BRÜMMER, VERA; SCHNEIDER, STEFAN; ABEL, THOMAS; VOGT, TOBIAS; STRÜDER, HEIKO KLAUS
Besides the positive effects of exercise on the cardiovascular and musculoskeletal systems, there is growing interest in the effect of exercise on the CNS. Although there are several studies testing the effect of different exercise modes and intensities (for review, see Crabbe and Dishman (3)), it is still unclear which type or intensity of exercise leads to increases or decreases in electrocortical activity and in which brain regions.
Studies examining the influence of exercise on brain cortical activity are rare, which is largely because of the difficulties of utilizing many brain-imaging modalities during exercise. High costs and limitations in feasibility and in temporal resolution make it difficult to use some imaging methods, such as positron emission tomography or functional magnetic resonance imaging. As such, investigations of changes in brain cortical activity induced by exercise have predominantly been made by electroencephalographic (EEG) recordings. Although results are inconsistent (3), three main hypotheses regarding the effects of exercise on brain cortical activity have emerged: 1) exercise has an effect on (pre-) frontal brain areas, which is expressed in the transient hypofrontality hypothesis and the dual-mode theory; 2) the effect of exercise depends on the chosen exercise intensity, which is expressed in the dose-response relationship; and 3) it depends on the chosen exercise type according to the individual bias as stated in the exercise preference hypothesis.
The transient hypofrontality hypothesis by Dietrich (5) and the dual-mode theory by Ekkekakis (6), Ekkekakis and Acevedo (7), and Ekkekakis et al. (9) assume that because of limited brain resources, intensive exercise is accompanied by a redistribution of the brain activity. The result is a shift of activity away from brain regions that are not involved in the task toward regions that are involved in planning and executing motor commands (mainly, the motor and sensory cortexes). Therefore, regions not involved in the demanding task of exercise (such as the prefrontal cortex) show less activity. Similarly, it has been shown that emotional brain areas in the prefrontal cortex are involved in learning a novel task and that activity shifts more and more toward parietal motor and somatosensory regions with practice (14,18,30).
A dose-response relationship was proposed by Ekkekakis et al. (10) and Ekkekakis and Petruzzello (11), regarding the effect of different exercise intensities. They suggested that the psychological effect of exercise (which is thought to be mediated by changes in brain cortical activity (22,33,34,38)) is intensity dependent. Whereas low-intensity exercise is thought to be insufficient to have a positive influence, high intensive exercise in turn is thought to be counterproductive. The optimal exercise intensity for the greatest psychological effect is somewhere in between these two extremes. This relationship between exercise intensity and brain cortical activity seems to be dependent upon the trained status of an individual. As shown by Schneider et al. (34), higher but not moderate exercise intensities resulted in decreased β activity and reduced psychological strain in recreational athletes. These studies highlight the importance of regarding the effect of different exercise intensities on changes in brain cortical activity as an underlying neurophysiologic mechanism.
Recently, we were able to show that decreased cortical activation in frontal cortex areas seems to be connected to exercise preferences (35). The "exercise preference hypothesis" is built on the assumption that the relaxation effects of exercise are linked to an individual's physical activity history and exercise preferences (2,36), where the "preferred exercise" represents the mode of exercise a subject is most familiar with (35). In combination with the dual-mode/hypofrontality theory, one would assume that exercise in accordance with an individual's physical activity history and preference would be accompanied by frontal deactivation. However, there are only a handful of studies on the changes in brain cortical activity induced by exercise, and none of those has taken exercise preference into account. Therefore, it seems essential to evaluate the role of different exercise modes and intensities in regard to individual preferences.
The aim of our study was to use source localization methods to examine the electrocortical changes in brain regions induced by different exercise modes (treadmill, bicycle, arm crank, and isokinetic wrist flexion exercises) and intensities (50% and 80% of individual aerobic or strength capacity). In addition, we aimed to investigate whether these cortical changes are modulated by exercise preferences, which has not previously been investigated. For a complete picture of electrocortical changes, we analyzed the frontal, temporal, parietal, and occipital cortexes of the brain. The two intensities were chosen for two reasons. First, on the basis of the findings by Ekkekakis et al. and the dose-response relationship (6,7,9-11), low-intensity exercise was expected to be an insufficient stimulus for recreational athletes to induce any changes in mood mediated by changes in brain activity. Second, a prior study testing a comparable pool of subjects showed that 80% of individual aerobic capacity is similar to an intensity voluntarily chosen by recreational runners (34), and this intensity has been shown to be sufficient to induce changes in brain cortical activity and mood. Therefore, changes in brain cortical activity are predicted with high- and not with moderate-intensity exercise. Because Woo et al. (38) showed an inverted-U relationship by systematically investigating exercise duration, revealing no effect after 15 min, a clear positive effect after 30 min, and a tendency toward a negative effect after 45 min, we chose 30 min for exercise duration in our study for the endurance exercise modes. To minimize individual variances, which may have led to previous heterogeneous results (3), we examined both exercise modes and intensities within the same group of participants. To localize changes in brain cortical function, we chose the method of EEG in combination with standardized low-resolution brain electromagnetic tomography (sLORETA).
On the basis of our previous findings (33-35), as well as the dual-mode (7) and transient hypofrontality theories (5), we hypothesized that specific changes in brain cortical activity, namely, a decrease in frontal β activity, are dependent on both exercise intensity and exercise preference. Regarding the dose-response relationship (10,11), we anticipated that a decrease in frontal cortex activity would only be induced by intense exercise (33). On the basis of the preference hypothesis (35), we expected this dose-response relationship to be strongest after the treadmill exercise with recreational runners. To confirm the model of exercise preferences, brain cortical activity was recorded in a second experiment comprising a group of semiprofessional hand cycling athletes (all spinal cord injured). Within this group of athletes, we hypothesized that we would find similar changes in brain cortical activity after hand cycling, compared with those found in the group of recreational runners after the intense treadmill exercise and thus a deactivation of the frontal brain areas.
Twelve healthy recreational runners (age 26.3 ± 3.8 yr; males: n = 8, 25.6 ± 4.2 yr; females, n = 4, 27.5 ± 2.9 yr; relative peak oxygen consumption on the treadmill = 46.47 ± 10.3 mL·min−1·kg−1) gave their informed written consent, and the study was approved by the local institutional ethics committee. Subjects were characterized to be recreational runners because they reported going for a run frequently, whereas they did not report performing bike frequently, nor arm crank exercise or isokinetic strength training. None of the subjects reported any psychological or physiological problems or previous head injuries, and none was prescribed medication. Subjects were randomly assigned to a quasirandom exercise sequence of different exercise modes and intensities.
For the second experiment, a group of five semiprofessional hand cycling athletes (n = 5 males, age 39 ± 7.9 yr) were recruited. Subjects had a traumatic spinal cord injury at least 12 months before the study (lesion level = Th6-11; all American Spinal Injury Association category A). All subjects gave informed written consent and provided a medical clearance for participation in this study.
Every subject in experiment 1 was tested on the treadmill, bicycle, and arm crank ergometer at 50% (moderate) as well as 80% (intense) of their V˙O2peak and on an isokinetic dynamometer at 50% (moderate) and 80% (intense) of their maximal strength. To calculate the individual performance intensity, V˙O2peak tests were performed on the treadmill, bicycle, and arm crank ergometer, each separated by at least 3 d. The stage length of all three incremental exercise tests was 3 min for each stage, and the test ended at volitional exhaustion. After each stage, a 20-μL capillary blood sample was collected from the ear lobe for blood lactate (LAC) analysis (BIOSEN C_line Glukose-/Laktat-Messsystem; EKF-Diagnostic, Barleben, Germany). HR (Polar S810i; Polar Electro, Buettelborn, Germany) was assessed at the end of each stage. Expired respiratory gasses were continuously collected breath by breath and analyzed at 30-s intervals (ZAN 600 ErgoTest; ZAN, Oberthulba, Germany). The highest 30-s V˙O2 value throughout the test was taken as V˙O2peak. The treadmill test (WOODWAY ELG 55; WOODWAY, Weil am Rhein, Germany), commenced at a velocity of 2 m·s−1 and increased every stage by 0.5 m·s−1. The starting resistance on the bicycle ergometer (ergoline ER 900; ergoline, Bitz, Germany) was set at 50 W and was increased by 50 W with every stage. Arm cranking (Cyclus2, Richter, Germany) started at 20 W and increased by 20 W with every stage. On the basis of the V˙O2peak, 50% and 80% intensities were calculated, and the accompanying HR was used to control exercise intensity during the respective experimental tests. For the isokinetic strength test (IsoMed 2000; D&R Ferstl GmbH, Hemau, Germany), three maximal isokinetic concentric contraction trials of wrist flexion were performed 5 min before testing to determine the average maximum force. Wrist flexion was performed at an angular velocity of 40°·s−1, commencing in the neutral position and ending at 30° flexion. After each contraction, the wrist was returned passively to the neutral position. To achieve the 50% and 80% target intensities, subjects were asked to contract until a visual force display followed a target curve that was displayed on a computer monitor.
Running and bicycling were performed for 30 min each. The arm crank exercise was performed for 3 × 10 min bouts, separated by 3-min breaks, and isokinetic concentric fatiguing contractions of 3 × 20 consecutive wrist flexions were repeated with a 1-min rest interval between each series. The interval protocols for arm cranking and wrist flexion were necessary because local muscle fatigue would prevent most subjects from completing 30 min continuously. We chose these four exercise modes to provide a wide spectrum of physical strain and cover endurance as well as strength characteristics.
The exercise protocol of the second experiment was an incremental arm crank test (Cyclus2, Richter, Germany), starting at 20 W and increasing by 20 W at each step, each for 5 min. The test ended at volitional exhaustion. After each stage, 20 μL of capillary blood was collected from the ear lobe for LAC analysis (BIOSEN C_line Glukose-/Laktat-Messsystem; EKF-Diagnostic). HR (Polar S810i; Polar Electro) was assessed at the end of each stage.
All exercise protocols and EEG recordings of experiments 1 and 2 took place in an air-conditioned laboratory to standardize testing conditions (room temperature = 22.22°C ± 1.84°C).
Before (PRE) and 17.6 ± 2.9 min after (POST) the exercise, brain cortical activity was recorded for 5 min with the subjects in a supine position. For EEG measurements, subjects were asked to close their eyes, relax, and avoid moving (resting EEG). The experimental environment around the subjects during the resting EEG was kept as quiet as possible to avoid distracting the subjects and associated artifacts. Half an hour before the PRE measurement, 19 Ag/AgCl electrodes (Fp1, Fp2, F3, F4, F7, F8, Fz, C3, C4, Cz, P3, P4, P7, P8, Pz, T7, T8, O1, O2) were mounted to the subject's head on the basis of the 10-20 system (21) with a flexible EEG cap (Electro-Cap International, Inc., Eaton, OH). The reference electrode, which was included in the cap design, was integrated in a triangle of Fp1, Fp2, and Fz. A midline position was used because it does not amplify the signal in one hemisphere or the other. The cap was suitable for different head sizes, breathable to avoid heat accumulation, and fixed to a strap around the chest by two strings to prevent shifting of the electrodes during exercise. Electro-Gel™ (Electro-Cap International, Inc.) inside the electrodes provided signal transduction. Electrodes with impedance values exceeding 10 kΩ were excluded from further analysis. A Braintronics ISO-1064CE box (Braintronics B.V., HL Almere, The Netherlands) was used to amplify and convert analog signals into digital signals. Data were stored with a frequency of 256 Hz on the hard disk of a Neurofile XP EEG System (IT-Med, Usingen, Germany).
For the group of hand cycling athletes, EEG was measured for 5 min PRE and immediately POST the exercise, with subjects on the arm crank ergometer in a seated position. The recently developed actiCAP provided by Brain Products GmbH (Munich, Germany) with 32 active Ag/AgCl electrodes was used. Electrodes Fp1, Fp2, F7, F3, Fz, F4, F8, FC5, FC1, FC2, FC6, T7, C3, Cz, C4, T8, TP9, CP5, CP1, CP2, CP6, TP10, P7, P3, Pz, P4, P8, PO9, O1, Oz, O2, and PO10 were also placed according to the 10-20 system. The advantage of these new active electrodes is their integrated noise subtraction circuits, which deliver lower noise levels compared with "normal" electrodes. Measurements were recorded by the BrainVision amplifier and the RecView software (Brain Products GmbH) with a frequency of 500 Hz.
Low-Resolution Brain Electromagnetic Tomography
Multichannel EEG measurements do not contain enough information to determine the neuronal activity in an exact three-dimensional distribution throughout the volume occupied by the brain. This is why it has not been possible to localize the original location and power of the EEG signals to date. Recently, researchers discovered that extracranial EEG signals are generated by neuronal postsynaptic potentials working in a highly synchronized manner and a dense cluster of neurons. This allows a back calculation by corresponding algorithm to locate the origin of these synchronized postsynaptic potentials. For this purpose, the traditional compartmentalization of the cortex in voxels (volume elements (37)) was used (1,16,19,28,31). On the basis of the coordinates of the digitized Talairach Atlas, these voxel values were allocated to their corresponding Brodmann areas (BA) to classify the cortical regions. The result is an authentic low-resolution brain electromagnetic tomogram, producing a blurred image of distribution and conserving the original location of the signal (29), because of the principles of superposition.
We compared HR and lactate concentration values at the end of each exercise mode using repeated-measures ANOVA (STATISTICA 7.1; StatSoft, Tulsa, OK), with the factors exercise mode (MODE; four levels: treadmill, bicycle, arm crank, isokinetic) and intensity (INT; two levels: 50%, 80%). The Tukey test was used for post hoc analyses. Significance threshold was P = 0.05, P = 0.01, and P = 0.001.
EEG/low-resolution brain electromagnetic tomography.
The first and last 30 s of the 5-min resting EEG recordings were taken out of the analysis because of possible artifacts due to, for example, lack of relaxation at the beginning or restlessness at the end of the measurement. By BrainVision Analyzer (Brain Products GmbH), the remaining 4 min of the resting EEG data were filtered with Butterworth zero-phase filters. A low cutoff was set at 0.5 Hz, and a high cutoff was set at 49.5 Hz, with a time constant of 0.31831 and 24 dB per octave. Afterward, the remaining frequency range was divided into 4-s segments with an overlap of 0.4 s and a corrected baseline. An automatic artifact rejection with a maximum voltage step/sampling point of 50 μV, an amplitude criterion of minimum 100 μV and maximum 100 μV, and a maximum − minimum difference of values of 200 μV marked and removed data sequences 200 ms before and after the event. In addition, data sets were visually screened, and distinct artifacts were manually sorted out.
To display regional changes in brain cortical activity from PRE to POST measurements, the coordinates of the EEG electrode positions were transformed into a digitized magnetic resonance imaging version of the Talairach Atlas first (McConnell Brain Imaging Centre, Montréal Neurological Institute, McGill University) to calculate the sLORETA transformation matrix. After, a transformation to average reference was done, at least twenty-five 4-s epochs of the above-described treated EEG data were averaged to create a cross-spectra in sLORETA for the α (7.5-12.5 Hz) and β (12.5-35 Hz) frequency bands according to every subject. To convert the cross-spectra to sLORETA files, a transformation matrix was used. Finally, three-dimensional images of the distribution from the original neuroelectrical activity were obtained, displaying the cortical neuronal oscillators in 6430 voxels, with a spatial resolution of 5 mm (27). A four-mode (treadmill, bicycle, arm crank, isokinetic) × two-intensity (50% + 80%) × two-EEG frequency band (α + β) × four-brain region (frontal, temporal, parietal, and occipital) model with repeated measures was used for experiment 1. For experiment 2, a one-mode (hand bike) × one-intensity (incremental test until subjective exhaustion) × two-EEG frequency band (α + β) × four-brain region (frontal, temporal, parietal, and occipital) model was applied. Differences in spectral activity between measurements were determined by a paired-samples t-test at each voxel for sLORETA power. Statistical analysis of sLORETA data has been carried out by a nonparametric test for paired variables in other studies already (26). The t values of brain cortical activation specific to the analyzed EEG frequency bands were analyzed on cluster level. The output shows exact probability values, which were corrected for multiple comparisons. These voxel-to-voxel values are displayed as statistical parametric maps. Significance threshold was P = 0.05, P = 0.01, and P = 0.001. The low number of EEG electrodes we used allows us to represent the results of the broad brain regions: frontal, temporal, parietal, and occipital cortexes.
An ANOVA, including the factors MODE and INT (n = 12), F3,33 = 9.2909, revealed a main effect of MODE with significantly (P < 0.05) higher lactate values (mmol·L−1) for POST of 50% arm crank exercise in comparison with all other moderate strains (with 50% treadmill, P < 0.001; with 50% bicycle, P < 0.001; and with 50% isokinetic, P < 0.001). In addition, it showed a main effect of INT with significantly higher values for POST of 80% arm crank exercise compared with the 50% arm crank exercise (P < 0.001). A main effect of INT was also obtained for the 50% treadmill and bicycle exercise compared with their 80% intensities, which had significant lower values for POST moderate exercise, respectively (bicycle 50% with 80%, P < 0.001; treadmill 50% with 80%, P < 0.001). The intensive arm crank exercise revealed significant higher lactate values compared with the intensive bicycle, treadmill, and isokinetic exercises (always P < 0.001), and the intensive bicycle exercise revealed significant higher values compared with the intensive treadmill and isokinetic exercises (P < 0.001 in both cases). For all absolute values and SE, please see Table 1.
Comparing HR values (beats·min−1) between exercise modes, a main effect of MODE was found: the 50% treadmill and bicycle exercises produced significantly higher rates compared with the 50% arm crank and isokinetic exercises (50% treadmill with 50% arm crank, P < 0.05 and with 50% isokinetic, P < 0.001; 50% bicycle with 50% arm crank, P < 0.05 and with 50% isokinetic, P < 0.001). For absolute values, please see Table 1. For the 80% isokinetic exercise, the HR was lower compared with all the other exercise modes (with 80% treadmill, P < 0.001; with 80% bicycle, P < 0.001; and with 80% arm crank, P < 0.001). After intensive arm crank exercise, the HR was lower than the post-treadmill and bicycle exercise values (P < 0.001 in both cases). In addition, a main effect of INT was revealed for all modes of strain (treadmill 50% with 80%, P < 0.001; bicycle 50% with 80%, P < 0.001; arm crank 50% with 80%, P < 0.001; isokinetic 50% with 80%, P < 0.05).
EEG α frequency range.
Comparing POST with PRE measurements after moderate exercise, sLORETA revealed a significant increase in EEG α frequency band for all four kinds of exercise (Fig. 1). The corresponding t and P values of changes in the α frequency range in the different brain regions are listed for the four exercise modes and both intensities in Table 2. After moderate treadmill and bicycle exercises, the increase of activity was found in the parietal cortex (treadmill: BA 1, 2; bicycle: BA 5, 7, 40). All other brain regions showed no significant changes in the α frequency range. Moderate arm crank and isokinetic exercises were shown to cause an increase of the frontal cortex activity (arm crank: BA 6, 9; isokinetic: BA 8, 9, 32). No changes were found in the other brain regions.
For the 80% intensity, only the isokinetic exercise protocol was shown to produce a significant increase of α activity in the frontal cortex (BA 10, 11, 44). For the intense treadmill exercise, results found a decrease of α activity in the frontal cortex. This just missed statistical significance. The other brain regions and the other modes, intensive bicycling and arm crank exercises, revealed no significant changes.
EEG β frequency range.
For the EEG β frequency range, all exercise modes performed with moderate intensity were shown to enhance brain cortical activity (Fig. 2). The corresponding t and P values of changes in the β frequency range in the different brain regions are listed for the four exercise modes and both intensities in Table 2. A significant increase of β activity was only found for the bicycle exercise in the parietal cortex (BA 30, 31). All the other 50% intensity exercise values and other brain regions missed significance.
FIGURE 2-Statistical...Image Tools
After intense treadmill running, a significant decrease in the β frequency band was found to occur in the frontal cortex (BA 11, 25, 47). No significant changes were found in the temporal, parietal, and occipital cortexes. Intense bicycling, arm cranking, and isokinetic exercise with 80% intensity were demonstrated to reveal no significant changes.
Mean and SD were calculated for the physiological values and duration of exercise for the hand cycling control group. Lactate values showed a mean of 13.0 ± 3.0 mmol·L−1 and an average exercise duration of 36.5 ± 3.4 min. Mean HR at the end of the incremental arm crank test was 198 ± 5 beats·min−1.
EEG α frequency range.
After the incremental hand bike test comparing PRE with POST measurements, EEG recordings revealed significant decreased α activity in the frontal cortex (BA 11, t = 21.137, tcritical = 21.137, P < 0.05) (Fig. 3). No significant changes were found in the temporal, parietal, and occipital cortexes. The corresponding t and P values of changes in the α and β frequency ranges in the different brain regions are listed in Table 3.
EEG β frequency range.
β activity decreased after exercise in the frontal cortex (Fig. 3, Table 3). Changes of β frequencies in the frontal cortex were not significant. The temporal, parietal, and occipital cortexes revealed no significant changes in β activity.
This study aimed to localize the exercise-induced effects on brain cortical activity by applying sLORETA. To test the dose-response relationship between exercise intensity and brain activity (10,11) as well as the proposed exercise preference hypothesis (34,35), we tested a group of recreational runners on four different exercise modes (treadmill, bicycle, arm crank, and isokinetic wrist flexions) at two exercise intensities (50% and 80% of individual capacity) as well as a second group of semiprofessional hand cycling athletes on a hand cycling ergometer. On the basis of the dual-mode theory (6,7,9) and the transient hypofrontality hypothesis (4,5), we hypothesized that for trained athletes, moderate-intensity exercise would not result in changes in brain activity, whereas intense workloads would lead to deactivation of the frontal brain regions (34,35). Furthermore, this deactivation of frontal brain cortical activity was expected to be more pronounced after preferred exercise modes, i.e., treadmill running for recreational runners and arm cranking for hand cycling athletes.
The results show that moderate exercise intensity leads to increases in α activity across all exercise modes. Less familiar exercise modes (arm crank and isokinetic exercise) lead to increased α activity in frontal brain regions, whereas familiar exercise modes (treadmill and bicycle) lead to increased α activity in the parietal cortex. In contrast, preferred exercise at high intensities (i.e., treadmill running) was accompanied by a reduction in β activity in the prefrontal cortex. Thus, our results lend some support for a dose-response relationship as well as the exercise-preference hypothesis, meaning that familiarization and adaptation to certain exercises and intensities influence the response to this stimulus and results in a specific brain cortical activation pattern, namely, a deactivation of frontal cortex activity.
In this study, moderate exercise intensity evoked an increase in α activity across all exercise modes. The observed increase in frontal and parietal α activity after moderate exercise intensity is in conflict with our hypothesis, which assumed, on the basis of the dose-response relationship (11), that 50% of individual maximal capacity would be insufficient to influence brain cortical activity. Thus, we need to reject our hypothesis.
Localized electromagnetic tomography revealed that α activity increased in the parietal brain regions related to somatosensory perception (BA 1, 2, 5, 7) after the treadmill and bicycle exercises and mainly in frontal brain regions related to emotional and sensory processes after the arm crank (BA 6, 9) and isokinetic exercises (BA 8, 9, 32). Because our subjects were most familiar with running and bicycling but not with arm crank and isokinetic exercises, these findings indicate that more familiar types of exercise lead to increases in α activity in somatosensory regions and that unfamiliar types of exercise lead to increases in α activity in emotion centers. An increase in frontal cortex activity has already been observed in motor learning studies during novel tasks (14,18,30). Furthermore, there is evidence that, with practice, activity in the somatosensory and motor areas in the parietal cortex is more and more pronounced (14,18,30).
For high-intensity exercise, an exercise-mode-dependent effect was observed. As expected, we found a reduced frontal β activity after intense treadmill running in the recreational runners. Intense bicycle, arm crank, and isokinetic exercises did not induce a change in frontal β activity. It is tempting to speculate that the observed changes in frontal brain activity during exercise are related to the metabolic or cardiovascular load of the exercise. This does not seem to be the case because peak HR was not different between the treadmill and bicycle exercises. Furthermore, the highest LAC values were observed after the intense arm crank exercise.
We attribute these findings to the proposed preference hypothesis and the accompanying familiarization/adaptation processes. This hypothesis is supported and may be explained by an internal or a functional mechanism.
First, Ekkekakis and Petruzzello (11) described many possible internal factors to explain individual differences in response to physical activity, for example, personality, self-efficacy, match of perceived exertion, and individual abilities. According to this approach and their dual-mode theory, we could conclude that experience in the preferred sport and at various intensity levels leads to higher placidity, a sense of well being, and/or positive emotions at higher exercise intensities, whereas unfamiliar strains or overload may produce strong interoceptive cues, insecurity, and fear of excessive demands inherently connected with negative affects (8), mediated by changes in cortical activity. Second, there is evidence that, with increasing automaticity of movements, less brain areas are activated and that activity in participating brain areas declines (12,20). Other studies indicate that learning a novel task involves higher activation in prefrontal brain areas related to emotion (14,18,30). Therefore, functional/structural changes or adaptations are thought to result in the different activation patterns after unfamiliar and familiar exercise modes. These two mechanisms might explain why we found increased α activity in the frontal brain regions (BA 10, 11) after the unfamiliar isokinetic exercise at high intensities and, in contrast, decreased β activity in the frontal brain regions (BA 11, 25, 47) after the preferred treadmill exercise at high intensities. The affected brain regions BA 10, 11, 25, and 47 are all thought to be implicated in emotion processing (13,15,32).
This exercise preference hypothesis is also supported by the significant decreased α activity in the frontal BA 11 and the decreased β activity, albeit not significant, in BA 32 after the arm crank exercise in the group of semiprofessional hand biking athletes. Again, both brain regions are thought to be involved in emotion processing. The small number of subjects in this second experimental group is likely to have led to the nonsignificant results in the β frequency range.
The activation patterns after the arm crank and isokinetic exercises were similar for moderate and high intensities, whereas brain activation patterns after the treadmill and bike exercises were different for both intensities. This indicates that a certain intensity is needed to evoke exercise mode-specific changes, in particular, frontal deactivation for preferred exercise, and this intensity threshold seems to be between 50% and 80% of the individual capacity. Therefore, preferred exercise (mode and intensity) has a different effect on brain cortical activity compared with nonpreferred exercise modes and intensities, and this should be taken into account during exercise recommendations.
Although the low-resolution brain electromagnetic tomography method has been questioned by some for its localization accuracy and electrophysiological and neuroanatomical constraints (17,23-25), there is strong evidence that validates this method against positron emission tomography and functional magnetic resonance imaging over a diverse range of physiological conditions (for review, see Pascual-Marqui et al. (28)). Furthermore, it also needs to be taken into account that these results are specific for the one postexercise measurement time point and groups of athletes tested, consisting of recreational runners or semiprofessional hand cycling athletes, including information about their age, training status, and psychophysiological conditions. Other groups of population might respond to other strains and intensities depending on their preconditions, which we, on the basis of our results, strongly recommend to respect. However, the results may point out an important factor for exercise recommendations in terms of fitness and health training, which is exercise preference. Further studies will be needed to clarify the relevance and benefit of different cortical activation patterns, for example, by testing effects of various higher intensities of preferred exercise modes together with assessments of mood, and cognitive performance as well as hormones, neurotransmitters, or neurotrophic factors. Furthermore, studies are required to test 1) various higher intensities to find the threshold for this frontal deactivation phenomenon and 2) athletes engaged in other exercise modes to see whether this is only true for aerobic exercise modes or also for sports like weight lifting.
In conclusion, these findings contribute to our understanding of the dose-response relationship between exercise intensity and brain cortical activity and imply that cortical activation patterns in response to exercise depend on both exercise mode and exercise intensity. The question of which brain regions are affected by exercise seems to be a matter of task familiarization. The individual exercise preference, including an adaptation to a specific mode and intensity, seems to result in a specific activation pattern, namely, a deactivation of frontal cortex activity. These findings support the notion that individualized exercise prescriptions are essential for achieving specific psychophysiological outcomes.
This study was supported by a grant for graduates of the German Sport University Cologne, Germany.
The authors thank their participants for committing a lot of time, effort, and patience toward this extensive study. Special thanks go to Petra Wollseiffen, Moritz Fölger, Sebastian Zeller, and Dominik Bonin for their help with data collection and to all who gave valuable comments by proofreading this article, including two unknown reviewers.
The results of the present study do not constitute endorsement by the American College of Sport Medicine.