The locomotion involves a set of structures hierarchically distributed. Skeletal muscles are needed to counteract gravity constraints and allow movements. The coordinated firing of the motoneurons controlling these muscles is organised at the level of the spinal cord by neural networks, called central pattern generators (CPG) [68]. A rostral CPG, located between C7-T1, controls the fore limbs [3], and a lumbar CPG, located between L1–L5 controls the hind limbs [13], [14] and [36]. CPGs are made of two lateral hemi-neural networks, that rhythmically stimulate a pool of motoneurons through glutamatergic synapses and inhibit the motoneurons of the ipsilateral antagonist muscles and the contralateral agonist muscles through glycinergic commissural interneurons (see [26] for a recent review). This peripheral organisation is capable of producing autonomously rhythmic patterns of movement, but a central command, involving various parts of the brain, initiates and controls the type of movement (speed, gait, etc.) by modulating CPG activities through four main descending pathways. The reticulospinal pathway mediates locomotor signals issued from the mesencephalic locomotor region (MLR), that receive axons of neurons coming from the basal ganglia; the vestibulospinal pathway mediates signals coming from the vestibular nucleus; the rubrospinal pathway mediates signals from the red nucleus; the corticospinal pathway originates form the motor cortical areas. The development of all these pathways is required to perform the whole adult behavioural repertoire, but they are not equally required to perform locomotion. The reticulospinal pathway adjusts postural muscles and produces basic motor activities, the vestibulospinal pathway is involved in the control of equilibrium, the rubrospinal pathway is in charge of precise and automated movements, the corticospinal pathway is in charge of the fine and precise adjustment of voluntary movement. The two former pathways are required for basic expression of movement; the two latter are involved in movements that are more complex and are not mandatory for basic motor outputs [45]. For instance, decerebrated animals with only the MLR activating the reticulospinal pathway are capable of automatic locomotion.

In addition to the motor command, the development of control sensory pathways needs to be achieved to allow the body stabilisation during movement [25].

This chain of motor output, motor control and postural control must be fully operational to allow the adult locomotion, and missing elements at any level can seriously alter the motor output. These various levels are not equally mature at birth, and in successive days, but they are at least partly functional, and the interactions of these differently matured structures on the motor abilities are not straightforward. The timetable of these various elements of the organization of locomotion is summarised below (Fig. 1).

The timetable of the development of the motor function in neonate rats refers to a set of studies that recorded the spontaneous activities of pups placed in a neutral enclosure (open field or corridor) for a few minutes [2], [5], [24] and [67]. These studies agreed on a progressive development based on the acquisition of motor capacities following an antero-posterior and proximo-distal gradient. Geisler et al. [24] distinguished three phases: an early postnatal period until PND4–5, when only head lifting and rooting movement occurred; from PND6 to PND12–13 rats were able to rise their ventral body from the floor while walking a few steps; in a third period, from PND16, complex motor patterns as locomotion, rearing and grooming acquired adult and fluent characteristics.

Whereas these studies clearly showed a progressive acquisition of locomotion in rats, the conditions for observations did not allow detecting what pups are potentially capable of performing. Recent observations of pups in relation to their mother showed that maternally directed behaviours involve, early after birth, righting and pivoting movements [20] as well as motor activities, including crawling, much earlier than previously assumed [52]. Altman & Sudarshan [2] themselves admitted that brief periods of patterned motor responses could appear earlier under high level of activation. They mentioned, for instance, that placing the pups on a refrigerated surface elicited a walking-type of locomotion with elevated trunk as early as PND4. These observations support the idea that spontaneous activities of pups do not reflect their potential capacities.

Electrophysiology studies of rats' spinal cords isolated early after birth showed the existence of rhythmic neuron firing in the CPGs. The recording focused on the lumbar CPG [14] and [36] and showed stable alternating bursts between agonist contralateral motoneurons and antagonist ipsilateral motoneurons. More recently, stable rhythms were also recorded in the forelimb CPG of mice, and links between rostral and caudal CPG were showed to be functional at birth [3]. The CPG activities recorded in vitro are defined as fictive rhythmic locomotion because they are produced without muscle activity [55], but they are commonly considered as an expression of potential locomotion even though the frequency is very low.

Early coordinated forelimb-hindlimb motor activities were also recorded in vivo early after birth (PND0–4) during swimming [4] and [15]. The swimming paradigm allows producing early alternate limb beating, by relaxing the postural constraints, but the meaning of these gaits suffers from a strong limitation, because swimming does not require the same fore- hind limbs coordination than walking. This could explain the higher variability and the progressive reduction in forelimb coupling in swimming neonate rats [15]. The ambiguity on the nature of the movement of pups after water immersion strongly reduces the pertinence of the swimming paradigm for studying the evolution of limb coordinations. An alternative protocol consists in placing the rat in a harness and stimulating the locomotion by a subcutaneous injection of L-Dopa [43], [56] and [63]. L-dopa induces regular rhythmic movements in both fore and hind limbs for about half an hour, revealing therefore the ability of the neuromotor control system to produce regular locomotion bouts. This protocol was widely used to investigate motor activities in rats aged from PND3 [32] to adultness [43]. The pharmacological L-Dopa injection probably stimulates dopaminergic locomotor pathways at the level of the brainstem [30] and is equivalent to a chemical decerebration [57].

A modified air-stepping protocol consisted in placing a tube containing material of the nest in front of the nose of the rat to stimulate an olfactory mediated homing behaviour (Fig. 2). The olfactory stimulation was used in rats as young as PND0 [22] and was efficient in premature rats whose birth date was experimentally anticipated by one day (E20). The kinematic study of the olfactory induced air stepping revealed that rostral and lumbar CPGs were able to modulate their respective frequencies as soon as PND0. The olfactory stimulation produced slower movements compared to L-dopa injection, and was capable of slowing down L-Dopa induced movements in PND3 rats [32]. As the olfactory stimulation involves most probably the cortical areas, this demonstrated that higher centres are capable of modulating the dopaminergic pathways.

The experiments on early air stepping showed that CPGs were capable of maintaining stable rhythmic movements over a long period of time; the contralateral and even the ispsilateral CPG reacted reciprocally, showing therefore that some connection existed between anterior and posterior CPGs, and cortical control was capable of modulating the motor activities. Altogether, these results argue for a capacity or the neuromotor system to control efficiently locomotor movements early after birth.

Air stepping releases the postural demand, however, and leaves open the possibility that postural control and musculoskeletal properties are not enough developed to support the body. Stimulating ground walking in pups by means of the olfactory stimulation answered this latter question [31]. The neonate rats show a strong tendency to move towards the nest odour presented in front of them. This reaction induced a walking-like movement characterized by an attempt to raise the belly off the ground and strong coordinated legs pushes, as early as PND3. This experiment demonstrated that neonate rats could show walking-like bouts very much earlier than previously suspected. When the nest odour was presented in a tube, the rat entered the nose into the tube, and pushed it. The range of speed and step amplitude measured between PND 4–9 corresponded to the performance previously recorded in PND 11–17 spontaneously walking rats [67], and showed that rats are potentially capable of performing much better than that they actually do. Placing the head in the tube allowed to get longer stepping sequences, by preventing the pup's lateral head oscillations to loose contact with the odour, but also improved equilibrium. At this age, the head amounts for more than 20% of the body weight and can unbalance the pups. This latter point can be a strong limiting factor for the pup to move spontaneously, especially as the trunk stabilisation is not achieved (see above), nevertheless, pups motor capabilities appear as much more developed as was previously assumed.

As was shown in this short review, the development of motor functions does not parallel the neuroanatomical changes. The development of the structures underlying locomotion is not fully achieved at birth, and will continue even after pups perform adult-like walking. On the other hand, the basic elements underlying locomotion are ready in a few days after birth, well before spontaneous walking is observed. Undoubtedly, the reorganization and refinement of initially established neural connections [44] and [65], and the inefficiency to cope with postural requirements [25] and [44] are limiting factors that contribute to the delayed motor output in neonate rats, but they cannot satisfactorily explain it, because pups are potentially capable of performing better than they spontaneously do. On the other hand, neural connections quickly develop around PND15–16 when a fluent walking pattern emerges and adult-like type of postural control starts to develop [25]. This coincident emergence of fluent walking and sudden change in the rate of neural connections was supposed to have various causes: either locomotion appears because the neural connections have reached a level of maturation, or connections correspond to a final tuning related to a new stage of the progressive maturation of locomotion. On the other hand, both structures and functions could also have their operational stage switched on by a trigger somewhere in the developmental process.

Behavioural observations support this latter interpretation. Olfactory stimulation with nest odour can activate air stepping and even ground walking in pups [22] and [31]. The olfactory stimulation is assumed to trigger a homing reaction. This behaviour allows bringing the pups back to the nest and has a high survival value. Besides, homing capacity was used to test various capacities in pups aged from 2 to 14 days (e.g., [34]). In these studies, pups were placed in the opposite corner of the cage and the percent of return was measured, however locomotion was not the main concern of these studies, and movement was not particularly studied.

Resolving the discrepancy between the development of structures and functions needs to refer to the ecological significance of altriciality. It would be ecologically irrelevant for neonate to move far from the nest: they take the risk of being lost and they would have difficulties to return in case of danger, while the mother would expend a lot of energy recovering dispersed pups. A more adaptive strategy consists in developing a control system to inhibit the pups' locomotor activities, while preserving an efficient homing behaviour. We can therefore hypothesize that the development of structure and their motor output are disconnected by a silencing mechanisms, of unknown nature, that inhibit motor activities, except for situation of survival emergency (e.g., homing). The sudden and almost concomitant apparition of exploratory behaviour, eye opening, and a burst in neural connections could result from this mechanisms' switch off at the end of the second postnatal week.

The idea that the delay in the motor output is not directly linked to the immaturity of motor properties, but is under a central control is reminiscent of a hypothesis presented earlier by Oakley and Plotkin [48]. On the assumption that maturation of mesencephalic arousal system results in a direct increase in general motor arousal, and is transiently reflected by a sharp increase in locomotor activity [11], these authors supposed that motor activity happens at different ages among species, depending on their degree of brain maturation at birth. To test their hypothesis they compared the timing of occurrence of the peak of spontaneous locomotor activity in three small mammal species with varying levels of maturity at birth, namely the rat, the rabbit and the guinea pig (Fig. 3). The peak did not appear in guinea pig for which mesencephalic maturation was supposed to happen in utero, and occurred at PND5 and PND18, in rabbit and rat, respectively. It is interesting to notice the comment by the authors that rabbits showed a poorly coordinated and clumsy movement during their peak phase of activity and therefore did not mirror the maturity of the motor system. Some prerequisites, like caudo-rostral development of the brain and comparable interspecies developmental timetable, were probably not met in their study. Their hypothesis, however, has an heuristic value, assuming that important changes in the motor output are not the image of the level of maturation of the motor system itself but correspond to a switch in the central command of movement that released the inhibition of motor output.

Another point was pertinently disputed by Henderson [27] and [28], who suggested that a high rate of locomotor activities would be maladaptive for altricial species. On the contrary, a quite strong selective pressure would have produced animals having a low activity level during their first weeks because of the high mortality risks early in life. Low activity and reliance on highly efficient parental retrieval response would be genetically selected. Henderson compared the early motor activity in wild mice, trapped outside and maintained in the laboratory for three generations, with the performance of inbred mice, with the underlying hypothesis that the selective fitness of low activity is relaxed in laboratory condition (Fig. 4). A significant difference in the early motility of wild and inbred mice suggested that low activity level was indeed selected under environmental pressure. It is also noticeable that mobility varies between pups of various inbred mice [27]. Likewise, the various laboratory rats' strains show different timetables of motor development [24]. These remarks are consistent with the hypothesis that protective conditions of laboratory relaxed the selective pressure on delayed motricity.

These two examples support the assumption that low motor performance in the altricial murid species does not reflect the immaturity of the motor system, but instead results from an adapted silencing of the motor output in view of maintaining the pups in the nest area. This does not mean that pups are early capable of walking, obviously they are not, and even the characteristics of walking-like behaviour induced by olfactory stimulation evolved with age [31]. It would be misleading, however, to speculate about a straight link between the occurrence of a motor function and the development of the underlying structures. The functions are not expressed when the underlying structures are ready, they are expressed when the context for use is appropriate. The development of locomotion is the result of a dialogue between structures and functions, in the context of adaptability. The neuromuscular activities are involved in the tuning of underlying structures. In the same way, the behavioural output is dependent both on the maturation of structures and the adaptive context for expression of a given behaviour.

From an evolutionary point of view, there is no qualitative difference between altrical and precocial strategy, both being successful in the context of different ecological strategies, where the trade-off among various life history traits parameters varies. Though the first mammals were supposed to be altricial, this probably corresponded to an adaptive solution for small species with a large reproduction rate, rather than a phylogenetic stage. On the other hand, it is noteworthy that the most evolved species (men and apes) mix both precocial and altricial characteristics giving them a small reproduction rate, but great autonomy, and high plasticity and ‘learning’ process in the young.