Agnew Relationship Measure – 5 (ARM-5)

Abstract

The Agnew Relationship Measure – 5 (ARM-5) is an ultra-brief, standardized self-report instrument developed to evaluate the quality of the therapeutic alliance in individual psychotherapy and psychological counseling. Constructed by Jane Cahill, William B. Stiles, Michael Barkham, Gillian E. Hardy, Gregory Stone, Roxane Agnew-Davies, and Gisela Unsworth (2012) as an abbreviated derivative of the comprehensive 28-item Agnew Relationship Measure (ARM), the ARM-5 addresses the clinical and operational demand for high-frequency, low-burden assessment within Routine Outcome Monitoring (ROM) and Feedback-Informed Treatment (FIT) paradigms. The instrument comprises five items mapped conceptually onto three foundational domains of the therapeutic relationship: Bond (item 1), Partnership (items 2 and 3), and Confidence in therapy (items 4 and 5). Responses are captured using a 7-point Likert-type scale ranging from 1 (“Strongly Disagree”) to 7 (“Strongly Agree”), yielding a cumulative total score between 5 and 35, or a session mean score ranging from 1 to 7. Psychometric evaluations of the ARM-5 demonstrate strong internal consistency (Cronbach’s alpha typically ranging between .80 and .87), exceptional alternative-forms reliability with the parent ARM-28 Core Alliance index (r ≥ .90), and robust predictive validity concerning symptom improvement and treatment retention. Exploratory and confirmatory factor analyses demonstrate that while the three theoretical sub-dimensions are conceptually distinct, the scale functions predominantly as a robust unidimensional core alliance measure. Clinicians and researchers must account for pronounced ceiling effects and negatively skewed response distributions characteristic of client-rated alliance instruments. This comprehensive review examines the theoretical heritage, structural validity, psychometric properties, scoring procedures, clinical utility, and implementation parameters of the ARM-5.

Keywords

Agnew Relationship Measure, ARM-5, therapeutic alliance, working alliance, Feedback-Informed Treatment, Routine Outcome Monitoring, psychometrics, psychotherapy research, clinical assessment, alliance ruptures, treatment outcomes, patient-reported experience measures

Authors

The ARM-5 was developed and psychometrically validated by a collaborative team of leading psychotherapy researchers from the United Kingdom and the United States:

• Jane Cahill, Ph.D. – Psychological Therapies Research Centre (PTRC), University of Leeds, Leeds, United Kingdom.
• William B. Stiles, Ph.D. – Department of Psychology, Miami University, Oxford, Ohio, United States; and Honorary Senior Research Fellow, University of Leeds.
• Michael Barkham, Ph.D. – Centre for Psychological Services Research (CPSR), Department of Psychology, University of Sheffield, Sheffield, United Kingdom.
• Gillian E. Hardy, Ph.D. – Department of Psychology, University of Sheffield, Sheffield, United Kingdom.
• Gregory Stone, M.Sc. – Psychological Therapies Research Centre, University of Leeds, United Kingdom.
• Roxane Agnew-Davies, Ph.D. – Original architect of the parent Agnew Relationship Measure (Agnew-Davies et al., 1998); clinical psychologist and research fellow associated with the Sheffield and Leeds psychotherapy research initiatives.
• Gisela Unsworth, Ph.D. – Psychological Therapies Research Centre, University of Leeds, United Kingdom.

Purpose

The primary purpose of the Agnew Relationship Measure – 5 (ARM-5) is to provide an empirically robust, psychometrically sound, and exceptionally brief metric for evaluating the quality of the client-therapist therapeutic alliance at the session level. The therapeutic alliance has consistently emerged as one of the most robust, cross-theoretical predictors of successful psychotherapy outcomes across diverse clinical populations, diagnostic presentations, and therapeutic modalities. Despite its recognized clinical value, the routine, session-by-session assessment of the alliance has historically been constrained by the administrative burden associated with legacy scales. Comprehensive instruments such as the 28-item Agnew Relationship Measure (ARM; Agnew-Davies et al., 1998), the 36-item Working Alliance Inventory (WAI), or the 24-item California Psychotherapy Alliance Scales (CALPAS) are often too extensive for routine clinical practice.

The ARM-5 was engineered specifically to overcome respondent fatigue, clinical friction, and logistical barriers inherent in longer inventories. Designed to be completed by the patient in under one to two minutes at the conclusion of a therapeutic encounter, the scale facilitates real-time tracking across consecutive sessions (e.g., every session, every second session, or every third session). This session-by-session implementation serves several distinct clinical and research functions:

• Early Detection of Alliance Deterioration: Alliance drops or sustained low ratings frequently precede premature client-initiated treatment termination (dropout). The ARM-5 serves as an early-warning signal, alerting practitioners to relational strain before it leads to disengagement.
• Facilitation of Rupture-Repair Sequences: Intersubjective strains, miscommunications, or disagreements regarding therapeutic tasks—collectively known as alliance ruptures—are pervasive in clinical practice. The ARM-5 offers an objective platform enabling clinicians to identify latent, unvoiced dissatisfaction, explore client experiences, and initiate collaborative repair sequences, processes empirically tied to enhanced therapeutic gains.
• Operationalization within Feedback-Informed Treatment (FIT): As modern behavioral health systems transition toward empirical accountability and measurement-based care, the ARM-5 supplies actionable session-level data that can be interpreted alongside standardized symptom tracking measures (such as the PHQ-9, GAD-7, or CORE-OM) to monitor systemic treatment responsiveness.
• Equivalence to Ultra-Brief Competitors: Methodological comparisons confirm that the ARM-5 performs with clinical efficacy comparable to the widely used Session Rating Scale (SRS; Miller et al., 2000), offering a Likert-style alternative to visual analogue scales while retaining explicit multidimensional referents (Bond, Partnership, and Confidence).

Psychological Construct

The ARM-5 operationalizes the therapeutic alliance not merely as mutual positive regard, but as a collaborative, multi-layered relational system. Grounded in both psychodynamic traditions and humanistic-pantheoretical formulations, the scale assesses three core interdependent dimensions that collectively constitute the therapeutic working relationship:

1. Bond

The Bond dimension reflects the socio-emotional climate established between client and clinician. Derived conceptually from the Rogerian core conditions of unconditional positive regard, empathy, and authenticity, as well as psychodynamic concepts of the “therapeutic working alliance” (Greenson, 1967), Bond encompasses feelings of warmth, psychological safety, mutual liking, and therapist supportiveness. In the ARM-5, this construct is represented by Item 1 (“My therapist is supportive”). An elevated Bond rating signifies that the patient perceives the clinical environment as an emotionally safe holding space where vulnerability is met with compassion, validation, and non-judgmental acceptance. Clinically, a deficit in this dimension suggests that the patient feels misunderstood, alienated, or emotionally distant from the practitioner, undermining the basic safety required for deep clinical exploration.

2. Partnership

The Partnership dimension assesses the degree of mutual agency, egalitarian collaboration, and consensus between client and therapist regarding the shared enterprise of therapy. This construct directly captures the “Task” and “Goal” dimensions articulated in Bordin’s tripartite alliance model. Therapeutic success requires that both parties work in active concert rather than engaging in asymmetrical compliance or passive resistance. The ARM-5 assesses Partnership via two complementary indicators:

• Positive Consensus (Item 2): “My therapist and I agree about how to work together” directly examines procedural clarity, strategic alignment, and joint ownership over the treatment agenda.
• Relational Friction / Dissension (Item 3, Reverse Scored): “My therapist and I have difficulty working jointly as a partnership” explicitly probes for perceived resistance, discord, non-collaborative interactions, or operational fragmentation.

Collectively, these items capture whether the client feels empowered as an active, respected co-investigator in their own healing process or conversely feels coerced into therapeutic modalities that do not resonate with their personal goals.

3. Confidence in Therapy

The Confidence dimension captures both treatment credibility and mutual technical optimism. Distinct from generic emotional attachment, Confidence reflects the client’s conviction that the therapeutic framework, interventions, and theoretical rationale are methodologically sound and potent enough to resolve their presenting difficulties. Research indicates that technical confidence often serves as the single strongest statistical predictor of symptom reduction, capturing the client’s expectancy of therapeutic gain alongside perceived therapist competence. In the ARM-5, Confidence is evaluated through a dual perspective:

• Client Trust in Treatment and Technique (Item 4): “I have confidence in my therapist and their techniques” gauges the patient’s subjective faith in the utility of the clinical strategies applied.
• Perceived Therapist Self-Efficacy (Item 5): “My therapist is confident in him/herself and his/her techniques” captures the clinician’s projected professional authority, security, and mastery as perceived by the patient.

When clients perceive a therapist as indecisive, anxious, or ungrounded, overall alliance degrades rapidly; conversely, projected calm mastery fosters client hope and collaborative engagement.

Theoretical Framework

The structural and conceptual architecture of the ARM-5 is rooted in decades of empirical and theoretical evolution within clinical psychology and psychotherapy process research. Its lineage synthesizes three primary intellectual foundations:

1. The Bordin Pantheoretical Alliance Model

The preeminent conceptual framework guiding modern alliance research is Edward Bordin’s (1979) pantheoretical model of the working alliance. Bordin departed from strictly psychoanalytic definitions of transference-based relationships by proposing that the alliance is an integrative, cross-theoretical vehicle central to all modalities (e.g., psychodynamic, cognitive-behavioral, humanistic, systemic). Bordin conceptualized the alliance as comprising three interdependent components: Bonds (the affective attachment and mutual trust between participants), Tasks (the specific activities and cognitive/emotional demands constituting the therapy sessions), and Goals (the overarching objectives and target outcomes negotiated between client and clinician). The ARM-5 explicitly preserves these Bordinian elements: Bond corresponds to Item 1, Tasks and Goals map onto the Partnership items (Items 2 and 3), and therapeutic efficacy expectations converge in the Confidence items (Items 4 and 5).

2. The Agnew Relationship Measure (ARM-28) Structural Taxonomy

In developing the original 28-item Agnew Relationship Measure, Agnew-Davies, Stiles, Barkham, Hardy, and Shapiro (1998) sought to establish a comprehensive scale that resolved empirical disagreements between competing alliance instruments (e.g., the WAI, CALPAS, and the Penn Helping Alliance Questionnaire). Through factor analytic exploration within the Second Sheffield Psychotherapy Project (SPP2), Agnew-Davies and colleagues delineated five specific alliance dimensions: Bond, Partnership, Confidence, Openness, and Client Initiative. Crucially, their structural investigations revealed that three of these dimensions—Bond, Partnership, and Confidence—consistently loaded onto a powerful, higher-order factor termed the Core Alliance. The ARM-5 was extracted directly from this Core Alliance factor, stripping away peripheral items to create an essentialized index of relational stability and alignment.

3. Common Factors Theory and Feedback-Informed Care

The theoretical rationale for condensing the ARM into a 5-item scale is heavily informed by Common Factors Theory (Lambert, 2013; Wampold & Imel, 2015). Research demonstrates that specific technical models (e.g., CBT vs. psychodynamic psychotherapy) account for far less variance in therapeutic outcomes than trans-theoretical relational common factors, foremost among which is the therapeutic alliance. However, research also reveals that clinicians possess poor intuitive sensitivity regarding their clients’ subjective perceptions of the relationship; therapists routinely fail to detect early alliance ruptures and systematically underestimate the risk of patient dropout. The ARM-5 provides an empirical feedback loop based on cybernetic principles of behavioral regulation, feeding observational relationship data directly back into the treatment dyad in real time.

Validity

The psychometric validity of the ARM-5 has been evaluated across large-scale clinical trials and routine service contexts, establishing its construct, convergent, discriminant, and predictive utility.

Construct and Structural Validity

In the seminal validation study conducted by Cahill et al. (2012), the psychometric properties of the ARM-5 were benchmarked against the full 28-item ARM using the extensive Sheffield Psychotherapy Project dataset (N = 1,073). Confirmatory correlation analyses demonstrated an alternative-forms correlation of r ≥ .90 between the ARM-5 total score and the original ARM Core Alliance composite score. These findings demonstrate that the reduction of the instrument from 28 items to 5 items preserved the primary measurement variance of the Core Alliance construct without significant loss of structural integrity.

Convergent Validity

The ARM-5 demonstrates robust convergent validity with alternative validated alliance scales. Independent investigations, including comparative psychometric analyses by Bouchard (2018), reveal substantial positive correlations between the ARM-5 and the Session Rating Scale (SRS; r values typically ranging between .72 and .84), as well as the Working Alliance Inventory – Short Revised (WAI-SR; r > .75). These elevated correlations indicate that the ARM-5 captures the established relational constructs measured by alternative gold-standard instruments while offering a standardized Likert-based scoring distribution.

Predictive Validity and Treatment Outcomes

Crucially for clinical practice, the ARM-5 demonstrates verified predictive validity with respect to symptom reduction and clinical outcomes. When administered longitudinally as part of cognitive-behavioral therapy (CBT) and psychodynamic-interpersonal therapy for depression and anxiety, early and mid-treatment ARM-5 scores significantly predict subsequent post-treatment symptom change as indexed by the Beck Depression Inventory-II (BDI-II) and the Clinical Outcomes in Routine Evaluation – Outcome Measure (CORE-OM). In alignment with broader psychotherapy literature, the Confidence sub-dimension (items 4 and 5) consistently emerges as the strongest individual statistical predictor of subsequent symptom alleviation.

Ceiling Effects and Distributional Skewness

A critical psychometric consideration identified by Cahill et al. (2012) is the presence of marked ceiling effects. In the Sheffield sample (N = 1,073), the normative mean session rating on the ARM-5 was 5.76 (SD = 0.91) on a 1-to-7 scale. Psychotherapy clients routinely assign exceptionally positive ratings to their therapists. As a consequence, the data distribution is significantly negatively skewed. Clinically, this means that even moderate alliance ratings (e.g., scores of 4 or “Neutral”) represent significant statistical deviations from the normative mean, frequently functioning as empirical indicators of latent alliance strain or rupture.

Reliability

The reliability of the ARM-5 has been confirmed across diverse therapeutic environments, demonstrating adequate internal consistency and temporal reliability suited to dynamic clinical tracking:

Internal Consistency

Despite containing only five items, the ARM-5 exhibits solid internal consistency across clinical trials. In the primary derivation and cross-validation cohorts examined by Cahill et al. (2012), Cronbach’s alpha (α) for the overall scale ranged between .80 and .87 across varied assessment points (early, middle, and late therapy sessions). The item-total correlations for each of the five items consistently exceed .50, reflecting high internal coherence within the composite scale.

Alternative-Forms Reliability

Alternative-forms reliability between the ARM-5 and the 12-item short form (ARM-12) as well as the parent 28-item ARM is exceptionally high. Cahill et al. (2012) observed correlations ranging from .89 to .94 between the 5-item version and the full 28-item Core Alliance composite, indicating that the reduction in item count does not lead to construct underrepresentation or measurement attenuation.

Temporal Stability vs. Process Sensitivity

Within psychometrics, traditional test-retest reliability metrics (such as stability coefficients measured across multi-week intervals) must be interpreted with caution when applied to alliance tracking tools. The therapeutic alliance is inherently a dynamic state variable rather than a static psychological trait; it fluctuates systematically across sessions in response to in-session interventions, discussions of emotionally challenging material, and interpersonal dynamics. Studies evaluating adjacent-session stability show moderate-to-high intra-class correlations (ICC ≈ .65 to .75), demonstrating sufficient baseline measurement stability alongside sensitivity to acute ruptures and subsequent collaborative repairs.

Factor Analysis

The latent structural properties of the ARM-5 have been rigorously examined using both Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) during its derivation from the 28-item instrument.

Exploratory Factor Structure

During initial scale development, Agnew-Davies et al. (1998) subjected the 28-item parent scale to principal components analysis and common factor analysis with oblique rotation, which revealed a distinct five-factor solution (Bond, Partnership, Confidence, Openness, and Client Initiative). When the items representing Bond, Partnership, and Confidence were re-analyzed separately, they converged strongly onto a single dominant primary factor with an eigenvalue exceeding 3.0, accounting for the vast majority of the shared variance. This statistical convergence led the authors to define this unified latent dimension as the “Core Alliance”.

Confirmatory Factor Analysis (CFA) and Model Fit

When Cahill et al. (2012) modeled the 5-item subset, they evaluated both a three-factor oblique model (differentiating Bond [1 item], Partnership [2 items], and Confidence [2 items]) and a strictly unidimensional one-factor model representing global Core Alliance. The structural findings revealed:

• Factor Inter-correlations: Latent factor correlations between the three theoretical domains were exceptionally high (typically r > .80), indicating substantial conceptual overlap and statistical collinearity among the sub-facets in such an abbreviated format.
• Unidimensionality: A single-factor confirmatory model exhibited acceptable to excellent goodness-of-fit indices across clinical validation datasets (Comparative Fit Index [CFI] > .95; Tucker-Lewis Index [TLI] > .93; Root Mean Square Error of Approximation [RMSEA] ≤ .06; Standardized Root Mean Square Residual [SRMR] ≤ .04).
• Factor Loadings: Standardized factor loadings across the five items onto the general Core Alliance construct are uniformly robust, ranging from .62 to .86. The negatively phrased, reverse-scored item (Item 3: “My therapist and I have difficulty working jointly as a partnership”) consistently demonstrates the lowest—yet fully acceptable—loading (approx. .62 to .68), an effect common to reverse-worded items due to method variance.

Consequently, the authors and subsequent psychometricians conclude that while the three subscales possess qualitative and clinical heuristic utility, the ARM-5 should psychometrically be treated and scored as a unidimensional scale representing overall Core Alliance.

Instrument / Measurement Tool

The operational specifications and structural profile of the ARM-5 are summarized below:

• Instrument Name: Agnew Relationship Measure – 5 (ARM-5)
• Alternative Title: Agnew Relationship Measure Ultra-Short Form
• Authors / Developers: Jane Cahill, William B. Stiles, Michael Barkham, Gillian E. Hardy, Gregory Stone, Roxane Agnew-Davies, and Gisela Unsworth (2012)
• Theoretical Precursor: Agnew Relationship Measure (ARM-28; Agnew-Davies et al., 1998)
• Target Respondent: Adult and adolescent psychotherapy clients (Client Version). Clinician and Observer adaptations exist for research contexts.
• Administration Format: Paper-and-pencil questionnaire, tablet-based clinical interface, or secure digital patient portal.
• Administration Timing: Completed immediately following the conclusion of a therapy session, ideally administered every session, every second session, or every third session.
• Time to Complete: Approximately 1 to 2 minutes.
• Item Count: 5 items.
• Response Scale: 7-point Likert-type agreement scale:
  • 1 = Strongly Disagree
  • 2 = Disagree
  • 3 = Slightly Disagree
  • 4 = Neutral
  • 5 = Slightly Agree
  • 6 = Agree
  • 7 = Strongly Agree
• Subscale Structural Allocation:
  • Bond: Item 1
  • Partnership: Items 2 and 3
  • Confidence in therapy: Items 4 and 5
• Reverse-Scored Items: Item 3 is reverse scored prior to calculating cumulative totals or calculating scale averages:
  • Raw response 1 → Recoded as 7
  • Raw response 2 → Recoded as 6
  • Raw response 3 → Recoded as 5
  • Raw response 4 → Recoded as 4
  • Raw response 5 → Recoded as 3
  • Raw response 6 → Recoded as 2
  • Raw response 7 → Recoded as 1
• Scoring Formulas:
  • Total Raw Score: Sum of Item 1 + Item 2 + Recoded Item 3 + Item 4 + Item 5. Score range: 5 to 35. Higher scores reflect a stronger therapeutic alliance.
  • Session Mean Score: Total Raw Score divided by 5. Score range: 1.00 to 7.00.
• Normative Reference Values: In the baseline validation sample (Cahill et al., 2012; N = 1,073), the normative client mean was 5.76 with a standard deviation of 0.91. Due to negative skew and ceiling effects, a mean score below 5.0 indicates an alliance quality falling below the 20th percentile, signaling potential therapeutic friction or risk of dropout.

Permissions & Fee and Test Year

The ARM-5 was officially published in 2012 in the peer-reviewed journal Psychotherapy Research (Cahill et al., 2012), following the development of the original 28-item instrument in 1998 (Agnew-Davies et al., 1998). The instrument is copyrighted by the Society for Psychotherapy Research (SPR) and the authors. However, in accordance with the authors’ intent to foster Feedback-Informed Treatment and clinical research, the ARM-5 is made available without licensing fees for non-commercial clinical, educational, and academic research purposes. Practitioners and healthcare organizations may reproduce and integrate the scale into paper workflows or electronic health records (EHRs) provided appropriate bibliographic citation is maintained. Commercial software vendors or proprietary platforms seeking to monetize the instrument should secure formal permission from the developer group or copyright holders.

References

• Agnew-Davies, R., Stiles, W. B., Hardy, G. E., Barkham, M., & Shapiro, D. A. (1998). Alliance structure assessed by the Agnew Relationship Measure (ARM). British Journal of Clinical Psychology, 37(2), 155–172. https://doi.org/10.1111/j.2044-8260.1998.tb01292.x
• 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
• Bouchard, D. (2018). Assessment of the Therapeutic Alliance Scales: A Reliability and Validity Evaluation (Doctoral dissertation, School of Psychology, University of Ottawa). https://doi.org/10.20381/ruor-22238
• Cahill, J., Stiles, W. B., Barkham, M., Hardy, G. E., Stone, G., Agnew-Davies, R., & Unsworth, G. (2012). Two short forms of the Agnew Relationship Measure: The ARM-5 and ARM-12. Psychotherapy Research, 22(3), 241–255. https://doi.org/10.1080/10503307.2011.643253
• Greenson, R. R. (1967). The technique and practice of psychoanalysis. International Universities Press.
• Horvath, A. O., Del Re, A. C., Flückiger, C., & Symonds, D. (2011). Alliance in individual psychotherapy. Psychotherapy, 48(1), 9–16. https://doi.org/10.1037/a0022186
• Lambert, M. J. (Ed.). (2013). Bergin and Garfield’s handbook of psychotherapy and behavior change (6th ed.). John Wiley & Sons.
• Miller, S. D., Duncan, B. L., & Johnson, L. D. (2000). The Session Rating Scale (SRS 3.0). Chicago, IL: Authors.
• Wampold, B. E., & Imel, Z. E. (2015). The great psychotherapy debate: The evidence for what makes psychotherapy work (2nd ed.). Routledge. https://doi.org/10.4324/9780203582015

13. Items of the Scale (Questionnaire)