As running increases in severity we are made consciously aware of mounting ‘threat’ through increasingly discomforting interpretations of arising sensory information (Marcora et al., 2009; Smirmaul, 2012). At the whole-body level growing discomfort influences psycho-emotional state, amplifying perceptions of anxiety, ‘pain’ and diminished attention which in turn intensify the inner conflict between motivational drive and perceived effort that, collectively, we interpret as mounting ‘fatigue’ (Marcora et al., 2009; Seay et al., 2011; Smirmaul, 2012). At the local level, muscle activation patterns are subtly modulated to offload sensitized tissues, thereby moderating regionalized discomfort and alleviating tissue irritation (Mizrahi et al., 2000; Gerlach et al., 2005; Seay et al., 2011). Through these mechanisms our interpretation of arising psychobiological discomfort informs us of increasing risk—of impending tissue damage, elevating metabolic costs, increasing neural processing demands and cognitive effort—, thereby providing a direct means through which the perceived relevance of changing sensation directly influences running behavior (Marcora et al., 2009; Wolpert et al., 2011).

Prompted by subtle, but persistent, sensory signals the CNS continually searches for economic trade-offs between desired outcomes, available resources and discomforting perceptions of ‘threat’ (Hodges and Tucker, 2011; Miller et al., 2012). As we accumulate running experiences, we learn to more precisely triangulate between sensory feedback, feedforward activation and desired running outcomes and gravitate toward coordinative solutions more satisfactorily resolving these multiple competing constraints. Progressively, with practice, sensory information and muscular activation strategies co-evolve into a seamlessly integrated sensorimotor system: whereby changes in sensation directly modulate muscular activations, and changes in activation directly modify sensation (Wolpert et al., 2011; see Figure 1). Through this elegantly efficient process, sensory feedback information and feedforward activation instructions become irrevocably mutually entangled: preserving running robustness within acceptable limits through an integrated sensorimotor process of ‘self-organizing optimality’ (Glazier and Davids, 2009).

Biological lifeforms are reflectively characterized as complex adaptive systems. Complex: as the behaviors of individual components are inextricably linked to those of multiple others through arrays of processes, cycles and regulatory feedback loops. Adaptive: as the behaviors and collaborative outputs of collections of components flexibly modify their concerted contributions to best fit current context (Manor and Lipsitz, 2013).

Each individual entity within the complex organism is linked, physically or functionally, to every other. Nevertheless there remains an evident modularity, whereby collections of elements are more densely networked to each other than to elements within other modules (Whitacre, 2010; Mason, 2015). All modules are inter-connected, yet are simultaneously partially insulated and functionally semi-autonomous. Modularity, accordingly, facilitates robustness as modules can evolve, reshape, rewire, and repair in tandem, or independently, without jeopardizing survivability of the entire organism (Maleszka et al., 2014; Mason, 2015).

Modularity is a fundamental neuro-biological organizing principle, greatly simplifying otherwise overwhelmingly disordered complexity. Related modules exhibit extensive functional overlap, such that alliances of neural networks and peripheral tissues can spontaneously modify behaviors to achieve equivalent ‘outputs’ through a multiplicity of pathways. This functional agility is often conflated with redundancy, but is perhaps more reflectively termed degeneracy (Glazier and Davids, 2009; Mason, 2015; Seifert et al., 2016). Degeneracy describes the ability of alternate structural pathways to achieve similar functional outcomes in one context, or dissimilar functional outcomes in divergent contexts (Seifert et al., 2016). Degenerate systems are composed of diverse elements, capable of alternately fulfilling similar or overlapping functions and are fundamental facilitators of complexity, robustness, and evolvability (Whitacre, 2010; Maleszka et al., 2014; Mason, 2015). Redundancies, in contrast, occur when sub-sets of identical elements combine to achieve similar outcomes and are subsequently rare, as there are few identical neural and/or biological entities. Degeneracy describes a more flexibly adaptive phenomenon, whereby collaborating communities of fundamentally distinct components produce reliably consistent outputs under fluctuating conditions (Mason, 2015; Seifert et al., 2016).

The human runner represents a highly degenerate system. Consider the phenomenon of leg stiffness during ground-contact—the accurate calibration of which facilitates the protective dampening and economic re-cycling of impact shocks. Our highly degenerate neuro-biological design can produce equivalent leg stiffness’s using diverse coordinative strategies: muscle-tendon units (MTU’s) vary individual contributions whilst, collectively, whole-leg power outputs remain consistent; individual MTU’s achieve similar force outputs by summating different muscle and tendon contributions; individual muscles vary activated motor unit populations under differing contractile conditions to produce identical tensions; multiple combinations of torso, leg and foot postural orientations and pre-set tensionings deliver equivalent propulsive and stabilization-enabling contributions (Wickham and Brown, 1998; Roberts and Azizi, 2011; Turvey and Fonseca, 2014). This option-rich, highly degenerate movement landscape provides a multiplicity of avenues through which collaborating modular alliances combine, and re-combine, to flexibly satisfy dynamically shifting demands.

This degenerate design offers multiple means to accomplish running objectives. Historically, the apparently overwhelming complexity presented by this proliferation of movement ‘options’ was famously interpreted as a control ‘problem’ (Bernstein, 1967). This potentially complex problem, however, is reduced by the gradual construction of synergies—coordinative structures comprised of highly context-specific, context-sensitive functional linkages serving to temporarily constrain collaborating elements such that they act as single coherent units (Latash et al., 2007; Wu and Latash, 2014). Through the formation of synergies the control ‘problem’ is greatly simplified, while simultaneously retaining the benefits of complexity and degeneracy. As such, more recently, the apparent problem of excessive choice has been reframed as the ‘bliss’ of motor abundance (Latash, 2012). When running, this abundance of potentially over-whelming movement ‘choice’ can be, through effective coordination, productively deployed to disperse the running work-burden among networks of collaborating tissues: thereby promoting efficiency and robustness.

Conventionally, we equate skilful running with metronomic regularity. As proficient runners achieve reliably consistent stride outcomes, it seems sensible to assume experts precisely replicate running stride characteristics. In recent years, however, close scrutiny of running behaviors illustrates that, even when experts run at steady paces, movement parameters persistently vary (Stergiou and Decker, 2011). Through the lens of traditional motor control paradigms such variability was initially interpreted as ‘noise’—meaningless error arising from the intricacies of the engineering challenge, measurement inaccuracies and fallible biological components. Intriguingly, however, more recent investigations reveal the structure of gait variability to be neither randomly erratic, nor independent of prior events. Instead, the architecture of past, current and future stride variability’s appear statistically linked through, as yet incompletely understood, long-range correlations (Hausdorff, 2007; Stergiou and Decker, 2011; Hamill et al., 2012).

As we move through a running life, accumulative cycles of ‘wear and tear’—of injury, overuse, misuse, and disuse—gradually degrade both the material integrity of biological components and the networked richness of neural connectivity (Elbert and Rockstroh, 2004; Taubert et al., 2010; Hoppeler et al., 2011). As neuro-biological complexity contracts, the landscape of viable degenerate permutations, capable of satisfying running demands, deteriorates. Now, the mechanical stress of running must be distributed amongst shrinking networks of collaborating components (Pelletier et al., 2015).

Reductions in viable degeneracies do not, however, inevitably decrease running variability. Instead, as the neuro-biological systems struggles to proficiently manage imposed loadings, mechanical stress becomes either more tightly focused on restricted populations of working tissues, or is erratically dispersed amongst expanded webs of unconditioned tissues (Van Orden, 2007; Hamill et al., 2012). As illustration, ACL-deficient knees typically exhibit reduced, whereas ACL-reconstructed knees exhibit dramatically expanded, inter-stride variability (Stergiou and Decker, 2011; Hamill et al., 2012). Such deviations from habituated variability ranges, oscillating between overly formulaic constancy and wild randomness, signify an impaired capacity to absorb, disperse, and purposefully recycle and re-direct impact momentums (Nakayama et al., 2010; Figure 3). As coordinative fluency deteriorates, vulnerability to Overuse syndromes and unexpected perturbations escalates.

The entangled nature of complex neurobiology ensures that when variability changes at any discrete location, accommodating compensatory behaviors occur elsewhere in the system. As illustration: active injuries typically reduce habitual running variability in the injured leg—constraining control to protect sensitized tissues—, while simultaneously inducing expansions of variability in the non-injured leg (Hamill et al., 2012). Such evidence illustrates that, although running injury is a site-specific event, the accommodation of injury is a system-wide phenomenon occasioning system-wide coordinative adjustments. Importantly, these behavioral modifications, although temporarily functional, inevitably expose compensating tissues to unhabituated loadings. What has not been discussed within the running-related literature, however, are the neural and biological mechanisms which structurally embed running habits and which must be micro-architecturally altered to support coordinative change.

As running experience accumulates, the sensorimotor apparatus becomes ever-more efficient at executing the running task. But neural resources are evolutionarily precious and fundamentally limited commodities and, as such, are persistently re-deployed to fulfill varying roles within diverse tasks. Such conflicting usage patterns drive competitive plasticity processes, as neural networks strive to persistently re-model neurological ‘form’ to best fit currently prioritized ‘function.’

Consequently, as we progress from novice to ‘skilful,’ a by-product of on-going neuroplastic refinement is that fewer networked collaborators are required to manage the evermore highly practiced running pattern (Coq and Barbe, 2011; Avanzino et al., 2014; Pelletier et al., 2015). Subsequently it becomes evolutionarily wasteful to continually dedicate expanded sensorimotor networks to task execution. Accordingly, when the range of running behaviors to which we are regularly exposed becomes monotonously stereotypical, evolutionary pressure to economize resource uptake ensures the landscape of conditioned neural and biological collaborators, dedicated to executing highly practiced running patterns, progressively diminishes as under-utilized resources are re-allocated elsewhere (Elbert and Rockstroh, 2004; Avanzino et al., 2014). A drawback, therefore, of engaging in only a narrow band of overly stereotypical running tasks, is that we become hyper-efficient at deploying reduced populations of degeneracies to execute a narrowing band of self-similar running patterns. As direct consequence, we become increasingly vulnerable to both overuse syndromes, and unhabituated challenges.

Prolonged abstinence from running drives a progressive loss of physiological conditioning, and also dims the regular flow of running-related sensorimotor information. Consequently, the cortical circuitry normally maintained by consistently processing running-related sensorimotor information is eroded as voracious competitive plasticity re-models neural connectivity to best fit current usage patterns (Elbert and Rockstroh, 2004; Pelletier et al., 2015). Subsequently when we return to regular running coordinative control is slightly less proficient and slightly less resilient.

When we run in injured or irritated states we subtly alter coordination patterns to divert discomforting mechanical stress away from sensitized tissues: alleviating negative sensation, tempering structural damage, and facilitating healing. If, however, we continue to run in compromised patterns for prolonged periods, newly adapted remedial strategies become progressively more plastically engrained within CNS and tissue architectures (Engineer et al., 2012; Avanzino et al., 2014).

The dynamic inter-play between experience-driven and competitive plasticity processes ensure the traces of temporarily functional coordinative compensations typically remain plastically embedded within neuro-biological structures: thereby becoming the new ‘normal,’ and exerting a legacy not easily erased within the abbreviated timeframes offered by conventional rehabilitation paradigms (Pelletier et al., 2015).

Ultimately plastic re-modeling, as it consumes precious material and energetic resources, is evolutionarily expensive (Merzenich et al., 2014; Clark et al., 2015). Within the adult-brain it is not evolutionarily economical to plastically adapt to all stimulation—valuable neural reserves would be immediately depleted. Accordingly, structures in the mature cortex plastically remodel only when specific criteria—regulated by modulatory neurotransmitters such as acetylcholine, dopamine, serotonin, and norepinephrine—are satisfied (Merzenich et al., 2014; Clark et al., 2015). Operating collectively these neuro-modulatory enablers act as “on–off” switches, engaging excitatory and inhibitory processes and temporarily opening plasticity-enabling ‘windows of opportunity’ within which sensorimotor inputs contributing to ‘success’ are selectively amplified; while signals from competing inputs, uncorrelated with that success, are selectively dampened (Merzenich et al., 2014).

Over time, the continued amplification of relevant sensorimotor inputs provides a competitive advantage; greatly enhancing the representational detail embedded in the cortical territory dedicated to processing running-related sensorimotor information (Wolpert et al., 2011; Engineer et al., 2012). Crucially, a core finding emanating from this research domain is that repetitively non-varying, non-challenging ‘mindless’ movements—those not demanding focused attention for satisfactory execution—are insufficiently stimulating to reliably release the cocktail of neuromodulating chemical catalysts necessary for plastic re-modeling within the mature motor cortex (Merzenich et al., 2014; Clark et al., 2015). In contrast, positive plastic re-modeling is optimized in response to behaviorally relevant intense practice, executed at the limits of current abilities and therefore demanding high attentional and motivational drives (Avanzino et al., 2014; Merzenich et al., 2014). Thus positive neural re-modeling is promoted only when tasks are neither so easy that they fail to stimulate focused attention, nor so difficult that continuous failure undermines motivation. In essence coordination improves through engaging challenge, not mindless routine. A rationale perhaps explaining why rehabilitation processes employing non-challenging coordinative tasks typically fail to generate optimal recovery (Elbert and Rockstroh, 2004; Merzenich et al., 2014; Clark et al., 2015).

Deeper appreciation of the various phenomena underpinning running coordination potentially informs many aspects of conventional theory and practice. The topics below are offered as tentative examples:

The perspective offered within this review is that coordination is the overarching super-capacity ultimately orchestrating how proficiently neural, muscular, cardiovascular, and metabolic reserves are purposefully harnessed, or wastefully squandered. In relation to running: coordination is the learned deployment of available neuro-biological resources to satisfactorily realize running objectives for an acceptable ‘cost’—in terms of depletion of energetic and neural reserves, and exposure to risk. It is the physical expression of a confluence of psychological, emotional, neural, and biological constraints emerging in response to the on-going interplay between intention, motivation, and perception of risk; informed by emerging sensory feedback; modulated by prior experiences and expectations; biased toward repeatedly re-employing plastically embedded coordinative solutions to current running ‘problems.’

Ultimately, running performance is underpinned by a conglomeration of assorted capacities—cardiovascular, neurological, psychological, physiological, anatomical, muscular, and biomechanical. Yet it is the super-capacity of coordination that regulates how proficiently these overlapping performance contributors are collaboratively expressed to generate propulsive power, promote efficiency, preserve robustness, and accomplish running objectives for an acceptable exposure to discomfort and risk. A deeper appreciation of the underpinnings of the running coordination phenomenon will hopefully enable practitioners to more judiciously design interventions to promote, nurture, and preserve coordinative proficiency in the face of the inevitably accumulating ‘wear and tear’ endured over the course of a running lifetime.