Effects of aerobic exercise on event-related potentials related to cognitive performance: a systematic review

Introduction

Aerobic exercise has been shown to improve brain health as well as cognitive functioning (Barha et al., 2017; Laurin et al., 2001; Pope, Shue & Beck, 2003; Smith et al., 2010). The physiological links between aerobic exercise and cognitive function may be facilitated through many different mechanisms, e.g., secretion of neuromodulators (such as brain-derived neurotrophic factor (BDNF)), neurogenesis, increased brain plasticity and increased brain blood flow (Waters et al., 2020), but so far evidence for the biological mechanisms underlying this relationship remain sparse. Moreover, it may be speculated that different underlying mechanisms may mediate the effects of acute exercise interventions, which may be immediate and short-lived, versus longer exercise interventions, which conversely may be slower to develop and more long-lasting. A number of studies have found an effect of long-term exercise on the hippocampus, white matter and the cortical mantle (Colcombe et al., 2006; Erickson et al., 2011; Pajonk et al., 2010). Although imaging studies of acute and immediate exercise effects are lacking, findings examining the effects of exercise on BDNF further indicate differential effects of acute and longer-term exercise. Specifically, BDNF was found to be elevated immediately following an acute bout of exercise, but not following 3 months of exercise (Krogh et al., 2014; Tsai et al., 2021). Nevertheless, both acute and chronic exercise seem to have beneficial effects on the brain. Acute aerobic exercise has been shown to facilitate learning mechanisms (Perini et al., 2016; Voss et al., 2021) and improve cognition in post-recovery period following exercise in healthy subjects (Erickson et al., 2019). A meta-analysis showed that acute bouts of aerobic exercise improved cognitive task performance (Lambourne & Tomporowski, 2010) and a large body of literature supports this notion (Ratey & Loehr, 2011). A systematic review concluded that aerobic exercise interventions exceeding one month are associated with modest improvements in attention and processing speed, executive functioning and memory (Smith et al., 2010). It could be theorized that chronic exercise enhances cognitive aspects by modulation of brain structure and vascular proliferation and perfusion, which develops over time, and acute exercise works by the immediate effects of neurosecretion related to exercise. However, differences in precisely which cognitive domains that are affected, and by which underlying mechanisms, by acute and chronic exercise respectively are yet to be investigated in more detail.

One method to quantify the impact of aerobic exercise on brain function and cognitive performance is by event-related potentials (ERPs), which has been widely used in studies investigating perception, attention and cognitive functioning (Helfrich & Knight, 2019). ERPs are small electrical potentials generated in the cortex (or subcortical generators) in response to a specific stimulus or event and can be measured noninvasively by electroencephalography (EEG) or magnetoencephalography (MEG) using scalp electrodes (Woodman, 2010). Information about neural activity such as early sensory perception processes and higher-level processing such as attention, cortical inhibition, response selection, error monitoring, memory update, and other cognitive functions (Duncan et al., 2009; Polich, 2007) can be obtained as different ERP components vary according to stimulus type and type of cognitive task. The most studied endogenous ERP is the P3 (300–500 ms post-stimulus), which is interpreted as an index of ability to sustain attention to a target. P3 is difficult to localize and most studies agree that P3 (P3b) has multiple dipole sources, e.g., the hippocampus and the parahippocampal areas, the insula, the temporal lobe, the occipital cortex and the thalamus (Sokhadze et al., 2017). Another frequently studied ERP is the N2, associated with categorization, perceptual closure, inhibitory control and attention focusing (Sokhadze et al., 2017) and is generated by frontal and anterior cingulate cortex (Heil et al., 2000). Due to the noninvasiveness, ease of use and temporal resolution of the technique and the fact that it can be applied immediately after an intervention, ERPs represent an attractive method of capturing neural effects of acute bouts of exercise (Pedroso et al., 2017). ERPs are in general considered to express different components of executive functions, such as processing time and the amount of cognitive resources allocated to the perception and processing of an event/task. Especially P3 seems significantly impacted by exercise in most studies.

The objective of the present study was to carry out a systematic review of studies reporting on the effects of both acute and chronic exercise on ERPs related to cognitive performance and associated behavioral measures such as accuracy and reaction time. Furthermore, we aimed to investigate whether exercise intensity was correlated with changes in ERPs.

Methods

Study selection, data extraction and data items

Three authors screened and selected the included studies (JG, MG, KF). Authors were blinded with regards to the results of each authors’ screening results. The authors initially screened articles on title and abstract level. Subsequently, full text articles for those identified in the first step were retrieved and assessed for final inclusion. Any disagreement with regards to whether a study could be included was resolved by consensus (JG and KF) (see Fig. 1 for flow-chart and supplementary table for more detailed explanations on exclusion reasons). Relevant data was extracted by the same author (JG) and reviewed by another author (MG). Data was extracted in an Excel data extraction sheet that was piloted using four studies before being applied to the rest of the studies. The following items were extracted: number of participants, gender, age, diagnosis, study design, characteristics of comparator and intervention (type, duration, exercise composition, intensity and how it was measured), methods (EEG/MEG used, cognitive paradigm and sensory modality, ERP outcome measurements), reported effect of the intervention on ERP measurements and behavioral results. A risk-of-bias assessment was carried out using Cochrane’s Risk of Bias version 1.

Results

Acute exercise intervention results

Due to the large heterogeneity across studies in terms of the paradigms used, ERPs investigated and methods employed, general trends in the data are difficult to extract. A total of 21 different ERPs were examined and seventeen different cognitive paradigms to elicit ERPs were used with flanker task as the most often used paradigm (12 studies). The most frequently reported ERP was the P3 (33 studies) followed by N2 and N140. Regarding P3, each study sometimes reported results for P3 amplitudes and latencies from more than one experiment as several paradigms were used to elicit the P3 or several exercise intensities were investigated, and thus numbers of experiments are reported in the following. A total of 27 experiments out of 41 experiments reported significant effects of the exercise intervention on amplitude (25 increased, two decreased) and 16 reported effects on latency (two increased, 14 decreased). Fifteen of the 27 experiments reported significant effects on both amplitude and behavioral results (mainly decreased reaction time, only one study reported effect on accuracy (increased)) and seven experiments out of the 16 experiments reported a significant effect on latency and behavioral results (decreased reaction time). Looking across exercise intensities, there was a tendency that effects on amplitudes were significant mainly in interventions using moderate intensity exercise across all ERPs and paradigms used (see Fig. 2). Twenty-six out of 31 studies that investigated moderate-intensity interventions reported significant increases in one or more ERP component post-exercise. Conversely, out of the eleven studies that did not report any effect on ERP amplitude, two studies were of moderate intensity exercise. See Table 3 for all the results of studies using acute aerobic exercise.

Table 3:

Acute aerobic exercise results.

References	Diagnosis	Recorded time after exercise	Outcome measures	Paradigm	ERP results	Behavioural results
Light-intensity exercise results
Kamijo et al. (2009)	Healthy	2 min	P3	Flanker task (visual)	Amplitude: P3↔	Accuracy:↔
					Latency: P3↓	Reaction time:↔
Kamijo et al. (2007)	Healthy	3 min	P3	Flanker task(visual)	Amplitude: P3↑	Accuracy:↔
					Latency: P3↓	Reaction time:↓
Moderate-intensity exercise results
Tsai et al. (2014)	Healthy	15-20 min	P3 and CNV	Visuospatial attention task(visual)	Amplitude: P3↑ CNV↑	Accuracy:↔
					Latency: P3↔ CNV↔	Reaction time:↓
Yagi et al. (1999)	Healthy	Within 10 min	P3	Oddball task (visual and auditory)	Amplitude: P3↔	Accuracy:↔
					Latency: P3↔	Reaction time:↔
Scudder et al. (2012)	Healthy	20,2 ± 6,4min	P3 and N2	AX-continuous performance tasks (visual)	Amplitude: P3↑ N2↔	Accuracy:↑
					Latency: P3↔ N2↔	Reaction time:↔
Akatsuka et al. (2015)	Healthy	5 min	N140	Go-/No-go task (somato-sensory)	Amplitude: N140↑	Accuracy:↔
					Latency: N140↔	Reaction time:↔
Kamijo et al. (2009)	Healthy	2 min	P3	Flanker task (visual)	Amplitude: P3↑	Accuracy:↔
					Latency: P3↓	Reaction time:↓
Kao, Wang & Hillman (2020)	Healthy	30 min	P3	Serial N-back task (visual)	Amplitude: P3↑	Accuracy:↔
					Latency: P3↔	Reaction time:↔
Aly & Kojima (2020)	Healthy	HR returned to within 10% of pre-ex	P2, N2c and P3	Flanker task (visual)	Amplitude: P2↑ P3↑ N2c↑	Accuracy:↔
					Latency: P2↔ P3↔ N2c↔	Reaction time:↓
Wen & Tsai (2020)	Healthy	HR returned to within 10% of pre-ex	P2, N2 and P3	Stroop task (visual)	Amplitude: N2↓ P3↑ P2↔	Accuracy:↔
					Latency: N2↓ P3↓ P2↔	Reaction time:↔
Chu et al. (2015)	Healthy	Within 10 min	P3 and N1	Stop-signal task (visual)	Amplitude: P3↑ N1↔	Accuracy: NR
					Latency: P3↑ N1↔	Reaction time:↓
Shibasaki et al. (2019)	Healthy	Right after	N140 and P300	Go-/No-go task (somatosensory)	Amplitude: N140↓ P3↔	Accuracy:↔
					Latency: N140↔ P3↓	Reaction time:↔
Ligeza et al. (2018)	Healthy	HR returned to within 10% of pre-ex	N2 and P2b	Flanker task (visual)	Amplitude: P3↔ N2↑	Accuracy:↑
					Latency: NR	Reaction time:↓
Pontifex et al. (2015)	Healthy	HR returned to within 10% of pre-ex	P3a and P3b	Oddball task (visual)	Amplitude: P3a↔ P3b↑	Accuracy:↔
					Latency: P3a↔ P3b↔	Reaction time:↔
Kamijo et al. (2007)	Healthy	3 min	P3	Flanker task (visual)	Amplitude: P3↑	Accuracy:↔
					Latency: P3↓	Reaction time:↓
Chu et al. (2017)	Healthy	NR	P3 and conflict SP	Stroop color-word task (visual)	Amplitude: P3↑ SP↔	Accuracy:↔
					Latency: P3↔ SP↔	Reaction time:↓
Zhou & Qin (2019)	Healthy	HR returned to within 10% of pre-ex	P2, N2, P3b and N450	Stroop color-naming task (visual)	Amplitude: P2↑ N2↔ P3b↔ N450↔	Accuracy:↔
					Latency: P2↔ N2↔ P3b↔ N450↔	Reaction time:↔
Dimitrova et al. (2017)	Healthy	Within 20 min	NR	Stroop task (visual)	Amplitude: ERP↑	Accuracy: n↔
					Latency: NR	Reaction time:↓
Won et al. (2017)	Healthy	HR returned to within 10% of pre-ex	P3	Stroop color-word conflict task (visual)	Amplitude: P3↑	Accuracy: NR
					Latency: NR	Reaction time:↓
Hwang et al. (2019)	Healthy	90 min	N2	Facial Go-/No-go task	Amplitude: N2↓	Accuracy:↔
					Latency:↔	Reaction time:↔
Kao et al. (2018)	Healthy	NR	P3	Flanker task (visual)	Amplitude: P3↑	Accuracy:↔
					Latency: P3↔	Reaction time:↓
Wu et al. (2019)	Healthy	30 min	P3b and N1	Task-switching test	Amplitude: P3↑ N1↔	Accuracy:↔
					Latency: NR	Reaction time:↓
Winneke et al. (2019)	Healthy	2,56 min (range 2 –3,10)	N2 and P3	Flanker task	Amplitude: N2↑ P3↔	Accuracy:↔
					Latency: N2↔ P3↓	Reaction time:↔
Takuro, Nishihira & Soung-Ryol (2009)	Healthy	Right after + when HR had returned to pre-ex values	P3, early and late CNV	Go-/No-go reaction time task	Amplitude: P3↑ early CNV↑ late CNV↑ (only right after)	Accuracy: NR
					Latency: P3↔ early CNV↔ late CNV↔	Reaction time: NR
Chacko et al. (2020)	Healthy	Right after	BP, pN, N1, pN1, pP1 and P3	Discriminative response task	Amplitude: N1↓ all other↔	Accuracy:↔
					Latency: pN1↓ all other↔	Reaction time:↔
Milankov et al. (2012)	Healthy	NR	P3	Oddball task (auditory)	Amplitude: P3↔	Accuracy: NR
					Latency: P3↑	Reaction time: NR
Bae & Masaki (2019)	Healthy	HR returned to within 10% of pre-ex	P3	Task-switching paradigm (visual)	Amplitude: P3↑	Accuracy:↔
					Latency: P3↓	Reaction time:↓
Drapsin et al. (2012)	Healthy	Right after	P3	Oddball task (auditory)	Amplitude: P3↑	Accuracy:↔
					Latency: P3↔	Reaction time:↔
Swatridge et al. (2017)	Chronic stroke	Both 0, 20 and 40 min after	P3	Flanker task (visual)	Amplitude: P3↑ (40 min post-ex)	Accuracy:↔
					Latency: P3↓ (20 min post-ex)	Reaction time:↔
Tsai et al. (2018)	Mild cognitive impair-ment	HR returned to within 10% of pre-ex	P3	Flanker task (visual)	Amplitude: P3↑	Accuracy:↔
					Latency: P3↔	Reaction time:↓
Wang et al. (2020)	Heroin addiction	HR returned to within 10% of pre-ex	N2 and N2d	Go-/No-go task (visual)	Amplitude: N2↑ N2d↑	Accuracy:↑
					Latency: N2↔ N2d↔	Reaction time:↔
Chang et al. (2017)	Healthy	15 min after	N1, N2, P3 and N450	Stroop task (visual)	Amplitude: P3↑ N450↑ N1↔ N2↔	Accuracy:↔
					Latency: N450↓	Reaction time:↓
Chang et al. (2015)	Healthy	Within in 10 minutes	P3	Attention network task	Amplitude: P3↑	Accuracy:↔
					Latency: P3↔	Reaction time:↑
High-intensity exercise results
Xie et al. (2020)	Healthy	Within 15 min	P3 and LPP	Flanker task (visual)	Amplitude: P3↔ LPP↑	Accuracy:↔
					Latency: NR	Reaction time:↓
Walsh et al. (2019)	Healthy	10 min	RewP	Novel gambling task (visual)	Amplitude: RewP↓	Accuracy:↔
					Latency: NR	Reaction time:↔
Ligeza et al. (2018)	Healthy	HR returned to within 10% of pre-ex	N2 and P2b	Flanker task (visual)	Amplitude: N2↔ P2b↔	Accuracy:↔
					Latency: NR	Reaction time:↔
Kamijo et al. (2007)	Healthy	3 min	P3	Flanker task (visual)	Amplitude: P3↔	Accuracy:↔
					Latency: P3↓	Reaction time:↓
Rietz et al. (2019)	Healthy	30 min	P3, CNV, N2	Continous performance task (visual)	Amplitude: P3↑ CNV↔ N2↔	Accuracy:↔
					Latency: P3↔ CNV↔ N2↔	Reaction time:↔
Rietz et al. (2019)	Healthy	41 min	P3, CNV, N2	Flanker task (visual)	Amplitude: P3↔ CNV↔ N2↔	Accuracy:↔
					Latency: P3↔ CNV↔ N2↔	Reaction time:↔
Rietz et al. (2019)	Healthy	54 min	P3, CNV, N2	Four-choice reaction-time task (visual)	Amplitude: P3↔ CNV↔ N2↔	Accuracy:↔
					Latency: P3↔ CNV↔ N2↔	Reaction time:↔
Kao et al. (2017)	Healthy	20 min	P3	Flanker task (visual)	Amplitude: P3↓	Accuracy:↔
					Latency: P3↓	Reaction time:↓
Kao et al. (2018)	Healthy	NR	P3	Flanker task (visual)	Amplitude: P3↔	Accuracy:↔
					Latency: P3↓	Reaction time:↓
Milankov et al. (2012)	Healthy	NR	P3	Oddball task (auditory)	Amplitude: P3↓	Accuracy: NR
					Latency: P3↔	Reaction time: NR
Themanson & Hillman (2006)	Healthy	HR returned to within 10% of pre-ex	Error negativity, error positivity and N2	Flanker task (visual)	Amplitude: all↔	Accuracy:↔
					Latency: all↔	Reaction time:↔
Drapsin et al. (2012)	Healthy	Right after	P3	Oddball task (auditory)	Amplitude: P300↔	Accuracy:↔
					Latency: P300↔	Reaction time:↔
Jain et al. (2014)	Healthy	HR returned to within 10% of pre-ex	P3	Oddball task (auditory)	Amplitude: P3↑	Accuracy: NR
					Latency: P3↓	Reaction time: NR
Unknown intensity results
Wollseiffen et al. (2016)	Healthy	NR	P2 and N1	Chalkboard challenge	Amplitude: P2↔ N1↔	Accuracy:↔
					Latency: P2↔ N1↔	Reaction time:↔
Magnié et al. (2000)	Healthy	When body temperature and HR had returned to pre-exercise levels	P3, P2, N1 and N2	Oddball task (auditive)	Amplitude: P3↑ P2↔ N1↔ N2↔	Accuracy:↔
					Latency: P3↓ P2↔ N1↔ N2↔	Reaction time:↔
Nakamura et al. (1999)	Healthy	10 min	P3, P2, N100 and N2	Oddball task (auditory)	Amplitude: P3↑ P2↑ N100↔ N2↔	Accuracy:↔
					Latency: P3↔ P2↔ N100↔ N2↔	Reaction time:↔
Yagi et al. (1998)	Healthy	Right after	P3	Oddball task (visual)	Amplitude: P3↔	Accuracy:↔
					Latency: P3↔	Reaction time:↔

DOI: 10.7717/peerj.13604/table-3

Notes:

Shows results for all ERPs investigated with different cognitive paradigms using acute aerobic exercise interventions. Arrows (↑) indicate increase in measure following exercise, arrows (↓) indicate decrease in measure following exercise and arrows (↔) indicate no difference.

HR

Heart rate

NR

not registered

CNV

contingent negative variation

LPP

late positive potential

RewP

reward positivity

Discussion

The aim of the present study was to carry out a systematic review assessing the impact of aerobic exercise on ERPs related to cognitive performance. Heterogeneity across studies regarding methodology limited the possibility to draw definitive conclusions but the most consistent findings were that acute aerobic exercise was associated with higher amplitudes, and to a lesser extent shorter latencies, of ERPs. Moderate-intensity exercise was the most effective exercise intensity across studies in terms of affecting ERPs. For chronic exercise only a few studies were identified and results were less consistent. Lastly, in about half of the studies reporting an effect on ERPs, behavioral outcomes were also affected by the interventions.

Our findings are consistent with previous findings in another systematic review that evaluated the influence of physical activity or exercise on P3 in elderly participants. The authors concluded that physical activity and physical exercise positively influences changes in amplitude (Pedroso et al., 2017). Findings also suggested that P3 latency was less sensitive to physical activity compared to amplitude (Pedroso et al., 2017), which also aligns with the findings of the present systematic review.

P3 was the most frequently reported ERP in both acute and chronic exercise interventions in the identified studies. P3 represents the amount of attentional resources that is allocated to a specific task. Shorter latencies represent faster processing and higher amplitudes may be associated with attentional functioning (Polich & Heine, 1996). Our findings suggest that acute and to a lesser extent chronic exercise interventions seem to affect P3 amplitude. This aligns well with other studies that have found that exercise had a positive impact on attentional functioning and cognitive performance (Northey et al., 2018; Radel, Tempest & Brisswalter, 2018; Waters et al., 2020).

In the identified studies, other ERPs were also investigated such as the N1, N140 and N2 with the latter being the most frequently reported. Here, findings were more conflicting with studies reporting increases, decreases and no effects on amplitude and latency. These discrepancies may be due to difference in terms of intervention, control condition, paradigm used and study population, and further conclusions regarding these ERPs in relation to exercise are not possible.

We found that effects on ERPs seemed dependent on the exercise intensity, as most significant results for amplitudes were found in studies using moderate-intensity exercise. It has been proposed that P3 amplitude changes may be described by an inverted U-shaped curve relative to exercise intensity (Kamijo et al., 2004) and results from our study support this. It is uncertain what may mediate the U-shaped relationship indicating that low-intensity is “not enough” whereas high-intensity is “too much”. Interestingly, in a meta-analysis of patients with dementia, lower-frequency exercise interventions were more effective in improving cognition than higher-frequency interventions (Groot et al., 2016a) also hinting at the concept that not all doses of exercise are beneficial. In an observational study, a differential effect of exercise on cognitive functions was found, as physical activity was found to be positively associated with executive function and processing speed and negatively with memory (Frederiksen et al., 2015). Further studies are needed however, as results are conflicting (Hoffmann et al., 2016), and studies comparing different exercise intensities directly are few (Kamijo et al., 2007; Kao et al., 2017; Wang et al., 2016).

The use of ERPs in measuring cognitive performance post-exercise is practical and informative as the method enables assessment immediately before and after exercise. Further, temporal sensitivity is high, so ERP components can be tracked during cognitive paradigms. However, the spatial sensitivity is lacking and the different ERP components are difficult to locate (Woodman, 2010). Linking ERP findings with structural and functional MRI would give valuable information in this regard. However, studies are lacking in which MRI pre- and post-exercise has been carried out, especially in studies using acute exercise interventions. Future research should focus on concomitant use of MRI and ERPs in the investigation of cognitive responses, as the methods are complementary.

A number of aspects that possibly affect ERPs in relation to cognitive performance include age, exercise modality and exercise duration. We therefore divided the studies in acute and long-term aerobic exercise, although it can be theorized that smaller distinctions in duration could as well affect ERPs differently, e.g., an exercise session under 20 min versus over 20 min. The majority of the studies included using acute exercise interventions had exercise duration between 9–40 min and no apparent difference was observed. Age related differences are also worth taking into account when interpreting ERP results, as studies have shown a latency increase and a P3 amplitude decrease in healthy senior individuals compared to younger individuals. Healthy seniors compared to individuals with dementia show further increase in latencies and decrease in amplitudes, which had led to the suggestion that P3 could be considered as a biological marker of cognitive impairment (Hedges et al., 2016; O’Mahony et al., 1996; Pedroso et al., 2012). Elderly individuals’ ERPs post-exercise could therefore be more susceptible to exercise than younger individuals (Hillman et al., 2002). Another ERP component that was frequently reported was N2, which is involved in inhibitory control. An increase in N2 amplitude is found to correlate with correctly inhibited no-go stimuli in both younger and older adults, whereas P3 amplitude in the same study showed an age-related decrease (Kardos, Kóbor & Molnár, 2020). This adds several variables to the issue, where it seems that some ERP components are age-dependent and some are not, but included studies in the systematic review investigating N2 were few and we refrain from concluding anything based on these results.

Several limitations were present in the studies included in the systematic review. There were concerns regarding risk of bias in all studies included. In study designs using aerobic exercise interventions it is almost impossible to blind participants, as aerobic exercise is not possible to mask and therefore performance bias is a risk. Furthermore, the studies were also difficult to compare, as ERPs were elicited through different sensory modalities by various cognitive tests that examined different aspects of cognitive performance. The investigators most often also examined different ERPs, such as P3 or N2, making direct comparisons between studies difficult.

The systematic review also has limitations. We chose to include studies on populations that were both healthy and with different diseases in our synthesis, which could have biased our results. Studies reporting on diseased participants were few and in general no discernable trends in the findings convincingly indicated a different response between healthy participants and diseased. We chose to focus on aerobic exercise interventions as these have shown more robust and consistent effects on cognition (Groot et al., 2016). However, other exercise types such as resistance training may have similar effects and therefore it cannot be ruled out that an effect on ERPs is also present for these types of exercise. We had very broad inclusion criteria in terms of paradigms and ERPs and having instead focused on one or two paradigms and ERPs would have perhaps left less uncertainty in terms of interpretation and may have enabled a meta-analysis. However, by including as many paradigms and ERPs as possible, we will enable researchers in having an overview of those used in exercise research. Thus we here present a wider review compared to the previous systematic review both in terms of population, intervention and ERPs examined (Pedroso et al., 2017).

In conclusion, we found that aerobic exercise, especially acute exercise, affected amplitudes and also to a lesser extent latencies of ERP components. Most studies focused on acute aerobic exercise in healthy participants and future research should focus more on (1) which role acute versus chronic exercise play in regards to ERP amplitudes and latencies, and (2) whether ERP amplitudes and latencies are dependent on exercise intensity. Future studies should focus on comparing these aspects of aerobic exercise directly.

Supplemental Information