Invited Commentary

The article by Kuo and Donelan1 raises a number of essential issues about our understanding of gait. In the larger picture, the article reminds us of the importance of theory in research and practice. Theories underlying gait have an extensive history with anthropologists, who have long debated the evolutionary details of upright walking adopted by humans more than 4 million years ago.2 More recently in a historical review, Baker3 described how scientists made numerous observations of gait and developed theories of human movement using Newtonian mechanics in the 1700s and 1800s, but there was little experimental work to substantiate these theories. The Berkeley Biomechanics Group led by Verne Inman and Howard Eberhart contributed to the creation of modern-day gait analysis, and their group's article, published by Saunders et al4 in 1953, was a major milestone toward the development of conceptual theories underlying gait. A position statement by the American Physical Therapy Association emphasized that, in research, theory can provide an understanding of observable phenomena, yield testable predictions, and motivate new lines of investigation, including novel interventions.5 The theories presented by Saunders et al4 were based on observations and measurements of gait and used general principles such as Newton's Laws. An important aspect of a theory is that it should generate testable hypotheses, which then may lead to refinement of existing theories or creation of new theories.5 Certainly, Saunders et al4 presented a number of hypotheses and predictions that were testable. Despite the passing of more than 50 years since Saunders and colleagues' article,4 Kuo and Donelan point out that there has been little validation of the theories presented in that article. One has to sympathize with the gait researchers of that time period because an analysis of a single stride required 14,000 numerical calculations done by hand, in addition to copious graphical plots.3 Nevertheless, it is somewhat surprising that such prominent theories of gait were not tested more extensively over the past 3 decades when computerized gait analysis came into existence. Thus, I applaud Kuo and Donelan's work to test the hypotheses generated from Saunders et al4 and, in particular, the hypothesis that “fundamentally locomotion is the translation of the center of gravity through space along a pathway requiring the least expenditure of energy.” Kuo and Donelan present a very convincing and elegant example that contradicts this hypothesis on energy expenditure, that is, walking with shorter, but faster, steps minimizes center of mass displacement but results in greater energy expenditure. They also describe another example—walking with the legs bent (crouch gait)—that minimizes center of mass displacement but results in higher levels of energy expenditure. To be fair, it is difficult to imagine that Saunders et al4 meant for their hypothesis to be interpreted using this last example (an extreme walking pattern, albeit common in cerebral palsy). In fact, Saunders et al4 recognized that deviations away from the normal center of mass path caused inefficiencies. In a description of the gait of individuals with below-the-knee amputations, they stated, “Because of the minor changes in direction from the smooth sinusoidal pathway of the center of gravity, high accelerations are required which dissipate energy and make such deviations very costly.”4 These examples serve as a reminder that a theory should permit generalization, but if the limits of generalizability are not defined, it leaves the theory open to debate and to potential misinterpretation. Kuo et al introduce the principle of the step-to-step transition cost of human walking, which is the combined costs of both limbs to redirect the center of mass. This principle can explain the trade-off between speed and step length not only in the anterior-posterior direction but, impressively, also in the medial-lateral direction (step width). Undoubtedly, we will see this model tested and refined to expand boundaries. There seem to be unlimited hypotheses in which the step-to-step transition cost of human walking can be tested in regard to aging and rehabilitation. For example, are the shorter steps observed in older adults a result of a need for less redirection of the center of mass velocity and potentially more stability? Dynamic walking models have been used to examine walking stability and balance deficits. From these models, Kuo et al suggest that external stabilization could be manipulated as part of a rehabilitation program. Clinicians would argue that such practice is already in place (manipulating the ground surface with foam, providing physical assistance from a therapist, and progressing to handrails and assistive devices). A useful application would be to use these dynamic models to inform clinicians of the calculated stability of one therapeutic practice (eg, walking using an overhead harness) versus another (eg, walking and grasping a rail). It also would be of interest to model the gains in walking stability derived from sensory information when simply touching a rail6 or when using a shoe with enhanced foot-sole properties,7 as these sensory protocols have been shown to improve walking stability. In earlier work, Bauby and Kuo8 added “noise” to their dynamic walking models, which generated step variability, and thus simulated a model for balance deficits. Interestingly, step variability was shown to discriminate between “fallers” and “nonfallers” 3 decades ago,9 but there has been a recent surge in the use of this measure. Gait researchers have always known that step-to-step variability was greater in their pathological populations, with one practical solution being to collect more trials to reduce variability. It is heartening to know that individual trial data have value, rather than simply compressing all this information into a single “average” step profile. It also is promising to see psychometric properties being established for these variables, with a recent article on the clinical meaningful change of gait variability.10 Dynamic models have the potential to explore mechanisms underlying this measure, and Kuo's group has attempted to understand the interactions of the active versus passive contributions underlying step-to-step variability. As with all models, the limits of generalizability must be defined. The dynamic walking models by Kuo and Donelan are applicable only where the limbs behave like pendulums. Many pathological gait patterns—such as genu recurvatum, crouch gait, and hip circumduction—that are common in neurological conditions might not emulate pendular activity, and thus the step-to-step transition cost might not be valid. In addition, dynamic models, especially of the ankle-foot complex, that utilize the nonsagittal planes are still in their infancy. Recently, there have been attempts to use multi-segmented, multi-directional ankle-foot models, which should provide new experimental data.11 As Kuo and Donelan highlight, computational models cannot be expected to make quantitative predictions about every case. We can expect that, together, computational and clinical researchers will generate new and pertinent theories about gait. The late Dr Jules Rothstein emphasized that theories must not become dogma seen as “truths.”12 The testing and refining of these theories will serve to develop and advance the field of gait research and its clinical applications.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.