Beyond the Romberg Ratio: Direction-Resolved Modeling Reveals Reduced Visual Stabilization on Compliant Surfaces in Aging

Background : The Romberg ratio is widely used to assess the visual contribution to balance but is vulnerable to denominator compression: elevated baseline sway in older adults can attenuate ratio-based estimates of visual stabilization. Standard analyses also rarely resolve sway by direction, despite mediolateral (ML) instability being a stronger predictor of falls than anteroposterior (AP) sway. We used direction-resolved mixed-effects modeling to test whether age-related reductions in visual stabilization under proprioceptive challenge are more pronounced in ML than AP. Methods : Center-of-pressure data from 48 younger adults (18–30 years) and 50 older adults (≥60 years) were obtained from a public dataset. Participants completed three 60-s trials under four Firm/Foam × Eyes Open/Closed conditions. Log-transformed peak sway velocity was analyzed at the trial level using separate ML and AP mixed-effects models with age, vision, and surface as fixed effects and by-subject random intercepts and slopes. A combined model added direction as a fourth factor. The velocity-domain ML outcome was specified a priori and supported by nested cross-validation feature screening. Results : On foam, younger adults showed higher visual eyes-closed/eyes-open peak-velocity sway ratios than older adults (ML: 1.43 vs 1.20, p < .001, Cliff’s δ = −0.61), despite lower absolute peak values, consistent with denominator compression. Modeling absolute sensory cost showed a significant Age × Vision × Surface interaction in ML (F(1,881.0) = 26.33, p < .001) but not AP (F(1,881.5) = 2.93, p = .087). On foam, eye opening reduced ML peak sway velocity by 30% in younger adults versus 16% in older adults. The combined model confirmed the pooled three-way interaction (p < .001), whereas the four-way interaction with direction did not reach statistical significance (p = .071). Conclusions : Older adults showed reduced visual stabilization under proprioceptive challenge, most clearly in the mediolateral direction. However, the combined model provided only trend-level evidence that this interaction differed by direction. Conventional ratios can attenuate age-related deficits through denominator compression, whereas direction-resolved modeling more directly isolates absolute sensory challenge costs. The ML foam-related increase in eyes-open peak sway velocity most clearly distinguished the age groups and warrants evaluation as a candidate marker of age-related postural change.