Methods

Subjects

For part 1 of the experiments, 12 well-trained, competitive-level cyclists were recruited. For part 2 of the experiments, six well-trained, recreation-level cyclists were recruited. Before participation in the study, subjects completed the physical activity readiness questionnaire to ensure that they were not at high risk of injury during the trials. Each was informed of the risks associated with the study. Informed consent was obtained in writing before the initiation of the study. All procedures conformed to the declaration of Helsinki.11 The Research and Ethics Committee of the Faculty of Health Sciences of the University of Cape Town Medical School approved the study.

Testing procedure

All subjects reported to the laboratory (which had stable climatic conditions (22.4°C, SD 1.4; 53% relative humidity, SD 5; 100.5 kPa, SD 8), for preliminary testing and subsequently at intervals to complete the experimental trials. During the first visit, subjects underwent anthropometric assessment and preliminary testing for measurement of maximal oxygen consumption (Vo_2max). The subject’s height (cm) and body mass (kg) were measured using a precision stadiometer and balance (model 770; Seca, Bonn, Germany; accurate to 10 g). Subjects were asked to refrain from eating or drinking for at least 2 h before each of the performance tests. Each subject was asked to perform a 90-minute low intensity “recovery ride” 24 h before the experimental trials. They were also asked to refrain from consuming any caffeine or other stimulants on the day of each performance test. Before each testing session subjects were questioned to confirm that they had adhered to these instructions.

Preliminary testing

All tests were performed on an electronically braked cycle ergometer (Computrainer Pro 3D; RacerMate, Seattle, USA), which allowed subjects to cycle on their own bicycles. The rear wheel was inflated to 800 kPa after which the system’s load generator was calibrated to a rolling resistance of between 0.88 kg and 0.93 kg. After a 15-minute standardised warm-up protocol, the device was recalibrated.

During the progressive exercise test the subjects were familiarised with both the Borg 15-point RPE scale (part 1) or the Borg category ratio scale (part 2) and a standard set of instructions was given for the subsequent trials.

Progressive exercise test

Testing for peak oxygen consumption (Vo_2peak) was performed at a starting work rate of 2.50 W.kg⁻¹ body mass. The load was increased incrementally at a rate of 20 W every 60 s until the subject could not sustain a cadence greater than 70 rpm. Ventilation volume (V_E), oxygen uptake (Vo₂) and carbon dioxide production (VCo₂) were averaged over 15-s intervals using an on-line breath-by-breath gas analyser and pneumotach (Oxycon; Viasis, Hoechberg, Germany). Vo_2peak was recorded as the highest Vo₂ reading recorded for 30 s during the test. Peak power output was calculated by averaging the power output for the final minute of the Vo_2peak test. Subjects were requested to refrain from standing on the pedals throughout the test.

Cycle trials

Part 1

Each cyclist completed a self-paced maximal 40 km time trial (TT) on a simulated flat 40 km course weekly for four consecutive weeks (TT1, TT2, TT3 and TT4). Subjects were allowed to consume water freely12 throughout each test and were asked to produce the fastest possible time. All feedback was withheld, except completed distance. Subjects were asked to report their RPE score at 5 km intervals.

Subjects completed a fifth 40 km time trial (UTT) during which all feedback, including completed distance, was withheld until the final kilometre, at which point they were informed that they had 1 km to complete. During this trial, subjects were asked to report their RPE at 5 km intervals. In addition, to cause confusion and uncertainty about the remaining distance, they were asked to report their RPE randomly at 18, 27, 33 and 38 km. After 39 km subjects were informed that they still had 1 km to complete.

Part 2

Following a 40 km familiarisation trial, subjects completed four more trials at 3–7 day intervals. These trials were 5, 10, 40 and 100 km in duration and were completed in random order. Subjects were informed about the trial distance before each trial.

During the trials, subjects were allowed to ingest a commercially available sports drink (8% carbohydrate content) at a rate of 600 ml/h, divided into 150 ml every 15 minutes, as this has been shown to prevent the development of hypoglycaemia during prolonged exercise bouts.13 Water was also freely available during the trials.

RPE scores were recorded at the start of the trial and at 10% distance intervals using the Borg category ratio scale.14

Statistical analysis

Data were analysed for statistical significance using STATISTICA version 7.0. Statistical significance was accepted when p<0.05. All data are expressed as means (SD).

Homogeneity of variance for all trials was measured using the Levene’s test. When this test showed that the variance differed, a non-parametric analysis of variance (ANOVA), Friedman test, which compared multiple dependant variables was used to identify significant differences between trials. When this analysis showed a statistical significance for all trials, a Wilcoxon matched pairs test for two dependent variables was used to confirm significant differences between trials for specific time points.

For parametric datasets, a two-way ANOVA with repeated measures was used to examine differences between trials. When a significant difference was found for either main effect (trial or time), a Tukey post hoc analysis was performed.

To examine the rate of increase in RPE scores, all RPE points were plotted against time using GRAPHPAD PRISM version 3.0. A fourth order polynomial function (determined as best fit for RPE vs time) as well as a linear regression plot of RPE versus time, passing through the final data point were plotted. Differences between these two plots were calculated using the calculated slope and intercept equations.

The sum of residuals (root of differences squared) for the differences between fourth order polynomial and linear plots was considered a deviation from linearity (linearity score), whereas the area under the linear plot but above the fourth order polynomial plot was considered a conservative deviation from the optimal RPE strategy (conservative score; fig 1).

Results

Subject characteristics

The descriptive characteristics and results of the research subjects are shown in table 1.

Part 1

RPE strategy

RPE increased significantly over time in the first four trials (p<0.001). A significant time by trial interaction effect was found. RPE was significantly greater at 20, 25, 35 and 40 km in TT4 in comparison with TT1 (p<0.05; fig 2).

The mean RPE for TT1 (17.18, SD 0.79) was significantly lower than during TT2 (17.68, SD 0.85), TT3 (17.75, SD 0.88) and TT4 (17.83, SD 1.05; p = 0.006; table 2). There were no differences between mean RPE scores for TT2, TT3 and TT4.

During the UTT (17.51, SD 1.14), mean RPE scores were lower than during TT4 (p = 0.02). RPE scores at 25, 30 and 35 km for the UTT were significantly lower than during TT4 (p<0.05; fig 3).

The linearity of RPE scores (linearity of relationship between RPE and distance) increased significantly from TT1 to TT4 (p = 0.02) and subsequently decreased but did not reach significance for the UTT in comparison with TT4 (table 2). RPE conservative scores decreased significantly from TT1 to TT4 (p = 0.03) and subsequently increased again in the UTT in comparison with TT4 (p = 0.02).

Mean power output during time trials was significantly related to mean RPE (fig 4A), to linearity of RPE progressions (fig 4B) and to the conservative component (fig 4C).

Power output

Power output during TT4 was significantly greater than during TT1 from 20 km to the end of the trial (p<0.05; fig 5). During the UTT, power decreased after 30 km and was significantly lower at 35 km and 40 km than in TT4 (p<0.05; fig 6).

Power/RPE ratio

To evaluate the interaction between the temporal changes in power output and RPE, a ratio of power to RPE was calculated over the course of all five trials. The results are depicted in fig 7. In all five trials, the ratio decreased significantly over time for the first half of each trial (p<0.03), but there was no statistically significant change during the second half of the 5 km, 10 km and 40 km trials (fig 10). There were no differences between trials for any time points. The relationship between change in the power/RPE ratio and time was best described as a one-phase exponential decay (R²  =  0.2187).

Part 2

RPE strategy

Figure 8 displays the RPE measured at 10% intervals over all four trials. The RPE in the 5 km trial was significantly higher than during the 40 km and 100 km trials, and the RPE in the 10 km trial was significantly higher than during the 100 km trial (p<0.001). There was no difference between average RPE in the 5 km and 10 km trials, the 10 km and 40 km trials or between the 40 km and 100 km trials.

RPE increased significantly over time in all four trials (p<0.001). There was a significant time by trial interaction effect (p = 0.01). The RPE at 20% of the 5 km trial was significantly higher than the 100 km trial and it remained higher until 80% of the trial (p<0.04). The RPE was significantly higher in the 5 km trial at the start compared with the 40 km trial and it remained higher until 30% of the trial and again from 60% until 80% of the trial (p<0.04).

Power

Trials of shorter distances had a significantly higher average power output compared with the longer trial durations (p<0.001; fig 9).

The power output during the 5 km trial was significantly higher than the 10 km trial during the first 20% of the duration (p<0.003) but there were no differences in power output over the remainder of the trial. The power output in the 5 km trial was significantly higher than the 40 km and 100 km trials throughout the entire duration of the trials (p<0.005).

There was a significant difference in power output in the 10 km and 40 km trials in the first 10% interval and final 10% interval (p<0.02); however, there were no differences in power output over the rest of the trial.

The 10 km trial had a significantly higher power output than the 100 km trial throughout the duration of the trial (p<0.02).

Power/RPE ratio

In all five trials, the ratio of power/RPE decreased significantly for the first half of each trial (p<0.001) but there was no statistically significant change in the power/RPE ratio for the second half of the 5 km, 10 km and 40 km trials (fig 10). The power/RPE ratio continued to decline in the second half of the 100 km trial (p<0.001). There were no differences between trials for any time points for the 5 km and 10 km trials and for the 5 km and 40 km trials. The power/RPE ratio was lower in the 100 km trial in comparison with the 5 km trial in the first 10% and last 10% (p<0.03). The power/RPE ratio was lower in the 100 km trial in comparison with the 40 km trial for the first 10% (p<0.001; fig 10). The change in the power/RPE ratio over time was best described as a one-phase exponential decay for all trials (R²  =  0.555).

Discussion

The results of this study indicate that during maximal exercise bouts of varying duration, increases in perceived exertion are proportional to the relative distance completed. These findings support similar findings in other studies.6 9 10 15

However, in contrast to the findings of Joseph et al9 and others,10 15 we showed that the rate of increase in perceived exertion is not always constant, but changes in relation to certainty about the endpoint of exercise (fig 2 and 3) as well as exercise duration (fig 8). When subjects were initially unfamiliar with the exercise bout, they chose a perceived exertion strategy that maintained a larger metabolic reserve, which was then accessed near the end of the bout. This strategy was associated with a non-linear growth in RPE over time and a conservative approach (table 2, fig 2), rather than a linear strategy. With increased familiarity with the required exercise task, the RPE strategy became more aggressive, linear and presumably with less metabolic and cardiorespiratory reserve.

Similarly, during exercise bouts of longer duration (40 km and 100 km), subjects started exercise bouts with lower RPE scores and maintained these lower RPE values until the final 20–30% of the exercise bout, after which they increased the RPE to values similar to those recorded during the shorter exercise trials (5 km and 10 km). Subjects therefore maintained a greater metabolic and cardiorespiratory reserve during longer trials in which there was an increased possibility that the pacing strategy was incorrect or when unforeseen factors might arise that might affect the athlete’s capacity to sustain that pre-chosen pace.

When knowledge of the endpoint was obscured by blinding subjects to the completed distance and confusing them by asking for RPE scores at random, subjects reverted to a more conservative RPE strategy and once again maintained a greater metabolic reserve (table 2, fig 3).

This indicates that the exercise intensity and duration are not absolutely fixed at the start of an exercise bout but rather that a feed-forward strategy is implemented that is open to alteration during the bout in response to environmental, physiological, psychological or other factors. In addition, RPE scores do not increase linearly for all exercise bouts, regardless of distance but rather increase in relation to the certainty of the metabolic requirements of the exercise bout and the athlete’s subconscious belief that he or she is able to meet those requirements.

We8 and others15 16 have shown that perceived exertion is directly related to afferent feedback and that during exercise that proceeds with an unknown endpoint (open loop exercise), the relationship between power output and RPE declines in a linear fashion as exercise progresses.8 16 In contrast, during this study, the relationship between power output and perceived exertion declined in an exponential decay fashion. Therefore, during the second half of all the exercise bouts the power/RPE ratio declined at a slower and generally non-significant rate (figs 7 and 10). This indicates that during exercise bouts in which subjects have knowledge of the exercise duration (closed loop), the workload/RPE relationship is attenuated as the endpoint approaches. This relationship is virtually superimposable when expressed in a scalar fashion (fig 10). During the longest bouts (100 km), the workload/RPE ratio was lower at the start and end of the exercise bout (fig 10). These findings indicate that afferent feedback and its effect on perceived exertion are attenuated over time in relation to knowledge of the upcoming endpoint of the exercise bout. Previous research confirms that confusion about the endpoint of an exercise bout is associated with higher RPE scores in relation to workload.4 As longer bouts are associated with a greater uncertainty, the attenuation of afferent feedback occurs later in absolute but similar in relative time in relation to the end of the exercise bout.

Conclusion

In summary, we have provided further evidence that exercise is controlled in a feed-forward and adaptive teleoanticipatory fashion. In addition, we have shown that increasing familiarity with the exercise bout and certainty about its endpoint are associated with a more aggressive strategy that produces a superior exercise performance. We have also shown that afferent feedback is attenuated in relation to the relative distance from the end of the exercise bout. These data show that during endurance exercise, performance is controlled centrally in the brain.8 This is achieved by altering the rate of progression of perceived exertion. During exercise, certainty about the endpoint and the duration of exercise that remains affect both the RPE strategy and the manner in which afferent feedback is interpreted by the brain.