Humans underestimate the movement range of their own hands

Internal representations of our limbs’ range of motion are essential for efficient and safe interaction with the environment. Although adults accrue extensive motor experience and proprioceptive feedback that could support accurate knowledge of their wrist rotation limits, whether such representations are veridical is unclear. Prior assumptions in motor planning and motor imagery posit accurate biomechanical constraints within internal models and shared neural substrates supporting planning, imagery, and conscious monitoring. However, recent work shows distortions in hand position, size, and weight representations, motivating examination of whether range-of-motion (ROM) representations are similarly biased. The study asks: How accurate are ROM representations of the wrist, and are they systematically biased? Across two experiments focusing on simple wrist rotations (abduction, adduction, flexion, extension), the authors tested estimation and motor imagery of maximal amplitudes for both hands to characterize accuracy and directional bias.

The introduction situates the work within literature asserting that motor planning incorporates biomechanical costs and that motor imagery reflects real movement constraints, with shared parietal-premotor substrates for planning, imagery, and awareness. Nevertheless, distortions have been documented in implicit body representations of hand position, size, joint locations, and systematic underestimation of hand weight, suggesting internal body models may be biased. Biases have historically illuminated perceptual and sensorimotor heuristics (e.g., visual illusions, action control). The authors propose that potential underestimation of ROM may reflect adaptive constraints balancing movement efficiency and safety under sensorimotor noise, extending findings from undershoot biases in uncertain motor planning contexts.

Two experiments assessed representation of wrist ROM. Participants: Experiment 1 included 60 volunteers (final N=59 after excluding one frequent flexibility practitioner; mean age 21.3, SD 2.5; 55 right-handed; 48 men, 11 women) with normal or corrected vision and no wrist injury history. Experiment 2 included 25 volunteers (age 21–40; 22 right-handed; 8 men, 17 women), similarly screened; none reported manual activities influencing estimates. Procedures were approved by the biomedical ethics committee of Cliniques Universitaires Saint-Luc, Brussels; informed consent obtained; convenience sampling; not preregistered.

Data analysis: Analyses and visualization were conducted in Jamovi (v2.3.28), figures in Inkscape. Normality assessed via Kolmogorov–Smirnov tests. In Experiment 1, within-subject ANOVA tested effects of hand (dominant/non-dominant) and movement type (adduction, abduction, flexion, extension). One-sample t-tests assessed absolute estimation errors; paired t-tests compared real vs estimated angles. In Experiment 2, one-sample t-tests examined absolute errors; paired t-tests tested directional bias.

Experiment 1 task and apparatus: Participants performed two tasks per condition: (1) estimate the maximal imagined wrist rotation; (2) execute the maximal rotation. Movements: four cardinal wrist rotations (abduction, adduction, flexion, extension) for each hand, yielding eight conditions. Setup: Participants seated at a table with a 360° protractor drawn on the surface. A standardized “resting position” aligned the arm in the sagittal plane to the protractor’s reference axis. Positioning aids included double-sided adhesive pads: (a) olecranon aligned to the protractor reference; (b) for adduction/abduction with pronated hand, pad on transverse carpal ligament aligned to the protractor center; (c) for flexion/extension with neutral hand, pad on the ulnar styloid aligned to protractor center. The hand was strapped to a rigid plate to minimize finger movements; fingers remained together. For each condition, a brief 2–4 s video illustrated the movement (small amplitude; clarified as non-representative of natural range). Estimation employed a staircase-like yes/no procedure: Half the participants started from a large angle (90° for abduction/adduction; 120° for flexion/extension) descending in 3° steps; the first “yes” was recorded as the estimated maximum. The other half started at 3° ascending in 3° steps; the last “yes” was recorded. Execution: Participants then performed the maximal rotation for each movement in the same order; goniometric measurement recorded the last reached dot on the protractor. Each movement was executed three times; the median angle was used as maximal amplitude.

Experiment 2 motor imagery: The hand was centered on an unlabelled protractor using palpation landmarks (crucifixion fossa posterior to the third metacarpal; anterior dorsal radius; medial to lunate); the depression and middle fingertip served as references with the palm prone. Participants imagined maximal abduction and adduction; the experimenter indicated left/right directions without performing the motion to avoid range cues. Participants pointed with the non-tested hand to the scale location where the middle finger would land after maximal imagined rotation. Each estimation was repeated twice; mean of repetitions used (differences <3°). After imagery, participants executed the relevant movements; middle finger positions were used to derive actual angles.

Experiment 1: Absolute discrepancies between actual and estimated maximal wrist rotation were large and significant across all movements (one-sample, two-tailed t-tests: all t(58) > 9.51, all p < 0.001, all d > 1.24). Movement type affected inaccuracy (within-subject ANOVA: F(1,174) = 27.21, p < 0.001, η² = 0.113). Post-hoc paired t-tests indicated greater inaccuracies for flexion/extension versus adduction/abduction (except abduction vs extension not significant). Average inaccuracy magnitudes for the dominant hand: flexion ≈ 35°, extension ≈ 28.5°, abduction ≈ 17°, adduction ≈ 15°. Directionally, paired comparisons showed estimated maxima were smaller than real maxima for three movements (dominant hand: abduction t=4.33, p<0.001, d=0.563; flexion t=5.25, p<0.001, d=0.683; extension t=2.77, p=0.008, d=0.36; adduction trend only t=1.97, p=0.054. Non-dominant: abduction t=5.44, p<0.001, d=0.708; flexion t=4.50, p<0.001, d=0.586; extension t=4.01, p<0.001, d=0.522; adduction ns t=−1.08, p=0.286). Underestimations were larger in the non-dominant hand (t(58)=2.87, p=0.006).

Experiment 2: Absolute errors in imagined ROM were significant in all conditions (one-sample t-tests: dominant adduction t=5.93, p<0.001, d=1.19; dominant abduction t=6.80, p<0.001, d=1.36; non-dominant adduction t=6.64, p<0.001, d=1.33; non-dominant abduction t=6.70, p<0.001, d=1.34). Directionally, participants underestimated abduction for both hands (dominant t=3.846, p<0.001, d=0.769; non-dominant t=4.753, p<0.001, d=0.951) and underestimated adduction for the non-dominant hand (t=2.317, p=0.029, d=0.463) but not for the dominant hand (t=0.957, p=0.348). Underestimations were larger for the non-dominant hand (t(24)=3.58, p=0.022). Overall, participants underestimated maximal ROM in three of four cardinal wrist directions, with stronger biases for non-dominant hands and for flexion/extension movements.

The study directly addresses whether internal representations of one's own wrist ROM are veridical. Across an estimation task and a motor imagery task, participants systematically underestimated their maximal ROM in three of four directions, indicating a biased internal model of biomechanical limits. The authors argue this bias is adaptive: motor planning must contend with sensorimotor noise (neural variability, fatigue, recruitment patterns), and planning to operate near true joint limits risks overshoot, potential action failure, or tissue strain. Underestimation imposes small efficiency costs (e.g., recruiting additional joints or corrective movements) while enhancing safety and robustness, thus optimizing the efficiency–safety trade-off. The larger underestimation for the non-dominant hand suggests a relationship between movement precision/experience and representational precision. An alternative account is that biased ROM arises from distorted somatosensory body metrics (position, size, shape) in cortex, which could propagate to motor planning, imagery, and monitoring that share parietal substrates and rely on internal models (efference copy/body schema). Future tests should directly examine how uncertainty and sensory weighting (vision vs proprioception) shape ROM representations and how these biases map onto individual differences in motor skill and proprioceptive acuity.

Representations of wrist movement range are not veridical: people tend to underestimate their true biomechanical limits in most directions. The authors propose this underestimation reflects an adaptive strategy that balances efficiency with safety under the stochastic nature of motor control. The work advances understanding of internal body models used in planning and imagery. Future research should delineate contributions of vision, proprioception, and internal models (e.g., body schema, efference copy) to ROM estimation, test the role of uncertainty directly, and investigate inter-individual variability linked to motor expertise and proprioceptive precision.

Underestimation was not uniform across all movements and conditions: wrist adduction showed no consistent group-level underestimation (trend in Experiment 1; significant only for non-dominant hand in Experiment 2). Bias magnitude tended to be larger for movements with larger ROM. There was notable inter-individual variability, with some participants overestimating ROM. Such variability may reflect differences in reliance on visual vs proprioceptive cues, limited daily-life use of full ROM (promoting conservative estimates), and individual differences in motor control skills, experience, proprioceptive ability, and personal trade-offs between undershoot cost and overshoot risk.