The present review focuses on the physiological states of spinal networks, which are stochastically modulated by continuously changing ensembles of proprioceptive and supraspinal input resulting in highly redundant neural networks. Spinal epidural interfaces provide a platform for probing spinal network dynamics and connectivity among multiple motor pool-specific spinal networks post-injury under in vivo experimental conditions. Continuous epidural low-frequency pulses at low intensity can evoke motor responses of stochastically changing amplitudes and with an oscillatory pattern of modulation. The physiological significance of this oscillatory pattern, intrinsic to "resting" spinal networks and observed in both uninjured and injured locomotor circuits, is unclear. This neural variability among spinal networks appears to be a fundamental mechanism of the network's design and not a "noise" interfering with movement control. Data to date also suggest that the greater the level of stimulation above motor threshold, the greater the loss of modulation over the motor output that is physiologically provided by interneuronal networks, which integrate naturally occurring proprioceptive and cutaneous input generated during movement. Sub-motor threshold spinal electrical stimulation experiments demonstrate a range of functional improvements of multiple physiological systems when used in concert with sensorimotor training after spinal cord injury. Although our understanding of the systemic, cellular and molecular modulatory mechanisms that trigger these activity-dependent adaptive processes remain incomplete, some basic physiological principles have evolved, at least at the systemic and neural network levels and to some degree at the cellular level.