In this study, our hypotheses were partially supported, as during the single-step trials, as the slip-like perturbation magnitudes increased, older people showed a small, yet significantly longer step initiation duration and step duration (R2 = 0.155 and 0.166, Fig. 1A, B). The spatial parameters, i.e., step length and step velocity, showed a larger increase than the temporal parameters of stepping as the perturbation magnitude increased (R2 = 0.272 and 0.254, respectively, Fig. 1C, D). This shows that the first recovery step, especially the spatial parameters, demonstrated a flexible behavior which we defined as the ability to adopt new kinematic movement patterns following changes in task requirements, i.e., the magnitude of perturbation [34]. This suggests that in the single-step trials, where the perturbation magnitudes were relatively low, an adaptive behavior was used, whereas longer and faster recovery steps were performed as the challenge of the test became greater, i.e., the perturbation magnitudes increased. This is supported by Pai and Patton [35] and Pai et al. [36], who demonstrated that the occurrence of a step depends on the interaction between the CoM position and its velocity. Following their model, stepping is necessary if there is a sufficiently high velocity of CoM displacement, even if the vertical projection of the CoM is located within the base of support (BoS) at step initiation.
The above results are in agreement with Vlutters et al. [25], who found that after exposing young participants to mediolateral pelvic pull perturbations at various magnitudes during treadmill walking, the step length, i.e., the foot placement after the perturbation, was adjusted proportionally to the mediolateral CoM velocity. McCrum et al. [36] revealed that a different mode of perturbation showed similar effects. They exposed their participants to repeated trip-like perturbations and found that after repeated perturbations of the left leg, older adults required fewer steps to recover their balance. Furthermore, Epro et al. [38] found that the neuromotor system in older adults shows rapid plasticity to repeated unexpected trip-like perturbations. Luchies et al. [39] suggested that the central nervous system estimates the level of instability following a balance perturbation, selects the appropriate balance recovery response, and pre-plans the stepping behavior, i.e., the use of one large step to recover balance or the use of several small steps, even before the first step is completed. This is supported by the results of Miyake et al. [39], who found that after exposure to repeated trip-like perturbations, the minimum toe clearance was modified toward more precise control and lower toe clearance of the swinging foot, which appears to reflect both the expectation of potential forthcoming perturbations and a quicker recovery response in cases of balance loss.
In the present study, we expand current knowledge by exploring balance reactive responses in cases where multiple steps were needed to recover balance, i.e., where the perturbation magnitudes were relatively high. Multiple-step responses were always performed at higher perturbation magnitudes than in the single-step trials; thus, they were more similar to a real-life balance loss threat. Interestingly, in the multiple-step trials, as the perturbation magnitudes increased, older people did not show changes in the timing of their first step initiation (Fig. 1A'), first step length (Fig. 1C'), or first step velocity (Fig. 1D'), and a small, yet significant increase was found for first step duration (Fig. 1B'). This suggests that at high perturbation magnitudes, i.e., during the multiple-step trials, when participants performed more than one step to recover their balance, their first recovery step performance exhibited a rigid behavior. We define this behavior as the inability to adopt new kinematic movement patterns following changes in task requirements, i.e., at increased perturbation magnitudes, suggesting a more automatic/stereotypical behavior during the first step of the multiple-step trials. However, the total balance recovery parameters, which represent the whole balance recovery, showed a significant increase (Fig. 2A-C), which suggests a flexible behavior. Thus, during the first recovery step of the multiple-step trials, older adults activate pre-programmed kinematic movement patterns of the extra steps, which are depended on the magnitude of perturbation, helping them to effectively recover balance trying to effectively "catch" the moving CoM over the BoS.
To fully understand the balance reactive response to increasing magnitudes, it must be noted that the kinematics of the first recovery step in the single- and multiple-step trials may be influenced by the strategy of the first recovery step. Our findings clearly show that during the single-step trials, as the perturbation magnitude increased, the LLSS was increasingly in use, from about 10% at low perturbation magnitudes to about 40% at the highest perturbation magnitudes (Fig. 2A). Meanwhile, the ULSS strategy decreased in use, from about 100% at the lowest perturbation magnitudes to about 10% at the highest magnitudes (Fig. 2B). With the LLSS strategy, there is a need to first unload the loaded leg, and then to swing the loaded leg sideways to perform the recovery step [2, 17]. It was reported earlier that the step initiation duration was delayed about 200 ms when an LLSS was performed compared to unloaded leg strategies [2, 17]. This, in fact, explains our findings that as the perturbation magnitude increased, the step initiation duration of the single-step trials also slightly increased (R2 = 0.155, Fig. 1A). Due to the unloading phase of the loaded leg in the LLSS, it took a longer time to initiate and complete the recovery step when the perturbation magnitude increased. This also resulted in a small, yet significant increase in the duration of step execution (R2 = 0.166, Fig. 1B). The different strategies that were used as the perturbation magnitudes increased further support the notion that the first stepping response in the single-step trials is flexible in nature and that participants selected a different strategy as the perturbation magnitudes increased (Fig. 2A). Since not only the timing of step initiation increased during the single-step trials, but also the strategies used changed as the magnitudes increased, it appears that older adults pre-plan their stepping performance, and perhaps a learning effect occurred during the experiment.
During the multiple-step trials, in which the older adults were exposed to higher perturbation magnitudes, as the magnitudes increased, they used similar first step strategies. The unloaded leg strategies, i.e., COS and ULSS, were the most commonly used, i.e., these made up about 70% of the strategies used across all perturbation magnitudes, while the LLSS strategy was rarely performed (16.7%) (Fig. 2B). This indicates that when older adults were exposed to large perturbation magnitudes, their strategies of balance responses also show rigid behavior. The COS strategy was performed because high perturbation magnitudes induce a faster CoM displacement to load the standing leg and unload the swing leg, allowing foot-lift of the unloaded leg to "crossover" the loaded leg. Since the initiation duration, step duration, and step length of these recovery strategies was somewhat similar across all perturbation magnitudes of the multiple-step trials (Fig. 1A'-D'), this supports the view that using the COS strategy was the best to control the moving CoM during the multiple-step trials at relatively high perturbation magnitudes, and that this response may not be under volitional control, thus, requiring an automatic response during the first recovery step. The present study results also indicate differences in balance recovery responses in walking [25, 27] versus standing. Both Vlutters et al. [25] and Nachmani et al. [27] found that as the perturbation magnitudes increased during self-selected treadmill walking, there were small yet significant decreases in the timing of the step response. In the present study, however, we found an increase in the timing of the step response as the magnitude increased, suggesting that there are differences in balance recovery responses in walking [25, 27] versus standing. We assume that these differences are related to specific learning effects to a different condition. One foot is usually more loaded during walking than the other leg, allowing the participants to easier react with the unloaded leg and learn how to perform the step even faster along with the experiment. In standing howver, both legs are equally loaded, and the delay in step initiation may be associated with learning how to better control, i.e., decelerate the moving CoM over the BoS.
Several limitations of this study should be acknowledged. First, the results are based on a sample of older people who had a relatively high function level, limiting generalization of these conclusions to frail older adults. Second, in the single-step trials, since the magnitude of unexpected perturbations always occurred in the same order (from low to high perturbation magnitudes), there may have been a learning effect that enabled some of participants to predict that the next perturbation would be higher and to "resist" stepping; thus, the duration of the first step initiation was somewhat delayed. This is supported by earlier studies that showed that when repeatedly applying perturbations, people show rapid plasticity and learning [25, 34,35,36,37,38, 40]. However, this would not have been the case in the multiple-step trials i.e., at relatively high perturbation magnitudes where the strategies and kinematics were similar across all perturbation magnitudes, and participants did not show different coordination patterns; thus, these results were not influenced by the perturbation order. Third, some may argue that the instructions we gave to the participants, i.e., "React naturally and try to avoid falling", may not have been appropriate, and that more constrained instructions such as "Try not to take a step" or "Step as rapidly as possible" would have been more appropriate. But we chose to use unconstrained instructions since these were likely to be more relevant for exploring the ability of the individual to respond in a real-life situation and are, thus, more ecologically valid. Fourth, out of 86 older adults, 18 asked to stop the experiment before completing all of the perturbation trials. Since these participants might have been "the weakest older adults" in our sample, this may have influenced the results in the high perturbation trials i.e., multiple-step trials as in Fig. 2, after magnitude 8, a small gradient decrease appears, especially in the step initiation duration as the magnitude increases.