Scand J Med Sci Sports 2011: 21: 73–78 & 2009 John Wiley & Sons A/S

doi: 10.1111/j.1600-0838.2009.01025.x

Modelling the determinants of 2000 m rowing ergometer

performance: a proportional, curvilinear allometric approach
A. M. Nevill1, S. V. Allen2, S. A. Ingham2
1
Department of Sports Studies, University of Wolverhampton, Walsall, UK, 2English Institute of Sport, Loughborough University,
Loughborough, Leicestershire, UK
Corresponding author: Alan M. Nevill, PhD, Research Institute of Healthcare Sciences, University of Wolverhampton,
Walsall Campus, Gorway Road, Walsall, WS1 3BD, UK. Tel: 144 01902 322 838, Fax: 144 01902 322 894, E-mail:
a.m.nevill@wlv.ac.uk
Accepted for publication 8 July 2009

Previous studies have investigated the determinants of realistically describes the greater increment in power re-
indoor rowing using correlations and linear regression. quired to improve a rower’s performance by the same
However, the power demands of ergometer rowing are amount at higher speeds compared with that at slower
proportional to the cube of the flywheel’s (and boat’s) speed. speeds. Furthermore, the fitted exponent, 0.28 (95% con-
A rower’s speed, therefore, should be proportional to the fidence interval 0.226–0.334) encompasses 0.33, supporting
cube root (0.33) of power expended. Hence, the purpose of the assumption that rowing speed is proportional to the cube
the present study was to explore the relationship between root of power expended. Despite an R2 5 95.3%, the initial
2000 m indoor rowing speed and various measures of power model was unable to explain ‘‘sex’’ and ‘‘weight-class’’
of 76 elite rowers using proportional, curvilinear allometric differences in rowing performances. By incorporating anae-
models. The best single predictor of 2000 m rowing erg- robic as well as aerobic determinants, the resulting curvi-
ometer performance was power at VO · 0.28
2max (WVO2max ) , linear allometric model was common to all rowers,
that explained R2 5 95.3% in rowing speed. The model irrespective of sex and weight class.

The rowing ergometer is a popular and convenient 1973), with the relative amount of energy derived
apparatus for indoor training. Not only does it from anaerobic metabolism being estimated at 21–
provide a valuable form of controlled, weight-sup- 30% (Secher, 1990) or possibly a little less (Spencer
ported exercise that promotes health and fitness, but and Gastin, 2001). It is likely therefore that the above
it is thought to provide a valid proxy for rowing on measures of power, together with a measure of
water (Mikuli et al., 2009) although some doubt as to maximum power lasting considerably o 3 min are
the validity of indoor rowing ergometers to simulat- likely to make a valuable contribution to indoor
ing rowing on water has been questioned (Nevill rowing performance. Invariably, however, many of
et al., 2009). Such is the popularity of rowing these studies have investigated the determinants of
ergometers, thousands of competitors from all over indoor rowing using correlation and linear regression
the world compete in the Concept II World Indoor techniques that assume the determinants are linearly
Rowing Championship held in Boston each year. related to 2000 m ergometer rowing performance.
A number of authors have explored the physiolo- Linear models, for example, will assume that for a
gical determinants of rowing ergometer performance similar increase in power, the same absolute improve-
(Secher, 1973; Ingham et al., 2002; Bourdin et al., ment in rowing speed will result irrespective of the
2004). Some authors have reported that ergometer rower’s level of performance (e.g. the speed of elite
rowing performance over 2000 m is best predicted by international rowers compared with club standard
peak or maximum-power output (Wmax) sustained rowers).
during a maximal incremental test, intervals lasting The relationship between indoor rowing ergometer
3 min (Bourdin et al., 2004), while others identified performance and power, however, is definitely not
the power associated with VO ·
2max (WVO2max ) as linearly related. Indeed, according to researchers at
among the best predictor (Ingham et al., 2002). Oxford University, UK (http://www.atm.ox.ac.uk/
Performance over 2000 m on a rowing ergometer is rowing/physics/ergometer.html), the power required
dependent upon the functional capacity of both the by the rower to rotate the air braked flywheel is
aerobic and anaerobic energy pathways (Secher, proportional to the cube of the flywheel speed. A

73
Nevill et al.
similar cube law relationship holds for the relation- syringe. Computer integration of volume and gas concentra-
ship between the dissipated power and boat speed/ tion signals accounted for the delay in gas passing through the
velocity. Consequently, the ergometer or boat speed capillary line. Respiratory gas exchange variables (oxygen
uptake, carbon dioxide production and minute ventilation)
achieved by the rower should also be proportional to were calculated and displayed for every breath and averaged
the cube root (0.33) of the power expended. Hence, over the final full minute of each exercise intensity. The
the purpose of this article was to explore the likely ·
VO 2max was defined as being the highest 30 s average achieved
curvilinear relationship between 2000 m indoor row- during the 4 min maximal test (coefficient of variation for this
ing speed and various measures of power output laboratory 5 5.5%). Oxygen cost of movement (ECON) was
assessed by calculating the mean oxygen uptake per watt (mL/
using a proportional allometric model to better W) of the submaximal stages (coefficient of variation 5 7.5%).
understand determinants and limitations of indoor Solving the regression equation describing VO·
2 and power
Concept II rowing performance. for the five incremental intensities of exercise calculated the
power associated with VO ·
2max (WVO2max ). A sample of capil-
lary blood drawn from the earlobe was taken at the end of
Methods each stage and assayed for lactate (Analox GM7, London,
UK). Plots of [La–]b against power were inspected for a non-
Following approval from the regional Ethics Committee, 76 linear increase in [La–]b taken as a power at lactate threshold
current or former senior or under 23 World rowing or sculling (WLT), The power outputs associated with [La–]b of 2, 3 and
finalists provided written informed consent to take part in 4 mmol/L were determined by interpolation. Heart rate was
progressive incremental rowing tests on the ergometer, a recorded using telemetry (Polar Electro, Oy, Kempele, Fin-
maximal ergometer power test and a maximal 2000 m erg- land). All 2000 m time trials on the ergometer were performed
ometer time trial. The physical characteristics of the rowers, on selected Concept IIC ergometers as a criterion assessment
grouped according to sex and rowing weight class [heavy for the domestic governing body, using a drag factor of 138–
weight (HWT) and light weight (LWT)], are given in Table 1. 140. All 2000 m tests were performed within 15 days of the
Height (stature) and body mass were measured using a laboratory visit.
stadiometer and beam-balance scales (Avery Berkel, Walsall,
UK), respectively. The percentage body fat was estimated
using the sum of four skinfold sites (Durnin & Womersley, Statistical methods
1974; Siri, 1956) using callipers (John Bull, British Indicators
Pearson’s product moment correlations was used to examine
Ltd., St Albans, UK).
the relationship between individual physiological variables
All evaluations were performed on a modified Concept II,
and 2000 m performance speed for men and women both
model C air braked rowing ergometer (Concept2, Nottingham,
separately and combined.
UK). The test system (Avicon II, Berlin, Germany) incorporated
As 2000 m rowing speed increases in proportion to the rate
a load cell (U9B, HBM Germany, Darmstadt, Germany) for
of energy expended by the rower but will also be limited by the
force measurement and a rotary transducer (ROD 454 M,
drag/resistance of the ergometer, our initial model to explain
Heidenhain, Germany). The subjects warmed up for 10 min,
2000 m rowing ergometer speed was based on the following
then rowed two build-up strokes, followed by five consecutive
proportional (curvilinear) allometric or power-function mod-
maximal rowing strokes at a fixed rate of 30 strokes/min. Max-
els (Nevill et al., 2006; Ingham et al., 2008),
imal force (Fmax), power (Wmax), work, stroke length and stroke
rate were averaged over the five strokes. Five 4 min increments Rowing speed ðm=sÞ ¼ a ðW VO2max Þk1 e ½1
were rowed on the ergometer, with 30 s rest between stages. Work
intensity was increased by 30 W for the men and by 25 W for the ·
where WVO2max is the power at VO2max, ‘‘a’’ is a constant and
women and by 2 strokes/min, at each stage. Following stage 5 ‘‘k1’’ is the exponents likely to provide the best predictor of
there was a 150 s rest followed by a 4 min maximal effort. rowing speed, and ‘‘e’’ is the multiplicative error ratio.
Pulmonary gas exchange was determined breath-by-breath. Further determinants, known to be proportional to 2000 m
Subjects wore a nose clip and breathed through a low dead rowing speed, can be added to the allometric model (eqn. [1])
space (90 mL), low-resistance (0.1 kPa/L/s at 15 L/s) mouth- and backward elimination (see Draper & Smith, 1981, Chapter
piece. Air was sampled through a 2 m small bore capillary line 6, for a discussion of this and other methods) will be used to
and analyzed for O2 concentration using a differential para- obtain the parsimonious model, i.e. at each step, the least
magnetic analyzer and CO2 concentration using a side-stream important variable is dropped from the current model until all
infrared analyzer (Oxycon Alpha, Viasys, Brighton, Sussex, remaining predictor variables make a significant contribution
UK), which were calibrated using gases of known concentra- to the final ‘‘parsimonious’’ model.
tion. Expiratory volumes were determined using a turbine The model (eqn. [1]) can be linearized with a log-transfor-
volume transducer (Viasys) that was calibrated using a 3-l mation, and linear regression can be used to estimate un-

Table 1. The physical characteristics (means  standard deviations) of the rowers, grouped according to sex and rowing weight class

HWT men  SD LWT men  SD HWT women  SD LWT women  SD

N 33 15 21 7
Age (years) 23.3 3.2 24.9 5.0 26.1 4.9 25.1 4.1
Height (cm) 192.4 5.4 181.3 4.1 179.4 4.9 167.8 1.6
Body mass (kg) 94.7 5.9 74.5 2.8 75.7 5.2 59.5 1.9
Body fat (%) 12.6 2.7 10.7 2.7 22.4 3.0 19.4 2.3

SD, standard deviation; HWT, heavy-weight rowers; LTW, light-weight rowers.

74
 Modelling 2000 m rowing ergometer performance
known parameters a and k1. The log-transformed model the fitted log-linear regression model, eqn [1]),
becomes, s 5 0.0167% or 1.69%, having taken antilogs.
loge ðspeedÞ ¼ loge ðaÞ þ k1 loge ðW VO2max Þ þ loge ðeÞ: ½2 To demonstrate the curvilinear nature of these
The parameter ‘‘a’’ can be allowed to vary between groups models, the rowing speeds for the male and female
(e.g. sex and weight class), thus conducting a form of analysis HWT and LWT rowers plus the fitted allometric
of covariance (ANCOVA). models are plotted in Fig. 2. Note the spread of data
around the fitted curvilinear models demonstrate a
‘‘shot-gun’’ effect, confirming the need to incorporate
Results a multiplicative error term ‘‘e,’’ that increases pro-
The mean 2000 m ergometer performance time (s), portionally with the size of WVO2max [see (eqn. [1])].
speed (m/s), VO· When all the variables identified in the correlation
2max, economy (ECON), peak power
(Wmax), peak force (Fmax), power at max VO · matrix (Table 3) were added to the model (eqn. [1]),
2max
(WVO2max ), power at 2 mmol/L (W2 mmol), power at backward elimination revealed the following allo-
3 mmols/L (W3 mmol), power at 4 mmols/L (W4 mmol) metric model to be the parsimonious solution to
and VO2 at lactate threshold (VO2LT), grouped by predict 2000 m rowing speed:
sex and weight class, are given in Table 2.
Table 3 presents the correlation coefficients for likely Rowing speed ¼ 0:842 ðW VO2max Þ0:22
½3
determinants with rowing ergometer speed for men  ðW max Þ0:073 VO2LT 0:072
and women separately and for all rowers combined.
ANOVA identified a difference in the mean (  SEE) that explained R2 5 96.16% (k1 5 0.22 SEE 5
rowing speeds by weight class (Po0.001) and sex  0.024, k2 5 0.073 SEE 5  0.019, k3 5 0.072
(Po0.001) but with no interaction (P40.05), Fig. 1. SEE 5  0.019) and the error ratio (the standard
As an initial exploration into the determinants of deviation of residuals about the fitted log-linear
2000 m rowing speed using the log-transformed regression model, eqn. [1]), s 5 0.0152% or 1.53%,
model (eqn. [2]), ANCOVA identified significant having taken antilogs. The above model was com-
differences in rowing speeds due to the main effects mon to all rowers irrespective of sex and weight class.
‘‘sex’’ and ‘‘weight class’’ (both P 5 0.001) but with
no sex-by-weight-class interaction. The ANCOVA Table 3. Correlation coefficients (r) for likely determinants with rowing
also identified power at VO ·
2max (WVO2max ) as a ergometer speed for men and women separately and for all rowers
significant covariate of 2000 m rowing speeds combined
(Po0.0001). The proportional allometric models
Women Men All
can be expressed as
·
Rowing speed ðHWT menÞ ¼1:079 ðW VO2max Þ0:28 ; VO2max (L/min) 0.74 0.82 0.94
ECON (ml/W)  0.46  0.02  0.33
Rowing speed ðLWT menÞ ¼1:058 ðW VO2max Þ0:28 ; WVO2max (W) 0.92 0.84 0.96
Wmax (W) 0.69 0.82 0.94
Rowing speed ðHWT womenÞ ¼1:039 ðW VO2max Þ0:28 ; Fmax (N) 0.69 0.81 0.93
W2 mmol (W) 0.78 0.77 0.92
Rowing speed ðLWT womenÞ ¼1:019 ðW VO2max Þ0:28 ; W3 mmol (W) 0.82 0.75 0.92
W4 mmol (W) 0.84 0.73 0.91
with R2 5 95.3% (k1 5 0.28, SEE 5  0.024) and the VO2LT (L/min) 0.45 0.83 0.92
error ratio (the standard deviation of residuals about
·
Table 2. The mean 2000 m ergometer performance time (s), speed (m/s), VO2max, economy (ECON), peak power (Wmax), peak force (Fmax), power at
VO2max (WVO2max ), power at 2 mmols/l (W2 mmol), power at 3 mmols/l (W3 mmol), power at 4 mmols/l (W4 mmol) and VO2 at lactate threshold (VO2LT),
grouped by sex and weight class

HWT men SD LWT men SD HWT women SD LWT women SD

2000 m time (s) 361.1 9.5 381.3 6.8 416.7 15.7 435.4 11.3
Speed
· (m/s) 5.5 0.1 5.2 0.1 4.8 0.2 4.6 0.1
VO2max (L/min) 5.84 0.45 5.08 0.40 4.13 0.30 3.71 0.19
ECON (ml/W) 15.31 0.92 14.98 0.69 15.74 0.72 15.71 1.19
WVO2max (W) 382.6 33.5 339.4 25.7 262.7 22.5 236.9 13.0
Wmax (W) 636.0 40.4 508.2 40.6 418.6 45.4 334.6 19.4
Fmax (N) 779.0 44.7 646.3 39.6 551.0 46.1 475.0 31.1
W2 mmol (W) 320.4 34.1 283.1 27.5 225.4 23.3 205.6 18.1
W3 mmol (W) 346.7 37.0 307.8 31.4 245.3 23.4 223.0 19.3
W4 mmol (W) 367.8 39.7 327.3 34.7 261.6 24.4 237.7 20.2
VO2LT (L/min) 4.5 0.4 3.8 0.4 3.1 0.3 2.8 0.2

SD, standard deviation; HWT, heavy-weight rowers; LTW, light-weight rowers.

75
Nevill et al.
Discussion ·
at VO 2max (WVO2max ) into the allometric model (eqn.
[1]) to predict rowing ergometer speed, the model was
The current findings provide a novel insight into the demonstrably curvilinear (Fig. 1), proportional to
relationship between physiological variables and (WVO2max )0.28, that explained 95.3% of the variance in
2000 m ergometer rowing performance. Previous rowing ergometer performance speed. At this early
work has adopted correlations or linear regression stage of the modelling process, significant differences
methods to identify either peak-power output (Wmax) remained between both levels of the main effects
·
(Bourdin et al., 2004) or power at VO2max (Ingham et ‘‘sex’’ and ‘‘weight class’’ using WVO2max as the only
al., 2002) as the best single predictors of 2000 m covariate.
rowing ergometer performance. The present study The model has a practical advantage for the coach
also adopted correlations to initially explore the and sports scientist. Given a rower’s power at
likely physiological determinants of 2000 m rowing ·
VO 2max (WVO2max ), we can use the curvilinear model
ergometer performance speed, identifying power at (eqn. [1]) to predict the rower’s average speed and
·
VO 2max (WVO2max ) as the best single predictor of hence his/her rowing time directly. To appreciate the
indoor rowing performance (Table 3). value of fitting the proposed curvilinear allometric
However, in contrast with previous linear meth- models, consider the following example. Suppose a
ods, the current study adopted more appropriate HWT male rower with a WVO2max of 300 W wishes to
proportional, allometric models to explore these increase his mean estimated rowing speed from 5.18
associations. As anticipated, when we entered power to 5.40 m/s, the model predicts that he would need to
increase his WVO2max from 300 to 350 W, an increase
5.8 of 50 W. In contrast, suppose a second HWT rower
HWT with a WVO2max of 400 W wishes to increase his esti-
5.6
LWT mated mean speed by the same amount (0.22 m/s)
5.4 from 5.61 to 5.83 m/s, he would require an increase in
5.2 WVO2max from 400 to 462 W, i.e., a greater increase in
speed (m.s–1)

5 WVO2max (24% more) of 62 W.

4.8
The above curvilinear power function exponent,
0.28 [SEE 5 0.027, 95% confidence interval (CI)
4.6
0.226–0.334], is similar to that derived from known
4.4 association between a rower’s power expended (P)
4.2 and the speed of the boat (u), given by P 5 cu3, i.e.
4 u 5 (P/c)0.33 where c is a constant depending on the
men women rowers weight, sex and boat type (http://www.at-
Fig. 1. The mean (  SEE) rowing speeds of heavy weight m.ox.ac.uk/rowing/physics/ergometer.html). Further
(HWT) vs light weight (LWT) rowers (Po0.001), by sex support for the fitted exponent comes from the
(Po0.001) but with no interaction (P40.05). known curvilinear association between cycling speed

5.8

5.6
HWT men
5.4
LWT men
speed (m.s–1)

5.2 HWT women

LWT women
5
HWT model men
4.8 LWT model men
HWT model women
4.6
LWT model women
4.4

4.2

4
200 250 300 350 400 450 500
WVO2max

Fig. 2. The rowing speeds for the male and female heavy weight (HWT) and light weight (LWT) rowers plus their fitted
allometric models.

76
 Modelling 2000 m rowing ergometer performance
and energy expenditure (Nevill et al., 2005). On level the stepwise linear regression model reported by
ground, the power demand of cycling (also predomi- Ingham et al. (2002) with R2498%, was also absent
nantly due to air resistance) is thought to be propor- of the grouping factors ‘‘sex’’ and ‘‘weight class.’’
tional to the cube of the cyclists’ speed (Olds et al., However, the fact that the model reports a rowing
1995). Consequently, the speed of a cyclist should speed intercept of 3.26 (m/s) for zero values of the
also be proportional to the cube root (0.33) of the predictors (e.g., power outputs) further highlights
power expended. the limitations of fitting and reporting such linear
One of the most insightful finding of the modelling models.
process reported above came from the backward A similar absence of the grouping effect ‘‘sex’’ was
elimination of likely determinants of indoor rowing identified by Nevill et al. (2008) when reporting
performance (variables reported in Table 2). The relative contributions of anaerobic and aerobic en-
parsimonious solutions (eqn. [3]) identified WVO2max ergy supply during 100, 400 and 800 m track running
as before, plus two additional covariates that made performance. Using accumulated oxygen deficit
significant contributions to rowing ergometer perfor- (AOD) and VO ·
2max to predict running performances,
mance speed, these being peak power (Wmax) re- the three regression models were able to confirm that
corded over five maximal strokes and VO2 at once the researchers had controlled for differences in
lactate threshold (VO2LT). the dominant energy supplies of AOD and VO ·
2max,
As described earlier, 2000 m rowing ergometer there was no significant difference between men and
performance depends on the functional capacity of women’s running speeds. It would appear that 100,
both the aerobic and anaerobic energy pathways 400 and 800 m track running performances as well as
(Secher, 1973). The parsimonious model (eqn. [3]) 2000 m rowing performances can be determined
would appear to confirm the need for high func- entirely by the appropriate contributions of aerobic
tionality from both pathways. The maximum power and anaerobic energy supply, irrespective of the
output (Wmax) averaged over the five maximal athletes’ gender and size.
strokes will provide an estimate of anaerobic cap- Recently, Nevill et al. (2009), was able to demon-
ability. The two components of WVO2max and VO2 at strate that in order to predict single sculling rowing
lactate threshold (VO2LT) will both provide esti- speed of elite junior male rowers on water, a greatly
mates of aerobic energy supply. Together, the improved prediction was obtained by including a
selection of aerobic and anaerobic variables in the combination (as a ratio) of both Concept II rowing
model that explained R2 5 96.16% of rowing erg- ergometer performance and body mass (m). The
ometer speed confirm the need and describe the authors found that water-based rowing speed (m/s)
interplay between the dual aerobic and anaerobic was proportional to the following ratio (allometric
energy pathways when performing a 2000 m rowing model),  [ergometer speed]1.87m  0.425. If, as reported
ergometer performance trials. We do recognize, above, the best single predictor of rowing ergometer
however, that the energy demands of ergometer speed is (WVO2max )0.28, we can estimate that single
rowing are very complex. For example, even the sculling rowing speed is likely to be proportional to
energy required to move the seat back and forth [ergometer speed]1.87m  0.425 5 [(WVO2max )0.28]1.87
 0.425
could help to account for some of the unexplained m 5 (WVO2max )0.52m  0.425. As reported above,
(3.84%) variance. 2000 m rowing ergometer speed is dependent on
A further reassuring aspect of the parsimonious absolute aerobic and anaerobic power output that
model (eqn. [3]) was the absence of the grouping are both known to benefit from greater body mass.
effects, ‘‘sex’’ and ‘‘weight class.’’ When only the In contrast, it would appear that single sculling
aerobic component of power at VO ·
2max (WVO2max ) rowing speed in water is likely to be dependent on
was incorporated into our initial exploration of the a greater power-to-weight ratio, given by
determinants of 2000 m rowing, the model was in- (WVO2max )0.52m  0.425 or (WVO2max m  0.82)0.52. We do
adequate, incapable of explaining the differences due recognize, however, that this power-to-weight ratio is
to ‘‘sex’’ and ‘‘weight class.’’ Simply by incorporating somewhat speculative, given that the ratio has been
further aspects of energy supply, in particular max- derived from data taken from two separate studies.
imum power (Wmax), the more complete or compre- In order to establish whether this power-to-weight
hensive model (eqn. [3]) were common to all rowers, ratio optimally predicts rowing in water, preferably
i.e. the models were able to explain the rowing all data (power output plus the ergometer and water
performances of all athletes, irrespective of sex and based rowing performances) should be obtained
weight class. from the same rowers.
Given that the present study predicts rowing Similar to our work, several researchers have
ergometer speed based on a range of different work proposed performance determinant models that con-
determinants, the absence of these group differences sider a contribution from a low-intensity metabolic
might well have been anticipated. We recognize that ‘‘threshold’’; a maximal and/or functional aerobic

77
Nevill et al.
capacity; and where possible an indicator of anaero- posed allometric model more realistically describes
bic/maximal power capability (in this study, VO ·
2LT, the greater increase in power required to improve a
WVO2max , Wmax, respectively), (Coyle, 1995; Jones & rower’s speed at an elite level, compared with that
Carter, 2000; e.g. Heugas et al., 2006). These ap- required of a club-standard rower at a slower speed.
proaches have shown some variation in the factors The best single predictor of Concept II rowing
deemed important and deterministic to middle dis- ergometer performance was found to be the power
tance performance. However, commonality is found ·
at VO 0.28
2max (WVO2max ) , that was able to explain
in the selection of parameters/capabilities from along 95.3% of the variance in rowing speed. The fitted
the breadth of the exercise intensity continuum. exponent, 0.28 (95% CI 0.226–0.334) is similar to the
Physiologically, the contribution of wide ranging exponent 0.33, based on the assumption that theore-
abilities makes particular logical sense considering tically a rower’s speed is proportional to the cube
the need for high capacity from aerobic and anaero- root (0.33) of the power expended. Furthermore,
bic ATP resynthesis pathways for overall middle simply, by incorporating both anaerobic as well as
distance performance. further aerobic determinants, the allometric curvi-
linear model identified power at VO ·
2max (WVO2max ),
plus peak power (Wmax) and VO2 at lactate threshold
Perspectives (VO2LT) as key proportional determinants of 2000 m
Previous studies have investigated the determinants rowing performance that was common to all rowers,
of indoor rowing using correlations and linear mod- irrespective of sex and weight class.
els. Linear models however assume that for a similar
increase in power, the same improvement in rowing Key words: rowing ergometer performance, power at
·
VO
speed will result irrespective of the rower’s level of 2max (WVO2max ), allometric models, curvilinear
performance (e.g. elite, club standards). The pro- power function.

References
Bourdin M, Messonnier L, Hager J-P, Ingham S, Whyte G, Jones K, Nevill A. predict cycling time-trial performance
Lacour J-R. Peak power output Determinants of 2000 m rowing in the field: a non-linear approach. Eur
predicts rowing ergometer performance ergometer performance in elite rowers. J Appl Physiol 2005: 94: 705–710.
in elite male rowers. Int J Sports Med Eur J Appl Physiol 2002: 88(3): 243– Nevill AM, Ramsbottom R, Nevill ME,
2004: 25: 368–373. 246. Newport S, Williams C. The relative
Coyle EF. Integration of the Jones AM, Carter H. The effect of contributions of anaerobic and
physiological factors determining endurance training on parameters of aerobic energy supply during track
endurance performance ability. Exerc aerobic fitness. Sports Med 2000: 29(6): 100-, 400- and 800-m performance. J
Sport Sci Rev 1995: 23: 25–63. 373–386. Sports Med Phys Fitness 2008: 49(2):
Draper NR, Smith H. Applied regression Mikuli P, Smoljanovi T, Bojani I, 138–142.
analysis, 2nd edn. New York: Wiley, Hannafin J, Pedii E. Does 2000-m Olds TS, Norton KI, Lowe EL, Olive S,
1981. rowing ergometer performance time Reay F, Ly S. Modeling road-cycling
Durnin JV, Womersley J. Body fat correlate with final rankings at the performance. J Appl Physiol 1995: 78:
assessed from total body density and its World Junior Rowing Championship? 1596–1611.
estimation from skinfold thickness: A case study of 398 elite junior rowers. Secher N. Development of results in
measurements on 481 men and women J Sports Sci 2009: 27: 4,361–4,366. international rowing championships
aged from 16 to 72 years. Br J Nutr Nevill AM, Beech C, Holder RL, Wyon 1893–1971. Med Sci Sports 1973: 5:
1974: 32: 77–97. M. Scaling Concept II rowing 195–199.
Heugas AM, Nummela A, Amorim MA, ergometer performance for differences Secher N. Rowing. In: Reilly T, et al: eds.
Billat V. Multidimensional analysis of in body mass to better reflect rowing in Physiology of sports. London: Spon,
metabolism contributions involved in water. Scand J Med Sci Sports 2009, 1990: 259–285.
running track tests. J Sci Med Sport doi: 10.1111/j.1600-0838.2008.00874.x. Siri WE. The gross composition of the
2006: 10(5): 280–287. Nevill AM, Jobson SA, Davison RCR, body. Adv Biol Med Phys 1956: 4:
Ingham SA, Nevill AM, Pedlar C, Whyte Jeukendrup AE. Optimal power-to- 239–280.
GP, Dunman N, Bailey DM. mass ratios when predicting flat and Spencer MR, Gastin PB. Energy
Determinants of 800m and 1500m hill-climbing time-trial cycling. Eur J system contribution during 200- to
running performance using allometric Appl Physiol 2006: 97(4): 424–431. 1500-m running in highly trained
models. Med Sci Sports Exerc 2008: Nevill AM, Jobson SA, Palmer GS, Olds athletes. Med Sci Sports Exerc 2001:
40(2): 345–350. TS. Scaling maximal oxygen uptake to 33: 157–162.

78