Item 19 (“At times the therapist seems distant”)<\/em> captures clinician detachment, coldness, or emotional unavailability.<\/li>\n<\/ul>\n6. Theoretical Framework<\/h2>\n
The Helping Alliance Questionnaire is anchored at the intersection of psychodynamic relationship theory, humanistic philosophy, and empirical common factors research.<\/p>\n
Psychoanalytic Foundations: From Freud to Greenson and Zetzel<\/h3>\n
Sigmund Freud (1912, 1913) initially acknowledged that before deep psychoanalytic interpretation could be tolerated, the analyst had to establish an affectionate, collaborative attachment with the patient\u2014a rapport rooted in reality rather than neurotic distortion. Decades later, ego psychologists such as Elizabeth Zetzel (1956) and Ralph Greenson (1965, 1967) formalized this concept into the “therapeutic alliance”<\/em> and “working alliance.”<\/em> Greenson argued that the patient\u2019s capacity to align their observing ego with the analyst\u2019s analyzing ego against their instinctual conflicts and resistances was the sine qua non of psychoanalytic cure. Luborsky expanded this clinical observation into an empirical paradigm through the Penn Psychotherapy Project, developing observational and self-report scales to quantify these psychoanalytic constructs systematically.<\/p>\nBordin\u2019s Pantheoretical Working Alliance Model<\/h3>\n
In 1979, Edward Bordin reformulated the working alliance as a pan-theoretical construct applicable not merely to psychoanalysis, but to all psychotherapeutic modalities (including behavioral, gestalt, and family therapies). Bordin posited that the alliance consists of three interdependent constituents:<\/p>\n
\n- Bonds:<\/strong> The interpersonal attachments, mutual trust, and positive affective regard linking client and therapist.<\/li>\n
- Tasks:<\/strong> Mutual agreement upon and engagement in the specific in-session activities, exercises, and behaviors that form the substance of the therapeutic work.<\/li>\n
- Goals:<\/strong> Shared consensus regarding the broad targets and intended outcomes of the therapeutic intervention.<\/li>\n<\/ol>\n
Although Luborsky\u2019s HAq-II maintains its distinct historical nomenclature (Type 1 Supportive vs. Type 2 Collaborative), its empirical item content maps directly onto Bordin\u2019s tripartite taxonomy. Items assessing interpersonal liking and dependability reflect Bonds<\/em>; items assessing procedural suitability and meaningful exchanges capture Tasks<\/em>; and items assessing agreement on problems and desired ends correspond directly to Goals<\/em>.<\/p>\nCommon Factors and the Contextual Model<\/h3>\n
In contemporary psychotherapy research, the theoretical framing of the HAq-II is reinforced by the Contextual Model<\/em> formulated by Bruce Wampold (2001, 2015). Drawing on the classic anthropological and psychological work of Jerome Frank (Persuasion and Healing<\/em>, 1961), the Contextual Model asserts that the specific technical ingredients of various psychotherapies account for a relatively small proportion of outcome variance, whereas common relational factors\u2014foremost among them the therapeutic alliance\u2014account for a substantially larger proportion. The HAq-II serves as the operational instrument verifying that an emotionally charged, confiding relationship, situated within a culturally sanctioned healing context, has been established.<\/p>\n7. Validity<\/h2>\n
The psychometric integrity of the Helping Alliance Questionnaire (specifically the HAq-II) has been extensively corroborated across diverse clinical populations, diagnostic categories, and research protocols (Luborsky et al., 1996; Barber et al., 1999, 2000).<\/p>\n
Construct and Factorial Validity<\/h3>\n
In the landmark scale revision study conducted by Luborsky, Barber, Siqueland, Johnson, Najavits, Frank, and Daley (1996), the HAq-II was administered to substantial cohorts of psychiatric outpatients, including patients participating in the National Institute on Drug Abuse (NIDA) Collaborative Cocaine Treatment Study ($N = 200$<\/span>) and depressed outpatients ($N = 82$<\/span>). Confirmatory and exploratory analyses supported a predominantly strong general alliance dimension that accounted for the vast majority of shared variance, demonstrating that the 19 items coherently reflect a unified underlying construct of relational collaboration and safety.<\/p>\nConvergent Validity<\/h3>\n
Convergent validity has been established through moderate-to-high correlations between the HAq-II and other gold-standard alliance measures:<\/p>\n
\n- Working Alliance Inventory (WAI):<\/strong> Correlations between patient-rated HAq-II and the WAI Total Score typically range between $r = .72$<\/span> and $r = .86$<\/span> ($p < .001$<\/span>), demonstrating exceptional alignment in assessing the relational bond, task alignment, and goal consensus.<\/li>\n
- California Psychotherapy Alliance Scales (CALPAS):<\/strong> Correlations between the HAq-II and CALPAS global scores range from $r = .68$<\/span> to $r = .80$<\/span>.<\/li>\n
- Penn Helping Alliance Rating Scale (Penn Method – Observer Rated):<\/strong> Significant correlations ($r = .45$<\/span> to $.60$<\/span>) have been demonstrated between patients\u2019 subjective HAq-II self-reports and external, blinded observer ratings of alliance derived from audio\/video recordings of clinical sessions.<\/li>\n<\/ul>\n
Discriminant and Divergent Validity<\/h3>\n
To demonstrate that the HAq-II does not simply measure generalized positive affectivity, neuroticism, or baseline symptom severity, researchers evaluated its discriminant validity:<\/p>\n
\n- Correlations with baseline intake symptom measures\u2014such as the Beck Depression Inventory (BDI), the Symptom Checklist-90-Revised (SCL-90-R), and the Hamilton Rating Scale for Depression (HRSD)\u2014are characteristically low ($r = -.10$<\/span> to $-.22$<\/span>, often non-significant), demonstrating that the HAq-II captures interpersonal process rather than initial psychiatric distress.<\/li>\n
- Correlations with social desirability indices (such as the Marlowe-Crowne Social Desirability Scale) remain low ($r < .20$<\/span>), indicating that high alliance scores are not merely artifacts of impression management.<\/li>\n<\/ul>\n
Predictive and Criterion Validity<\/h3>\n
The hallmark of the HAq-II is its robust predictive validity regarding psychotherapy outcomes:<\/p>\n
\n- In the NIDA Collaborative Cocaine Treatment Study, early-treatment HAq-II scores (measured at Session 2 and Session 5) significantly predicted subsequent drug abstinence, reductions in craving, and treatment retention ($p < .01$<\/span>).<\/li>\n
- Meta-analytic syntheses (e.g., Horvath & Symonds, 1991; Martin, Garske, & Davis, 2000; Fl\u00fcckiger et al., 2018) encompassing HAq instruments establish an overall average alliance-outcome correlation of $r = .26$<\/span> to $.28$<\/span>, which translates to roughly 7% to 8% of the total variance in clinical outcome directly attributable to the alliance\u2014an effect size rivaling or exceeding that of specific treatment modalities.<\/li>\n<\/ul>\n
8. Reliability<\/h2>\n
The Helping Alliance Questionnaire demonstrates high empirical reliability across multiple operational parameters, satisfying the psychometric standards required for both research investigations and individualized clinical assessment.<\/p>\n
Internal Consistency<\/h3>\n
In the seminal psychometric re-evaluation by Luborsky et al. (1996), the HAq-II displayed outstanding internal consistency:<\/p>\n
\n- Patient Version:<\/strong> Cronbach\u2019s alpha ($\u0007lpha$<\/span>) coefficients have been consistently reported between .90 and .93<\/strong> across diverse clinical samples (e.g., major depressive disorder, generalized anxiety disorder, substance dependence).<\/li>\n
- Therapist Version:<\/strong> Cronbach\u2019s alpha coefficients for the therapist parallel form range between .88 and .91<\/strong>.<\/li>\n
- Item-Total Correlations:<\/strong> In the revised 19-item version, Luborsky and colleagues eliminated items from the original 24-item HAq-I that exhibited low item-total correlations. All remaining 19 items on the HAq-II exhibit corrected item-total correlations exceeding $r = .40$<\/span>, with the majority exceeding $r = .55$<\/span>.<\/li>\n<\/ul>\n
Test-Retest Reliability and Stability<\/h3>\n
Assessing the temporal stability of the therapeutic alliance requires careful psychometric calibration, as the alliance is theoretically expected to fluctuate in response to clinical events, session content, and developmental phases. When measured across brief intervals where treatment conditions remain stable (e.g., between Session 2 and Session 3):<\/p>\n
\n- Test-retest reliability coefficients for the HAq-II typically fall in the range of $r = .75$<\/span> to $.82$<\/span><\/strong> ($p < .001$<\/span>).<\/li>\n
- Over longer intervals (e.g., between early treatment at Session 3 and late treatment at Session 20), stability coefficients diminish moderately ($r = .50$<\/span> to $.65$<\/span>), reflecting authentic dynamic shifts, relational deepening, or mid-treatment alliance ruptures and repairs.<\/li>\n<\/ul>\n
Inter-Rater Reliability (Observer Contexts)<\/h3>\n
While the HAq-II is designed primarily as a client- and therapist-completed self-report measure, its precursor and companion observational systems (the Penn Helping Alliance Rating Scale, Penn Global Rating Scale) demonstrate intraclass correlation coefficients (ICC) ranging between .78 and .88<\/strong> among trained clinical raters evaluating videotaped sessions.<\/p>\n9. Factor Analysis<\/h2>\n
The structural dimensionality of the Helping Alliance Questionnaire has been the focus of extensive psychometric exploration using both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA).<\/p>\n
Exploratory Factor Analysis (Luborsky et al., 1996)<\/h3>\n
During the development and revision of the HAq-II from the original 24-item HAq-I, Luborsky and his team conducted principal components analyses (PCA) with both orthogonal (Varimax) and oblique (Promax) rotations:<\/p>\n
\n- The scree plot and eigenvalue examination ($\\lambda_1 > 8.0$<\/span>) revealed a massive first unrotated factor<\/strong> accounting for approximately 45% to 52% of the total variance across clinical samples.<\/li>\n
- All 19 items loaded substantially on this primary factor (factor loadings ranging from $.45$<\/span> to $.82$<\/span>).<\/li>\n
- Although Luborsky historically distinguished Type 1 (supportive) from Type 2 (collaborative) alliance at a conceptual level, EFA results demonstrated that in self-report questionnaires, patients do not sharply dissociate feeling supported from working collaboratively. Instead, the patient\u2019s experience forms a cohesive, unitary Gestalt of relational safety and cooperative striving.<\/li>\n<\/ul>\n
Subsequent Confirmatory Factor Analyses and Structural Modeling<\/h3>\n
Subsequent psychometric examinations (e.g., Fenton et al., 2001; Crits-Christoph et al., 2011) have evaluated bifactor and multidimensional models of the HAq-II:<\/p>\n
\n- Two-Factor Oblique Model:<\/strong> Several investigators have identified a robust two-factor solution distinguishing Positive Alliance \/ Collaboration<\/em> (comprising positively worded items such as 1, 2, 3, 5, 6, 7, 9, 10, 12, 13, 14, 15, 17, 18) and Negative Alliance \/ Relationship Strain<\/em> (comprising the reverse-scored items: 4, 8, 11, 16, 19). Model fit indices for this two-factor correlated structure frequently outperform the strict single-factor model: Comparative Fit Index ($\text{CFI}) ge .94$, Tucker-Lewis Index ($<\/span>\text{TLI}) ge .93$, and Root Mean Square Error of Approximation ($<\/span>\text{RMSEA}) le .058$.<\/li>\n
- Bifactor Model:<\/strong> A bifactor configuration\u2014consisting of a general Helping Alliance factor ($G$<\/span>) capturing universal relationship quality, alongside orthogonal group factors representing specific collaborative work and negative process\u2014yields an exceptional fit ($\text{CFI} = .97<\/span>$,$<\/span>\text{RMSEA} = .044<\/span>$), supporting the reporting of a single composite total score while retaining clinical utility for examining negative strain items separately.<\/li>\n<\/ul>\n
10. Instrument \/ Measurement Tool<\/h2>\n\n- Instrument Name:<\/strong> The Helping Alliance Questionnaire (Revised Version: HAq-II).<\/li>\n
- Instrument Type:<\/strong> Standardized psychological self-report inventory; dyadic psychotherapy process measure.<\/li>\n
- Versions Available:<\/strong>\n
\n- Patient Version:<\/em> Completed by the individual undergoing psychotherapy to evaluate their subjective experience of the therapist and collaborative process.<\/li>\n
- Therapist Version:<\/em> Completed by the treating clinician to rate their perception of the client\u2019s engagement, shared understanding, and dyadic bond.<\/li>\n<\/ul>\n<\/li>\n
- Item Count:<\/strong> 19 items (identical core semantic content across both parallel forms, adjusted for grammatical perspective).<\/li>\n
- Administration Format:<\/strong> Paper-and-pencil, computerized assessment system, or integrated electronic health record (EHR) rating portal.<\/li>\n
- Administration Time:<\/strong> Approximately 3 to 5 minutes.<\/li>\n
- Target Population:<\/strong> Adolescent and adult psychotherapy outpatients and inpatients across all diagnostic categories; suitable for individual, couples, and group psychotherapeutic interventions.<\/li>\n
- Response Format:<\/strong> 6-point Likert rating scale:\n
\n- 1<\/strong> = Strongly disagree<\/span><\/li>\n
- 2<\/strong> = Disagree<\/span><\/li>\n
- 3<\/strong> = Slightly disagree<\/span><\/li>\n
- 4<\/strong> = Slightly agree<\/span><\/li>\n
- 5<\/strong> = Agree<\/span><\/li>\n
- 6<\/strong> = Strongly agree<\/span><\/li>\n<\/ul>\n<\/li>\n
- Scoring Methodology:<\/strong>\n
\n- Positively Worded Items (14 items):<\/strong> Items 1, 2, 3, 5, 6, 7, 9, 10, 12, 13, 14, 15, 17, and 18 are scored directly as marked (1 to 6).<\/li>\n
- Negatively Worded \/ Reverse-Scored Items (5 items):<\/strong> Items 4, 8, 11, 16, and 19 reflect therapeutic discord, distrust, or procedural dissatisfaction and must be reverse-coded prior to computing composite totals:
\n
$$6
\nightarrow 1,\\quad 5
\nightarrow 2,\\quad 4
\nightarrow 3,\\quad 3
\nightarrow 4,\\quad 2
\nightarrow 5,\\quad 1
\nightarrow 6$$<\/span>
\n
(Formula: $\\text{Reversed Score} = 7 – \\text{Raw Response}$<\/span>).<\/li>\n- Total Alliance Score:<\/strong> The grand score can be expressed as either:\n
\n- Summed Score:<\/em> The sum of all 19 items (ranging from a minimum of 19 to a maximum of 114 points). Higher scores signify stronger, more collaborative helping alliances.<\/li>\n
- Mean Item Score:<\/em> The total summed score divided by 19 (ranging from 1.00 to 6.00), which retains the interpretable metric of the original Likert response scale.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<\/li>\n
- Clinical Interpretation Guidelines:<\/strong>\n
\n- Mean Score $ge 5.0$<\/span> (Sum $ge 95$<\/span>):<\/em> Strong to exceptionally high alliance. Reflects deep mutual trust, clear procedural consensus, and warm emotional bonding.<\/li>\n
- Mean Score $4.0 – 4.9$ (Sum $76 – 94$):<\/em> Moderate alliance. Represents adequate working collaboration but may contain areas of minor friction or ambivalent engagement.<\/li>\n
- Mean Score $< 4.0$<\/span> (Sum