Contemporary accounts of control of pacing during endurance exercise focus on physical limitations, generally assuming humans work to that physical limit. Conceptually, control is ceded to the body at the beginning of an exercise bout and is returned to central cognition upon achieving a state of exhaustion. We advance an alternative decision-making model of control of pacing, where the decision whether to persist in the effort is revisited continuously, and cessation of the exercise bout is an explicit, cognitively controlled decision. Our model depends on the following assumptions and features: 1) decisions are made in discrete cycles, 2) repetitive bodily motions depend on a central pattern generator, 3) afferent physiological feedback produces a sense of perceived exertion, 4) central cognition mediates between perceived exertion and the value of persisting (motivation) to perform an ongoing cost-benefit analysis, and 5) cessation of exercise occurs when an explicit decision is made to discontinue the effort.
Why do we continue in the face of fatigue, and when and why do we give up? This question is so fundamental, an entire body of research in the field of exercise physiology has been dedicated to answering it. Much of the field has focused on physical determinants of eventual performance, whether it be maximal oxygen consumption, hydration, or mechanical factors such as leg length or body composition. Motivational factors are often assumed away, frequently by studying world-class athletes who can be broadly characterized as exceptionally motivated, and cognitive factors are easily neglected. The physical performance is treated much as one would test an internal combustion engine, with maximal performance determined by the physical properties of the engine. That these athletes would persist during testing is a given, and that they produce a maximal effort is one of the further assumptions that defines the implications of the studies. Cognition is given short shrift, both in terms of theory and methodology.
Common protocols consist of exercising to exhaustion at a specified intensity, or completing a set distance in a minimal time. Pacing models have been proposed to describe human behavior; these are predicated on an often hidden assumption that physiology determines pacing (via physical fatigue), rather than that behavior drives pacing (via cognition). This assumption is rarely exposed in constant load exercise because experimenters have worked to remove cognition from the performance and isolate the physical aspect, often through the use of expert athletes who can push themselves to the limits of physiology. A maximal physical effort, if it is purely limited by bodily constraints, has little need to involve cognition. A sufficiently motivated participant is assumed to produce a maximal effort, and cognition is reduced to a simple on/off switch, where it invokes the required effort.
Pacing strategy is a matter of identifying the maximal
power output that can be sustained across the expected time
interval. For the elite athletes commonly used in pacing
studies, the self selection of a maximal pace is done with
relative ease, resulting in models that have slight variations
from constant power output, but which can be explained
through simple physiological explanations such as reserves
of anaerobic energy. One commonly used model, known as
the Hill model (after A.V. Hill), is based on the idea that
exercise produces linear changes in metabolism, until
demand exceeds capacity, resulting in fatigue and cessation
of exercise. This simple, conventional model of exercise
performance also produces some surprising predictions,
however, which have been justifiably criticized
Putting the Brain Back in Charge of Pacing
The trend toward removing cognitive aspects of “exercise to
fatigue” has been turned on its head in several studies,
however, where deception about pacing has been used to
examine the cognitive inputs to sustained physical exertion.
For example,
It is exactly those theoretical gaps that we intend to address here, by making our model computationally explicit. We assert that, while participants may have a rough goal to do their best, they are engaged in an ongoing comparison between the expected duration of the work bout and their current intensity of effort with reference to their prior experiences. That is, they are retrieving prior events from memory for comparison, where the content of this memory includes aspects such as effort, duration, environmental conditions, and, critically, sustainability of the effort.
The participant need not balance, nor even be aware of, most of the factors that define endurance performance during the majority of work bouts. The surprising concordance of physiological limitations (where body temperature, energy reserves, and cardiac output, for example, simultaneously reach their limits during work to exhaustion) can be explained largely by physiological adaptations: systems that fail often adapt first, until all are roughly on par with each other. For example, the ability to handle heat stress can change dramatically with only a few weeks of training. There is no need for the athlete to attempt to optimize these factors individually, much less be aware of them in many cases. On the other hand, a participant is likely to be acutely aware of any single system (whether body cooling, oxygen delivery, or bodily afferent feedback such as muscular pain) that signals an impending or realized failure. Thus, by reacting to the system that corresponds to the weakest link and matching the current effort level based on personal history, cognitive control can give the appearance of near-optimality without reference to a preconceived plan.
In the remainder of this paper we will 1) formalize this theory within the framework of a cognitive architecture, and 2) demonstrate its utility through a computational model of exercise pacing.
Using a Unified Theory of Cognition to Constrain a
Theory of Cognitive Control of Pacing
Unified theories of cognition
When running or cycling, ~150-200 individual leg movements are typically made per minute. This automaticity requires a helper system capable of regulating repetitive motions without the need to burden central cognition. That is, without such a helper system, one would be unable to do anything other than perform the exercise itself because there would be no free cycles to devote anywhere else. Thus, it is apparent, even without recourse to a UTC-based analysis, that because endurance exercise does not overwhelm central cognition, it follows that it must primarily be handled elsewhere.
Turning back to UTCs, modeling highly interactive
realtime tasks often requires helper systems running at higher
frequencies than central cognition. For example,
The implication of this analysis is that, in the context of endurance exercise, there must be a “helper” that conducts and regulates much of the activity involved in endurance exercise, and central cognition can be expected to primarily interact with that helper.
One candidate slave system that might provide a link
between central cognition and endurance exercise is the
Central Pattern Generator, a spinal network capable of
producing rhythmic limb movements in the absence of
cognitive control
Within the central executive, this model of control is based on the retrieval of relevant templates and comparison to ratings of perceived exertion (RPE). Given a target time/distance, memory can be scanned for a relevant effort that was successfully made at that time/distance. Given ongoing RPE feedback, the effort can be increased or reduced relative to the current effort.
Using this model, undershooting and overshooting of pacing efforts are both possible and likely. Overshooting has worse outcomes (failure), while undershooting can result in less than optimal performance. Specifically, undershooting leaves energy to be spent more rapidly at the end (end spurt), but due to task limitations, it may not be possible to spend all of the available energy. In all cases, these experiences are learned and stored, resulting in accumulated knowledge with experience. In the absence of experience, pacing can be expected to fail often, since there are no successes to draw from. This naturally produces learning predictions as well, since lack of experience should result in many more failures in pacing.
Specific Theoretical Predictions The constraints discussed above result in the following theoretical predictions: • Completely inexperienced athletes (as young athletes often are) are likely to have more failures of overpacing and under-pacing • Experienced athletes, after learning to avoid failures, will pace conservatively, and will need to spend more of their energy budget at the end of the effort, producing an end spurt (increase in output intensity near the end of a work bout) • Attentional manipulations that divide attention will negatively impact pacing (often through a reduction in cadence that reduces work rate) • Attentional manipulations that focus attention will positively impact pacing (through maintenance of an even pacing strategy) • Perceived Effort interacts by way of interruption, focuses attention on perception of effort (pain, discomfort, effort) • Central Executive needs to increase effort signal to CPG as fatigue occurs to maintain the same muscular input • Pacing should not be natural, but should emerge with learned experience over time • With experience, cadence will tend toward just under 90, but while learning it will be lower. This is not automaticity or power law speedup, but rather a removal of the cognitive effort of deliberative processing that is replaced with the natural frequency (speedup beyond a cadence of 90 should not happen with greater experience).
We will next examine whether these predictions are sustained or contradicted by the existing literature, and we will explore a computational model that implements this model, providing a proof of the theoretical concept. Constraints of Human Physiology on Endurance
The preceding discussion focuses on the cognitive control of pacing. There are also hard limits on endurance performance imposed directly by human physiology. An individual's capacity to perform endurance exercise is characterized by many features; chief among these are: • Aerobic capability: the ability to metabolize oxygen to produce work, at lower intensities and long time scales. This ability may be defined in terms of critical power (the maximal work rate that can be indefinitely sustained), and is commonly expressed in units of oxygen consumed per measure of body weight. • Anaerobic work capacity (AWC): the conversion of stored chemical energy to work without the use of oxygen), at higher intensities and shorter time scales. This anaerobic work creates an oxygen debt that must be repaid through respiration. • Heat tolerance: the ability to maintain homeostasis in response to heat produced through exercise, primarily through sweat production and diversion of blood to surface skin capillaries to radiate heat. • Maximal power output: the greatest work rate that can be sustained regardless of the timespan. • Muscular fatigue: the generation of chemical waste products that inhibit further muscular action.
While there are many other factors that influence the
ability to perform endurance exercise, these factors capture
many of the main constraints that impact pacing. Aerobic
capability and anaerobic work capacity are the two factors
used in the critical power model of
Graphically, the CP model traces an asymptotic hyperbolic curve, predicting work rates that approach infinity as the time approaches 0, and work rates that never decrease beyond asymptote as time approaches infinity. Despite this limitation, the CP model provides an excellent account of performance from durations of several minutes to several hours. Figure 1 below depicts the CP model in relation to actual historical performances for one individual athlete, showing the ability of this model to predict limitations while taking individual differences into account. 1200 1000 ) t(W 800 u p t ruO 600 e w oP 400 200 CP Model Performance 0.5 1.0 1.5 2.0 2.5 Time: Log(seconds) 3.0 3.5 4.0
Our modeling goal is to predict these work rate – duration
curves through a model of cognitive control, given a set of
inputs available to the cognitive system. This requires
interaction with a physiology module that incorporates the
modeling of heat generation and dissipation, oxygen
consumption (as a function of resting metabolism, aerobic
exercise, and repayment of oxygen debts), anaerobic energy
use and reserves, and the provision of a perceived exertion
signal. This last quantity, perceived exertion, is primarily
based on cardiac output (the original RPE scale, in fact, was
a linear transformation of heart rate), but also includes heat
stress, and muscular fatigue components
Cognitive Control of Pacing Behavior
The preceding discussion outlines the effort-duration
relationship for endurance exercise. The cornerstone of our
model is the use of memory for historical efforts as the basis
for establishing and refining a current effort. Those
memories, or their absence, are exactly the source of
predictions of expert-novice differences. Given a specific
duration, an athlete will gauge their effort based on
historical experience. Specifically, we suggest that a
blended memory retrieval
The conventional orient-decide-act cycle, embedded within cognitive architectures, is also a critical component of our model. As we previously pointed out, if endurance exercise required constant attention, it would overwhelm attentional resources. The central pattern generator may largely be responsible for handling the continuance of exercise under non-challenging circumstances. The critical question relates to when and how often attention focuses the individual on reconsidering the decision to either persist, increase, decrease, or entirely cease the effort, during challenging efforts.
Fatigue and pain are both implicated in this refocusing of
attention. Anyone who has ever engaged in repetitive
physical activity knows that the activity will become more
difficult during a work bout, no matter the intensity. One
important reason for this is the fatigue of neuromuscular
circuits. Specifically, achieving the same muscular output
(measured via EMG) requires stronger and stronger neural
input, which corresponds to an increasing perception of
effort. Voluntary muscle activation, which is measured as
the percentage of neural activity achieved under voluntary
control when compared to direct electrical stimulation, is
approximately 90% prior to fatigue, but drops to less than
75% under conditions of fatigue. This mechanism protects
muscle from catastrophic damage, regulating exercise by
adjusting output from the brain
Exercising muscles also experience microscopic damage and produce metabolic waste products, leading to a perception of muscular pain and fatigue. This everincreasing pain signal further attracts attention. The neural fatigue protective mechanism also interacts with regulation of activity via the CPG indirectly. At exactly the time when an athlete needs to focus attention on making the decision to maintain an effort (to increase the neural output signal to the exercising muscles), they may be interrupted to process the urgent sensory input of a pain signal.
Finally, while the CPG might allow the repetition of rhythmic activities, it will not drive effortful performance on its own. In the absence of a decision to continue, CPGdriven behavior will revert to the natural resonant frequency of the CPG, possibly reducing cadence during high effort times, and the same CPG signal will result in a decreasing output effort due to fatigue of neural pathways, despite any conscious decision to continue.
The decision to persist during endurance exercise to exhaustion, thus, must be revisited more and more frequently as the interval continues through the involuntary