Locomotor Primitives in Newborn Babies and Their Development

See allHide authors and affiliations

Science  18 Nov 2011:
Vol. 334, Issue 6058, pp. 997-999
DOI: 10.1126/science.1210617


How rudimentary movements evolve into sophisticated ones during development remains unclear. It is often assumed that the primitive patterns of neural control are suppressed during development, replaced by entirely new patterns. Here we identified the basic patterns of lumbosacral motoneuron activity from multimuscle recordings in stepping neonates, toddlers, preschoolers, and adults. Surprisingly, we found that the two basic patterns of stepping neonates are retained through development, augmented by two new patterns first revealed in toddlers. Markedly similar patterns were observed also in the rat, cat, macaque, and guineafowl, consistent with the hypothesis that, despite substantial phylogenetic distances and morphological differences, locomotion in several animal species is built starting from common primitives, perhaps related to a common ancestral neural network.

Rich repertoires of complex behaviors are created from the flexible combination of a small set of modules (19). A locomotor module is a functional unit—implemented in a neuronal network of the spinal cord—that generates a specific motor output by imposing a spatiotemporal structure to muscle activations (1, 3, 6). Each module involves a basic activation pattern (temporal structure) with variable weights of distribution (spatial structure) to different muscles (Fig. 1A). Thus, the electromyographic (EMG) activity of trunk and leg muscles during human adult locomotion is explained by few basic patterns independent of locomotion mode, direction, speed, and body support (3, 4, 8). These patterns may be regarded as locomotor primitives in a computational sense, because they are the building blocks from which locomotor activities are constructed. But are they primitives also in a developmental sense—that is, are they related to precursors present at or before birth?

Fig. 1

(A) Schematic of motor modules. Simulated example of muscle activity profiles as weighted sum of basic patterns: mi(t)=∑jpj(t)wij. The outputs of the first (green), second (blue), and third (magenta) modules are summed together to generate overall muscle activation (black envelope). (B) Illustration of a step cycle in a neonate.

There are alternative hypotheses about the development of motor patterns (5, 912): (i) Adult-like stereotyped patterns are selected from an initially much larger number of motor patterns. (ii) Primitive patterns are discarded, replaced by entirely new patterns. (iii) Primitive patterns are retained and tuned, while new patterns are added during development. Clearly, these hypotheses imply different constraints on the development of the underlying neural networks.

To discriminate among these possibilities, we compared the locomotor output of neonates with those of toddlers, preschoolers, and adults (13). Stepping can be evoked in newborns, but it is very irregular and typically disappears ~4 to 6 weeks postnatally (unless trained or supported by water buoyancy) to reappear at ~6 to 8 months, evolving into intentional walking (1012). Meanwhile, size, mass, and proportions of the body segments have changed dramatically, as have the biomechanical requirements for locomotion.

We elicited stepping in neonates supported on a table (Fig. 1B and movie S1) and recorded kinematics, contact forces, and EMG activity from up to 24 muscles simultaneously. EMG was modulated sinusoidally with the step cycle, with many extensor muscles coactivated over most of stance, and flexor muscles coactivated mainly during swing (Fig. 2A). Toddlers exhibited more complex, individuated EMG profiles, which were similar to those of preschoolers and adults, but sinusoidal modulation persisted in some muscles. Motoneuron outputs gradually became pulsatile, as in adults.

Fig. 2

Recorded EMG profiles and derived basic patterns. (A) Ensemble-averaged (across all subjects of each group) EMG profiles during the step cycle, aligned with stance onset in the right leg. Shaded areas are the experimental data, and black traces the profiles reconstructed as weighted sum of the patterns extracted from the ensemble. ES, erector spinae; GM, gluteus maximus; TFL, tensor fasciae latae; Add, adductor longus; HS, hamstrings; VM, vastus medialis; VL, vastus lateralis; RF, rectus femoris; MG, gastrocnemius medialis; LG, gastrocnemius lateralis; Sol, soleus; TA, tibialis anterior. (B) Basic patterns from averaged (across steps) EMG profiles in 10 subjects of each group (black). Patterns from ensemble EMG averages (colored). (C) Normalized weights of the ensemble patterns in color scale.

To quantify the basic patterns underlying bilateral muscle activations, we applied a nonnegative-matrix-factorization (1314) to the EMGs pooled across steps. Because the EMG profiles were similar relative to a cycle regardless of its period (fig. S1), all data were normalized to the cycle. In each group of subjects, we could reproduce the EMG profiles of all recorded muscles by combining two to four basic patterns (Fig. 2B) with appropriate weights (Fig. 2C). Patterns were ordered on the basis of the relative timing of the peak. Two sinusoidal-like patterns accounted for 89 ± 6% (mean ± SD) of variance across neonates (fig. S2A). One pattern peaked at ~30% of the cycle, and the other one at ~75% (fig. S2C). For both, the duration at half-maximum was ~35% of the cycle (fig. S2B). In toddlers, we found two kinds of basic patterns, and a total of four patterns accounting for 90 ± 4% of variance. The peak durations of patterns 1 and 3 were ~20% of the cycle. The shape of patterns 2 and 4 was similar to that of the neonate (neonate-toddler correlation r = 0.88 and 0.98, respectively), as were peak duration (~35%) and timing (25 and 75%, respectively). Therefore, we infer that patterns 2 and 4 are retained from the stage of newborn stepping, whereas patterns 1 and 3 develop after that stage. In preschoolers, all four patterns (accounting for 91 ± 2% of variance) showed transitional shapes, the average peak timing being intermediate between that of the toddler and the adult. Patterns 2 and 4 were quite variable across preschoolers, with a time shift of the peak relative to the cycle that correlated with age (r = 0.86 and 0.79 for patterns 2 and 4): the older the child, the closer the pattern to the adult. These transitional patterns strongly suggest a continuous development of the corresponding motor modules. In adults, we also found four patterns explaining 89 ± 4% of variance, with peak duration 15 to 20% of the cycle. The timing of patterns 1 and 3 was similar to that in toddlers and preschoolers. Patterns 2 and 4 were timed on each foot contact, instead of midstance or midswing as in neonates and toddlers. The results depended little on the specific EMG decomposition technique (fig. S3).

The neural underpinnings of these developmental changes remain speculative (Fig. 3A). Neonate stepping mainly reflects spinal and brainstem control, as shown by stepping anencephalic infants (10). Subsequent development stems from a growing integration of supraspinal, intraspinal, and sensory control (1012). The lack of a specific activation pattern timed on foot contact in neonates could depend on immature sensory and/or descending modulation of stepping. Indeed, in the absence of sensory and descending modulation (e.g., during fictive locomotion), the spinal circuitry of animals may produce sinusoidal-like patterns (15), similar to those of human neonates. The addition of basic patterns in the first months of life may imply a functional reorganization of interneuronal connectivity, additional functional layers in the spinal central-pattern-generators (CPGs), and/or more powerful descending and sensory influences on CPGs (9, 15, 16).

Fig. 3

Relationship between neural control modules and key biomechanical features of locomotion. (A) Motor rhythms and patterns generated by CPGs under descending and sensory influence (conceptual scheme). Activation patterns are distributed to different motoneuronal pools via a premotor network, dynamically reconfigurable through flexible weights. Intersegmental coordination (B), stick diagrams (C), and shifts of the center of pressure (D) are color-matched to the corresponding activation patterns. In toddlers, the first pattern (red) is timed at foot strike, the second (violet) at weight acceptance, the third (cyan) at forward propulsion, the fourth (green) at lift-off. In newborns, there are only two patterns, corresponding to the second and fourth of toddlers. Planar covariation of thigh elevation angle versus shank and foot angles identifies counterclockwise loops, with foot strike and lift-off at the top and bottom (B).

We found a good correlation between the developmental changes of the patterns and parallel changes in locomotion biomechanics (Fig. 3, B to D). In neonates, two activation patterns were sufficient for planar covariation (4) of segment motion (Fig. 3B), but posture was flexed, feet were lifted high during swing (Fig. 3C), and stereotyped heel-to-toe shifts of pressure during stance were absent (Fig. 3D). Pattern 2 provided (partial) body support during stance while pattern 4 drove the limb during swing, but there was no specific activation at either touch-down or lift-off. In toddlers, the new patterns (1 and 3) were timed at touch-down and lift-off, providing shear forces to decelerate and accelerate the body. The other two patterns (2 and 4) were similar to those of the neonate, as were the corresponding kinematic and kinetic events. In adults, the four patterns were accurately timed around the four critical events of the gait cycle: heel strike, weight acceptance, forward propulsion, and lift-off (3, 4, 8). Accordingly, the legs were kept much straighter than in children, and the planar covariation of segment motion was adjusted to fully exploit the inverted pendulum mechanism, with minimum muscle activity to support and propel the body. Center of pressure shifted smoothly heel-to-toe.

Habitual erect, bipedal mode of locomotion sets humans apart in the animal kingdom and may have been a crucial initiating event in human evolution. Given the unique biomechanical features of human locomotion, it is not surprising that its muscle activity profiles differ markedly from those of other animals. Also, the developmental time course can differ (17): Small-brained animals tend to walk independently shortly after birth, whereas independence is achieved by human infants only after ~1 year. These observations beg the question: Are the basic motor patterns unique to humans, or are they shared by other vertebrates with legged terrestrial locomotion? We applied the same analysis used in humans to the published recordings available from a few other mammals (rat, cat, rhesus monkey) and a bird (guineafowl) (13). Guineafowls are bipedal, whereas the other three species are quadrupedal. In neonate stepping rats, we found two patterns (accounting for 81% of variance) nearly identical to those of human neonates (human-rat correlation r = 0.94 and 0.98 for patterns 2 and 4, Fig. 4A). In all examined adult animals, we found four patterns with two types of modulation (accounting for 90 to 93% of variance, Fig. 4B): The two shorter pulses overlapped patterns 1 and 3 of human toddlers, whereas the two longer pulses overlapped patterns 2 and 4 of toddlers (toddler-animal correlation r = 0.94 ± 0.04). The human developmental path appears to diverge from that of other animals after the stage of independent locomotion in toddlers, perhaps to accommodate discrete arm movements (such as reaching and grasping an object) within rhythmic locomotion (4). Upper limb and trunk muscles can be engaged independently of locomotor activations (involving some of the same muscles) only if the latter occur as the brief events of adults (Fig. 4C).

Fig. 4

Comparison of activation patterns for locomotion in humans and other vertebrates. (A) Average patterns of human newborns are superimposed on those of neonatal rats; (B) patterns of human toddlers are superimposed on those of adult rats, cats, monkeys, and guineafowls; and (C) patterns of human adults stand alone.

Our observations are consistent with the idea that motor patterns are highly conserved across substantial phylogenetic distances and morphological differences of vertebrates (18). Thus, we identified similar activation patterns in animal species, some of which probably diverged more than 100 million years ago (17). Locomotion in humans and other species may be built starting from common, largely inborn primitives. CPGs coordinating muscle activity in legged vertebrates may have emerged during evolution from a common ancestral network (19). Conserved locomotor primitives may be a systems-level analog of the core modules of evolutionary biology (16, 20). Genes, proteins, cell types, and networks are conserved across several species with different lineages, constructing diverse natural forms and functions. Old elements are not discarded in favor of a totally new design, but they become adapted to solve novel problems efficiently.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S4

Table S1

Movie S1

References (2127)

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: We thank W. Gerstner, M. Molinari, P. Viviani, and M. Zago for helpful comments on earlier versions of the manuscript. This work was supported by the Italian Ministry of Health, Italian Ministry of University and Research (PRIN grant), Italian Space Agency (DCMC and CRUSOE grants), and European Union FP7-ICT programs (MINDWALKER grant 247959 and AMARSi grant 248311).
View Abstract

Stay Connected to Science

Navigate This Article