Perceiving inter-leg speed differences while walking on a split-belt treadmill

Abstract

Walking is one of the most common forms of self-motion in humans. Most humans can walk effortlessly over flat uniform terrain, but also a variety of more challenging surfaces, as they adjust their gait to the demands of the terrain. In this, they rely in part on the perception of their own gait and of when it needs to be adjusted. Here, we investigated how well N = 48 participants detected speed differences between two belts of a split-belt treadmill. As participants walked at a constant speed, we either accelerated or decelerated one of the belts at quasi-random intervals and asked participants to judge their relative speeds in a two-alternative forced-choice task. Using an adaptive psychophysical procedure, we obtained precise perception-threshold estimates for inter-leg speed differences after accelerating or decelerating one belt. We found that most participants could detect even very small speed differences, with mean threshold estimates of just over 7% for both perturbation types. These were relatively stable within, but highly variable across participants. Increased-speed and decreased-speed thresholds were highly correlated, indicating that despite different biomechanics, the detection mechanisms might be similar. This sheds light on how perceiving their own motion helps humans manage interlimb coordination in perturbed walking.

Keywords: Just-noticeable differences; Perception and action; Self-motion; Sensorimotor adaptation; Walking.

Fig. 1

Experimental setup and procedure. (a): Virtual environment with a participant walking on the split-belt treadmill along an endless road scene, secured with a safety harness while holding the response controller. Infrared cameras around the treadmill recorded marker positions. The question used for the 2AFC task was displayed on the screen (German: „Links oder rechts schneller? “, translating to “left or right faster?”), indicating participants to give a response. (b): Procedure of the experimental blocks, each starting with a baseline phase, then altering between perturbations (block-wise either speed increases or decreases) and short baseline periods for 5 min and ending with baseline walking. The number of perturbed steps depended on the response timing with a maximum of 10 steps, baseline steps were randomized between 6 to 9 steps between perturbation trials.

Fig. 2

Mean JNDs per participant for increased and decreased speeds. Black semi-transparent dots show the individual data, the larger red dots the overall mean. Inline descriptions at the red lines indicate the 5% and the 95% quantiles, respectively. Shaded lines connecting threshold of the same participant.

Fig. 3

Correlations of mean JNDs for first and second occurred blocks of increased-speed perturbations (panel 1) and decreased-speed perturbations (panel 2) as well as for overall increase and decrease blocks (panel 3). Each dot represents one participant, the red lines indicate the Deming corrected regression line. Solid black line indicates unity.

Fig. 4

Trajectories of threshold estimates and mean steps to response. (a): Mean trajectories of the absolute threshold estimates for increased-speed perturbations (green) and decreased-speed perturbations (red) for each trial per block. The x axes end with the minimum number of trials presented to any participant in any block (this number could vary as we fixed the time and not number of perturbations per block). Shaded areas indicate ± 1 between-participant SEM. Threshold estimates started at 10% and rapidly and roughly asymptotically approached range of the final JND for both perturbation types. Dashed lines show the mean width of the 95%-confidence intervals of the estimates for each trial, computed from the QUEST’s probability density function. (b): Mean steps to response for each trial, depending on the perturbation type. Trajectories did not differ between increases (green) and decreases (red).