**I**n recent years, interest in anaerobic energy production in sports has increased because of the realization that most team sports and individual events rely heavily on anaerobic power in addition to aerobic power^{(1,5,12)}. Relatively little attention, however, has been directed to the anaerobic fitness tests that may reflect individual ability to do supramaximal work for a short period of time^{(1,16,18,22,29)}.

Laboratory measures to estimate anaerobic capabilities have included assessment of adenosine triphosphate (ATP), phosphocreatine (PC), muscle glycogen, and lactate levels ^{(1,12,14)}. To avoid invasive techniques, a number of short-term, high-intensity performance test methods have been published using either cycle ergometer, track, or treadmill to predict anaerobic capabilities^{(1,19,23)}. One test of mechanical power utilizing the cycle ergometer, the Wingate Anaerobic Capacity (AC) test, has been used more extensively than others because it has been shown to be objective, reliable, and simple to administer ^{(2)}. Questions about its validity, however, have been raised that are related to an aerobic contribution to the 30-s exercise bout and the inability of the test to utilize maximally the AC ^{(28)}.

It has been proposed that an indirect assessment of AR capacity could be obtained by performing a series of three exhaustive work bouts at three different power outputs on a cycle ergometer ^{(24)}. The exercise trials yield a limit work and limit time for each of the three exhaustive work bouts. This technique is based on the Monod and Scherrer Critical Power test ^{(23)} yielding a very linear relationship for the three plotted data points and provides two separate measures of physiological performance: 1) the slope parameter “b”, or critical power that represents aerobic power, and 2) the y-intercept“a”, that represents anaerobic reserve or AR^{(5,23,24)}. AR, unlike the slope parameter, is unaffected by hypoxia and occluded circulation ^{(24)}. Although the AR test has withstood several physiological challenges to validity, both the Wingate and Critical Power tests and results are open to interpretation ^{(12,28)}. Therefore, the primary objectives of this study were to investigate the relationship between the anaerobic components of the Wingate test and the Critical Power test, and since little data is available on female subjects, to describe differences in AC and AR that may be attributable to gender.

#### METHODS

**Subjects.** Thirteen male and 16 female subjects volunteered for this investigation. Subjects were chosen randomly from undergraduate and graduate college students and were not preselected for fitness or anaerobic training, although some were active in various sports and others were sedentary. The project was approved by the Institutional Review Board and each subject gave informed consent prior to testing.

**Wingate Anaerobic Capacity.** AC was determined on a Monark cycle ergometer. Prior to initiation of the test, the seat height of the bicycle was adjusted to allow for near full extension of the subject's legs while pedaling. The subject warmed up for 2-4 min at an intensity sufficient to cause heart rate increases to 150-160 bpm. Then following a 2-min rest, the subject began pedaling as fast as possible for 30 s while resistance was quickly adjusted to the nearest 0.5-kp setting corresponding to 0.090 kp·kg^{[minus 1} of body weight for men and 0.075 kp for women^{(2,29)}. A rapid workload setting during all tests was achieved by presetting workload prior to the start of the test and making final adjustments after the start. The required settings were always achieved in less than 2 s. At the moment the final load was achieved, an electronic pedal revolution counter was zeroed and allowed to count total completed revolutions over 30 s. AC was calculated as the total work performed during the 30-s test period in which constant resistance was maintained.

**Critical Power Testing.** Each subject performed three exhaustive work bouts on a Monark cycle ergometer at three different power loadings determined according to the subject's physical fitness. Each exhaustive exercise bout resulted in a unique time limit and work limit for each exercise bout. A plot of the three limit work (y-axis) vs limit time (x-axis) yielded a linear regression determining y-intercept (AR) and slope (critical power)(Fig. 1). In general, a first and highest work rate was selected that could be maintained by each subject for a minimum of 45 s, as recommended by Monod and Scherrer ^{(23)}. Thereafter, subsequent trials were completed at lower work rates lasting longer time periods. The lightest resistive loads were selected to produce limit times of approximately 6 min or less. On the average, the highest, medium, and lowest resistive loads resulted in limit work times of 1-2 min, 2-4 min, and 4-10 min, respectively.

The load selection process was by trial and error, taking into account subject weight, fitness, fatigue rate on the Wingate test, and performance(limit time) on the previous higher work rate (i.e., resistance). Power output was maintained nearly constant by monitoring pedal speed, and resistance values were chosen sufficiently high as to lead, eventually, to the onset of muscular fatigue.

Each of the three trials for each subject was followed by a sitting or supine 20- to 30-min rest period that allowed the heart rate to return to within 10 bpm of resting value in every case. This time period also allowed for near total recovery of muscle phosphagen stores ^{(8)}. Prior to the first work bout, the seat height was adjusted to allow for near full extension of the subject's legs while pedaling. The subject performed a 2- to 4-min warm-up by pedaling against a resistance of 0.5 kp. Following a 2-min rest, the test was initiated by having the subject begin pedaling at a rate of 69 rpm (25 km·h) against an unloaded flywheel. The predetermined loading was then applied, as in the AC test, and when the appropriate power loading was attained, a stopwatch and an electronic pedal revolution counter were activated. Immediately upon loading the ergometer(applying resistance), the revolution rate decelerated (involuntary) to 60 rpm and was maintained at 60 ± 5 rpm to achieve maximal sustainable power^{(4,6)}. The subject was encouraged to maintain the required pedaling rate throughout the entire work bout. The exercise bout was terminated immediately when the subject could no longer maintain a pedal rate above 55 rpm (20 km·h^{-1}). Anaerobic reserve (AR) was determined to be the amount of work corresponding to the y-intercept of the linear relationship between Work Limit (total work to exhaustion) and Time Limit(total time to exhaustion). The y-intercept at zero time predicts stored work capacity (energy) where the appropriate units are joules. Critical Power was determined to be the power output (watts) corresponding to the slope of the linear relationship between Work Limit and Time Limit and has also been termed maximal sustainable power ^{(17)}. Internal reliability of AR and the slope parameter was monitored using a Critical Power retest performed by six subjects 1-3 d apart.

**Aerobic Power (˙VO2max).** Aerobic power was determined on an electric cycle ergometer (Quinton Inst. Model 844, Bothell, WA) by using a graded work protocol. Power loads were increased 16.3 W·min^{-1} to exhaustion. Oxygen consumption was measured by using a metabolic cart(Sensormedics Model 4400, Yorba Linda, CA) calibrated for CO_{2} and O_{2} before each test using certified gas mixtures. Maximal aerobic power was recorded as the power load (W or kg-m·min^{-1}) or as˙VO_{2max} when oxygen consumption plateaued and R was at least 1.2 or greater.

**Statistical Procedures.** Means, standard deviations, Pearson product-moment correlations, simple linear regressions, and dependent paired*t*-tests were obtained by using STATMOST® (Data Most Corp., Salt Lake City, UT). To examine the important factors contributing to AC and AR test estimates, forward stepwise regression analysis was performed using absolute ˙VO_{2max}, weight, height, age, and gender as independent variables. To further assess the influence of weight or fitness on gender differences, analyses of covariance (ANCOVA) were performed using weight or˙VO_{2max} (l or ml·kg^{-1}·min^{-1}) as the covariates. Statistical significance was tested at the *P* < 0.05 level.

#### RESULTS

The descriptive data for the male and female subjects' fitness and physical characteristics are presented in Table 1. The female subjects were significantly shorter, lighter, and had a lower˙VO_{2max} (35.4 ± 1.5 vs 41.7 ± 2.4 ml·kg^{-1}·min^{-1}, *P* < 0.05). The pooled data show that subjects' Critical Power tests yielded highly linear relationships between Work Limit and Time Limit for all subject with a range of values for “r” being 0.98 to 1.00.

Test-retest correlation for AR was r = 0.81, SEE = 792 J (*N* = 6,*P* < 0.05, unidirectional), with mean ARs on the test-retest 12459± 2650 and 12517 ± 2305 J, respectively (t = -0.09, *P*> 0.05). Test-retest correlation for the slope parameter was r = 0.89, SEE= 21 W (*N* = 6, *P* < 0.05). The mean slopes on the test-retest were 187 ± 38 and 183 ± 46 W, with no significant difference between the means (t = 0.39, *P* > 0.05).

AC (240.2 ± 30.5 J·kg^{-1}) and AR (184.0 ± 51.2 J·kg^{-1}) were not well-correlated (r = 0.07, *P* > 0.72). When expressed as joules and not adjusted for body weight the correlation was statistically significant but low (r = 0.41, *P* < 0.02). In all cases (male, female, or total), the anaerobic estimates derived from the Critical Power test were 33.2, 14.9, and 24% lower (*P* < 0.05), respectively, than the Wingate test (Table 2). When comparing male and female performances for each test, males were 20%(J·kg^{-1}) higher than females for the Wingate test, but 6% lower for the Critical Power test (Table 2). To see if other gender differences existed or may have been confounding, we repeated the correlational analyses and found a significant but low correlation between AC and AR (J·kg^{-1}) in females (r = 0.52, *P* < 0.03). However, the same comparison for males showed a negative but nonsignificant relationship (r = -0.13, *P* > 0.66). When comparing total capacity in joules (not body-weight adjusted), the female value was lower (r = 0.14,*P* > 0.59), whereas the male value was nearly identical (r =-0.04, *P* > 0.90). Finally, although weight and ˙VO_{2max} differences may have influenced the gender-based correlational comparisons, ANCOVA analyses using ˙VO_{2max} or weight as covariates had no significant effect on the quantitative AR and AC gender comparisons reported in Table 2. However, when the total subject sample was analyzed for the influence of height, weight, gender, ˙VO_{2max}^{(1)}, and age on AC (J) or AR (J), stepwise regression analysis revealed that only gender (F = 4.72, *P* < 0.04) significantly contributed to AR (J) prediction with all other independent variables not meeting inclusion criteria (*P*'s > 0.32 for all). In contrast, the same stepwise regression showed an ordered inclusion of weight(*P* < 0.001), age (*P* < 0.001), and sex (*P*< 0.001), which contributed significantly to predicting AC (J). These stepwise regression data are only illustrative because the sample in this study is too small (needed *N* > 75 for five independent variables) to allow meaningful interpretation of the regression results. Total and gender-specific comparisons for the Wingate and Critical Power tests are presented in Table 2.

#### DISCUSSION

Monod and Scherrer ^{(23)} suggest that the y-intercept of the linear relationship between Work Limit and Time Limit represented the AR specific to the muscles used in performing the test. The test as used in the present study is reliable and compares favorably with reliability coefficients and coefficients of variation reported in similar studies^{(9,25,29)}. Jenkins and Quigley have shown that AR can effectively “track” or measure changes in AC^{(18)}. The significant correlation between AR and AC (J) in the present study provides limited support for the validity of AR as a measure of anaerobic capability. Although these data are consistent with a recent report ^{(28)}, they are physiologically marginal. Although this relationship is generally supported by the literature^{(4,10,12)}, in the present study only 16% of the variance between AR and AC was shared. It appears that other factors play an important role in their relationship. One such factor was probably the aerobic component of the Wingate test (related to the oxygen kinetics), which the literature estimates at between 13% and 44% of the total energy supply^{(3,27)}. In the present study the correlations between maximal aerobic power (MAP) and AC was r = 0.38 (r^{2} = 0.14), accounting for 14% of the unexplained variance between AR and AC. Another factor that may affect the relationship is that the Wingate test is too short to exhaust the anaerobic energy stores and may underestimate AC^{(29)}. It has been estimated that during exhausting exercise, muscle concentrations of high-energy phosphates (PC and ATP) decrease about 18 mmol·kg^{-1} wet weight muscle^{(7,13,20,21,26)}. Also, assuming a 40% body weight involvement, exhaustive/intense cycle ergometer exercise yields a 7.2-mmol high-energy phosphates/kg body weight utilization. When these values are converted to volume of 0_{2}, the breakdown of PC and ATP contributes 25.6 ml·kg^{-1}. An additional 6.0 ml·kg^{-1} is contributed by 0_{2} stored in blood and myoglobin ^{(9,23)}. The total 31.6 ml O_{2}·kg^{-1} is in good agreement with 35 ml O_{2}·kg^{-1} oxygen deficit calculated from 184 J·kg^{-1} for AR in the present study. The current data, when adjusted for body weight and fitness, are in perfect agreement with Medbo et al., who have reported that AC can be adequately represented by maximal accumulated oxygen deficit ^{(22)}. These values, however, are 24% lower than AC measured by the Wingate test. This may be due to an overestimate related to aerobic contribution error (15%) of the Wingate test^{(28)}. A major unresolved point is the missing contribution of anaerobic glycolysis to our estimates of AR-related O_{2} deficit calculations. If this is an oversight, adding an estimated 44 ml O_{2}·kg^{-1} ^{(16,22)} would yield 65-75 ml O_{2}·kg^{-1} total deficit depending on the assumptions for activated muscle mass (range, 25-40%). The preceding estimate is clearly double the deficit of 35 ml O_{2}·kg calculated from the 184 J·kg^{-1} mean AR. Since AR values were maximal and at least equal to or greater than values reported in the literature for trained subjects ^{(5,12,18)}, it seems the AR energy reserve does not represent conventional AC ^{(11)}. Furthermore, following 60-s supramaximal cycle ergometer exercise (150% MAP, 90-104 rpm) in well-trained subjects, Gollnick et al. report only 20% reduction in muscle glycogen with only 5 of 359 sampled FT fibers showing moderate reductions in glycogen content ^{(10)}. Therefore, it is possible that the extrapolated “time zero” AR derived from the Critical Power test may not account for glycolytic capacity, as is the case for most other noninvasive measures of AC. Proper interpretation of this issue will require further research.

**Gender Comparisons.** The mostly average fitness levels of the subjects is confirmed by the ˙VO_{2max} values reported inTable 1. As would be predicted, the females were significantly lower than the males in aerobic estimates (15%) and in the absolute anaerobic estimates as well (38%). When adjusted for body weight, however, the significant gender difference in AR disappeared. In fact, the female values were slightly higher than the males (188 vs 177 J·kg^{-1}; *P* = not significant). The present data suggest that AR assessment is influenced by body (muscle) mass. This finding is confirmed by ANCOVA where previously gender-specific AR (J) differences on the Critical Power test (*P* = 0.015) become insignificant when weight is used as a covariate. These findings did not extend to the Wingate AC assessment where ANCOVA adjustment for body weight did not affect AC differences between males and females. The greater muscle mass distribution in the lower body of females seems to have contributed to a higher relative anaerobic work output for AR (J) during cycle ergometer exercise^{(30)}. Thus, expressing AR relative to“total” body weight may have resulted in the higher relative AR(J·kg^{-1}) in females. Why these factors did not affect the Wingate results similarly is not presently clear.

The correlational analyses showed interesting gender-related variations between AC and AR. These correlations were affected by body weight adjustments in females but not males (r = 0.14 in males to r = 0.52 in females). It has been reported that a higher aerobic contribution (25% vs 20%) to the AC is achieved by females than males ^{(15)}. This may explain the body weight adjusted increases in correlations with AR in females. Since females' relative leg muscle masses are not appreciably different in strength from males ^{(30,31)}, it is hard to explain the observed discrepancies. Moreover, a nonsignificant negative correlation trend(r = -0.14, *P* > 0.50) between AR (J) and body weight (r = 0.01,*P* > 0.97 for males) clouds the issue further. Since these findings were not anticipated, higher anaerobic energy (J) being expected in subject with greater body weight (i.e., gross muscle mass), we are unable to further explain some of these gender differences. Further research on this problem is needed.

Since fitness (˙VO_{2max}) and body weight are both known to be key contributors to gender differences in AC with ˙VO_{2} specifically influencing Wingate test estimates of AC, ANCOVA analyses were performed to assess the effects of ˙VO_{2max} on observed gender differences in AC and AR. The statistical analyses showed that the greater male AC (J) assessments were unaffected by ANCOVA adjustment for ˙VO_{2max}^{(1)}. In contrast, the AR (J) gender differences became nonsignificant when covaried for ˙VO_{2max} ^{(1)}. Again, it appears that the Wingate and Critical Power tests are not assessing the same anaerobic compartment. Two key variables (weight and fitness) related to anaerobic capacity/reserve, which are major determinants of gender-based work capacity difference, affect the two tests differently.

It appears that the differences in AR and AC when expressed in similar units can be explained by proposing different definitions for “anaerobic capacity” and “anaerobic reserve.” If a definition of“anaerobic reserve” is limited to energy release (energy stores) from muscle ATP, PC, and oxygen stores in blood and muscle (the alactacid anaerobic energy), as seems the case for AR estimation, time limits for work approach 10-20 s. The energy associated with this work would be lower than estimated from the Wingate test (aerobic overestimation and anaerobic glycolysis). This argument is well-supported by Jenkins and Quigley, who show progressively greater total anaerobic work estimates from five repeated 60-s exercise bouts resulting in progressively greater correlations with AR^{(17)}.

In conclusion, the results of the present study demonstrate that the Critical Power test is a reliable and apparently valid method for the measurement of AR. As implemented in this study, the Critical Power test provides, theoretically and experimentally, a good estimate of alactacid AR. The criticism of the Critical Power test's inability to maximally account for stored anaerobic energy when compared with the Wingate test may be a function of comparing different anaerobic energy compartments. We propose that AR may not include anaerobic glycolysis. Also, the study provides preliminary data on gender differences as assessed by the Critical Power test. These data suggest that neither the Wingate or Critical Power tests represent anaerobic capacity as presently defined in the literature ^{(12)}.