Back

On attaining and estimating steady walking speed

Wong, J. D.; Darici, O.; Kuo, A.

2026-03-03 neuroscience
10.64898/2026.03.01.708843 bioRxiv
Show abstract

Gait speed is widely used to quantify mobility and functional status, compare how interventions work, and predict health outcomes. Yet despite its ubiquity, the common assumption that short walking tests capture a persons "steady" speed has rarely been verified. A lack of standardization also makes it difficult to compare gait test results across protocols. Here we show that most gait test distances, typically 10 m or less, are too short for young healthy adults (N = 10) to attain steadiness. Using gait measurements for a range of ten short distances, we found that peak speed increases systematically with total walking distance and only gradually approaches the individuals preferred speed at about 10 m. How the individual accelerates is characterized by a "distance constant", also identified from data and only loosely correlated with preferred speed ({rho} = 0.55). The speed trajectories also show how most conventional gait tests systematically underestimate steady speed by about 30%, while dynamic start and stop conditions partially but not fully corrected the bias. Gait speed is not a single fixed property but a dynamic process that depends on the distance available for acceleration and deceleration. Although conventional gait tests remain valuable, most characterize transient rather than steady walking performance. We further show that measurements at three to five distances are sufficient to predict steady-state speed in young adults with zero average bias. This correction requires only a stopwatch and modest additional testing time, making it readily adoptable in both clinical and research settings.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.