Documente Academic
Documente Profesional
Documente Cultură
R. Hugh Morton
The critical power and related whole-body bioenergetic models
Accepted: 13 October 2005 / Published online: 12 November 2005
Springer-Verlag 2005
Abstract This paper takes a performance-based ap-
proach to review the broad expanse of literature relating
to whole-body models of human bioenergetics. It begins
with an examination of the critical power model and its
assumptions. Although remarkably robust, this model
has a number of shortcomings. Attention to these has
led to the development of more realistic and more de-
tailed derivatives of the critical power model. The
mathematical solutions to and associated behaviour of
these models when subjected to imposed exercise can
be applied as a means of gaining a deeper understanding
of the bioenergetics of human exercise performance.
Keywords Anaerobic work capacity Energy
Endurance Exercise Metabolism Performance
Introduction
Over recent decades there appears to have been a drift
towards a more micro view of human bioenergetics,
focussing more towards the cellular and even molecular
end of the spectrum. Although the macro view has not
been neglected, it is nevertheless timely to examine
models of human bioenergetics from the whole-body
perspective. The human body is a highly complex
organism and has been the subject of intense scientic
scrutiny for several centuries. Despite considering a
broad picture of only one aspect of its functioning, I do
not make any special attempt to be comprehensive in
this review. Rather, I attempt to blend the essential
theoretical elements of these models and their mathe-
matical solutions with the practical and performance-
related interpretations of their properties. In so doing, I
hope their value as an investigative research vehicle or as
a framework for practical aspects of the study of human
performance, or even as a teaching tool for undergrad-
uate or postgraduate students, can be evaluated.
The review begins with an examination of the simple
critical power model and its assumptions. When sub-
jected to scrutiny, this model is remarkably robust, but it
nevertheless has a number of shortcomings. Remedial
attention to these has led to the development of more
realistic and more detailed derivatives of the critical
power model. The mathematical solutions to these
models under a variety of exercise scenarios provide
fertile ground for gaining a better understanding of the
bioenergetics of human exercise performance.
This review is not intended to update the literature on
critical power (Hill 1993; Vandewalle et al. 1997) nor that
on the relationship between power output and endurance
time (Billat et al. 1999a; Morton and Hodgson 1996) nor
its relationship to whole-body fatigue (Walsh 2000), but
rather to set the critical power model and its derivatives in
the wider bioenergetic modelling context.
The critical power model
Origins, assumptions and development
It was Monod and Scherrer (1965) who rst observed a
hyperbolic relationship between the level of constant
power output (P) and corresponding time to exhaustion
(t) in a single muscle group. This relationship, when
transformed knowing that total work performed (W) is
given by the product Pt, can be expressed in a linear form.
The intercept of this line was termed the anaerobic work
capacity (AWC) and its slope termed the critical
power (CP). Mathematically the two equations:
W AWC CP t 1a
and
t AWC=(P CP) 1b
R. H. Morton
Institute of Food, Nutrition and Human Health,
Massey University, Private Bag 11-222,
Palmerston North, New Zealand
E-mail: h.morton@massey.ac.nz
Tel.: +64-6-3504265
Fax: +64-6-3505781
Eur J Appl Physiol (2006) 96: 339354
DOI 10.1007/s00421-005-0088-2
are equivalent. There are in fact four other equivalent
equations:
P AWC=t CP, 1c
a transformation rst proposed by Whipp et al. (1982)
t (W AWC)=CP 1d
and
P W CP=(W AWC), 1e
W AWC P=(P CP) 1f
although only the rst three are in common use. Given
that t is the most natural dependent variable, and P the
most natural explanatory variable, Eq. 1b is the pre-
ferred version from an exercise performance viewpoint,
though a similar case could be made for Eq. 1f.
Moritani et al. (1981) extended those ideas for the rst
time to whole-body exercise on the cycle ergometer. The
four essential assumptions of the whole-body CP concept
as it became known (Hill 1993) are the following:
1. There are only two components to the energy supply
system for human exercise, termed aerobic and
anaerobic.
2. The aerobic supply is unlimited in capacity, but is
rate limited. This limiting parameter is the critical
power, CP.
3. The anaerobic supply conversely is not rate limited,
but is capacity limited by the parameter AWC.
4. Exhaustion, and by implication termination of exer-
cise, occurs when all of AWC has been utilised.
From these assumptions it can be deduced that for
constant power P CP, the notion of exhaustion is
vacuous and t is undened or innite, whereas for
P>CP, Eq. 1b and its equivalents follow directly. Over
the last 20 or more years, a large number of studies have
found good ts of data to Eqs. 1a, 1b, 1c, at least within
certain bounds.
It is important nevertheless to realise that implicitly
embedded in these assumptions are a number of others,
specically:
5. Aerobic power is available at its limiting rate CP the
moment the exercise begins and remains so right up
until the end of the exercise.
6. The power domain over which the model applies is all
of CP<P<.
7. The time domain over which the model applies is all
of 0<t<, and endurance at CP is innitely long.
8. The eciency of transformation of metabolic energy
to mechanical energy is constant across the whole
power (and time) domain(s).
9. CP and AWC are constants, independent of P (and/
or of t).
All models of real processes are necessarily simpli-
cations and the above assumptions are indicative of
them. There are of course unrealities to varying degrees
in all of them, and these will be discussed in Extending
the critical power model when extensions to the CP
model and the reasons behind them will be considered.
Parameter interpretation and estimation
The assumptions above suggest clearly enough that CP
is aerobic in nature and that AWC is anaerobic, but
empirical verications are vital. It was Moritani et al.
(1981) who rst provided evidence for these interpreta-
tions. They found CP to be highly correlated with the
ventilatory anaerobic threshold and AWC to be unaf-
fected by hypoxia. Several other early studies (Green
and Dawson 1993; Green et al. 1994; Green 1994, 1995;
Housh et al. 1991; Jenkins and Quigley 1990, 1991;
Talbert et al. 1991) have provided similar evidences,
though the evidence is not all conclusive (Housh et al.
1992). Perhaps the most convincing evidence of the
aerobic and anaerobic natures of these parameters can
be found amongst the collection of studies reporting
interventions and their impact on the CP model, which
will be discussed in Interventions and their impact on
the model.
In order to estimate the parameters CP and AWC for
any subject, it is necessary to complete at least two
separate trials, cycling to exhaustion at two dierent but
constant power outputs. In practice four or more are
common, spread across a spectrum, though six or more
are uncommon. On each occasion P and t (and hence
W = Pt also) are recorded. Any one of the six equation
forms can be tted using any standard statistical curve-
tting software. It is important to choose these experi-
mental design aspects carefully, for the CP and AWC
estimates may dier according to how many trials are
performed (Housh et al. 1990; Taylor and Batterham
2002), what durations (and therefore also what power
selections) are planned (Bishop et al. 1998; Jenkins et al.
1998; Poole et al. 1986) and which of the mathematical
forms of the model may be used (Bull et al. 2000;
Gaesser et al. 1990, 1995; Hill et al. 1993; Hill 2004;
Housh et al. 2001). Furthermore, the rest interval be-
tween trials may aect the estimates also (Bishop and
Jenkins 1995; Hill 1993) varying between 30 min and
over 24 h in dierent studies.
It has been observed too that variations in repeated
performances are larger when durations are longer
(McLellan et al. 1995; Poole et al. 1988). For this reason
the exact nature of the curve-tting routine employed
may also aect the estimates. It may be statistically
advisable to use weighted regression, giving less weight
to long-duration observations and vice versa (Morton
and Hodgson 1996).
Performance implications
Application of the CP model using data from any sub-
ject clearly provides an assessment of two dimensions of
the bioenergetic characteristics of that individual. Then,
340
employing Eqs. 1b, 1f for example and substituting any
given value of P provides predictions of that individuals
performance (i.e. endurance) time and amount of work
that could be accomplished, respectively. Such perfor-
mance aspects are obvious and have been employed in
determining the ability of humans generally as sources of
mechanical power (Wilkie 1960) or specically in activ-
ities such as human-powered ight (Nadel 1988, 1996).
Of far more relevance in todays world of competitive
sports performance is the question of optimal strategy.
For example, can performance time (and the amount of
work accomplished) be improved by working at other
than the constant power throughout determined by the
solution to the CP model equations for that individual?
Fukuba and Whipp (1996, 1999) have solved this
problem (albeit for the running analogue of the CP
model). The answer, or at least part of it, is No. They
also demonstrate that any one of an innite number of
strategies involving stagewise steps of periods of con-
stant power produces an identical endurance time, pro-
vided P never drops below CP. This can also be shown
to be true when P changes (always above CP) in any
smooth and continuous fashion (Morton 2004).
Conceptualisation as a hydraulic model
This conceptualisation is originally due to Margaria
(1976) and can be described as follows. If we envisage
uid contained in a vessel or vessels to represent meta-
bolic energy and the ow of uid therefore to represent
power, the CP model can be diagrammatically repre-
sented as in Fig. 1.
The system is comprised of two vessels representing
the aerobic (A
e
) and anaerobic (A
n
) energy supply
components (assumption 1). A
e
is of innite capacity
(assumption 2). R
1
is a connecting tube of xed diame-
ter, which determines the limit to the uid ow from
A
e
to A
n
. This limit is denoted by CP (assumption 2).
Vessel A
n
is of limited capacity, denoted by AWC
(assumption 3). T is a tap which can be opened to allow
an unlimited positive ow P (assumption 6). This in turn
(since CP is nite) implies that the rate of decrease of the
uid level in A
n
is also unlimited (assumption 3). Pro-
vided P>CP (assumption 6), the uid level in A
n
must
drop to make up the dierence PCP, and P can no
longer be maintained when A
n
becomes empty
(assumption 4). This event occurs at time t>0
(assumption 7). The ow through R
1
is always equal to
exactly CP, determined only by the unchanging depth of
vessel A
e
(assumption 5). When P=CP or less, uid
from A
n
is not required and A
n
will never be emptied,
i.e. t is innite (assumption 7). Since the whole hydraulic
model is a physical construct with xed dimensions and
containing an incompressible uid, the volumetric and
rate (and therefore also eciency) parameters are xed
and independent of P (assumptions 9 and 8).
All the above lead directly to the derivation of Eq. 1b
from rst principles. Knowing also that W = Pt enables
all other mathematical equivalences to Eq. 1b to be
derived algebraically.
Interventions and their impact on the model
The CP model has been the object of a large number of
intervention studies designed to discover what happens
when it is exposed to deliberate changes in conditions.
These studies have assisted signicantly in our under-
standing of the model. In general, their results conform
to what physiological logic might lead us to expect and
arms the overall nature of the concept and its com-
ponent structure.
An early study of Jenkins and Quigley (1990) showed
that a group of trained subjects had a signicantly
higher average CP and AWC than another group of
untrained subjects. Since that time, more specically
designed training studies have examined the direct ef-
fects of various training strategies on changes in CP and/
or AWC in the same subjects. Given the aerobic nature
of CP, it would be expected that endurance or pre-
dominantly aerobic training would increase the CP of
those undertaking such training. A number of studies
have conrmed that this indeed occurs (Gaesser and
Wilson 1988; Jenkins and Quigley 1992; Jones and
Carter 2000; Poole et al. 1986). Likewise it would be
expected that strength or predominantly anaerobic
training would increase AWC and, from the outset,
studies do conrm this (Jenkins and Quigley 1992; Poole
et al. 1990; Stokes et al. 1993). Interval training, in which
periods of low-intensity training alternate with periods
of high intensity, is generally regarded as a particularly
eective overall training strategy (Billat 2001; Laursen
and Jenkins 2002) because it taxes energy supply
mechanisms across the spectrum. Inferring from the
above, it would be expected and has been shown that
interval training improves both CP and AWC (Gaesser
and Wilson 1988; Poole et al. 1990).
From the hydraulic representation of Fig. 1, aerobic
training appears to increase the diameter of pipe R
1
. It
would not likely increase the depth of vessel A
e
, as this Fig. 1 The CP concept as a hydraulic model
341
depth could be viewed as the analogue of the O
2
partial
pressure in the inspired air. Strength training on the
contrary appears to increase the volume of vessel A
n
,
probably via a cross-sectional increase rather than a
depth change.
Again, given the aerobic nature of CP, it would be
expected that subjects tested under hypoxic conditions
would evidence reduced CP, while comparative tests
under hyperoxic conditions would evidence an elevated
CP. This appears to be the case (Moritani et al. 1981;
Whipp et al. 1982), whereas AWC is unchanged. From
the hydraulic model perspective similarly therefore, hy-
poxia probably represents reduced depth of vessel A
e
rather than a constriction of R
1
, whereas hyperoxia
represents the converse.
Since AWC is regarded as an anaerobic store repre-
senting energy derived from high-energy phosphates and
glycolysis leading to net lactate formation, it may be
expected that creatine and/or carbohydrate supplemen-
tation would increase AWC. Studies show that this is the
case (Miura et al. 1999; Smith et al. 1998; Stout et al.
1999). Creatine supplementation has no eect on CP,
but its eects on AWC are augmented by carbohydrate
supplementation. No studies have reported the eects of
carbohydrate supplementation alone. Like strength
training, these types of supplementation appear to in-
crease the volume of vessel A
n
.
Likewise, if prior exercise is undertaken with little or
no recovery or if the timing is only started once VO
2
max is reached, a diminution of AWC would be ex-
pected. Once again, empirical evidence conrms these
expectations (Billat et al. 1999a, b; Heubert et al. 2003,
2004; Swanson et al. 1992).
Diering pedal cadences have been considered as a
confounding factor with the potential to cause changes
in the CP and/or AWC estimates. The available evidence
is to some extent contradictory (Carnevale and Gaesser
1991; Green et al. 1995) but it is believed that it is best to
allow subjects to select a cadence at which they are
comfortable, at least between the limits of 60100 rpm
(Hill et al. 1994; Hill 2004).
Another confounding factor which does appear to
aect CP and/or AWC estimates is a learning eect,
manifesting under repeat performance or test/retest
conditions (Scarborough et al. 1991; Smith et al. 1991).
As would be expected, performances are better and
estimates higher on subsequent testing occasions.
In summary, this array of interventions serves to
strongly arm the robust nature of the concept in
assigning an aerobic character to CP and an anaerobic
character to AWC. The interpretation of these param-
eters is now rmly established.
Applications to other constant power exercise modalities
Although Lloyd (1966) rst suggested application of
these ideas to running, it was Hughson et al. (1984) who
rst published the details in which critical velocity (CV,
measured in metre/second) and anaerobic distance
capacity (ADC, measured in metres) were considered
and evaluated as analogues of CP and AWC, respec-
tively. While not subject to quite the same degree of
scrutiny as the original form of the CP concept, studies
do in general conrm those results reported previously.
For example, CV has been found no dierent from the
maximal lactate steady-state running velocity or from
the lactate turn point velocity (Smith and Jones 2001),
and to be a threshold velocity above which exercise of
sucient duration will lead to attainment of VO
2max
(Hill and Ferguson 1999). The success with which the
running version of the concept is able to predict times at
running various distances does, however, seem in some
doubt (Kolbe et al. 1995; McDermott et al. 1993) and
the choice of which equation of the model to use also
seems unclear (Housh et al. 2001).
In the same way as for running, CV and ADC can be
evaluated for swimming (Biggersta et al. 1992; Steward
et al. 1994; Wakayoshi et al. 1992a, b). Although subject
to even less scrutiny than running, the critical swimming
velocity corresponds to the maximal lactate steady-state
swimming velocity (Wakayoshi et al. 1993) and corre-
lates signicantly with swimming VO
2
at the anaerobic
threshold and the onset of blood lactate accumulation
(Wakayoshi et al. 1992a, b) and with the lactate
threshold (Martin and Whyte 2000). Critical swim
velocity has been found to correspond to a maximal
stroke rate which can theoretically be maintained con-
tinuously without exhaustion (Dekerle et al. 2002). The
choice of equation for swimming, however, seems to be
of little consequence (Lane et al. 1994), but there is
conicting evidence on whether the anaerobic swimming
distance capacity provides a reliable estimate of anaer-
obic capacity or not (Biggersta et al. 1992; Dekerle
et al. 2002; Lane et al. 1994).
In theory of course, any exercise modality in which
physical work can be accomplished, or which involves
some form of locomotion, oers scope for application of
the CP concept. Several of these have been briey re-
ported, such as kayaking (Clingeleer et al. 1994),
wheelchair propulsion (Arabi et al. 1999), weightlifting
(Morton et al. 2005) and rowing (Hill et al. 2003; Ken-
nedy and Bell 2000). Cycling when regarded as a means
of locomotion appears to have not been considered.
In all these cases, whatever the form of locomotion,
the relationship between energy cost of transport and
speed over the applicable range must be known and
explicitly introduced into the calculations (di Prampero
1999).
Applications to variable power output exercise
There is nothing particularly special about continuous
exercise at constant power, except perhaps that the
mathematical solutions (and the associated curve tting
to data) to the CP model under such conditions are
extremely simple. One exercise protocol common in
342
laboratory testing settings is ramp exercise, where power
output starts at zero (though in general this need not
necessarily be the case) and increases linearly at some
xed rate, referred to as the ramp slope, s. The CP model
has been adapted for ramp exercise (Morton 1994), ex-
cept that endurance time becomes a function of s, albeit
mathematically slightly more complex than linear or
hyperbolic. Like the original CP model, it too is avail-
able in several dierent mathematically equivalent forms
(Morton 1997). Its validity has been questioned (Van-
dewalle 1995) despite the fact that its assumptions are
precisely the same as for the constant power CP model
(Morton 1995). Indeed, investigation has shown that
estimates of both CP and AWC from ramp tests are not
signicantly dierent from those derived from constant
power tests in the same subjects (Morton et al. 1997).
Similarly, there is nothing special about continuous
exercise per se. Many sports (particularly team games)
involve exercise of an intermittent nature, where periods
of relative rest and intense exercise are interspersed. In-
deed intermittent exercise (Christensen 1960) has for long
been regarded as a particularly benecial formof exercise
(Billat 2001; Laursen and Jenkins 2002). Can the CP
concept therefore be applied to intermittent exercise?
Examination of the hydraulic representation of the
CP model depicted in Fig. 1 suggests that if tap T were
to be closed, corresponding to the cessation of work (or
a period of relative rest), then vessel A
n
would rell by
virtue of the ow through R
1
. Only when A
n
had been
completely relled would the ow through R
1
cease.
Thus for intermittent exercise one can regard the intense
exercise periods as drains on A
n
and the relative rest
periods as relling A
n
. This is exactly the approach ta-
ken by Morton and Billat (2004), who nd good ts of
the derived model equation to data, but the CV esti-
mated from continuous running was signicantly higher
than the CV estimated from intermittent running in the
same subjects. This discrepancy may be due to increased
sensitivity of the model to the implicit assumption of the
innitely fast step increases and decreases in VO
2
occurring many times in this mode or exercise. In
addition, it should be noted that this particular appli-
cation of the CV concept is quite dierent to that of
Kachouri et al. (1996) and others, who simply apply the
standard model approach to aggregated intermittent
running times. Rather than being an example of the
limits of the CP concept as claimed, Kachouris appli-
cation is perhaps more an example of an inappropriate
or inaccurate procedure, not based on proper consider-
ation of the CP concept assumptions detailed above.
Extending the critical power model
Questioning the assumptions
All models purporting to represent aspects of the human
system are necessarily simplications, and the CP model
is no exception. All nine of the assumptions of Origins,
assumptions and development can be questioned. In
brief, the following critiques can be made of them.
1. There are only two components to the energy supply
system for human exercise, termed aerobic and
anaerobic. While it may be true that there are only
two biochemical pathways, aerobic and anaerobic, to
human energy metabolism, there are certainly more
than two components to the energy source or supply.
The aerobic pathway taps into both glycogen and fat
as fuel sources. The anaerobic pathway comprises a
lactic component in which glycogen metabolises
resulting in net lactic acid production and an alactic
component (essentially the high energy phosphate
compounds or phosphagens) whose metabolism does
not result in lactic acid production. Incorporating all
these aspects realistically will of necessity complicate
the modelling. It has been attempted in a simple
fashion, described in Applications and developments
of the 3-component hydraulic model.
2. The aerobic supply is unlimited in capacity, but is rate
limited. This limiting parameter is the CP. Since the
aerobic pathway requires an energy substrate to be
oxidised, the amount of which is of course limited in
any human, it is clearly untrue to assert that the
aerobic capacity is unlimited. It may well be relatively
large (in energy units) in comparison to other avail-
able energy substrates by an order of magnitude or
more, and this may well explain the robust character
of the CP model at least for exercise of moderate
duration. However, it is well established that the
aerobic supply is rate limited. Whether this rate is
called the critical power, maximum aerobic power
(MAP) or VO
2max
, or indeed some other convenient
term, is not the issue. Some upper limit clearly exists.
3. The anaerobic supply conversely is not rate limited, but
is capacity limited by the parameter AWC. The
anaerobic supply component likewise requires energy
substrates, and so it is quite reasonable to assume a
limited capacity. While explosive muscular power
may well be quite high, it is clearly not unlimited.
Humans cannot run with instantaneous velocity of
more than about 13 m/s, and if the resistance on the
pedal cranks of a cycle ergometer is too high, the
pedals simply cannot be turned. No physical work is
possible and no power is generated. The notion of an
endurance time at such intensities is clearly vacuous.
4. Exhaustion, and by implication termination of exercise,
occurs when all of AWC have been utilised. While it
may be reasonable for a mechanical engine to cease
when its usable fuel is fully depleted (provided there is
no other malfunction), the human engine appears to
be dierent. It has for long been known that at
exhaustion, signicant amounts of unused glycogen
remain, and that the higher is the power requirement
at exhaustion, the more it is that remains (Saltin and
Karlsson 1971). Furthermore, it is well established
that if the power requirement at the point of
exhaustion were to be signicantly reduced, exercise
343
could continue. It is clear therefore that some linkage
or dependency exists between the power demanded at
the point of exhaustion and the residual unused fuel
supply available at that time. Thus while it may be
reasonable to assert that the anaerobic capacity is of
a xed and limited value, the usable amount clearly is
not. Consequently any estimates of this value based
on the CP model assumptions will obviously be
underestimates.
5. Aerobic power is available at its limiting rate CP the
moment exercise begins and remains so right up until
the end of exercise. The kinetics of oxygen uptake,
whether regarded as mono- or bi-exponential and
with or without any delay(s) (Barstow and Mole
1991), clearly dictate that the establishment of the
aerobic supply at any given rate (not just at CP) is not
instantaneous. Three minutes or so may be required.
What this means is that the capacity-limited anaero-
bic supply as reasoned in the CP model must be in
reality larger than the model prescribes. Again, as
above, AWC must be underestimated by the CP
model.
6. The power domain over which the model applies is all of
CP<P<. If P CP, the anaerobic supply is never
required, since under assumption 5, aerobic power at
CP (or less) is immediately available. The left-hand
inequality therefore is obviously reasonable. At the
other end of the scale, human muscle is not all-
powerful and, as indicated above, a nite upper limit
to the power that can be developed must exist and to
the velocity of running, etc. This upper limit could be
denoted P
max
and regarded as a maximum instan-
taneous power.
7. The time domain over which the model applies is all of
0<t<, and endurance at CP is innitely long. The
notion of innite endurance is a somewhat nebu-
lous one. It is not uncommon that long-duration
continuous exercises of moderate power output are
terminated for a variety of non-energy-substrate re-
lated reasons, such as boredom, psychological issues,
the necessity to eat, drink or toilet. This therefore
places an imprecise upper limit to the time domain
over which the CP model applies. In similar vein very
short time intervals at very high power (even if less
than P
max
) are not realistic either. Factors like over-
coming inertia and acceleration come into consider-
ation. Strictly speaking of course, if exercise at P
max
or more was to be attempted, endurance time should
be zero.
8. The eciency of transformation of metabolic energy to
mechanical energy is constant across the whole power
(and time) domain(s). This assumption, implying
that the energy cost of locomotion per unit distance is
constant, seems to depend critically on the form of
locomotion (di Prampero 1999). It seems a reason-
able assumption for running and cycle ergometry (at
least within the sorts of bounds discussed at various
places above), but it changes substantially with speed
in swimming and cycling on the road or track. If the
form of such dependency was known, it would need
to be explicitly introduced into the model. While this
could be done, the result would be cumbersome.
9. CP and AWC are constants, independent of P (and/or
of t). The constancy (or otherwise) of these two
parameters of the model introduces issues of inter-
pretation. For example both CP and AWC are not
constant over time with respect to changes in exer-
cising habits (training in particular is discussed in
Interventions and their impact on the model above)
and, in any event, are nevertheless probably subject
to the usual intra-subject variations (other factors
being equal) typically inherent in any biological
organism. In another sense, Bishop et al. (1998) have
shown that the estimates of CP and AWC dier when
diering sets of P are used in the testing sessions.
While in some sense this non-constancy might seem
disturbing, relative stability of CP and AWC does
seem reasonable over a short term and indeed with-
out such an assumption not much progress can be
made.
Towards a more realistic critical power model
Bearing in mind the many deciencies in the assump-
tions as described above, the CP model can nevertheless
be improved without a great deal of added complication.
In several instances this has been done.
For example, Wilkie (1980) focussed on assumption
5, introducing a correction factor for oxygen uptake
kinetics based on a single exponential with time constant
s and without delay. Wilkies version of Eq. 1a now
becomes:
W a bt bs(1 e
t=s
), 2a
where a and b correspond to AWC and CP, respectively,
and the third term represents the amount of energy re-
leased from anaerobic sources before the attainment of
an aerobic steady state at CP. This factor is the degree
by which AWC underestimates the true anaerobic
capacity and makes up for the dierence in energy terms
between the assumed innitely fast (square wave) VO
2
kinetics of the CP model and real VO
2
kinetics with a
time constant equal to s. Wilkie also presents a version
of Eq. 1c with a corresponding correction term, but
derivations of corresponding equations 2a to 1b, 1d, 1e
and 1f are mathematically not straightforward.
The bioenergetics of Wilkies correction can be
demonstrated by Fig. 2. Here the real anaerobic
capacity is not a as it would have been under the CP
model, but a plus the approximately triangular area
below the dotted line and above the curve (the correction
factor in Eq. 2a).
Wilkie showed his correction to be appropriate for
times to exhaustion between 50 s and 10 min, giving s a
value of about 10 s. Although 10 s is rather too fast
(30 s might be more reasonable), this term does adjust
344
appropriately for oxygen kinetics in exercise of fairly
short duration. For exercise of longer duration, its eect
is relatively lessened.
Wilkies correction can be modelled by a simple
change to the hydraulic model presented in Fig. 1. All
that is required is to lower the vessel A
e
and its con-
necting tube R
1
such that when A
n
is full, the uid levels
in A
n
and A
e
are in alignment (Fig. 3)
Following Conceptualisation as a hydraulic model,
this model operates as follows. When T is opened to
some value P>0 the level in A
n
immediately begins to
drop, creating a dierential in levels between A
n
and A
e
.
This dierential induces a ow through R
1
directly
proportional to h, corresponding to a rise in VO
2
above
rest. This ow in turn slows the rate at which the level in
A
n
drops, slowing the rate at which VO
2
began to rise. If
P CP, then h and VO
2
will reach a steady state with
the level in A
n
no lower than the connecting tube R
1
.
That is, the height h is exactly sucient to induce a ow
through R
1
exactly equalling the ow P through T.
Mathematically, the ow through R
1
, and hence VO
2
,
can be shown to be mono-exponential, whether or not a
steady state is reached.
Of course if P>CP, h will increase beyond the depth
of vessel A
e
, the ow through R
1
will have reached its
limit at CP, and A
n
will continue to empty in order to
sustain the ow P>CP. In the model context, exhaus-
tion occurs when A
n
is fully depleted as P can no longer
be sustained. Mathematically, this can be shown to be
precisely the Wilkie formulation of the CP model.
While other corrections may be feasible focussing on
any one or more of the problems associated with the
original CP model, probably the most important from a
physiological point of view is the way in which AWC
appears to not necessarily all be usable.
A feedback control system and the 3-parameter model
It is well recognised that the last 100 m of a longer race
(say 5,000 m) cannot be run at the same speed as that
which could be achieved had the runner been sprinting
100 m afresh. Taking this in conjunction with the work
of Saltin and Karlsson (1971), there appears to be some
form of control or feedback system at work, which
operates to limit the maximum power output achievable
inversely according to how fresh the runner is or more
precisely according to how much energy store remains.
Therefore we can retain assumption 2 and modify
assumption 6 such that CP<P<P
max
and introduce a
simple linear feedback control system in accordance
with the amount of the A
n
store remaining.
This means that the maximum achievable power at
any time becomes a function of the amount of A
n
, spe-
cically:
CP + A
n
(P
max
CP)=AWC. 3
If the individual is fully rested and replete,
A
n
= AWC and P
max
can be achieved. On the contrary
if A
n
is empty then only CP can be achieved, and as
before the notion of exhaustion in vacuous unless
P>CP. This assumption of a linear control feedback
forms the basis of the 3-parameter CP model and is fully
described by Morton (1996).
It is immediately apparent that this model avoids, or
at least mitigates, the problems associated with
assumptions 3, 4 and 6. Research at the time (Gaesser
et al. 1995; Morton et al. 1997) and subsequently for
example Bull et al. (2000) and Hill et al. (2003) suggest
that AWC estimates are higher and CP estimates lower
than those obtained from the 2-parameter model, and
that endurance at CP (while not innite) is signicantly
longer than previously (assumption 7).
Diagrammatically, the 3-parameter model can be
represented by the hydraulic model of Fig. 1, but with
the addition of an appropriate linkage between the uid
level in A
n
and the tap T in accordance with Eq. 3
above.
Further developments
The most obvious development of the above discussion
is to combine Wilkies correction with the feedback
Fig. 3 Wilkies correction to the CP hydraulic model
Fig. 2 Wilkies correction to the CP model
345
assumption of the 3-parameter CP model. The details of
this combination have not previously been published, so
it can be achieved quite simply as follows.
di Prampero (1999) gives a version of Wilkies cor-
rection as his Eq. 3 on p. 163, which since W = Pt can
be adapted to
A
n
P CPt CPs1 e
t=s