The leading framework in this area is the Risk-Need-Responsivity (RNR) model which has been widely adopted in correctional treatment and assessment globally (Bonta & Andrews, 2023). A well-established research literature shows that programs adhering to RNR principles achieve significantly better outcomes, with community recidivism rates reduced up to 35% (Bonta & Andrews, 2007, 2023). The risk principle asserts that the intensity of treatment should be aligned with the individual’s risk level (Bonta & Andrews, 2007). Individuals at high risk require more intensive interventions, and grouping individuals according to their risk levels is crucial to maximize resource use and treatment gain (Bonta & Andrews, 2007). This approach is not fundamentally different from standard psychotherapy, where individuals with more severe symptoms may be matched to more intensive interventions, sometimes referred to as triage (Walton & Grenyer, 2002).

The need principle notably departs from traditional psychotherapy. Addressing criminogenic needs, or dynamic factors associated with increased risk of criminal behavior, in order to reduce criminal behavior is often the primary goal of forensic/correctional treatment (Bonta & Andrews, 2023). This distinguishes forensic/correctional treatment from traditional psychotherapy, where symptom relief (e.g., depression, anxiety) is often the main objective (Bonta & Andrews, 2023). Although treating such non-criminogenic needs may be done on humanitarian grounds and/or to promote engagement, per the responsivity principle (see below), this may not necessarily translate into decreased recidivism (Bonta et al., 2013; Bonta & Andrews, 2023). As such, the goals and tasks outlined in Bordin’s model differ within a correctional framework, focusing on criminogenic rather than symptomatic, non-criminogenic needs.

Finally, the responsivity principle includes a general component that finds that correctional populations tend to respond better to cognitive behavioral therapy, as well as behavioral, social learning and skills building approaches (Bonta & Andrews, 2007). These approaches are also effective for other client populations (Bonta & Andrews, 2023); however, there is also a specific responsivity component which entails tailoring the services to the individual characteristics of justice-involved clients, including their developmental stage, cognitive ability, cultural background, and many other factors (Bonta & Andrews, 2007; Bonta & Andrews, 2023). This individualized approach is also a defining feature of traditional psychotherapy, where therapists who are more flexible to client’s needs have better results (B. Blasko et al., 2018); however, the adjustments that are made may be quite different especially given the unique characteristics of correctional populations (e.g., low levels of insight and/or motivation, antisocial personality features).

The Working Alliance in Correctional Contexts: Characteristics and Challenges

The unique characteristics of correctional populations can present challenges to alliance building. Meta-analytic research demonstrates that high risk, high need clientele, and particularly those presenting with prominent features of psychopathy are at elevated risk for treatment noncompletion (Olver et al., 2011). High psychopathy persons often form weaker alliances, show less progress, and exhibit more treatment-interfering behavior (Ogloff et al., 1990; Olver & Wong, 2011). Further, correctional populations are more likely to have traumatic histories and insecure attachment styles, which can hinder trust and relational engagement (Gobin & Freyd, 2014; T. Ross & Pfäfflin, 2007). Trust is especially crucial when addressing offending behavior, as shame and stigma toward offending often complicate this process (Bourgon & Guiterrez, 2013). Effective alliance-building in this setting requires a non-judgmental stance, that does not condemn the individual, yet addresses the offense behavior (B. Blasko et al., 2018; Youssef, 2017).

In correctional settings, service providers must also balance institutional priorities of safety and security with clients’ therapeutic needs—often resulting in dual-role conflicts and ethical dilemmas (Ward, 2013). For instance, therapists may be required to report disclosures that could lead to punitive consequences, discouraging openness and undermining the trust essential for a strong working alliance (Bourgon & Guiterrez, 2013; Kozar, 2013). Compounding these issues, forensic/correctional treatment is often mandated, which can lead clients to feel coerced into treatment (Bourgon & Guiterrez, 2013; Hachtel et al., 2019). These dynamics can contribute to fundamental differences in how working alliances are formed in forensic and correctional settings versus voluntary treatment contexts, leading some scholars to argue that alliance theories developed in voluntary settings may not generalize to mandated environments (Bourgon & Guiterrez, 2013). This is notable given that Flückiger et al. (2018) did not account for mandated status and included few studies from mandated contexts in their meta-analysis.

The influential meta-analysis by Flückiger et al. (2018) included only three studies on forensic clients, limiting any direct application to correctional settings. The extant literature suggests a varied and at times complex association between alliance and correctional treatment outcomes. For instance, several studies demonstrate weaker alliances to be associated with increased correctional program attrition from sexual offense programs (D. R. DeSorcy et al., 2016), high intensity violence reduction programs (Polaschek & Ross, 2010), youth custodial programs (Fitch, 2012; Florsheim et al., 2000), and community forensic mental health and legal settings (Miles-McLean et al., 2019; Trawver, 2011). Moreover, within these settings, ethnocultural minority groups (e.g., Indigenous clients), have demonstrated weaker alliances on average (but not uniformly), and higher rates of attrition than non-minority clientele (D. R. DeSorcy et al., 2016). By contrast, the picture seems to be less consistent for working alliance associations in community and custodial programs with future crime and violence, with several studies demonstrating associations with decreased recidivism (e.g., Florsheim et al., 2000; Gutierrez, 2010; Holmqvist et al., 2007; Kennealy et al., 2012), and others finding no such association (e.g., D. R. DeSorcy et al., 2016; Quick-Parikh, 2024; A. Tatman et al., 2024; Trawver, 2011).

Rationale for the Present Study

Given the contextual, practice-related, and philosophical differences between forensic and non-forensic psychological interventions, a comprehensive review of the working alliance and its relationship to correctional outcomes in a forensic context is needed. While research links the working alliance to treatment outcomes across general psychotherapy settings (Flückiger et al., 2018), comparably fewer studies focus specifically on forensic samples and have yet to be aggregated. To address this gap we conducted a meta-analysis of the working alliance and its relationship to three sets of correctional outcomes in forensic samples; treatment dropout, community recidivism, and institutional misconduct. These three outcomes were chosen because they represent important goals of forensic treatment and have adequate data available. In addition, we examine moderators that may influence the relationship between the working alliance and treatment outcomes (e.g., measure type, administration method) which may have implications for forensic service delivery.

Method

Search Process

To identify relevant studies, we utilized a multi-step process, beginning with a selection from existing reviews and followed by systematic searches across several databases. Searched databases included PsycINFO, ProQuest Theses and Dissertations Global, Web of Science, MEDLINE, Criminal Justice Abstracts with Full Text, Social Services Abstracts, and Academic Search Complete. The searches utilized predefined sets of terms, referred to here as “concepts.” The first concept targeted studies referencing the working alliance, while concepts three, four, and five focused on the outcomes of community recidivism, institutional misconduct, and treatment dropout respectively. The concept of the working alliance was combined with each of these outcome-related concepts to find studies examining the link between the working alliance and these outcomes. For searches looking at treatment attrition, a set of terms specifying a forensic context (concept 2) was also combined, given that treatment attrition has also been examined in non-forensic settings. All searches were conducted on September 2, 2024. Following the initial collection of studies, reference lists were examined to identify additional relevant articles, all of which helped ensure a comprehensive and thorough literature search.

After the search, identified studies were evaluated for inclusion based on specific criteria. Eligible studies must: (a) utilize any measure of the working alliance; (b) assess the working alliance between a mental health professional or probation officer and the client; (c) examine community recidivism, treatment completion, or institutional misconduct as an outcome variable; (d) provide descriptive or inferential statistics sufficient for calculating an effect size; and (e) focus on a correctional or forensic mental health sample. Studies that focus on residential treatment samples, substance use treatment samples, or other non-forensic mental health populations were excluded. Studies were excluded if they did not involve either a mental health intervention or probation. Studies were also excluded if they were not available in English. These inclusion and exclusion criteria were applied independently by the first author with guidance from supervisors about any uncertainties to ensure consistency and rigor in the selection process.

Effect Size Coding

Included studies were systematically coded using a comprehensive data coding protocol. The first section of the protocol contained foundational details about the study, including its design (e.g., randomized controlled trial, quasi-experimental) and treatment specifics, such as program intensity and the type of service provider involved. A second section focused on measures of the working alliance, addressing key aspects such as the rater of the working alliance measure (e.g., client or therapist), and key outcome variables such as community recidivism, treatment dropout, and institutional misconduct.

Community recidivism was defined as any rearrest, charge, or conviction occurring after an individual’s release into the community, including continuous measures such as the number of offenses committed while in the community. Treatment dropout was defined as whether a participant successfully completed a treatment program, though if this information was unavailable, time spent in treatment and its relationship to the working alliance was used. Institutional misconduct encompassed any form of rule-breaking, violence, or misconduct within an institution, with both binary and non-binary measures (e.g., number of offenses committed while institutionalized) included when available.

The entire process of searching the literature, applying inclusion/exclusion criteria, and data extraction/coding followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to promote transparent communication of the review’s purpose, findings, and methodology (Page et al., 2021). Each step is documented in a PRISMA flow diagram (see Figure 1), allowing for full replication of the search process and ensuring the same studies are identified for inclusion based on the specified criteria. In alignment with PRISMA, we have detailed the characteristics of each included study and present outcome-level data for each study in (see Table 2). Ethics approval was not required for this project, as it involved only the synthesis of publicly available data.

Figure 1.

PRISMA diagram.

To maintain rigor and accuracy, the first two articles identified as meeting inclusion criteria were coded collaboratively. The remaining articles were coded independently, with the first author consulting other members of the project team as needed to resolve any uncertainties. All available results were also recorded by the first author including all time points and measures. If the correlation between the working alliance and a particular outcome was reported across multiple time points or through different modes these values were averaged to yield a single effect size per study. For instance, effect sizes for client and service provider rated working alliance measures were averaged to yield an overall effect, and then independently examined through moderator analyses (see below). Averaging was also used when multiple operationalizations of a given recidivism criterion were used, or when a study employed multiple working alliance measures, which were later separately examined through moderator analyses. These were reviewed and double-checked for accuracy by members of the project team.

Statistical Analyses

Effect size data were analyzed using Comprehensive Meta-Analysis Software (Version 3.0) to calculate and aggregate effect sizes quantifying the strength of the relationship between two variables in a population (Cohen, 1988) and allows the analysis of moderators (Brüggemann & Rajguru, 2022). We planned effect size calculations for moderators (i.e., working alliance measure and mode of administration), the correlation between different variables and the working alliance, and for all outcome variables. Relevant formulae were used to convert all effect sizes to a standardized mean difference (Cohen’s d), categorized as follows: small (0.20), medium (0.50), and large (0.80; Cohen, 1988). Effect sizes were aggregated into a common effect through both fixed effects (d_FE) and random effects (d_RE) models; the latter introduces an additional term accounting for between study variance that results in relatively greater weight to smaller studies, and wider confidence intervals, although estimates become less stable with a small numbers of studies (Borenstein et al., 2007). Effect size heterogeneity was assessed using the Q statistic and I² statistic. The Q statistic measures the extent to which study results vary beyond what would be expected by random chance, while the I² statistic quantifies the proportion of variation that can be attributed to true heterogeneity rather than random chance at 25% (low), 50% (medium), and 75% (large; Higgins & Thompson, 2002).

Results

Search Results

Through the systematic study selection process described above and depicted in Figure 1, 48 potentially relevant articles were identified and assessed for eligibility. Of these, 24 met all inclusion and exclusion criteria; however, one sample was represented in three eligible articles (D. DeSorcy, 2013; D. R. DeSorcy et al., 2016, 2020), and another in two articles (Sturm et al., 2021a, 2021b). For data and statistical calculations, we used the largest sample in DeSorcy et al., and we averaged the results for Sturm et al. As a result, the final analysis included 21 unique studies. These studies and their samples reflected diverse characteristics across key variables provided in Table 1 and study level data provided in Table 2.

Table 1.

Descriptive Characteristics of Included Studies.

Variable	M/n(%)
Total N	3,226
Gender
Male	2,907
Female	317
Age (years)	31.14
Country
United States	14 (70%)
Canada	4 (20%)
Netherlands	1 (5%)
Sweden	1 (5%)
New Zealand	1 (5%)
Setting
Parole/Probation office	6 (30%)
Community mental health facility	5 (25%)
Juvenile residential treatment facility	3 (15%)
Juvenile outpatient facility	1 (5%)
Hospital	2 (10%)
Prison	3 (15%)
Court	1 (5%)
Sample
Juvenile (general)	6 (30%)
General unspecified correctional	5 (25%)
Substance abuse	3 (15%)
Sexual offending	2 (10%)
Mental illness	2 (10%)
Juvenile (mental illness)	1 (5%)
Intimate partner violence	1 (5%)
Violent (primarily nonsexual) offending	1 (5%)
Working alliance administration
Throughout treatment	11 (55%)
Beginning of treatment	2 (10%)
Middle of treatment	2 (10%)
End of treatment	1 (5%)
Unknown	4 (20%)
Working alliance measure
Working Alliance Inventory variant	16 (64%)
Dual-Role Relationship Inventory	4 (16%)
Other measures	5 (20%)

Table 2.

Characteristics of Included Studies.

Author (year)	Country	Setting	Sample type	Sample size	Working alliance measure	Time of measurement	Measurer	Outcomes
Beyko and Wong (2005)	Canada	Hospital	Sex offenders	64	Working Alliance Inventory	Unknown	Client	d = 0.370 (institutional misconduct); d = 0.022 (treatment dropout)
Brocato (2004)	United States	Community mental health facility	Substance abusing offenders	140	Working Alliance Inventory	Unknown	Client	d = 0.279 (treatment dropout)
Brochu et al. (2006)	Canada	Community mental health facility	Substance abusing offenders	124	California Psychotherapy Alliance Scale	Beginning	Client	d = 0.350 (treatment dropout)
Broome et al. (1997)	United States	Community mental health facility	Substance abusing offenders	250	Therapist Competence Subscale Client Evaluation of Self and Treatment	Throughout	Client	d = 0.280 (community recidivism)
Cosgrove (2020)	United States	Community mental health facility	Juvenile (general)	654	Family Self Report	Throughout	Client	d = 0.173 (community recidivism)
D. R. DeSorcy et al. (2020)	Canada	Hospital	Sex offenders	427	Working Alliance Inventory	Middle	Client	d = 0.450 (treatment dropout); d = 0.020 (community recidivism)
Fitch (2012)	United States	Prison	Juvenile (general)	59	Working Alliance Inventory – Short	Unknown	Client	d = 1.118 (treatment dropout)
Florsheim et al. (2000)	United States	Juvenile residential treatment facility	Juvenile (general)	121	Working Alliance Inventory	Throughout	Client and treatment provider	d = 0.327 (community recidivism)
Gutierrez (2010)	Canada	Parole/probation office	General unspecified offender sample	72	Working Alliance Inventory and Dual-Role Relationship Inventory	Throughout	Observer	d = 0.712 (community recidivism)
Holmqvist et al. (2007)	Sweden	Juvenile residential treatment facility	Juvenile (general)	59	Helping Alliance Questionnaire	Unknown	Client and treatment provider	d = 0.368 (community recidivism)
Kennealy et al. (2012)	United States	Probation office	General	109	Dual-Role Relationship Inventory	Throughout	Client	d = 0.670
Litter (2005)	United States	Juvenile outpatient	Juvenile (general)	74	Vanderbilt Psycho Scales	End	Client and treatment provider	d = 1.010 (community recidivism)
Miles-McLean et al. (2019)	United States	Community mental health facility	Domestic batterers	293	Working Alliance Inventory–Short	Throughout	Client and treatment provider	d = 0.309 (community recidivism)
Polaschek and Ross (2010)	New Zealand	Prison	Violent (primarily nonsexual) offenders	70	Working Alliance Inventory–Short	Throughout	Observer, client, and treatment provider	d = 0.470 (treatment dropout)
Quick-Parikh (2024)	United States	Parole/probation office	Mentally disordered offenders	71	Working Alliance Inventory–Short	Beginning	Client	d = −0.245 (treatment dropout); d = −0.138 (community recidivism)
Savicki (2007)	United States	Prison	Juvenile (general)	114	Working Alliance Inventory Adolescents	Throughout	Client	d = 0.473 (institutional misconduct)
Skeem et al. (2007)	United States	Parole/probation office	General unspecified offender sample	90	Working Alliance Inventory and Dual-Role Relationship Inventory	Unknown	Observer, client, and treatment provider	d = 0.105 (community recidivism)
Simpson et al. (2013)	United States	Juvenile residential treatment facility	Juvenile (mental illness)	58	Working Alliance Inventory	Unknown	Client and treatment provider	d = 0.160 (institutional misconduct)
Sturm et al. (2021a, 2021b)	Netherlands	Parole/probation office	General unspecified offender sample	199	Working Alliance Inventory for Mandated Clients	Throughout	Client	d = 0.159 (community recidivism)
A. Tatman et al. (2024)	United States	Parole/probation office	General unspecified offender sample	145	Working Alliance Inventory – Short	Throughout	Client	d = 0.110 (community recidivism)
Trawver (2011)	United States	Court	Mentally disordered offenders	148	Dual-Role Relationship Inventory	Middle	Client	d = 0.866 (treatment dropout); d = −0.045 (community recidivism)

Note. Holmqvist et al. (2007) and Sturm et al. (2021a, 2021b) each represent two sets of studies with their outcomes averaged out.

Working Alliance Measurement

We observed considerable variation in study characteristics, including sample types, settings, and measurement approaches (Table 1). A total of six distinct measures of the working alliance were used, with the Working Alliance Inventory (WAI; Horvath & Greenberg, 1989) and its variants (e.g., Tatman & Love, 2010) being used the most frequently followed by the Dual-Role Relationship Inventory (DRI-R; Skeem et al., 2007). (A total of 15 measures were used when counting variations of tools separately). Most measurements were administered throughout treatment (12), with the remainder occurring at the beginning (2), middle (2), or end (1). Geographic diversity was notably limited with 14 of 21 unique studies conducted in the US, and the remainder from other Westernized countries.

Working Alliance and Correctional Outcomes: Aggregate Associations

Significant small to medium in magnitude associations were found between the working alliance and the three correctional outcomes. We report fixed effects for space considerations and given the relative instability of random effects with small k. For treatment dropout (k = 9), the effect size was Cohen’s d_FE = 0.390 (SE = 0.067), p < .001 (see Table 3 for study-level results). Heterogeneity for this outcome was not significant (Q = 13.65, I² = 41.40), suggesting consistent findings across studies. For institutional misconduct (k = 3), better working alliances were significantly inversely associated with outcomes, d_FE = 0.374 (SE = 0.124), p = .003; heterogeneity was low and non-significant (Q = 0.940, I² = 0.00; see Table 4 for Study level results). Finally, WAI total scores (k = 15) were significantly inversely associated with decreased community recidivism (d_FE = 0.196, SE = 0.027) p < .001 (see Table 6 for study-level results). The heterogeneity for this outcome was also non-significant (Q = 24.73, I² = 43.39), supporting the overall consistency of findings (see Table 5 for study level results).

Table 3.

Treatment Dropout Forest Plot.

Study name	Statistics for each study							Std diff in means and 95% CI
Study name	Std diff in means	Standard error	Variance	Lower limit	Upper limit	z-Value	p-Value	Std diff in means and 95% CI
Beyko	0.022	0.250	0.063	−0.468	0.512	0.088	.930
Brocato	0.279	0.183	0.033	−0.080	0.638	1.525	0.127
Brochu	0.350	0.134	0.018	0.087	0.613	2.612	0.009
DeSorcy	0.450	0.145	0.021	0.166	0.734	3.103	0.002
Fitch	1.118	0.740	0.548	−0.332	2.568	1.511	0.131
Florsheim	0.325	0.208	0.043	−0.083	0.733	1.563	0.118
Polaschek	0.470	0.324	0.105	−0.165	1.105	1.451	0.147
Quick-Parikh	−0.245	0.345	0.119	−0.921	0.431	−0.710	0.478
Trawver	0.866	0.189	0.036	0.496	1.236	4.582	0.000
	0.390	0.067	0.005	0.258	0.522	5.785	0.000

Table 4.

Institutional Recidivism Forest Plot.

Study name	Statistics for each study							Std diff in means and 95% CI
Study name	Std diff in means	Standard error	Variance	Lower limit	Upper limit	z-value	p-Value	Std diff in means and 95% CI
Beyko	0.370	0.222	0.049	−0.065	0.805	1.667	.096
Savicki	0.473	0.180	0.032	0.120	0.826	2.628	.009
Simpson	0.160	0.268	0.072	−0.365	0.685	0.597	.550
	0.374	0.124	0.015	0.131	0.617	3.016	.003

Table 5.

Community Recidivism Forest Plot.

Study name	Statistics for each study							Std diff in means and 95% CI
Study name	Std diff in means	Standard error	Variance	Lower limit	Upper limit	z-value	p-Value	Std diff in means and 95% CI
Broome	0.280	0.062	0.004	0.158	0.402	4.516	.000
Cosgrove	0.173	0.038	0.001	0.099	0.247	4.553	.000
DeSorcy	0.020	0.100	0.010	−0.176	0.216	0.200	.841
Florsheim	0.327	0.182	0.033	−0.030	0.684	1.797	.072
Gutierrez	0.712	0.348	0.121	0.030	1.394	2.046	.041
Holmqvist S1	0.386	0.268	0.072	−0.139	0.911	1.440	.150
Holmqvist S2	0.350	0.388	0.151	−0.410	1.110	0.902	.367
Kennealy	0.670	0.199	0.040	0.280	1.060	3.367	.001
Litter	1.010	0.314	0.099	0.395	1.625	3.217	.001
Miles-McLean	0.309	0.258	0.067	−0.197	0.815	1.198	.231
Quick-Parikh	−0.138	0.314	0.099	−0.753	0.477	−0.439	.660
Skeem	0.105	0.112	0.013	−0.115	0.325	0.938	.349
Strum	0.159	0.105	0.011	−0.047	0.365	1.514	.130
Tatman	0.110	0.169	0.029	−0.221	0.441	0.651	.515
Trawver	−0.045	0.212	0.045	−0.461	0.371	−0.212	.832
	0.196	0.027	0.001	0.143	0.248	7.327	.000

Table 6.

Working Alliance Measure and Moderator Associations With Correctional Treatment Outcomes.

Measure/moderator	Outcome	Random effects					Fixed effects				Q	I ²
Measure/moderator	Outcome	k	d	SE	95% CI	p	d	SE	95% CI	p	Q	I ²
All measures
Total score	TDO	9	0.377	0.095	0.190, 0.563	<.001	0.390	0.067	0.258, 0.522	<.001	13.65	41.40
Total score	IR	3	0.374	0.124	0.131, 0.617	.003	0.374	0.124	0.131, 0.617	.003	0.94	0.00
Total score	CR	15	0.216	0.048	0.122, 0.311	<.001	0.196	0.027	0.143, 0.248	<.001	24.73	43.39
Version
Client	CR	14	0.221	0.056	0.111, 0.331	<.001	0.199	0.027	0.146, 0.251	<.001	30.93**	57.97
SPO	CR	6	0.272	0.153	−0.028, 0.572	.075	0.155	0.080	−0.002, 0.312	.053	16.31*	63.25
Measure type
WAI
Total	CR	7	0.138	0.061	0.018, 0.258	.024	0.135	0.059	0.020, 0.250	.022	6.27	4.29
Task	CR	4	0.070	0.083	−0.092, 0.233	.397	0.065	0.144	−0.222, 0.342	.397	1.86	0.00
Bond	CR	4	0.161	0.137	−0.108, 0.429	.241	0.092	0.136	−0.017, 0.515	.260	5.70	47.36
Goal	CR	4	0.065	0.083	−0.097, 0.227	.433	0.070	0.133	−0.132, 0.389	.433	2.16	0.00
DRI-R
Total	CR	4	0.314	0.181	−0.041, 0.669	.083	0.224	0.086	0.056, 0.392	.009	9.89*	69.68
Developmental group
Adult	CR	11	0.187	0.063	0.064, 0.309	.003	0.196	0.039	0.119, 0.273	<.001	16.46	45.34
Juvenile	CR	4	0.354	0.127	0.106, 0.603	.005	0.196	0.036	0.125, 0.267	<.001	8.27	51.60
Setting
Community	CR	11	0.231	0.057	0.119, 0.342	<.001	0.204	0.028	0.148, 0.259	<.001	20.38*	50.93
Custody	CR	4	0.163	0.101	−0.034, 0.361	.105	0.130	0.081	−0.030, 0.289	.111	3.62	17.03
Relationship
Therapeutic	CR	8	0.235	0.092	0.116, 0.353	<.001	0.201	0.030	0.143, 0.260	<.001	13.34	47.60
CJS role	CR	7	0.193	0.092	0.013, 0.374	.036	0.173	0.061	0.054, 0.292	.004	11.20	46.43
Country
US	CR	10	0.236	0.060	0.118, 0.354	<.001	0.207	0.029	0.150, 0.264	<.001	18.51*	51.38
Other	CR	5	0.168	0.089	−0.007, 0.343	.060	0.137	0.067	0.004, 0.269	.043	5.31	24.64
Offense group
General	CR	11	0.252	0.065	0.125, 0.379	<.001	0.193	0.031	0.132, 0.255	<.001	18.47*	45.84
Specific	CR	4	0.157	0.099	−0.036, 0.350	.111	0.203	0.051	0.103, 0.302	<.001	6.24	51.92

Note. Significant p-values in bold font. Outcome: TDO = treatment dropout; I = institutional recidivism; R = community recidivism; SPO = service provider (therapist, probation officer) or observer. WAI = Working Alliance Inventory; DRI-R = Dual-Role Relationship Inventory. CJS role = criminal justice/supervision role.

Moderators of Working Alliance Associations With Community Recidivism

The data catchment enabled the development of several key moderators of working alliance-community recidivism associations (Table 6). Among measure type, 7 studies examined variations of the WAI (total score) with recidivism, and four examined its subscales. The WAI total score (d_FE = 0.135, SE = 0.059, p = .022) predicted community recidivism, however, none of the subscales did; however, DRI-R total score was significantly inversely associated with decreased recidivism with a small effect across four studies (d_FE = 0.224, SE = 0.086, p = .009). In terms of administration modality, client rated measures of the alliance had small significant associations with decreased community recidivism (d_FE = 0.199, SE = 0.027, p < .001), while service provider or observer (SPO) ratings had small non-significant effects (d_FE = 0.155, SE = .080, p = .053).

Moderator analyses based on sample composition, setting, service provider, and jurisdiction also yielded illuminating findings; not uncommonly, fixed effects analyses yielded generally identical findings among moderators, while some differential effects (non-significant) were observed for random effects analyses. For instance, juvenile and adult studies generally yielded identical working alliance associations with recidivism (d_FE = 0.196, both subgroups, p < .001), although random effects demonstrated larger associations in youth samples. Community settings also yielded stable, significant effects for working alliance associations with decreased recidivism (d_FE = 0.204, SE = .028, p < .001), and were higher than those observed for custodial settings (d_FE = 0.130, SE = .081, p = .111).

The type of client-service provider relationship demonstrated significant effects between alliance and decreased recidivism, whether the relationship was client and therapist (d_FE = 0.201, SE = .030, p = < .001) or a criminal justice/supervision role (e.g., probation officer, case manager, therapeutic court judge; d_FE = 0.173, SE = .061, p = .004), although the association was slightly higher for the former. Further, generally identical small significant effects were observed between general correctional samples (d_FE = 0.193, SE = .031, p < .001) and those with a specific offending pattern (d_FE = 0.203, SE = .051, p < .001); although again, greater disparities were observed with random effects analyses. Finally, two-thirds of studies were in US-based settings, which in turn yielded the strongest working alliance associations with decreased recidivism (d_FE = 0.207, SE = .029, p < .001) compared to those from international jurisdictions (d_FE = 0.137, SE = .067, p = .043).

Discussion

We conducted a meta-analysis of the working alliance correctional outcome literature, which predominantly featured variations of the WAI measures. We examined associations with treatment dropout, institutional misconduct, and community recidivism. Primary themes and their implications for service delivery in correctional and forensic mental health contexts in light of study findings are discussed.

Working Alliance Measurement

The distribution of working alliance administrations, with most occurring throughout treatment, is advantageous, as repeated measurements enable a more dependable assessment of alliance development over time (Crits-Christoph et al., 2011). Accuracy in predicting treatment outcomes is further improved when client ratings are used, rather than ratings provided by observers or treatment providers (Doran, 2016), a finding observed in the current review, and every study but one (Gutierrez, 2010) used some form of client rating; although a sizeable minority examined both client and service provider or observer ratings. Thus, most studies assessed the working alliance using methods that align with best practices in the literature while also demonstrating diversity in their measurement strategies.

The limited geographic diversity and lack of cultural representation have important implications for interpreting our findings, although US studies tended to yield stronger effects, compared to other westernized non-US counties (Canada, Sweden, Netherlands). As previously discussed, relationship-building processes—and by extension, the therapeutic alliance—can vary significantly across cultural contexts (Lukowitsky et al., 2024). For instance, while Western cultures tend to value direct and explicit communication, collectivist cultures often favor more gradual, indirect approaches to establishing trust (Sanchez-Burks et al., 2003). Minority populations in Western countries can also differ in communication (Park & Kim, 2008). These cultural nuances are not captured by existing alliance measurement tools, which take a Western lens (Buono et al., 2024). As a result, they may misrepresent the formation and function of the working alliance in non-Western contexts, or in minority populations, potentially overlooking culturally distinct expressions of alliance and their unique impact on treatment outcomes.

The Working Alliance and Treatment Noncompletion

Our findings regarding alliance and treatment dropout closely align with existing literature. We observed a fixed effect of d = 0.39 (random effect, d = 0.38) which is consistent with prior meta-analyses conducted primarily in non-forensic contexts, such as d = 0.55 reported by Sharf et al. (2010) and d = 0.37 reported by Flückiger et al. (2018). Importantly, we observed a stable relationship between studies, with very low to moderate heterogeneity. Interestingly, our analysis demonstrated substantially lower heterogeneity compared to previous meta-analyses on treatment outcomes more broadly (I² = 70.8%; Flückiger et al., 2018) and treatment dropout specifically (cf. Sharf et al., 2010). This consistent and medium effect underscores the role and relevance of the working alliance in treatment noncompletion. Per the general responsivity principle, a strong working alliance, however operationalized (e.g., WAI vs. DRI-R), translates into a positive relationship characterized by warmth, shared understanding of goals and objectives, and mutual agreement on intervention or supervision related activities, all of which should increase client buy in, receptiveness to recommended interventions, and retention in services. This interpretation is supported by qualitative research on client and therapist experiences of dropout. For instance, a qualitative review found that forensic clients who drop out of treatment often report perceiving their therapist as disingenuous, punitive, incompetent, or untrustworthy (Sturgess et al., 2016). These clients frequently express concerns that the therapist will judge or expose them, fostering a deep sense of mistrust. Treatment is viewed as irrelevant to their personal goals, a waste of time, or merely a “box ticking” exercise to gain parole or early release, indicating a significant misalignment in therapeutic goals and tasks. Conversely, therapists often view clients who dropout as having unrealistic expectations and may perceive them as difficult to engage (Roos & Werbart, 2013). These patterns suggest that the decision to discontinue treatment could be strongly shaped by both clients’ and therapists’ perceptions of the relationship and their satisfaction with treatment.

The correctional environment appears to contribute to greater consistency in this relationship, however. This may be attributable to reviews of non-forensic populations encompassing a comparably broader range of treatment settings, client characteristics, and therapeutic goals, all of which can introduce variability (Flückiger et al., 2018; Sharf et al., 2010). In contrast, correctional treatment is typically delivered in structured, often mandated group settings, with a primary focus on reducing recidivism rather than a focus on a wide array of clinical outcomes (Hachtel et al., 2019; E. C. Ross et al., 2008). As such, correctional interventions could be more uniform in design and delivery, leading to more consistent effects across studies compared to general psychotherapy approaches.

The Working Alliance and Institutional Misconduct

The results for institutional misconduct had a small to medium effect with working alliance with low variability (k = 3), demonstrating a strength of association that is roughly consistent with what has been reported in the broader psychotherapy literature with outcomes (Flückiger et al., 2018). In addition to unexpectedly strong results, this finding showed substantially less variability than our other outcomes. This outcome was also defined quite differently in each study. While constructs such as treatment dropout or recidivism may vary in operationalization (e.g., rearrest vs. adjudication, time in treatment vs. premature termination), they are basically describing the same act. By contrast, definitions of institutional misconduct in this study encompass fundamentally different types of behavior such as a 4-point scale assessing disruptive behavior (Beyko & Wong, 2005), a binary variable measuring any involvement in a violent incident (Simpson et al., 2013), versus a sum score of rule violations (Savicki, 2007). We should expect more variation because prior research on treatment dropout has shown that differences in operational definitions can moderate relationships with dropout rates (Swift & Greenberg, 2012). Similarly, the variation in definitions of institutional misconduct likely contributed to greater heterogeneity in findings compared to other outcomes.

Working Alliance Community Recidivism Associations

The working alliance demonstrated a significant small effect with decreased community recidivism. This association was notably weaker than those we observed with institutional misconduct and treatment dropout rates, indicating that the influence of the working alliance on recidivism may be indirect. While it would seem plausible that greater recidivism risk could mediate this relationship, studies linking recidivism risk and working alliance report an inconsistent relationship (Beyko & Wong, 2005; B. L. Blasko & Jeglic, 2016; D. R. DeSorcy et al., 2016; A. Tatman et al., 2024). Again, per general responsivity, the working alliance-recidivism link is likely an indirect one, in which the working alliance serves as a conduit to risk reduction by way of promoting treatment buy-in, retention, and therapeutic gain, thereby reducing recidivism.

This interpretation has some mixed evidence. Researchers in the general psychotherapy literature have theorized that the working alliance may cause better outcomes through greater therapeutic progress; that is, the relationship is in and of itself healing (Crits-Christoph et al., 2006). Evidence for this perspective comes from studies indicating that increases in the strength of the working alliance temporally precede improvements in symptoms (Falkenström et al., 2016). However, there is also evidence that increases in the working alliance proceed an improvement in symptoms, potentially due to an increased bond caused by an improvement in symptoms (DeRubeis & Feeley, 1990). In correctional literature, studies assessing the temporal relationship between the working alliance and risk reduction are limited; however, existing research suggests that greater changes in working alliance over time may be associated with reduced recidivism (Aylwin, 2010; Sturm et al., 2021a), which is consistent with the general responsivity explanation advanced above that the working alliance could have an indirect effect on recidivism by enabling therapeutic progress.

Working alliance associations with community recidivism had broadly moderate heterogeneity, which may stem from its inclusion of a broader range of settings, including parole, inpatient, and outpatient populations, across different measures, developmental groups, and jurisdictions. Moderator analyses were informative in this regard; briefly, working alliance recidivism associations tended to be strongest in community settings, with juvenile and general correctional samples, completed in the form of client ratings, and in the context of a therapeutic relationship; although some significant working alliance outcome associations were observed at different levels of other moderators (e.g., adult samples, to criminal justice supervisory roles). These findings make sense conceptually; a community environment is less restrictive and closer proximity to community outcomes; clients as the recipients of services at some level know best if they have a positive working relationship with a service provider; therapists are trained in developing and managing alliances and using these as vehicle of change; and youth are in a developmentally ripe period to intervene, the power of a warm and trusting prosocial adult relationship for a youth who may struggle with relationship and attachment experiences cannot be understated. That said, we strongly argue that working alliance matters for everybody; it is a necessary, but not sufficient condition in our view to promoting treatment buy in and gain toward risk reduction and ultimately successful community reintegration.

Limitations

First, a key limitation of the present review was the limited available data from the extant literature, which placed constraints on the scope of our analyses to yield potentially stronger findings or more novel conclusions. This limitation is notable given that client and therapist variables are known to influence the working alliance (Muran et al., 2010; Watson & Kalogerakos, 2010). In forensic contexts, the determinants may differ from those observed in general clinical settings due to unique environmental and relational dynamics. Similarly, moderating variables—studied in non-correctional populations—may exert distinct effects within correctional settings. Nonetheless, we were able to extract a number of broad moderators, such as variants on the measures used and their administration, developmental group, the context and setting of services provision, and the relationship to the service provider, which help extend our knowledge about factors that could moderate working alliance associations with outcome.

Second, incomplete reporting of outcome data may have influenced findings. Several identified studies collected working alliance and outcome data but did not report their statistical association, while others presented associations in formats unsuitable for effect size calculation, resulting in their exclusion (per Figure 1). Among the included studies, many reported outcomes based on regression models that adjusted for multiple covariates, which may have obscured the direct effect sizes necessary for meta-analytic synthesis. More consistent and comprehensive reporting of outcome data would have strengthened the current meta-analysis and allowed for greater confidence in findings. Finally, the associations examined are correlational in nature; no causal inferences can be drawn about working alliance associations with outcome; we know that it matters, vis-à-vis general responsivity, but absent a control group, causal inferences between working alliance and improved correctional outcomes cannot be drawn.

Future Directions

Our analysis indicates that transparent and comprehensive statistical reporting in this field is in need of improvement; although several additional studies exist, working-outcome associations were not reported, or they were reported in a non-convertible metric, or they reported only complex multivariable associations that confounded extraction of simple zero order bivariate associations. In turn, this may have biased overall findings, as studies without significant results or findings in the expected direction could have been less likely to report their outcomes. This issue is particularly relevant given the growing demand in the social sciences on reporting all results, including non-significant results to support meta-analytic efforts (Lakens & Etz, 2017). In turn, more comprehensive statistical reporting would permit more nuanced examination of the predictors and moderators of the working alliance-outcome relationship, and deepen understanding of the challenges associated with working alliance building in a forensic context (e.g., incorporating culturally responsive approaches to alliance-building in treatment).

These issues notwithstanding, the findings underscore the validity and utility of the working alliance in forensic and correctional treatment contexts and have clinical implications. For one, given both the distinct challenges of these environments and the relationship we found with treatment outcomes, clinician training in forensic settings should place greater emphasis on the development and maintenance of a strong working alliance in clinical service delivery. Such training could place greater emphases on forensic context-specific considerations such as rapport building with mandated clients, navigating dual-role tensions, or addressing the mistrust that often characterizes correctional populations. Structured alliance training has been associated with the development of stronger alliances in the general psychotherapy literature (Eubanks-Carter et al., 2010), and by extension, in the correctional literature, interventions that target related concepts of treatment motivation and readiness in forensic populations are commonly used (Kozar, 2013). Building on this foundation, future interventions should be designed to foster strong, individualized alliances in ways that are responsive to client needs, offense-specific criminogenic needs, and the institutional setting. Addressing these challenges may enhance engagement, reduce dropout, and ultimately improve long-term outcomes for forensic clients.

Conclusions

While existing research consistently links the working alliance to treatment outcomes in general psychotherapeutic settings (Flückiger et al., 2018), the present study aimed to determine whether this relationship extends to forensic settings through meta-analysis. Our meta-analytic findings revealed significant and meaningful associations in the expected direction between the working alliance and all three outcomes examined: treatment dropout, institutional misconduct, and community recidivism. Although future research is required, these results provide evidence that the working alliance matters and has at least some bearing on treatment outcomes in forensic contexts. Clinically, this underscores the importance of fostering strong therapeutic relationships in correctional settings and has direct implications for the training and intervention strategies used by professionals working with justice-involved populations. By fostering stronger working alliances, we can move closer to achieving better outcomes for clients, safer, more rehabilitative correctional settings, and increased public safety.

Footnotes

ORCID iDs

Mark E. Olver

Keira C. Stockdale

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Effect size data available from authors on request.*

References

Aylwin

A. S.

(2010). The therapeutic alliance in sex offender treatment: The juxtaposition of violence and care. Unpublished dissertation, University of Alberta.

*Beyko

M. J.

Wong

S. C. P.

(2005). Predictors of treatment attrition as indicators for program improvement not offender shortcomings: A study of sex offender treatment attrition. Sexual Abuse A Journal of Research and Treatment, 17(4), 375–389. https://doi.org/10.1007/s11194-005-8050-8

Blasko

Serran

Abracen

(2018). The role of the therapeutic alliance in offender therapy. In Jeglic

E. L.

Calkins

(Eds.), New frontiers in offender treatment: The translation of evidence-based practices to correctional settings (pp. 87–108). Springer. https://doi.org/10.1007/978-3-030-01030-0_5

Blasko

B. L.

Jeglic

E. L.

(2016). Sexual offenders’ perceptions of the client-therapist relationship: The role of risk. Sexual Abuse, 28(4), 271–290. https://doi.org/10.1177/1079063214529802

Bonta

Andrews

D. A.

(2007). Risk-need-responsivity model for offender assessment and rehabilitation. Die Rehabilitation, 6(1), 1–22. https://www.publicsafety.gc.ca/cnt/rsrcs/pblctns/rsk-nd-rspnsvty/index-en.aspx

Bonta

Andrews

D. A.

(Eds.). (2023). The Psychology of criminal conduct (7th ed.). Routledge. https://doi.org/10.4324/9781003292128

Bonta

Blais

Wilson

H. A.

(2013). The prediction of risk for mentally disordered offenders: A quantitative synthesis. Public Safety Canada.

Bordin

E. S.

(1979). The generalizability of the psychoanalytic concept of the working alliance. Psychotherapy: Theory, Research & Practice, 16(3), 252–260. https://doi.org/10.1037/h0085885

Borenstein

Hedges

Rothstein

(2007). Meta-analysis: Fixed vs. random effects. www.meta-analysis.com

10.

Bourgon

Guiterrez

(2013). The importance of building good relationships in community corrections: Evidence, theory and practice of the therapeutic alliance. In Raynor

Ugwudike

(Eds.), What works in offender compliance: International perspectives and evidence-based practice (pp. 256–275). Palgrave Macmillan. https://doi.org/10.1057/9781137019523_15

11.

*Brocato

(2004). Predictors of client retention in alternative-to-prison substance abuse programs. Unpublished dissertation, Florida International University.

12.

*Brochu

Cournoyer

L. G.

Tremblay

Bergeron

Brunelle

Landry

(2006). Understanding treatment impact on drug-addicted offenders. Substance Use & Misuse, 41(14), 1937–1949. https://doi.org/10.1080/10826080601025995

13.

*Broome

K. M.

Knight

D. K.

Knight

Hiller

M. L.

Simpson

D. D.

(1997). Peer, family, and motivational influences on drug treatment process and recidivism for probationers. Journal of Clinical Psychology, 53(4), 387–397. https://doi.org/10.1002/(SICI)1097-4679(199706)53:4%3C387::AID-JCLP12%3E3.0.CO;2-C

14.

Brüggemann

Rajguru

(2022). Comprehensive meta-analysis (CMA) 3.0: A software review. Journal of Marketing Analytics, 10(4), 425–429. https://doi.org/10.1057/s41270-022-00184-5

15.

Buono

Sassier

Thulesius

Hoffman

Nabbe

Petek

Le Reste

J. Y.

(2024). Translation of the working alliance inventory short revised into Italian using a Delphi procedure and a forward-backward translation. Frontiers of Medicine, 10, 1236273. https://doi.org/10.3389/fmed.2023.1236273

16.

Cohen

(Ed.). (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Routledge. https://doi.org/10.4324/9780203771587

17.

*Cosgrove

J. A.

(2020). Fidelity and alliance effects on youth outcomes of functional family therapy. Unpublished dissertation, University of Maryland, Baltimore.

18.

Crits-Christoph

Gibbons

M. B.

Hamilton

Ring-Kurtz

Gallop

(2011). The dependability of alliance assessments: The alliance-outcome correlation is larger than you might think. Journal of Consulting and Clinical Psychology, 79(3), 267–278. https://doi.org/10.1037/a0023668

19.

Crits-Christoph

Gibbons

M. B. C.

Hearon

(2006). Does the alliance cause good outcome? Recommendations for future research on the alliance. Psychotherapy: Theory, Research, Practice, Training, 43(3), 280–285. https://doi.org/10.1037/0033-3204.43.3.280

20.

DeRubeis

R. J.

Feeley

(1990). Determinants of change in cognitive therapy for depression. Cognitive Therapy and Research, 14, 469–482. https://doi.org/10.1007/bf01172968

21.

DeSorcy

(2013). Working alliance and its relationship with aboriginal ancestry, psychopathy, treatment completion, and recidivism in a sample of federal sex offenders. Unpublished dissertation, University of Saskatchewan.

22.

*DeSorcy

D. R.

Olver

M. E.

Wormith

J. S.

(2016). Working alliance and its relationship with treatment outcome in a sample of aboriginal and non-aboriginal sexual offenders. Sexual Abuse A Journal of Research and Treatment, 28(4), 291–313. https://doi.org/10.1177/1079063214556360

23.

DeSorcy

D. R.

Olver

M. E.

Wormith

J. S.

(2020). Working alliance and psychopathy: Linkages to treatment outcome in a sample of treated sexual offenders. Journal of Interpersonal Violence, 35(7–8), 1739–1760. https://doi.org/10.1177/0886260517698822

24.

Doran

J. M.

(2016). The working alliance: Where have we been, where are we going? Psychotherapy Research, 26(2), 146–163. https://doi.org/10.1080/10503307.2014.954153

25.

Eubanks-Carter

Muran

J. C.

Safran

J. D.

(2010). Alliance ruptures and resolution. In Muran

J. C.

Barber

J. P.

(Eds.), The therapeutic alliance: An evidence-based guide to practice (pp. 74–95). The Guilford Press.

26.

Falkenström

Ekeblad

Holmqvist

(2016). Improvement of the working alliance in one treatment session predicts improvement of depressive symptoms by the next session. Journal of Consulting and Clinical Psychology, 84(8), 738–751. https://doi.org/10.1037/ccp0000119

27.

*Fitch

K. D. P.

(2012). The effect of self-identified stage of change and therapeutic alliance on attrition in treatment in incarcerated juvenile males. Unpublished dissertation, Fielding Graduate University.

28.

*Florsheim

Shotorbani

Guest-Warnick

Barratt

Hwang

W. C.

(2000). Role of the working alliance in the treatment of delinquent boys in community-based programs. Journal of Clinical Child Psychology, 29(1), 94–107. https://doi.org/10.1207/s15374424jccp2901_10

29.

Flückiger

Del Re

A. C.

Wampold

B. E.

Horvath

A. O.

(2018). The alliance in adult psychotherapy: A meta-analytic synthesis. Psychotherapy, 55(4), 316–340. https://doi.org/10.1037/pst0000172

30.

Gobin

R. L.

Freyd

J. J.

(2014). The impact of betrayal trauma on the tendency to trust. Psychological Trauma Theory Research Practice and Policy, 6(5), 505–511. https://doi.org/10.1037/a0032452

31.

*Gutierrez

(2010). Examining the relationship factor in a criminal justice setting: Therapeutic alliance, the dual-role and principles of effective correctional counselling. Unpublished master’s thesis, Carleton University.

32.

Hachtel

Vogel

Huber

C. G.

(2019). Mandated treatment and its impact on therapeutic process and outcome factors. Frontiers in Psychiatry, 10, 219. https://doi.org/10.3389/fpsyt.2019.00219

33.

Hatcher

R. L.

(2010). Alliance theory and measurement. In Muran

J. C.

Barber

J. P.

(Eds.), The therapeutic alliance: An evidence-based guide to practice (pp. 7–28). Guilford.

34.

Higgins

J. P. T.

Thompson

S. G.

(2002). Quantifying heterogeneity in a meta-analysis. Statistics in Medicine, 21(11), 1539–1558. https://doi.org/10.1002/sim.1186

35.

*Holmqvist

Hill

Lang

(2007). Treatment alliance in residential treatment of criminal adolescents. Child & Youth Care Forum, 36, 163–178. https://doi.org/10.1007/s10566-007-9037-z

36.

Horvath

A. O.

Greenberg

L. S.

(1989). Development and validation of the Working Alliance Inventory. Journal of Counseling Psychology, 36(2), 223–233. https://doi.org/10.1037/0022-0167.36.2.223

37.

*Kennealy

P. J.

Skeem

J. L.

Manchak

S. M.

Eno Louden

(2012). Firm, fair, and caring officer-offender relationships protect against supervision failure. Law and Human Behavior, 36(6), 496–505. https://doi.org/10.1037/h0093935

38.

Kozar

(2013). Treatment readiness and the therapeutic alliance. In Day

Casey

Ward

Howells

Vess

(Eds.), Transitions to better lives offender readiness and rehabilitation (pp. 195–213). Willan. https://doi.org/10.4324/9781843927204

39.

Lakens

Etz

A. J.

(2017). Too true to be bad: When sets of studies with significant and nonsignificant findings are probably true. Social Psychological and Personality Science, 8(8), 875–881. https://doi.org/10.1177/1948550617693058

40.

*Litter

(2005). Relationship-based psychotherapy with court-involved youth: The therapy relationship’s effect on outcome. Unpublished dissertation, The Wright Institute.

41.

Lukowitsky

M. R.

Winseman

J. S.

Feld

Kam

S. M.

Balkoski

V. I.

(2024). Patients’ perceptions of their therapist: Effects of race, culture, and cultural competency on the working alliance. American Journal of Psychotherapy, 77(4), 154–159. https://doi.org/10.1176/appi.psychotherapy.20230056

42.

Messer

S. B.

Wampold

B. E.

(2002). Let's face facts: Common factors are more potent than specific therapy ingredients. Clinical Psychology Science and Practice, 9(1), 21–25. https://doi.org/10.1093/clipsy.9.1.21

43.

*Miles-McLean

H. A.

LaMotte

A. D.

Semiatin

J. N.

Farzan-Kashani

Torres

Poole

G. M.

Murphy

C. M.

(2019). PTSD as a predictor of treatment engagement and recidivism in partner abusive men. Psychology of Violence, 9(1), 39–47. https://doi.org/10.1037/vio0000161

44.

Muran

J. C.

Safran

J. D.

Eubanks-Carter

(2010). Developing therapist abilities to negotiate alliance ruptures. In Muran

J. C.

Barber

J. P.

(Eds.), The therapeutic alliance: An evidence-based guide to practice (pp. 191–209). Guilford.

45.

Ogloff

J. R. P.

Wong

Greenwood

(1990). Treating criminal psychopaths in a therapeutic community program. Behavioral Sciences & the Law, 8(2), 181–190. https://doi.org/10.1002/bsl.2370080210

46.

Olver

M. E.

Stockdale

K. C.

Wormith

J. S.

(2011). A meta-analysis of predictors of offender treatment attrition and its relationship to recidivism. Journal of Consulting and Clinical Psychology, 79(1), 6–21. https://doi.org/10.1037/a0022200

47.

Olver

M. E.

Wong

(2011). Predictors of sex offender treatment dropout: Psychopathy, sex offender risk and responsivity implications. Psychology Crime and Law, 17(5), 457–471. https://doi.org/10.1080/10683160903318876

48.

Page

M. J.

McKenzie

J. E.

Bossuyt

P. M.

Boutron

Hoffmann

T. C.

Mulrow

C. D.

Shamseer

Tetzlaff

J. M.

Akl

E. A.

Brennan

S. E.

Chou

Glanville

Grimshaw

J. M.

Hróbjartsson

Lalu

M. M.

Loder

E. W.

Mayo-Wilson

McDonald

. . . Moher

(2021). The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ, 10(1), 89. https://doi.org/10.1136/bmj.n71

49.

Park

Y. S.

Kim

B. S.

(2008). Asian and European American cultural values and communication styles among Asian American and European American college students. Cultural Diversity & Ethnic Minority Psychology, 14(1), 47. https://doi.org/10.1037/1099-9809.14.1.47

50.

*Polaschek

D. L.

Ross

E. C.

(2010). Do early therapeutic alliance, motivation, and stages of change predict therapy change for high-risk, psychopathic violent prisoners? Criminal Behaviour and Mental Health, 20(2), 100–111. https://doi.org/10.1002/cbm.759

51.

*Quick-Parikh

(2024). The impact of therapeutic alliance and perceived respect on recidivism in an alternative to incarceration program. Unpublished dissertation, Fordham University.

52.

Roos

Werbart

(2013). Therapist and relationship factors influencing dropout from individual psychotherapy: A literature review. Psychotherapy Research, 23(4), 394–418. https://doi.org/10.1080/10503307.2013.775528

53.

Ross

E. C.

Polaschek

D. L. L.

Ward

(2008). The therapeutic alliance: A theoretical revision for offender rehabilitation. Aggression and Violent Behavior, 13(6), 462–480. https://doi.org/10.1016/j.avb.2008.07.003

54.

Ross

Pfäfflin

(2007). Attachment and interpersonal problems in a prison environment. Journal of Forensic Psychiatry & Psychology, 18(1), 90–98. https://doi.org/10.1080/14789940601063345

55.

Sanchez-Burks

Lee

Choi

Nisbett

Zhao

Koo

(2003). Conversing across cultures: East-Wes t communication styles in work and nonwork contexts. Journal of Personality and Social Psychology, 85(2), 363–372. https://doi.org/10.1037/0022-3514.85.2.363

56.

*Savicki

K. L.

(2007). Working alliance, readiness for change, and theory of change as predictors of treatment success among incarcerated adolescents. Unpublished dissertation, University of Texas at Austin.

57.

Sharf

Primavera

L. H.

Diener

M. J.

(2010). Dropout and therapeutic alliance: A meta-analysis of adult individual psychotherapy. Psychotherapy: Theory, Research, Practice, Training, 47(4), 637–645. https://doi.org/10.1037/a0021175

58.

*Simpson

Frick

Kahn

Evans

(2013). Therapeutic alliance in justice-involved adolescents undergoing mental health treatment: The role of callous-unemotional traits. International Journal of Forensic Mental Health, 12(2), 83–92. https://doi.org/10.1080/14999013.2013.787559

59.

*Skeem

J. L.

Louden

J. E.

Polaschek

Camp

(2007). Assessing relationship quality in mandated community treatment: Blending care with control. Psychological Assessment, 19(4), 397–410. https://doi.org/10.1037/1040-3590.19.4.397

60.

Sturgess

Woodhams

Tonkin

(2016). Treatment engagement from the perspective of the offender: Reasons for noncompletion and completion of treatment—A systematic review. International Journal of Offender Therapy and Comparative Criminology, 60(16), 1873–1896. https://doi.org/10.1177/0306624x15586038

61.

*Sturm

de Vogel

Menger

Huibers

M. J. H.

(2021a). Changes in offender-rated working alliance in probation supervision as predictors of recidivism. Psychology Crime and Law, 27(2), 182–200. https://doi.org/10.1080/1068316x.2020.1793986

62.

*Sturm

de Vogel

Menger

Huibers

M. J. H.

(2021b). The price of mistrust: A study into the working alliance as predictor for recidivism. Journal of Police and Criminal Psychology, 37(3), 576–586. https://doi.org/10.1007/s11896-021-09450-1

63.

Swift

J. K.

Greenberg

R. P.

(2012). Premature discontinuation in adult psychotherapy: A meta-analysis. Journal of Consulting and Clinical Psychology, 80(4), 547–559. https://doi.org/10.1037/a0028226

64.

*Tatman

Butler

H. D.

Zenz

(2024). Working alliance, risk, and recidivism: A community-based corrections analysis. Criminal Justice and Behavior, 51(6), 934–948. https://doi.org/10.1177/00938548241234366

65.

Tatman

A. W.

Love

K. M.

(2010). An offender version of the Working Alliance Inventory-Short Revised. Journal of Offender Rehabilitation, 49(3), 165–179. https://doi.org/10.1080/10509671003666560

66.

*Trawver

K. R.

(2011). Who succeeds in mental health courts? Identifying predictors related to program retention and legal recidivism. Unpublished dissertation, University of Texas at Austin.

67.

Walton

C. J.

Grenyer

B. F. S.

(2002). Prioritizing access to psychotherapy services: The Client Priority Rating Scale. Clinical Psychology & Psychotherapy: An International Journal of Theory & Practice, 9(6), 418–429. https://doi.org/10.1002/cpp.346

68.

Ward

(2013). Addressing the dual relationship problem in forensic and correctional practice. Aggression and Violent Behavior, 18(1), 92–100. https://doi.org/10.1016/j.avb.2012.10.006

69.

Watson

J. C.

Kalogerakos

(2010). The therapeutic alliance in humanistic psychotherapy. In Muran

J. C.

Barber

J. P.

(Eds.), The therapeutic alliance: An evidence-based guide to practice (pp. 320–340). Guilford.

70.

Youssef

(2017). The importance of the therapeutic alliance when working with men who have committed a sexual offence. Journal of Criminal Psychology, 7(3), 206–220. https://doi.org/10.1108/jcp-09-2016-0027