Active recovery affects the recovery of the corticospinal system but not of muscle contractile properties

Purpose Active recovery is often used by athletes after strenuous exercise or competition but its underlying mechanisms are not well understood. We hypothesized that active recovery speeds-up recovery processes within the muscle and the central nervous system (CNS). Methods We assessed muscular and CNS recovery by measuring the voluntary activation (VA) in the vastus lateralis muscle with transcranial magnetic stimulation (VATMS) and peripheral nerve stimulation (VAPNS) during maximal voluntary contractions (MVC) of the knee extensors in 11 subjects. Measurements were performed before and after a fatiguing cycling time-trial, after an active and a passive recovery treatment and after another fatiguing task (1 min MVC). The measurements were performed a second time 24 h after the time-trial. Results We observed a time × group interaction effect for VATMS (p = 0.013). Post-hoc corrected T-tests demonstrated an increased VATMS after active recovery when measured after the 1 min MVC performed 24 h after the time-trial (mean ± SD; 95.2 ± 4.1% vs. 89.2 ± 6.6%, p = 0.026). No significant effects were observed for all other variables. Conclusions Active recovery increased aspects of central, rather than muscle recovery. However, no effect on MVC was seen, implying that even if active recovery speeds up CNS recovery, without affecting the recovery of muscle contractile properties, this doesn´t translate into increases in overall performance.

Introduction A faster recovery after competition or an exhaustive training session allows an individual to again perform sooner at his or her best. This is not only important in enabling a better overall performance in competitions that can last several days but also in limiting susceptibility to injury and overtraining [1][2][3][4]. However, the underlying mechanisms of performance recovery and recovery treatments are still not fully understood, which makes it difficult to tailor training programs for the purposes of speeding up recovery processes [4].
Active recovery is a popular recovery treatment used by athletes [5], and is defined as physical activity performed at low intensity following a fatiguing exercise. It has been proven to be effective in increasing subsequent performance compared to passive recovery [6][7][8][9][10]. It has been suggested that the higher performance after active, compared to passive recovery, is based on a faster recovery of muscle performance, as active recovery can affect muscle lactate metabolism and muscle oxygenation [7,8,11].
Beside the processes taking place within the muscle, it has been demonstrated that the central nervous system (CNS) plays an important role in the regulation of performance during exercise [12,13]. The neuromuscular function following locomotor fatiguing exercise can remain impaired, from minutes up to days, depending on the duration of the task. The observed reduction in maximal voluntary force can be explained by a persisting presence of peripheral and central fatigue [14]. Interestingly, it appears that recovery over 48 h of impaired performance following intermittent sprint-exercise correlates better with the recovery of impaired neural drive to the muscles than with muscle damage markers [15]. Therefore, it could be suggested that the CNS plays an important role in the process of performance recovery [14][15][16]. Considering this, we hypothesized that active recovery may accelerate not only the recovery of the muscle itself, but also of the CNS, and thereby increase neuromuscular performance during the recovery period after an exhausting exercise.
Indeed, active recovery could possibly affect the recovery of CNS structures by several means. For example, during a fatiguing task, feedback from muscle afferents may reduce the neural drive and thus, reduce muscle activation [17]. Active recovery could speed up the clearance of the intramuscular metabolic perturbations induced by exercise [11], and thus limit the action of the afferent inhibitory feedback system on the neural drive.
Although the effect of active recovery has been studied mostly with short intermittent high intensity efforts so far [6][7][8], we selected a maximal continuous cycling exercise (time trial around 1 h) in the present study to induce a higher proportion of central fatigue [14,18]. Following an active or passive recovery treatment, subjects performed a one minute long MVC. This maximal task was performed in order to observe the effect of the recovery treatment on performance. To observe the effect of active recovery on muscle and CNS recovery, we measured the potentiated twitch at rest (Ptw) and the voluntary activation (VA) of the vastus lateralis muscle. It has been shown that after a strenuous locomotor exercise, neuromuscular recovery can take several days to fully occur [14,15] and that recovery of muscles and CNS may not follow the same trend over time [15]. Therefore, we reiterated our measurements 24 h after the end of the time-trial.
We hypothesized in the present study that active recovery, compared with passive recovery, improves parameters of muscle and CNS recovery as well as performance during a 1 min MVC test.

Participants
Experiments were approved by the ethics committee of the University of Konstanz and conducted in accordance with the declaration of Helsinki. 12 healthy male subjects were recruited and participated in the present study after signing the written informed consent. 1 subject suffered from a leg injury which was not related to the experiments and dropped out of the study. Therefore 11 subjects remained in the study (mean ± SD; age = 27.7 ± 4.3 years; height = 182.6 ± 3.9 cm; weight = 78.9 ± 5.9 kg). All subjects confirmed to be free of a lower limb injury at the time of the study and in the previous year and have no contra-indication for TMS. The subjects were asked to avoid caffeine and alcohol from the evening before until completion of testing.

General experimental design
The subjects were asked to participate in 4 experimental sessions, separated by 1 week of rest. The experimental sessions for one subject were always performed the same day of the week and at the same time of the day (see Fig 1). During the first session, subjects performed a step test to assess the power output at their individual anaerobic threshold. After 20 min of rest, subjects were familiarised with the neuromuscular measurements and the MVC test. In the second session, subjects had to complete the time-trial with the goal of minimizing the overall ride time. For this familiarization time-trial, subjects were instructed to cycle with a constant power output, equivalent to or just below their anaerobic threshold power. A tactical debriefing took place following the end of the time-trial to advise subjects how to increase their timetrial performance for the next session. During the 3 rd and 4 th session, neuromuscular parameters were assessed at baseline prior to the time-trial. After the initial measurements, subjects then performed the time-trial. Upon time-trial completion, neuromuscular parameters were re-assessed (starting 2 min after the end of the time-trial). After post-time-trial measurements had been taken, subjects performed either an active or a passive recovery treatment in a randomised and counter-balanced order. Following recovery treatment, subjects went through the post-recovery neuromuscular measurements (starting 3 min after the end of the recovery treatment). Twenty-four hours after the 3rd and 4th session, subjects completed the post 24 h neuromuscular measurements.

Cycling simulation
We used a custom-built cycling video simulation system, which consisted of a cycle ergometer driven by custom-coded cycling simulation software, that was faced toward a projection screen [19]. The cycle ergometer was a custom-made bicycle frame (made by the workshop of the university) coupled with a Cyclus2 brake system (RBM elektronik-automation GmbH, Leipzig, Germany). The projection on the screen consisted of a cyclist filmed during a ride on a specific road and any relevant information for the experiment (power output, map and gradient of the path, distance and time, current position of the cyclist on the map). The software controlled the speed of the video and the Cyclus2 braking system, forcing the subject to provide the power output necessary to cycle the virtual path. Moreover, the speed of the video was dependant on the subject's current power output, giving the impression of cycling on the real path.
The software displayed an avatar on the road, just in front of the rider. This avatar was programmed to correspond with the performance of a previous time-trialof the same subject. The rider was able to see the current distance between himself and his avatar online throughout the time-trial. During the 3 rd experimental session, the avatar was paced with the speed that the subjects achieved during the familiarization session, in order to encourage them to beat their previous time. During the 4 th session, the avatar was paced with the speed that the subjects achieved during the 3 rd session. To assure similarity between the 3 rd and the 4 th session, subjects were instructed during the 4 th session to stay within −5/+5 m of the avatar from the 3 rd session. To circumvent possible effects from the following of an avatar on fatigue, the recovery treatment after the 3rd or 4th time-trial session counter-balanced among subjects.

Step-test
The step-test was performed in power increments of 20 W every 3 min, starting at 100 W, until volitional exhaustion (mean ± SD; maximal power: 311 ± 62 W). We have used the official step-test protocol of the German Cycling Federation (BDR) documented in their youth program from 2012. Just before the end of each increment, blood was sampled at the ear lobe, in order to obtain a measure of blood lactate concentration (Lactate Pro 2, Arkray Factory TT is the acronym for time-trial and NM the acronym for neuromuscular. The overall protocol consisted in 4 experimental sessions separated by 1 week of rest. In session 1, subjects performed a step-test and then were familiarized with NM tests. In session 2, subjects were familiarized with the time-trial. In session 3 and 4, the time-trial was followed by either the active or passive recovery treatment (random order). NM tests were performed before the time-trial, 2 min after the end of the time-trial, following the recovery treatment, after the 1 min MVC and again 24 h hours after the time-trial, before and after the 1 min MVC.
https://doi.org/10.1371/journal.pone.0197339.g001 inc.). Anaerobic threshold power was estimated as the power at which blood lactate concentration rose significantly (visual discrimination from the lactate blood concentration/power curve). We have used the method and the terminology proposed by Skinner & McLellan [20]. Maximal power reached during the step-test (maximal step-test power) corresponded to the maximal power reached by the subject at exhaustion. However, if the subject stopped at the beginning of a step, i.e. in the first min, the maximal step-test power corresponded to the power of the previous step. Pure water was supplied at the convenience of the subject throughout the whole experimental session.

Time trial
A 9 min warm up that corresponded to the 3 first steps of the step-test protocol, preceded all time-trials. The time-trial was a simulation of the Alpine pass Flüela (western route, Switzerland). The time-trial distance was 12,397 m, with a rise of 923 m and an average gradient of 8.1% (see Fig 2A). To assess whether the time-trial induced a similar level of exertion during the 3rd and 4th experimental sessions, we measured blood lactate concentration at kilometre 2, 4, 9 and during the last 200 m of the session to estimate the level of exertion [21] (see Fig  2C). Heart rate was measured via an ANT+ heart rate belt (HRM1B, Xplova, Taipei City, Taiwan) and a Laptop with ANT+ Dongle. The Golden Cheetah Analysis Software (Version 3.3) was used to sample heart rate data in intervals of 1 Hz. Participants rated their perceived exertion (RPE; Borg 10 scale) every kilometre and at the end of the time-trial (see Fig 2B).

Recovery treatment
Recovery treatments were performed immediately following the completion of the post timetrial neuromuscular measurements. The active recovery treatment consisted of cycling at 100 W at preferred pedalling cadence for a duration of 15 min. During passive recovery participants were asked to sit in a chair for the same time. Blood lactate concentration was measured at the end of both the active and the passive recovery treatment.

Neuromuscular measurements
General procedure. Before the pre and post 24 h measurements, we measured blood lactate concentration at rest. Following the blood drawings, the right leg of each subject was prepared for the application of surface EMG sensors (Trigno wireless EMG system, Delsys Inc.) We placed the EMG electrodes according to SENIAM recommendations. The respective areas on the muscle bellies of the vastus lateralis (VL), rectus femoris, vastus medialis and biceps femoris (BF) were shaved, abraded with sand paper and cleaned with alcohol. EMG sensors were then applied to and taped to assure firm contact. EMG signals were amplified (× 1000), high-pass and low-pass filtered (20 ± 5 Hz and 450 ± 50 Hz respectively), sampled (4000 Hz) and registered on a computer through a Power 1401 interface (Cambridge Electronic Design) with the Signal software (Cambridge Electronic Design).
A mark was drawn with a waterproof pen on the right ankle, 2 cm higher than the line passing below the lateral malleolus, to ensure identical placement of the force transducer and knee angle throughout the entire experiment, as well as reapplication occurring 24 h later and the next week.
The subject was then seated in a custom-made chair that had been manufactured to measure isometric force during knee extensions. Participants were strapped at the chest and hip level, with the right ankle tightly fixed to a force transducer (Model 9321A, Kistler, Wintherthur, Switzerland) holding the knee at around 90˚of flexion in a non-compliant way.
Force signals were sampled and registered with the same interface and software as EMG signals.
To stimulate the femoral nerve (Stimulator DSH7, Digitimer), a cathode (copper, circular, 2 cm diameter, wrapped in a water soaked sponge) was taped in the femoral triangle and an anode (copper, 7 × 5 cm, wrapped in a soaked sponge) was taped over the gluteus maximus. Positions of both stimulation electrodes were marked on the skin with a waterproof pen to ensure similar and fast electrode repositioning after the time-trial and the recovery treatments. To ensure that the cathode was positioned as close as possible to the femoral nerve throughout the entirety of the experiment, an investigator held a small sandbag over its position and applied pressure. The pulse duration of the electrical stimulation was set to 1 ms and the intensity of the stimulation was gradually increased until reaching Mmax in the VL at rest and the maximum un-potentiated twitch at rest of the knee extensors, without any visual detectable motor response in the BF EMG. The intensity required for reaching Mmax was multiplied by 1.5 and remained constant during the entire experiment (final intensity: 73 ± 13 mA).
The subject completed a warm up of around 15 contractions, with gradual increases in intensity and rest periods, until reaching an intensity equal to around 90% of the perceived maximal voluntary contraction (MVC). After a rest of 2 min, the subject performed an initial MVC trial, followed by a second trial occurring another 2 min later. In the second trial the force did not increase more than 5% in none of the subjects, therefore a third trial was not performed. All MVCs were completed with a force plateau of 2-3 s. The investigator encouraged and motivated the subject verbally before and during all maximal MVC trials.
After the initial MVCs, we searched for the best position of the transcranial magnetic stimulation (TMS) coil to induce a motor evoked potential (MEP) in the VL during a 10% MVC contraction. The coil (MC-B70, MagVenture) and the stimulator (MagPro R30, MagVenture) was set to produce biphasic single pulse (butterfly-coil oriented in the posterior anterior axis of the head). The optimal position of the coil was drawn directly on the scalp to ensure identical coil positioning throughout the experiment. The stimulation intensity was set to evoke the largest possible MEP (~10-50% Mmax amplitude) in the VL, with the lowest response in the BF (visual control of BF EMG) at 10% MVC (90-100% maximal stimulation output).
During the experiment, we measured voluntary activation via 2 different methods, TMS (VA TMS ) and peripheral nerve stimulation (VA PNS ). VA TMS and VA PNS were always measured during separated muscle contractions (see Fig 3). We measured VA PNS following the interpolated twitch method protocol [22], i.e. by stimulating the femoral nerve during the force plateau of MVC and at rest 2 s after the end of the contraction and VA TMS following the guidelines of Sidhu and colleagues [23] for the quadriceps, i.e. TMS at the plateau of MVC, of a 50% MVC and 75% MVC contraction (see Fig 4), with 3 muscle contractions being separated by 10 s of rest. During the VA TMS protocol, 2 s after the first MVC, we stimulated the femoral nerve with peripheral nerve stimulation to obtain a potentiated twitch at rest (Ptw) in order to estimate peripheral fatigue as soon as possible in the post cycling and post recovery neuromuscular measurements since VA TMS was measured before VA PNS (Fig 3). Contrary to Sidhu and colleagues [23] we interspaced VA TMS and VA PNS measurements by one minute in order to ensure that the MVC required for the VA PNS procedure was not affected by potential fatigue induced by the earlier VA TMS procedure. During all procedures involving a muscular contraction, we provided visual feedback of all the target forces on a computer screen that was placed in front of the subjects. Neuromuscular measurements. Pre time-trial measurements: we measured VA TMS first. Then, 1 min after the last contraction of the VA TMS procedure, we measured VA PNS . After 2 min of rest, this sequence was repeated (Fig 3A). Post time-trial measurements: 2 min after the end of the time-trial, we measured VA TMS , and 1 min later we measured VA PNS (Fig 3B). Post recovery treatment measurements: we measured VA TMS , and 1 min later we measured VA PNS . One minute of rest later, the subject had to perform a 1 min MVC, with the following instructions: "Go full gas, reach MVC and maintain it for 1 min". Five seconds after the end of the 1 min MVC, we measured VA PNS and 5 s later, VA TMS (Fig 3C). This sequence of measurements after the 1 min MVC seems to be reasonable, as central fatigue may recover faster after a 1 min MVC than after a 1 h time-trial. Post 24 h measurements: we measured VA TMS and VA PNS twice, as in the pre-measurements. One min later, the subject had to perform a 1 min MVC and we measured 5 s later VA PNS and 5 s later VA TMS (Fig 3D).

Analysis
MVC was determined as the maximal value of the force measured during the maximal isometric leg extensions. Ptw values corresponded to the maximal amplitude of the potentiated twitch at rest.
For the pre and the beginning of the post 24 h neuromuscular measurements, MVC, VA TMS , VA PNS and Ptw corresponded to the highest values measured during both trials. For the post time-trial and post recovery measurements, MVC and Ptw were determined only during the first MVC trial, which we obtained during the VA TMS protocol. After the 1 min MVC (in the post recovery and post 24 h neuromuscular measurements), we averaged the MVCs and Ptws obtained during VA PNS and VA TMS procedures. It must be noted that one subject could not perform the VA TMS measurement during one session, due to stomach issues just after the end of the time-trial. However, he was able to perform the following VA PNS procedure. Therefore, when considering the effect of training over time, the VA TMS , MVC and Ptw measures were analysed only for the remaining 10 subjects. To quantify the effort produced during the 1 min MVC, we measured the mean force and the peak force (i.e. maximum amplitude) produced during the 1 min MVC.
We measured the MEP amplitude in the VL (normalized to Mmax) and in the BF (nonnormalized) during the VA TMS procedure. We also indicated the value of the estimated resting twitch, which is calculated as the y-intercept of the linear relation between the voluntary activation force amplitude and the interpolated twitch amplitude [24]. Additionally, we measured the correlation coefficient (Pearson's r) of this linear regression to give the reader an estimate of the success of the VA TMS procedure [25].
Mean power and mean standard deviation corresponded to the mean of the power output and the standard deviation of the power respectively during the time-trial followed by the active or the passive recovery treatment. Mean heart rate corresponded to the averaged heart rate measured during the entire time-trial.

Statistics
Shapiro-Wilk normality tests were performed on each data set. Two-way, within-subject ANOVAs were completed on lactate, RPE, MVC, Ptw, VA PNS , VA TMS values. Mauchly's tests of sphericity were performed, and when significant, a Greehouse-Geisser correction was used. When relevant, post hoc tests were completed using a pairwise comparison with paired T-Tests and a Holm-Bonferroni correction. Post-hoc tests for time effects were conducted against each time point, with data from both treatments pooled. To better understand group × time effect, t-tests were conducted between groups at each time point.
Paired T-tests were used to compare mean power, mean standard deviation, time, mean heart-rate during the time-trial, mean and peak force during the 1 min MVC, with the active or passive recovery treatment. We calculated Cohen's d for paired t-test (d = mean of paired difference /sd of paired difference).

Neuromuscular measurements
As displayed in Fig 5,  However, for VA TMS , we observed a time (p < 0.001), group (p = 0.022) and group × time effect (p = 0.013). Post hoc tests with Holm-Bonferroni correction revealed that the interaction effect could partly be explained by a higher VA TMS with the active recovery treatment measured after the 1 min MVC performed 24 h after the time-trial (95.2 ± 4.1% vs. 89.2 ± 6.6%, t 9 = 3.8, p = 0.026, Cohen's d = 1.2, see Fig 5). The neuromuscular values measured 24 h after the time-trial (post 24 h) were not significantly different from the pre values measured before the time-trial, with the exception of the Ptw, which was significantly lower (Fig 5).

Control variables for the VA TMS measurements
The estimated resting twitch (in N), the estimated resting twitch correlation coefficient, the MEP obtained in the VL at 100, 75 and 50% MVC (normalized to Mmax), and the MEP in the BF at 100, 75 and 50% MVC (in mV) are displayed in Table 1. It must be noted that a treatment effect was observed for the estimated resting twitch (p = 0.042), explained by slightly lower values for the passive recovery treatment (post-hoc T-tests were not significant). Active recovery modulates corticospinal recovery
For the RPE and blood lactate concentration (see Fig 2B and 2C respectively), there was a time effect (p < 0.001 for both), but no group effect (p = 0.39 and 0.25 respectively) or time × group interaction (p = 0.54 and 0.10).

Discussion
In the present study, we observed that active recovery, compared to passive recovery, prevented the reduction of VA TMS after the 1 min MVC performed 24 h after the time-trial. However, active recovery had no more effect than passive recovery on the recovery of Ptw and on the performance during MVCs.

Active recovery affected central but not peripheral recovery
Active recovery, when compared to passive recovery, had a large effect on VA TMS but no effect on VA PNS and Ptw. The relevance of active recovery on VA TMS is strengthened by the fact that its effect size is comparable to the one observed for the decreased VA TMS after the time-trial compared to VA TMS before the time-trial in the active recovery session (VA TMS of 97.5 ± 1.4% pre time-trial vs. 92.7 ± 5% post time-trial, d = 1.13). We suggest that, in our experimental context, active recovery increased the recovery of corticospinal voluntary activation but not the recovery of the whole CNS voluntary activation or of muscle contractile properties. This result is strengthened by our control measurements: i) The time-trial induced the same amount of fatigue in the 2 experimental sessions, and therefore cannot explain the observed differences in VA TMS . ii) The control variables of the VA TMS measurements showed no variation across time or between treatment that was similar to the one observed for VA TMS and therefore cannot explain the observed difference of VA TMS [25]. Thus, we conclude that the interaction effect observed for VA TMS was most probably the result of a difference in corticospinal voluntary activation of muscles indicating that active recovery possibly accelerated the recovery of corticospinal structures when compared to passive recovery.

Active recovery did not increase performance
Despite the effect of active recovery on VA TMS , we could not observe an effect on maximal performance as measured during MVCs and the 1 min MVC. This observation appears to be contradictory with the abundant evidences that the voluntary activation level is an important contributor to force production following exercise [14]. How to explain the lack of effect on maximal performance? It has to be mentioned that, in the present study, we detected a difference of VA TMS only during measurements while performance was heavily impaired (i.e. after the 1 min MVC), thus in situations where we also observed a large amount of peripheral fatigue. We propose that, due to the orderly motor units recruitment [26], the difference of VA TMS induced by the recovery treatments mostly affected the recruitment and/or firing rate of the high threshold motoneurons of the pool. During the 1 min MVC, these motor units were presumably fatigued in the first few seconds of the task [27], and after that, were not contributing anymore to force production. Then, their activation with active recovery, or their lack of activation with passive recovery, would not change the force output. Such effect can explain why a larger corticospinal voluntary activation could not induce a larger mean force in the 1 min MVC or maximal force in the following MVC. In a previous study, after a fatiguing task, we also observed an attenuated reduction in VA TMS , probably induced by brain stimulation, but not an attenuated reduction in Ptw. Similarly to the present study, the attenuation in VA TMS reduction was not accompanied by an immediate effect on MVC [28]. Thus, in the present experimental context, the main limiting factors of performance following the timetrial seem to have more peripheral than central origin.
In contrast to the results of the present study, it has been discussed that the CNS could be an important player in overall performance recovery [14][15][16]. However, the studies mentioned present mainly indirect and sometimes contradictory results. For example, in several studies, the time course recovery of MVC follows more the recovery of neural than muscles markers, indicating that performance might be driven more by central than peripheral substrates [14,15]. On the other hand, Thomas and colleagues have shown that after a simulated football game, VA PNS and VA TMS recovered faster than Ptw and MVC, indicating that muscle function recovery primarily depended on peripheral aspects of fatigue [29]. The mechanisms regulating performance recovery seems to be highly specific to the previous fatiguing task [14]. In the present study, we were able to target specifically the corticospinal performance recovery. This particular situation offered us the possibility to assess with direct observation the role of the CNS in maximal force production following a fatiguing task. The lack of effect on performance from the larger corticospinal voluntary activation suggests that after a cycling time-trial or a 1 min MVC, the main limitation in overall performance recovery is more muscular than central.
Since we only made measurement during maximal efforts, we cannot conclude about the possible effects of a larger VA TMS on submaximal tasks. We propose that a submaximal task could benefit more from a larger VA TMS , since in this situation, the relative main contributor to muscle fatigue may be more central than peripheral [14].

Limitations
We found an effect of active recovery on VA TMS but not on VA PNS . We have concluded that active recovery affected more corticospinal structures than the whole CNS. However, such conclusions must be considered cautiously as they are not free from methodological pitfalls. Indeed, the difference of effect between the two VA procedures may come from the possible sigmoidal relationship between the excitation of the motoneuron pool and the size of the interpolated twitch during VA PNS [30] and the more linear relationship observed during VA TMS [23,24,31,32]. Additionally, the recovery treatment may have had a different effect on the nervous structures and synergist or antagonist muscles that are activated in different proportions during VA TMS and VA PNS [31]. Moreover, the different ways to produce interpolated twitches during VA PNS and VA TMS may also have been affected differently by recovery treatments, e.g. spinal reflexes that may have influenced the twitch size during VA PNS [30].
We have used an active recovery treatment with a fixed power (i.e. 100 W). A treatment tailored to the aerobic capacity of each subject could have provided less variation in the effect of active recovery on neuromuscular performance. However, the 100 W treatment was well under the anaerobic threshold for all the subjects, and therefore we do not expect a negative influence on the recovery processes following the treatment.
In general, due to the complexity of the present study with many variables measured over several time points and a rather low number of subjects, statistics can be biased. On the one hand statistical power could have been too low to detect existing differences, and on the other hand the high number of comparisons could have provided false positive results. Therefore, the effect of active recovery on corticospinal voluntary activation should be confirmed in a further study tailored to the specific needs of this research question.

Conclusion
The present study shows that an active recovery treatment performed after an exhausting endurance cycling task can have an effect on corticospinal structures without having an effect on muscle contractile properties or performance. This indicates that in our experimental paradigm (i.e. after a long duration cycling uphill time-trial), the main limiting factor of overall performance recovery takes place more at peripheral than central level. The results of the present study support the need of a better understanding of the underlying mechanisms of recovery treatments in order to tailor the recovery strategy specifically to the previous and upcoming effort.
Given that an active recovery treatment has the capacity to accelerate the recovery or to prevent the reduction of corticospinal voluntary activation after a fatiguing task, its application would be optimal in a situation where performance is mainly limited by central components.
In that context, Husmann et al. have shown that knee extensor MVC following a short duration rowing time-trial in elite rowers was mostly limited by an impaired central drive [33]. Therefore, we propose that active recovery could have a greater effect in a short duration rather than in a long duration exhausting exercise. We suggest that a rowing time-trial, as used by Husmann et al. or a similar exercise could provide an excellent experimental paradigm to further study the role of the CNS in overall performance recovery.