Fig. 3

Proprioceptive accuracy at each tested robot distance. In the left column (A, C, E, G), the box plots display proprioceptive error in each of four parameters (EPE, RL, PSR, IDE) for age-matched controls (black) and individuals with stroke (gray). For stroke participants, left-facing triangles indicate individual stroke participants with the left side of the body as their more affected side, and right-facing triangles indicate individuals with the right side of the body as their more affected side. We observed that across each of the parameters in the left column, that stroke participants showed significantly more error compared to age-matched controls at each of the tested robot distances. In the right column (B, D, F, H) data was fit using ordinary least squares for individual participants to determine the intercept (\({\beta }_{0}\)) and slope (\({\beta }_{1}\)) of the average resultant behavior within participant group. Here, the goal was to examine changes in pattern and/or magnitude of proprioceptive error as a function of robot distance. Solid lines indicate the bootstrapped average fit for controls (black) and individuals with stroke (gray). Insets indicate fit coefficients (\({\beta }_{0}\) and \({\beta }_{1}\)), with individual participant coefficient data included on the box plot. Overall, we found that for most parameters tested, the pattern (slope) was similar between age-matched controls and individuals with stroke, but that the magnitude (intercept) was significantly different for all stroke participants suggesting that while stroke increases the overall magnitude, the pattern of error response is preserved when the distance of the reference movement is changed. Interestingly, our measure of how accurately participants matched the length of the movement (PLR, F) showed that stroke participants overestimated the distance of the passive robotic movement at shorter distances [7.5 cm, 10 cm] more than at longer distances [15 cm, 17.5 cm] compared to controls. *p ≤ 0.05, **p ≤ 0.01