Perhaps the best example in sport of where breathing frequency is naturally reduced is during front crawl swimming. During such exercise breathing (and specifically inspiration) must be coordinated with stroke mechanics and as a result tidal volume (V_T) is increased to compensate for the reduced breathing frequency (RBF) (Dicker et al., 1980). Such a restricted breathing pattern during front crawl swimming limits minute ventilation (V_E) and can lead to hypercapnia i.e. carbon dioxide retention (Cordain and Stager, 1988). The breathing pattern that swimmers adopt during front crawl largely depends on the swimming distance. For example, it is usual practice for swimmers to breathe every second stroke cycle during events of 200 metres or more, but to adopt a more restricted pattern during shorter front crawl distances (Maglischo, 2003). Consequently, it is not surprising that coaches have included RBF sets i.e. taking a breath every fourth, fifth, sixth or eighth stroke cycle during training: a practise which has been used in swimming since the 1970’s. Several studies have examined the influence of RBF on different physiological parameters in swimming since then. RBF during swimming reduced V_E, increased the alveolar CO₂ concentration (Cordain and Stager, 1988; Holmér and Gullstrand, 1980; Peyrebrune et al., 2003; Town and VanNess, 1990) and induced higher partial pressures of carbon dioxide in capillary blood (Kapus et al., 2002; 2003). Due to the technical limitations of measuring respiratory and blood parameters during swimming, the impact of RBF on physiological parameters has been investigated during cycle ergometry test (Kapus et al., 2009; 2010a; 2010b; Sharp et al., 1991; Yamamoto et al., 1987; 1988) and treadmill running (Matheson and McKenzie, 1988): here breathing frequency can be modified and respiratory and blood parameters measured with greater ease. These studies confirmed the presence of marked hypercapnia as a result of RBF during exercise. In addition, these studies also examined hypoxemia by measuring capillary blood Po₂ and oxygen saturation during exercise with RBF (Kapus et al., 2009; 2010a; 2010b; Matheson and McKenzie, 1988; Sharp et al., 1991; Yamamoto et al., 1987).

All of the above studies investigated the acute effects of RBF. To our knowledge only one study has examined the influence of RBF training on exercise performance and did so in swimming. After four weeks of RBF training (breathing every fourth stroke cycle during front crawl) swimmers reduced their breathing frequency from 32 ± 5 breaths per minute to 25 ± 7 breaths per minute during a maximal 200 meter front crawl swim (Kapus et al., 2005). Although the mechanisms responsible for this training adaptation are not clear, it is possible that RBF training increased tidal volume and did so sufficiently to the extent that V_E was increased. Given that RBF creates a hypercapnic training stimulus (Dicker et al., 1980; Peyrebrune et al., 2003) it is possible that the adaptation to hypercapnia could be the result of RBF training. Consequently the purpose of this study was twofold: firstly, to investigate the influence of RBF training on V_T during incremental exercise where breathing frequency is restricted. Secondly, to investigate the effect of RBF training on the ventilatory response during exercise when breathing a 3% CO₂ mixture, which is thought to be an indicator of ventilatory sensitivity to hypercapnia (Florio et al., 1979). Considering suggestions from previous studies, we hypothesise that RBF training will increase V_T during incremental exercise where breathing frequency is restricted and decrease ventilatory sensitivity during exercise when breathing a 3% CO₂ mixture.

Twelve healthy male participants volunteered to participate in this study. As students of the Faculty of Sport, they were active but none were currently participating in a regular training programme. During the study period, all participants stopped their usual physical activity (recreational) and only exercised during the training sessions as part of the experiment. Descriptive measures of participants and training groups are presented in Table 1.

None of the participants were smokers and all were free from respiratory diseases. They were fully informed of the purpose and possible risks of the study before giving their written consent to participate. The study was approved by the National Medical Ethics Committee of Slovenia.

All testing and training was performed on an electromagnetically braked cycle ergometer (Ergometrics 900, Ergoline, Windhagen, Germany) with a pedal cadence of 60 rpm. Participants completed the following exercise tests in the same order pre- and post-training: 1) an incremental test to obtain peak power output with spontaneous breathing (SB) (PPO_SB) and peak oxygen uptake (VO₂ peak); 2) an incremental test to obtain peak power output with RBF (PPO_RBF); and 3) a constant load test with SB to determine the ventilatory response during exercise when breathing a 3% CO₂ mixture (ventilatory sensitivity). At pre-training testing, two additional tests were performed: 1) a constant load test with SB at PPO_SB; 2) a constant load test with RBF at PPO_RBF. The data obtained from these tests were used for determining the training interval sets only. Each exercise test was performed on a different day. All testing took place under controlled environmental laboratory conditions (21°C, 40-60% RH, 970-980 mbar) and at the same time of day. Participants were asked to maintain their usual eating habits and to avoid consuming food 2 hours before testing. Post-training testing began 2 days after the last training session.

RBF was defined as 10 breaths per minute and was regulated by a breathing metronome. The breathing metronome was composed of a gas service solenoid valve (24 VDC, Jakäa, Ljubljana, Slovenia) and a semaphore with red and green lights. Both were controlled by micro-automation (Logo DC 12/24V, Siemens, Munich, Germany). The participants were instructed to exhale and inhale during a two second period of open solenoid valve (the green semaphore light was switched on) and to hold their breath, using almost all lung capacity (holding breath near total lung capacity), for four seconds when the solenoid valve was closed (the red semaphore light was switched on). Prior to testing and training, the participants were familiarized with cycling on the cycle ergometer and breathing in time with the metronome.

Incremental exercise test: The participants initially performed an incremental exercise test with SB to obtain VO₂ peak and PPO_SB. The test began at an intensity of 30 W and increased by 30 W every two minutes until volitional exhaustion. VO₂ peak was defined as the highest oxygen uptake averaged over a 60 second interval. On the next testing day, the participants performed an incremental exercise test with RBF to obtain PPO_RBF. With the exception of breathing pattern, the exercise protocols of the two tests were identical. In both incremental exercise tests peak power output (thus PPO_SB and PPO_RBF) was defined as the highest work stage completed (the last work stage that was actually sustained for two minutes), and hence power output obtained.

Constant load test with SB to determine ventilatory sensitivity: Ventilatory sensitivity was calculated from data obtained during 30 min cycle ergometery exercise at 50 W with a pedal cadence of 60 rpm. Participants started this test by breathing room air (initial 15 minutes) and then switched to breathing a humidified gas mixture containing 3% carbon dioxide, 21% oxygen and 76% nitrogen for a further 15 minutes (Kelley et al., 1982). The breathing mixture was directed to the inspiratory side of a Hans Rudolph respiratory valve (Hans Rudolph, Kansas City, Mo.). Ventilatory sensitivity was calculated using the following equation (Kelley et al., 1982):

Constant load test: Following a 5 minute warm-up at 50 W participants completed a constant load cycle test with a pedal cadence of 60 rpm to volitional exhaustion on two separate occasions. On one occasion the peak power output achieved during the PPO_SB test was adopted along with an SB breathing pattern. On another occasion the peak power output achieved during the PPO_RBF test was adopted along with an RBF breathing pattern. Time to complete each test (Tmax) was used to define the interval duration in the respective training programmes.

Measurements during cycle ergometry tests: Participants breathed through a mouthpiece attached to a pneumotachograph during each cycle ergometry test. Expired air was sampled continuously by a metabolic cart(V-MAX29, SensorMedics Corporation, Yorba Linda, USA) for breath-by-breath determination of V_E, V_T, P_ETCO₂, and the end-tidal partial pressure of oxygen (P_ETO₂). The pneumotachograph and O₂ and CO₂ analysers were calibrated with a standard three-litre syringe and precision reference gases, respectively. For further statistical analysis breath-by-breath data were averaged for each 10 second interval. During the incremental exercise tests oxygen saturation (SpO₂) was measured using a TruStatTM Pulse Oximeter (Datex-Ohmeda, Madison, USA). The pulse oximeter is an indirect oximetry measuring instrument, which displays SaO₂ every 4 s. The ear probe was attached to the earlobe after cleaning the area with alcohol.

Pulmonary function: A pneumotachograph spirometer (Vicatest P2a, Mijnhardt, Netherlands) was used to measure resting flow-volume profiles i.e. vital capacity (VC) and forced expiratory volume in one second (FEV_1.0). Pulmonary function measurements were made according to European Respiratory Society recommendations (Miller et al., 2005).

Participants were randomly assigned to either the experimental group (Group E; n = 6) or the control group (Group C; n = 6). Both groups undertook three cycle ergometry interval training sessions per week for six weeks under supervision in the laboratory.

Group E completed two to eight RBF training intervals per session. Interval durations were derived using manipulations of each participant’s Tmax value (between 60% and 75%) obtained during the constant load test with RBF (Laursen and Jenkins, 2002; Laursen et al., 2002). Interval intensities, however, were based on each participant’s PPO_RBF. Group C completed similar interval sets but did so with SB. Similar to Group E, interval durations ranged between 60% and 75% of each participant’s Tmax value which was obtained during the constant load test with SB. Once again, interval intensities were based on each participant’s PPO_SB. A work-to-rest ratio of 2:1 or 1:1 was adopted; however, the resting periods between the intervals did not exceed five minutes. The number of intervals completed during each training session was set according to each participant’s ability to complete the prescribed training programme along with consideration of the duration of each interval and the resting period. Each training session lasted approximately 60 minutes. At the fifth (first mid testing), tenth (second mid testing) and fifteenth (third mid testing) training sessions, participants performed a further constant load test with the same breathing pattern as they used during training. The interval and rest durations and the number of intervals for subsequent training sessions were then adjusted to reflect each participant’s improved Tmax. The training programmes are summarized in Table 2. The study design is summarized in Figure 1.

The results are presented as means and standard deviations. Intra-group differences between the pre- and post-training values were calculated with a paired, two-tailed t-test. Analysis of covariance (ANCOVA), with pre-training values as covariates and post-training values as dependent variables, was used to test for differences between the groups resulting from different training interventions. Statistical significance was accepted at the p ≤0.05 level. Effect sizes were calculated using Cohen’s d statistics to assess the magnitude of the treatment with 0.2 deemed small, 0.5 medium and 0.8 large (Cohen, 1988). All statistical parameters were calculated using the statistics package SPSS (version 15.0, SPSS Inc., Chicago, USA) and the graphical statistics package Sigma Plot (version 9.0, Jandel, Tübingen, Germany).

All participants completed the prescribed training programme. The parameters which determined the intensity and duration of the training interval sets are shown in Table 3. According to the experimental protocol, the intensity of the interval sets was defined with the resistance of constant load tests with different breathing patterns (pre-training PPO_RBF for Group E and pre-training PPO_SB for Group C) and was unchanged throughout the training. By contrast, the interval and rest durations and the number of intervals were adjusted after each mid testing according to each participant’s new Tmax value.

As shown in Table 4, VC, ventilatory sensitivity and VO₂ peak differed between groups in response to training (p < 0.05). A higher VC (p = 0.02, d = 0.36) and lower ventilatory sensitivity (p = 0.03, d = 0.81) were observed after training in Group E compared with pre-training values. By contrast, these parameters were unchanged in Group C.

PPO_RBF was enhanced with training in both groups (p < 0.01, d=2.81 in Group E and p = 0.03, d = 0.66 in Group C; Figure 2(a)). Importantly, this enhancement was greater (for 275%) in Group E than in Group C (p < 0.01). PPO_SB was increased with the training in Group C only (p < 0.01, d=2.02; Figure 2(b)).

Respiratory parameters and SpO₂ during the incremental test with RBF did not change (neither at submaximal intensities nor at PPO_RBF) throughout the training in Group C. On the other hand, there were significant training changes in these parameters in Group E. As shown in Figure 3 and 4, V_E (p = 0.01, d = 2.4), V_T (p = 0.01, d = 2.4), VO₂ (p = 0.01, d = 3.2), VCO₂ (p = 0.01, d=3.2) and P_ETco₂ (p = 0.03, d = 0.86) obtained at PPO_RBF was higher post-training in Group E. In addition, there were significant differences in V_E (p = 0.05), V_T (p = 0.05) and VO₂ (p = 0.04) obtained at post-training PPO_RBF between Group E and Group C. The lowest PPO_RBF obtained in pre- and post-training conditions was 150 W. Therefore, the average data of respiratory parameters and SpO₂ measured to this work stage and at PPO_RBF during the incremental test with RBF for Group E only are presented in Figure 3 and 4.

Respiratory parameters and SpO₂ during the incremental test with SB did not change (neither at submaximal intensities nor at PPO_SB) throughout the training in both groups. As shown in Table 4, only VO₂ peak (obtained at the incremental test with SB) increased in Group C following training (p = 0.01, d = 1.05).

The present study demonstrated that training with RBF during cycle ergometry exercise not only increased V_T in the face of RBF but also decreased ventilatory sensitivity in the presence of hypercapnia. Training with RBF also induced significantly larger improvement in PPO_RBF in Group E (42 ± 11%) compared with Group C (11 ± 9%; Figure 2(a)). Due to large (Group E) and medium (Group C) effect sizes in both tests, it could be suggested that the training intervention per se was primarily responsible for increased PPO_RBF. However, according to the training changes observed in spirometry parameters, ventilatory sensitivity, VO₂ peak and V_T at PPO_RBF, it does appear that training with different breathing patterns (i.e. spontaneous vs. RBF) induces specific adaptations. Given the necessity to coordinate breathing with stroke mechanics during front crawl swimming (Kapus et al., 2008; Town and VanNess, 1990), our findings are of particular relevance to this group of athletes.

When compared to spontaneously breathing exercise, cycle ergometry exercise with a similar reduction in breathing frequency (10 breaths per minute) is associated with a lower V_E by a magnitude of 25% (Sharp et al., 1991) to 49% (Kapus et al., 2009; Kapus et al., 2010b; Yamamoto et al., 1987). Even greater reductions (55%) have been observed in the swimming literature where breathing frequency is reduced from breathing every second stroke cycle to breathing every sixth (Town and VanNess, 1990) or every eight (West et al., 2005) stroke cycle. As a reduction in V_E is unlikely to be beneficial for exercise performance, it follows that an increase in V_E (via V_T) would be advantageous. Our results indicate that RBF training does increase V_T during incremental exercise with RBF i.e. training reduced breathing restriction. Specifically, V_T increased by 41 ± 19% following RBF and exercise training (Figure 3(b)). It is interesting that there were no significant differences between pre- and post-training V_T measured at submaximal work stages. It seemed that the obtained levels of V_T were sufficient to enable successful regulation of blood gases during submaximal exercise with RBF. In addition, these results indicated that the ventilatory level reached with RBF was a limiting factor only for maximal performance. Due to the prescribed and unchanged breathing frequency in the present study, an increase in V_T was the only mechanism available for increasing V_E (Figure 3(a)). This is consistent with reports that V_T increases in proportion to the magnitude of breathing frequency restriction (Town and VanNess, 1990) as well as exercise intensity (Sharp et al., 1991).

In addition to the observed increase in V_T following RBF and exercise training, VC was also increased in response to RBF and exercise training by 8 ± 8% (Table 4). This could be due to an increase in inspiratory muscle force production (Clanton et al., 1987) as an anther possible effect of breathing with higher V_T during training with RBF. Indeed, Jakovljevic and McConnell, 2009 found that inspiratory muscle fatigue was greater during front crawl swimming when breathing frequency was reduced from its already restricted pattern of breathing every second stroke cycle to breathing every fourth stroke cycle. They suggested that V_T would increase to compensate and consequently inspiratory muscle work would rise for two reasons. Firstly, the elastic load of breathing increases as lung volume increases, and secondly, a functional weakening of inspiratory muscles occurs due to the relationship between lung volume, compliance and the length-tension characteristics (Rahn et al., 1946). Considering our observations and those of Jakovljevic and McConnell, 2009, it is possible that RBF training could have an effect on the strength of inspiratory muscles also. However, this suggestion requires further study.

The hypoventilation which occurs during RBF exercise is associated with a reduction in CO₂ elimination. This increases the alveolar partial pressure of CO₂ (Dicker et al., 1980; Holmér and Gullstrand, 1980), the fractional expired CO₂ (West et al., 2005), P_ETco₂ (Kapus et al., 2010a; 2010b), and the capillary blood partial pressure of CO₂ (Kapus et al., 2009; 2010a; 2010b; Sharp et al., 1991; Yamamoto et al., 1987). In light of such findings it is not surprising that RBF training induces hypercapnic adaptations such that a higher level of CO₂ can be tolerated i.e. a “hypercapnic training” effect (Dicker et al., 1980; Holmér and Gullstrand, 1980; Town and VanNess, 1990). Data from the current study supports this view. For example, participants in Group E showed an attenuated ventilatory sensitivity by 45 ± 27% (Table 4) during the hypercapnic exercise test, a response which was not observed in Group C. However, it should be emphasised that the hypercapnic stimulus was reducing throughout the training period. Considering that breathing restriction during RBF exercise was reducing due to training, it could be assumed that hypercapnic stimulus was higher in the first few training sessions and lower at end of the period of RBF training. Such an adaptation should reduce the urge to breath (Ferretti and Costa, 2003), which may partly explain why swimmers have been found to take fewer breaths during front crawl swimming following RBF training (Kapus et al., 2005). Increased tolerance to CO₂ following RBF training is not restricted to swimmers, however. In activities where breath-holding occurs and thus so too does frequent exposure to hypercapnia, one would expect a blunted sensitivity to CO₂. This is indeed the case in breath-hold divers (Delapille et al., 2001; Florio et al., 1979; Masuda et al., 1982) and under water hockey players (Davis et al., 1987; Lemaitre et al., 2007) as well as swimmers (Ohkuwa et al., 1980).

Due to the additional stress caused by RBF, the workload during the interval training bouts in the current study was not identical between breathing conditions. Specifically, the workloads experienced by Group E were approximately 40% lower than in Group C (Table 3). These conditions might lead to different training effects during incremental exercise with SB (Figure 2(b)) and on VO₂ peak (Table 4). According to the findings of Laursen, Shing, Peake, Coombes and Jenkins (2002) the interval training workloads adopted in the present study were high enough to increase VO₂ peak during the SB condition i.e. Group C (Table 4) but were unlikely to in the RBF condition i.e. Group E. Furthermore, the increased VO₂ peak in Group C might explain their improved PPO_RBF post-training (Figure 2(a)), even though V_T was unchanged during exercise with RBF and when breathing a 3% CO₂ mixture (Table 4).

Possible study limitations: The present study has limitations. Firstly, because the participants were not blinded to the breathing intervention during training, one may argue that this might have influenced our results. However, participants were naäve and were not informed about the potential effect of training with different breathing patterns (i.e. spontaneous vs. RBF) on exercise performance. Furthermore, both trainings induced different, very specific adaptations. Therefore, the potential bias was minimised. Secondly, in previous studies CO₂ sensitivity has been determined using the rebreathing method, which was first employed by Read, 1967, and the steady-state CO₂ inhalation method (Kelley et al., 1982). The latter method provides two or three increments of inhaled CO₂ and the composition of inhaled gas is adjusted by mixing compressed air with pure CO₂ by using a precision gas bender. In the present study the ventilatory response to CO₂ during exercise was determined by using only one predetermined gas mixture, which was previously mixed and analysed and provided from a high-pressure cylinder. This could be the reason for the higher absolute ventilatory sensitivity in the present study compared to data of previous studies using the steady-state CO₂ inhalation method. The average ventilatory sensitivity values in the present study pre- and post-training were 4.7 l·min^-1·mmHg^-1 and 3.1 l·min^-1·mmHg^-1, respectively. These values are higher than the 3.25 l·min^-1·mmHg^-1 and 2.16 l·min^-1·mmHg^-1 obtained in divers and non-divers, respectively (Florio et al., 1979) but fall within the upper limit of those reported by McConnell and Semple, 1996. In addition, due to practical constraints, it was not possible to blind the participants to the start of the ventilatory sensitivity test. However, to minimise this influence, participants inhaled the breathing mixture for 15 minutes until V_E and P_ETco₂ were constant (Kelley et al., 1982). Only the data measured during the last three minutes were further analysed. Indeed, the data of both parameters were stable without additional noise during this period. Thirdly, it was not possible to standardize the training interventions between groups because of the demands associated with RBF. However, the design of the training programmes was based upon the established relationships of peak power output and Tmax on endurance performance (Laursen et al., 2002). Furthermore, both groups experienced similar and adequate levels of familiarisation. This was defined as cycling on the cycle ergometer and breathing in time with the metronome without problem for 15 minutes at a load of 100 W on at least two separate occasions. Consequently, it is unlikely that a learning effect accounted for the differences observed between groups. Indeed, substantial improvements in Tmax became evident after the ninth training session in Group E.

In conclusion, the effect of training with RBF during cycle ergometry exercise was twofold. It increased V_T during incremental exercise where breathing frequency was restricted. In addition, RBF training decreased the ventilatory response to exercise when breathing a 3% CO₂ mixture, which suggests an increased tolerance to CO₂ following such training. These data are relevant to those athletes whose training or competitions involve hypoventilation such as swimmers. It could be suggested that swim training with RBF might permit swimmers to take fewer breaths and to hold their breath for longer, which would be of particular benefit during the underwater phases (flip turns, gliding, and underwater strokes) as well as providing an important biomechanical advantage (Pedersen and Kjendlie, 2003). However, training with RBF could not be realized during high velocity swimming due to the additional stress caused by such a breathing pattern. Therefore, it could be speculated that the combination of the two forms of intense front crawl training under different breathing conditions (swimming with RBF at a lower velocity, and swimming with the usual breathing frequency at higher velocities) would be beneficial for competitive swimmers.