Evaluating feedback time delay during perturbed and unperturbed balance in handstand

•Estimates of feedback time delay were evaluated against EMG latencies in handstands.•Only delayed regression models provided estimates that were in agreement with EMG.•Cross correlations provided unrealistically low estimates and are not recommended.•Feedback time delay was much larger in unperturbed handstand than in perturbed.

Feedback delays in balance are often assessed using muscle activity onset latencies in response to discrete perturbations. The purpose of the study was to calculate EMG latencies in perturbed handstand, and determine if delays are different to unperturbed handstand. Twelve national level gymnasts completed 12 perturbed and 10 unperturbed (five eyes open and five closed) handstands. Forearm EMG latencies during perturbed handstands were assessed against delay estimates calculated via: cross correlations of wrist torque and COM displacement, a proportional and derivative model of wrist torque and COM displacement and velocity (PD model), and a PD model incorporating a passive stiffness component (PS-PD model). Delays from the PD model (161 ± 14 ms) and PS-PD model (188 ± 14 ms) were in agreement with EMG latencies (165 ± 14 ms). Cross correlations of COM displacement and wrist torque provided unrealistically low estimates (5 ± 9 ms). Delays were significantly lower during perturbed (188 ± 14 ms) compared to unperturbed handstand (eyes open: 207 ± 12 ms; eyes closed: 220 ± 19 ms). Significant differences in delays and model parameters between perturbed and unperturbed handstand support the view that balance measures in perturbed testing should not be generalised to unperturbed balance.