Action Research Arm Test

Abstract

The Action Research Arm Test (ARAT) is an internationally recognized, performance-based observational assessment designed to evaluate upper extremity motor function, dexterity, and functional capacity following neurological impairment, most notably stroke, traumatic brain injury, and multiple sclerosis. Developed by Ronald C. Lyle in 1981 as a streamlined modification of the Upper Extremity Function Test (UEFT), the ARAT comprises 19 standardized items organized into four discrete hierarchical subscales: Grasp (6 items), Grip (4 items), Pinch (6 items), and Gross Movement (3 items). Each item evaluates an individual’s ability to manipulate objects of varying size, weight, and shape, or execute proximal motor patterns within a standardized spatial setup and predefined execution time limits.

Performance on each item is quantified using a 4-point ordinal rating scale ranging from 0 (can perform no part of test) to 3 (performs test normally within the normal time limit), yielding subtest scores of 0–18 for Grasp, 0–12 for Grip, 0–18 for Pinch, and 0–9 for Gross Movement, with a composite maximum score of 57 points. The instrument operationalizes Guttman scaling principles, enabling rapid administrative completion (typically under 10 minutes) by employing strict item progression and termination rules based on hierarchical task difficulty. Psychometric evaluations demonstrate exceptional internal consistency (Cronbach’s alpha > .95), superior inter-rater and test-retest reliability (Intraclass Correlation Coefficient > .98), robust concurrent and construct validity against the Fugl-Meyer Assessment and Box and Block Test, and high responsiveness to clinically meaningful recovery across acute, subacute, and chronic neurorehabilitation phases.

Keywords

Action Research Arm Test, ARAT, upper extremity assessment, stroke rehabilitation, motor recovery, hemiparesis, prehension, manual dexterity, Guttman scale, psychometrics, neurorehabilitation

Authors

The Action Research Arm Test was originated and published by Ronald C. Lyle in 1981 during his tenure at the Department of Occupational Therapy and Rehabilitation Research units in the United Kingdom. Lyle adapted the instrument from the 33-item Upper Extremity Function Test originally developed by Douglas Carroll in 1965.

Subsequent psychometric standardization, administration protocol refinement, and modern clinical calibration have been led by international neurorehabilitation researchers, including:

• Jan H. van der Lee, MD, PhD — Department of Rehabilitation Medicine, VU University Medical Center, Amsterdam, Netherlands.
• Nuray Yozbatiran, PT, PhD — Department of Physical Medicine and Rehabilitation, McGovern Medical School, UTHealth Houston, Texas, USA.
• Steven C. Cramer, MD, MMSc — Department of Neurology, University of California, Irvine, and David Geffen School of Medicine at UCLA, USA.
• Ching-yi Wu, ScD, OTR and Keh-chung Lin, ScD, OTR — Department of Occupational Therapy, College of Medicine, National Taiwan University, Taipei, Taiwan.

Purpose

The primary clinical and scientific objective of the Action Research Arm Test is to provide an objective, rapid, standardized, and criterion-referenced assessment of paretic upper limb activity capacity in individuals with central nervous system lesions. Hemiparesis is among the most pervasive and disabling consequences of cerebrovascular accidents and focal brain lesions, directly degrading reaching, grasping, manipulation, and self-care independence. Traditional clinical examinations often rely either on subjective assessments of tone and reflexes or on extensive, time-exhaustive functional inventories. The ARAT addresses this clinical dilemma by measuring focal functional execution within an ecologically valid framework.

The ARAT is explicitly mapped onto the “Activity” dimension of the World Health Organization’s International Classification of Functioning, Disability, and Health (ICF). While impairment-level tools—such as the Fugl-Meyer Assessment—measure isolated joint kinematics, synergy emergence, and reflex regulation, the ARAT measures the functional synthesis of these motor elements during objective task completion. The test examines whether an individual can physically grasp, lift, transport, release, and manipulate objects of diverse dimensions (e.g., wooden blocks, ball bearings, tubes, marbles) against gravity across a dedicated testing shelf situated 37 cm above table height.

In research contexts, the ARAT serves as an international gold-standard primary or secondary endpoint in randomized controlled clinical trials investigating novel motor restoration interventions. These therapies include Constraint-Induced Movement Therapy (CIMT), upper-limb robotic exoskeleton training, neuromuscular electrical stimulation, transcranial magnetic stimulation (TMS), stem cell therapies, and intensive task-specific motor practice. Because the instrument incorporates a standardized setup and explicit timing cutoffs, it eliminates examiner subjectivity and provides high longitudinal sensitivity to therapeutic gains.

Psychological Construct

The psychological, neuromotor, and biomechanical construct operationalized by the ARAT is upper extremity functional manual capacity. This multi-tiered construct reflects the patient’s capacity to orchestrate sensorimotor planning, corticospinal fractionation of movement, visuospatial target acquisition, and proximal-distal motor integration. Reaching and prehension require real-time processing of sensory feedback, cognitive intent, and mechanical execution. The ARAT breaks down this complex construct into four distinct functional dimensions:

1. Grasp

The Grasp subscale measures the capacity to form whole-hand gross cylindrical and spherical prehensions around solid objects of varying volume and mass, elevate them against gravitational load, transport them over an obstacle or vertical elevation, and execute a controlled release. It evaluates motor recruitment in the finger flexors (flexor digitorum profundus and superficialis), wrist extensors (extensor carpi radialis longus/brevis, extensor carpi ulnaris) to prevent wrist flexion collapse during grasp, and shoulder/elbow musculature during the transport phase. The progression across block sizes (ranging from 2.5 cm to 10 cm) assesses the hand’s biomechanical capacity to modulate digit aperture based on visual spatial cues.

2. Grip

The Grip subscale evaluates functional prehension requiring combined mechanical holding, object orientation, and coordinated proximal-distal interactions. Tasks such as pouring water between two glasses demand isometric digital stabilization coupled with dynamically modulated forearm pronation and supination without accidental spillage. Similarly, transporting horizontal and vertical tubes over a standardized peg assesses dynamic cylinder grip combined with accurate spatial targeting and visual-proprioceptive feedback.

3. Pinch

The Pinch subscale evaluates fine motor dexterity and the selective activation of individual digits, known neurophysiologically as fractionated finger movements. Fractionation represents the hallmark of direct cortico-motoneuronal projections from the primary motor cortex (Betz cells) to spinal alpha motor neurons. The subscale evaluates distinct pad-to-pad and tip-to-tip pinch configurations across the thumb and index finger, thumb and middle finger, and thumb and ring finger using tiny ball bearings (6 mm diameter) and marbles (1.5 cm diameter). Impairments in this construct indicate residual synergistic coupling, cortical disinhibition, or sensory-tactile feedback deficits.

4. Gross Movement

The Gross Movement subscale measures proximal multijoint motor control of the shoulder complex and elbow in the absence of fine distal object handling. It tests the patient’s ability to position the hand in functional body spaces: placing the hand behind the head, touching the top of the head, and placing the hand to the mouth. These movements represent non-synergistic, anti-gravity motor control, requiring coordinated activation of the rotator cuff, deltoids, biceps, and triceps, and serving as biomechanical prerequisites for grooming, hygiene, and self-feeding.

Theoretical Framework

The ARAT is rooted in the intersection of systems theory of motor control, neurodevelopmental hierarchy, and psychometric scaling theory.

Systems and Neurodevelopmental Foundations

Classic neurorehabilitation paradigms—including those formulated by Signe Brunnstrom and Thomas Twitchell—documented that motor recovery following an upper motor neuron lesion follows a stereotypical pattern. Patients typically progress from initial flaccid paralysis to involuntary flexor and extensor synergies, followed by gradual synergy breakdown, emergence of voluntary isolated movements, and final restoration of fine, fractionated distal coordination. The ARAT reflects this neurodevelopmental progression. Gross proximal movements (such as touching the head or mouth) usually recover prior to isolated fine digit manipulation (such as a 6 mm ball bearing pinch). The ability to isolate the ring finger and thumb requires sophisticated central nervous system inhibition to suppress involuntary mass flexion of adjacent digits.

Guttman Scaling and Algorithmic Efficiency

A foundational theoretical innovation introduced by Lyle (1981) was structuring the ARAT according to the mathematical properties of a Guttman scalogram. In a true Guttman scale, items exist along a deterministic continuum of difficulty. A subject who successfully passes a highly difficult item is statistically predicted to pass all less difficult items within that domain. Conversely, a subject who fails an item of minimal difficulty is predicted to fail all items of greater difficulty within that domain.

Lyle operationalized this principle within each subscale through explicit testing rules:

• The examiner first administers the most difficult item of the subscale (Item 1). If the patient performs this item normally (achieving a score of 3), it is assumed that they can perform all subsequent items in that subscale normally. The examiner awards full credit (3 points) to all remaining items in that subscale and advances immediately to the next subtest.
• If the patient scores less than 3 on Item 1, the examiner administers the easiest item of the subscale (Item 2). If the patient scores 0 on this easiest item, it is assumed that they are incapable of performing the intermediate items. The examiner scores all remaining items in that subscale as 0 and proceeds to the next subtest.
• Only if the patient fails Item 1 (score < 3) but demonstrates partial or full capacity on Item 2 (score > 0) does the examiner need to administer the remaining items within that subscale.

This Guttman-based structural design reduces assessment time from 20–30 minutes to under 8 minutes in many stroke survivors, reducing patient fatigue and clinical burden.

Validity

The psychometric properties of the ARAT have been thoroughly investigated across diverse neurological populations, establishing high construct, concurrent, predictive, and discriminant validity.

Concurrent and Criterion Validity

Extensive validation studies have examined the relationship between ARAT scores and other established upper limb measurement instruments. In a landmark study by van der Lee et al. (2001) involving chronic stroke patients undergoing intensive rehabilitation, the ARAT demonstrated profound concurrent validity with the Upper Extremity Motor component of the Fugl-Meyer Assessment (FMA-UE), yielding Pearson correlation coefficients consistently ranging between r = .85 and r = .94. When correlated with manual dexterity instruments, such as the Box and Block Test (BBT), ARAT performance yields strong correlations ranging from r = .80 to r = .92.

Furthermore, strong associations have been demonstrated between the ARAT and real-world functional inventories, including the Motor Activity Log (MAL; r = .70–.84 for the Amount of Use scale and Quality of Movement scale) and the functional independence measure of self-care subscores.

Construct and Discriminant Validity

Construct validity is substantiated by the instrument’s capacity to discriminate between varying severities of neurological impairment. Patients stratified clinically into mild, moderate, and severe hemiparesis demonstrate statistically distinct ARAT score distributions with minimal distribution overlap (Yozbatiran et al., 2008). Kinematic validation using optical motion-capture systems confirms that higher ARAT scores correlate significantly with reduced trunk compensatory forward displacement, improved peak reach velocity, smoother wrist trajectories, and normalized joint angle synergies.

Predictive Validity and Responsiveness

ARAT scores captured in the acute and early subacute stages post-stroke (e.g., at 1 to 3 weeks) show high predictive validity for long-term functional arm use at 6 and 12 months. Patients achieving an initial ARAT score > 10 within the first several weeks post-stroke have a substantially higher probability of regaining useful functional hand capacity compared to those scoring < 10.

Regarding responsiveness to therapeutic change, the ARAT demonstrates moderate-to-large effect sizes (Cohen’s d > 0.80; Standardized Response Mean > 0.90) in intervention studies. Clinical trials establish a Minimal Clinically Important Difference (MCID) of approximately 5.7 points on the 57-point scale in chronic stroke patients (van der Lee et al., 2001) and between 12 and 17 points (or approximately 10% of total scale range) in acute-to-subacute stroke cohorts where spontaneous biological recovery intersects with behavioral training (Lang et al., 2008).

Reliability

The ARAT displays outstanding test-retest, intra-rater, and inter-rater reliability across clinical and laboratory settings.

Inter-Rater and Intra-Rater Reliability

Because the ARAT incorporates standardized object dimensions, designated table and shelf heights, and concrete operational scoring anchors, inter-examiner variance is minimal. Lyle (1981) reported initial inter-rater correlation coefficients exceeding .99. Modern evaluations utilizing video-recorded administrations and multiple independent blinded raters confirm that total ARAT scores maintain an Intraclass Correlation Coefficient (ICC) between .98 and .99. Subtest-specific ICC values remain similarly robust: Grasp (ICC = .97–.99), Grip (ICC = .96–.98), Pinch (ICC = .95–.99), and Gross Movement (ICC = .93–.97).

Test-Retest Reliability and Measurement Error

In stable, chronic stroke populations evaluated over intervals spanning 2 to 14 days without intervening therapeutic change, test-retest reliability estimates range from ICC = .96 to .99. The Standard Error of Measurement (SEM) for the total ARAT score is consistently calculated between 1.10 and 2.41 points. Consequently, the Minimal Detectable Change (MDC) at the 95% confidence level (MDC₉₅)—representing the statistical threshold that must be surpassed to ensure change is not due to measurement error—is estimated between 3.0 and 5.6 points on the total scale.

Internal Consistency

Internal consistency analyses across the 19 items indicate exceptional scale coherence. Cronbach’s alpha coefficients for the full instrument consistently surpass .96 (often reaching .98), indicating high item interrelatedness without excessive functional redundancy.

Factor Analysis

Structural evaluations of the ARAT have utilized both classical Exploratory and Confirmatory Factor Analysis (EFA/CFA) and modern Item Response Theory (Rasch Analysis).

Dimensionality via Factor Analysis

While Ronald Lyle clinically configured the ARAT around four intuitive functional categories (Grasp, Grip, Pinch, and Gross Movement), structural factor analytic studies typically reveal that the ARAT is dominated by a single, strong primary latent factor. In exploratory factor analyses, the first unrotated general factor accounts for upwards of 75% to 84% of total variance, with an eigenvalue exceeding 14.0. Factor loadings for individual items onto this primary latent factor are consistently high (ranging from .72 to .94). When multi-factor solutions are forced, items separate into distal fine-motor control (Pinch) versus proximal/gross manual control (Grasp, Grip, Gross Movement); however, high inter-factor correlations (r > .85) support utilizing the composite sum score (0–57) as a valid unidimensional index of upper limb capacity.

Rasch Measurement Model Insights

Modern psychometric investigations (e.g., Lin et al., 2010; Koh et al., 2006; Hsueh et al., 2002) have scrutinized the ARAT using Rasch rating scale modeling:

• Item Infit and Outfit Statistics: Mean square infit and outfit statistics for nearly all 19 items consistently fall within the acceptable psychometric window of 0.60 to 1.40, verifying that the items conform systematically to a single underlying continuum of difficulty.
• Item Difficulty Hierarchy: Rasch calibration confirms the hierarchical distribution of items. The Gross Movement items (e.g., placing hand to mouth or top of head) represent the least difficult items on the latent logit scale. Grasp items occupy intermediate difficulty. Pinch items (particularly pinching the 6 mm ball bearing with the ring finger and thumb) exhibit the highest logit difficulty, requiring supreme corticospinal integrity.
• Category Functioning: Rasch analyses have critically reviewed the 4-point rating scale (0 to 3). Studies demonstrate that while the scale functions well overall, Category 1 (“performs test partially”) and Category 2 (“completes test, but takes abnormally long time or has great difficulty”) are occasionally conflated across borderline functional performers. Nonetheless, collapsing the scale reduces clinical utility, leading researchers to retain Lyle’s original 4-point categorization.

Instrument / Measurement Tool

• Standardized Assessment Name: Action Research Arm Test (ARAT)
• Alternative Titles: Action Research Arm test (ARA-test)
• Primary Developer: Ronald C. Lyle (1981); derived from Douglas Carroll (1965)
• Assessment Type: Performance-based observational functional rating scale
• Target Population: Adults and older adults with neurological disorders (primarily stroke, traumatic brain injury, incomplete spinal cord injury, multiple sclerosis) experiencing upper extremity hemiparesis or dyspraxia.
• Administration Time: Approximately 5 to 15 minutes (shortened when applying Guttman hierarchical skip rules).
• Test Configuration and Standardized Equipment:
  • Rigid, level testing table and standardized adjustable chair (without armrests) ensuring the subject sits upright with trunk supported against chair back.
  • Wooden testing apparatus/shelf elevated exactly 37 cm above the table surface.
  • 19 specialized physical test objects: four square wooden blocks (10 cm, 2.5 cm, 5 cm, and 7.5 cm), cricket ball (7.5 cm diameter), sharpening stone (10 x 2.5 x 1 cm), two standardized drinking glasses and water jug, hollow metal tubes (2.25 cm and 1 cm diameter), pegboard post, washer (3.5 cm diameter) and bolt, small alloy ball bearings (6 mm diameter), glass marbles (1.5 cm diameter), and an open tin/cylinder.
  • Stopwatch calibrated in seconds.
• Number of Subtests: 4 subtests (Grasp, Grip, Pinch, Gross Movement)
• Total Number of Items: 19 standardized performance items
• Authentic Response Scale: 4-point ordinal rating scale per item:
  • 0 = Can perform no part of test
  • 1 = Performs test partially
  • 2 = Completes test, but takes abnormally long time or has great difficulty
  • 3 = Performs test normally (completed within normal time limit)
• Time Limits: Tasks are judged against standardized normative execution limits, typically within 5 seconds for normal execution (Score 3); completion between 5 and 60 seconds or with severe compensation earns a Score of 2; inability to finish within 60 seconds is scored as 1 or 0 depending on whether initial displacement occurred.
• Scoring and Computational Rules:
  • Subtest 1 (Grasp): Sum of items 1–6 (Score range: 0 to 18)
  • Subtest 2 (Grip): Sum of items 7–10 (Score range: 0 to 12)
  • Subtest 3 (Pinch): Sum of items 11–16 (Score range: 0 to 18)
  • Subtest 4 (Gross Movement): Sum of items 17–19 (Score range: 0 to 9)
  • Total Score: Sum of all 19 items (Score range: 0 to 57), where higher scores indicate superior upper limb motor performance and dexterity.

Permissions & Fee and Test Year

The Action Research Arm Test was originally published in 1981 by Ronald C. Lyle in the International Journal of Rehabilitation Research. The instrument and its administrative framework were placed in the public domain for clinical, scientific, and academic non-commercial research use. No clinical licensing fees or copyright royalties are required to score or administer the test protocol.

However, administering the ARAT requires standardized physical testing materials manufactured to exact dimensional tolerances (e.g., the 37 cm shelf, wooden blocks, metal tubes, marbles, and ball bearings). Standardized physical ARAT testing kits can be acquired commercially from medical rehabilitation equipment manufacturers or constructed independently using standardized manuals and specifications published by rehabilitation researchers (e.g., Yozbatiran et al., 2008).

References

• Carroll, D. (1965). A quantitative test of upper extremity function. Journal of Chronic Diseases, 18(5), 479–491. https://doi.org/10.1016/0021-9681(65)90030-5
• Hsueh, I. P., Hsu, M. J., Sheu, C. F., Lee, S., & Hsieh, C. L. (2002). Psychometric comparisons of 2 versions of the Action Research Arm Test in patients with stroke. Neurorehabilitation and Neural Repair, 16(4), 349–358. https://doi.org/10.1177/154596830201600404
• Koh, C. L., Hsueh, I. P., Wang, W. C., Sheu, C. F., Chai, H. M., & Hsieh, C. L. (2006). Validation of the full-length and abbreviated Action Research Arm Tests in patients with stroke. Journal of Rehabilitation Medicine, 38(4), 234–240. https://doi.org/10.1080/16501970600632613
• Lang, C. E., Edwards, D. F., Birkenmeier, R. L., & Dromerick, A. W. (2008). Estimating minimal clinically important differences of upper-extremity measures in subacute stroke. Archives of Physical Medicine and Rehabilitation, 89(9), 1693–1699. https://doi.org/10.1016/j.apmr.2008.02.029
• Lang, C. E., Wagner, J. M., Dromerick, A. W., & Edwards, D. F. (2006). Measurement of upper-extremity function early after stroke: Properties of the Action Research Arm Test. Archives of Physical Medicine and Rehabilitation, 87(12), 1602–1608. https://doi.org/10.1016/j.apmr.2006.09.003
• Lin, K. C., Chuang, L. L., Wu, C. Y., Hsieh, Y. W., & Chang, W. Y. (2010). Responsiveness and validity of three upper extremity outcome measures in stroke rehabilitation. Stroke, 41(6), 1286–1291. https://doi.org/10.1161/STROKEAHA.110.580118
• Lyle, R. C. (1981). A performance test for assessment of upper limb function in physical rehabilitation treatment and research. International Journal of Rehabilitation Research, 4(4), 483–492. https://doi.org/10.1097/00004356-198112000-00001
• van der Lee, J. H., Beckerman, H., Lankhorst, G. J., & Bouter, L. M. (2001). The responsive conduct of the Action Research Arm test: A comparison with the Fugl-Meyer Assessment and the Motricity Index. Archives of Physical Medicine and Rehabilitation, 82(1), 8–13. https://doi.org/10.1053/apmr.2001.18051
• van der Lee, J. H., De Groot, V., Beckerman, H., Wagenaar, R. C., Lankhorst, G. J., & Bouter, L. M. (2001). The intra- and interrater reliability of the Action Research Arm test: A study in patients with long-standing hemiparesis. Archives of Physical Medicine and Rehabilitation, 82(8), 1029–1034. https://doi.org/10.1053/apmr.2001.24296
• Yozbatiran, N., Der-Yeghiaian, L., & Cramer, S. C. (2008). A standardized approach to performing the Action Research Arm Test. Neurorehabilitation and Neural Repair, 22(1), 78–90. https://doi.org/10.1177/1545968307305353

13. Items of the Scale (Questionnaire)