Acute exercise performed before and after motor practice enhances the positive effects on motor memory consolidation

Performing a single bout of exercise can enhance motor learning and long-term retention of motor skills. Pa-rameters such as the intensity and when the exercise bout is performed in relation to skill practice (i


Introduction
The effects of acute exercise on learning and memory-related mechanisms have received increased attention in the last decade (Blomstrand & Engvall, 2021).By performing acute exercise before or after a learning session, it is possible to facilitate memory functions across distinct memory systems (Roig et al., 2013).This includes the encoding and consolidation of memory underlying motor skill learning (Wanner et al., 2020).Memory processes cannot be observed directly, but changes in performance can, however, serve as behavioral readouts enabling the quantification of motor memory formation.As such, exercise may improve motor learning by (1) facilitating within-session (or online) improvements in skilled performance that have been related to the effectiveness of memory encoding or (2) improving between-session (or offline) changes that have been proposed to reflect the effectiveness of memory consolidation (Kantak & Winstein, 2012).
Performing an exercise bout immediately after practicing a novel motor task is accompanied by robust effects on motor memory consolidation inferred by enhanced long-term retention of the acquired skill (Wanner et al., 2020).These effects are likely modulated by exercise parameters such as the intensity of the exercise bout and when it is performed (i.e., the timing of the exercise bout) (Roig et al., 2016).For instance, the positive effects of aerobic exercise on offline learning are more pronounced when performed at high intensity compared to moderate intensity (Thomas, Johnsen, et al., 2016).Moreover, the effect diminishes as the temporal gap between the exercise bout and motor practice session increases (Thomas, Beck, et al., 2016).Finally, exercising immediately after motor practicerather than beforeappears to have the most substantial effect on long-term retention (Wanner et al., 2020).This suggests that acute exercise promotes consolidation processes in a timing-and intensity-dependent manner, potentially by helping the transformation of memory traces from a labile state to a stable state that is robust to interference and retained without decay (Beck et al., 2020).
Effects of acute exercise on memory do not solely relate to postpractice interventions targeting consolidation processes.Positive, albeit less robust, effects have also been demonstrated when acute exercise is administered just before motor practice (Perini et al., 2016;Snow et al., 2016;Statton et al., 2015).Performing exercise prior to motor practice may facilitate encoding processes (Roig et al., 2016), quantified as the change in performance across the practice session (Kantak & Winstein, 2012).While some studies have shown immediate performance improvements or enhanced online learning when motor practice is preceded by exercise (Chartrand et al., 2015;Hübner et al., 2018;Moriarty et al., 2022;Perini et al., 2016;Snow et al., 2016;Statton et al., 2015), others have not (Helm et al., 2017;Mang et al., 2016;Singh et al., 2016).Differences between studies in exercise parameters such as intensity may explain ambiguous findings.For instance, studies showing a positive effect of preceding exercise on the change in motor performance during practice have often used moderate rather than highintensity exercise (Moriarty et al., 2022;Perini et al., 2016;Snow et al., 2016;Statton et al., 2015).Indeed, the impact of preceding highintensity exercise on motor performance might be detrimental and potentially mask short-term gains due to contaminant factors (e.g., exercise-induced fatigue and over-arousal) (McMorris et al., 2015).Additionally, moderate-intensity exercise has been shown to be more effective in improving cognitive elements of performance than highintensity exercise (McMorris & Hale, 2012).Thus, moderate-intensity exercise before motor practice may be a more suitable protocol to enhance online learning and the underlying encoding processes.On the contrary, high-intensity exercise performed after motor practice may be more effective in triggering peripheral and neurobiological events that potentially favor the underlying plasticity of motor memory consolidation (Andrews et al., 2020;El-Sayes et al., 2019;Taubert et al., 2015).The notion that different exercise protocols can affect distinct motor memory processes led us to hypothesize that certain exercise protocols performed both before and after motor practice could act in synergy to enhance learning.
In summary, the primary objective of the present study was to assess how different exercise protocols aimed to target memory encoding (i.e., moderate-intensity exercise before motor practice) and consolidation (i.e., high-intensity exercise after motor practice) affect distinct phases and aspects of motor skill learning.Notably, the study is the first to investigate whether acute exercise performed both prior to and following motor practice leads to additional learning improvements.As a secondary aim, we also sought to investigate which components of skill learning were affected by exercise.For this purpose, participants practiced a Sequential Visuomotor Accuracy Tracking (SVAT) task that allowed us to investigate sequence-specific and general learning components via sequential and non-sequential performance tests.More specifically, we assessed whether the effects of exercise were specific to the practiced motor sequence (i.e., motor sequence learning) or whether the effects generalized to non-sequential components (i.e., general motor learning).Effects of acute exercise on online and offline learning were assessed using immediate and long-term retention tests providing behavioral readouts likely reflecting memory encoding and consolidation processes, respectively.We hypothesized that implementing a bout of moderate-intensity exercise prior to motor practice would enhance motor memory encoding (indexed as a greater performance increase from baseline to immediate retention, i.e., online learning), whereas implementing a high-intensity bout after motor practice would promote motor memory consolidation (indexed as a greater or more stable performance from immediate retention to the long-term retention assessed seven days after, i.e., offline learning).Finally, we hypothesized that acute exercise performed both prior to and after motor practice would promote both memory encoding and consolidation and lead to the greatest increase in performance from baseline to seven days after (i.e., total learning).

Participants
Sixty-seven able-bodied male participants between the age of 18 and 35 were recruited to participate in the study.All participants were naïve to the motor task used to investigate learning and memory formation and practiced it with their dominant hand.The Edinburgh Inventory determined the participants' handedness (Oldfield, 1971).Participants were excluded if they had a neurological or psychiatric disorder, injury, or impairments that could affect motor performance or background as professional musicians or competitive gamers.Participants were instructed to avoid intense physical activity 24 h before and two hours after the experimental sessions.They were also instructed not to consume caffeine two hours before the sessions.Participants received written and oral information about the study procedures and gave written informed consent before participation.The study was approved by the ethics committee for the Greater Copenhagen Area (H-7019671).

Experimental design
The participants visited the laboratory for three separate sessions.An overview of the experimental design is shown in Fig. 1A.The first session involved a screening of the participants, questionnaires, and a graded maximal exercise test.At least three days later and no more than three weeks after the first visit, the participants returned to the laboratory for the second session (the main experiment).Participants were randomly assigned to one of four separate groups during the session before practicing the SVAT task.One group performed moderateintensity exercise before practice (PRE MO ), a second group performed high-intensity exercise after practice (POST HI ), a third group exercised at moderate intensity before and high intensity after practice (PRE MO + POST HI ), and a fourth group did not exercise during these periods (CON).Exactly seven days later, participants returned for the third session to perform a retention test, determining the effect of the different exercise interventions on long-term skill retention.The third session also included neuropsychological measures of working memory and sustained attention using the Cambridge Neuropsychological Test Automated Battery (CANTAB) tests.The main experiment and retention test began at the same time of the day (±2 h) to ensure that the circadian rhythm did not influence neuroplasticity and behavioral measures (Salehinejad et al., 2021).
Session 1: Screening In this session, a graded exercise test was performed on a bicycle ergometer (MONARK Ergomedic 839 E) to assess maximal oxygen consumption (VO 2peak ), fitness level, and peak power output (W peak ).The graded exercise test corresponds to protocols used in several previous studies investigating the effects of acute exercise on motor learning (Beck et al., 2020;Roig et al., 2012;Thomas, Beck, et al., 2016;Thomas, Johnsen, et al., 2016).The following procedure was used to obtain these measures.Participants performed a warm-up at 75 W (W) for five minutes and were instructed to maintain a cadence between 70 and 90 revolutions per minute (rpm).Next, the workload was increased to 100 W for three minutes and gradually raised by 50 W every third minute until exhaustion.Heart rate, pulmonary ventilation, oxygen consumption, and respiratory exchange ratio were measured every 15 s using a gas analyzing system (MasterScreen CPXH, Carefusion, Germany).Peak power output was quantified as the highest workload reached during the graded exercise test.This allowed us to titrate the exercise intensity administered during the main experiment for each participant in the intervention groups (PRE MO , POST HI , and PRE MO + POST HI ).
Session 2: Main experiment The main experiment was conducted on the participants' second visit to the laboratory.Participants were required to rest for 30 min upon their arrival.During this time, they filled out a sleep log providing information on the quantity of sleep the night before the main experiment.Then, participants were introduced to the SVAT task.To familiarize them with the task and setup, they first performed a short version of the task, followed by an assessment of their initial level of performance (baseline).Next, two groups (PRE MO and PRE MO + POST HI ) performed a single bout of moderate-intensity exercise (45 % of W peak ) on a bicycle ergometer for 20 min, while the remaining groups (POST HI and CON) rested on the bicycle during this period.Five minutes after exercise or rest, participants began practicing the SVAT task.Motor practice was followed by an immediate retention (IR) test that was used to quantify online learning (IR vs. baseline), mainly reflecting processes related to motor memory encoding (Kantak & Winstein, 2012).Following the IR test, two groups (POST HI and PRE MO + POST HI ) performed a single bout of high-intensity exercise (90 % of W peak ) for 20 min while the remaining groups (PRE MO and CON) rested on the bicycle for this duration.Finally, the participants remained in the laboratory for 30 min before leaving.Thus, we ensured that all participants were subject to the same conditions in the early period before and after the main experiment.
Session 3: Long-term retention Exactly one week after the main experiment, participants returned to the laboratory to complete a 7-day retention (7R) test.The 7R test procedures were comparable to those administered at baseline and IR.Performance changes from IR to the 7R were used to quantify offline learning, reflecting the effectiveness of memory consolidation processes.The net change from baseline to the 7R was used to quantify total learning, including both online and offline learning contributions.All Participants visited the laboratory on three separate occasions involving a screening session (session 1), the main experiment (session 2), and a long-term retention test (session 3).During the main experiment, participants practiced the SVAT task.Exercise intervals prior to motor practice were performed at moderate intensity (45 % of W peak ).Exercise intervals following motor practice were performed at a high intensity (90 % of W peak ).Rest conditions administered before or after motor practice consisted of seated rest on the bicycle.(B) Illustration of SVAT task.Participants controlled a red cursor in the vertical direction by increasing or decreasing the force applied to a load cell with their thumb and index finger.By moving the red cursor up and down, the participants were instructed to track rectangular target boxes, which continuously appeared on the screen one at a time for two seconds.At baseline, immediately after practice, and at the 7-day retention, the participants performed two test blocks containing different target types.One followed a predefined sequential order (S), depicted by the blue arrows, and the other a pseudorandom non-sequential order (N).The order of sequential and non-sequential blocks was counterbalanced between subjects.During the six training blocks, participants only practiced sequential trials.Motor performance was quantified as the percentage of the time spent inside the targets.Online learning was used as a marker of memory encoding, offline learning was used as a marker of memory consolidation, and online + offline learning was used to quantify total learning.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)participants rested for 30 min before the 7R test was performed, and filled out the sleep log during this period.Before being dismissed, standardized neuropsychological tests of Spatial Working Memory and Sustained attention were completed using the CANTAB (Seidler et al., 2012).

Exercise interventions
Each exercise intervention consisted of 20 min of bicycling either before (PRE MO ) or after (POST HI ) motor practice or both (PRE MO + POST HI ).In the main experiment, participants were randomly allocated to the intervention groups (PRE MO , POST HI , or PRE MO + POST HI ) or a resting control group that was to remain seated on the bicycle without exercising.All exercise interventions followed the same procedure: a five-minute warm-up at a workload of 50 W and three 3-min exercise intervals.All three exercise intervals were separated by two minutes of active rest at 50 W and ended with two minutes of active cooldown at the same intensity.A cadence between 70 and 90 rpm was required during the exercise bout.Exercise intervals performed before practicing the SVAT task were completed at moderate intensity (45 % of W peak ), whereas exercise intervals performed after motor practice were completed at a high intensity (90 % of W peak ).Heart rates were measured using chest monitors (Polar) during the exercise intervention.

Sequential visuomotor accuracy tracking (SVAT) task
We used an SVAT task to investigate the effects of acute exercise on motor learning and memory processes.The SVAT task was designed based on visuomotor tasks used in previous studies investigating the effects of acute exercise on motor skill learning and memory formation (Beck et al., 2020;Ostadan et al., 2016;Statton et al., 2015;Thomas, Beck, et al., 2016;Thomas, Johnsen, et al., 2016).These tasks require fast and accurate force control and rely on the refinement of motor execution.In addition, the SVAT task used in this study was further tailored to enable an assessment of sequential and non-sequential skill learning.
In the SVAT task, participants were asked to guide a cursor into rectangular target boxes appearing successively at different vertical positions on a computer monitor in front of them (Fig. 1B).This was achieved by controlling the force applied to a load cell (Dacell, AM210, Dacell Co. LTD) that participants held in a pincer grip between their index finger and thumb on their dominant hand (Fig. 1B).Increasing the pinch force moved the cursor upwards in the vertical direction while lowering the force to resting levels left the cursor at the bottom of the screen.Participants were instructed to keep the cursor inside each target for as long as possible while the target was presented.When the cursor was guided inside a target box, the color of the cursor turned blue, signifying correct positioning.Target boxes were presented one at a time for two seconds and separated by an inter-target interval of 200 ms.Force signals were amplified (x10), low pass filtered (10 Hz), and sampled (1000 Hz) on a PC via a USB6008 DAQ board (National Instruments, Inc).Raw force signals were sampled (1000 Hz) in Spike 2.0 via a 1401 CED DAQ board (Cambridge Electronics Design, Cambridge, UK).Customized software (Python 2.9) was used to run the SVAT task.Motor performance was quantified as the percentage of Time on Target (ToT) requiring participants to move the cursor rapidly between target locations and hold it within each target with high accuracy to improve their score.
Before starting the SVAT task, participants were given standardized verbal and visual instructions.Participants also completed two short versions of the task (25 and 45 s) to familiarize them with the setup.Next, two baseline tests were conducted, each containing 25 targets displayed in approximately one minute.The two baseline tests followed different target orders to distinguish between sequence-specific and general motor learning.Unbeknownst to the participants, one test block repeated five pre-defined target positions sequentially (i.e., sequential blocks), while the other followed a pseudorandomized order (i.e., nonsequential blocks).Test blocks were matched in terms of increasing/ decreasing isometric contraction levels, and target positions were equally distributed across the vertical axis of the screen to ensure that demands on pinch force production and precision were comparable.The sequential and non-sequential blocks were counterbalanced such that half of the participants performed the tests in the order presented in Fig. 1B and vice versa.The two tests were repeated immediately after motor practice and again seven days later and did not provide any augmented feedback besides online visual information on the cursor position.The motor practice consisted of six training blocks (B1-B6) of approximately four minutes, separated by two minutes of rest.Each block included 100 targets that repeated the same five targets as the sequential test blocks.After each training block, participants were given augmented feedback in the form of a score representing the average ToT during the entire training block.
In summary, both the sequential and non-sequential blocks required exquisite force control.Thus, motor performance could be improved through practice by general motor learning components encompassing improvements in speed and/or accuracy of the motor output, also referred to as improved motor execution.However, the sequential element embedded within the sequential test blocks also enabled participants to improve performance via sequence-specific learning.This allowed us to investigate whether potential interactions between acute exercise and motor learning were specific to the practiced motor sequence or generalized across both sequential and non-sequential blocks.

Data analysis and statistics
To address any potential influence from instances where participants encountered grip loss or the need for grip repositioning during the SVAT task, we removed data points (targets) where performance displayed deviations beyond ±2 standard deviations from the mean within each block, accounting for approximately 2.98 % of all targets.The following statistical analysis was performed using R (R Core Team, 2022) with the R-package lme4 (Bates et al., 2015) to fit the ToT data as the dependent variable in a linear mixed effect model with three-way interactions between the independent fixed factors GROUP (4 levels: CON, PRE MO , POST HI , PRE MO + POST HI ), BLOCK (3 levels: baseline, IR, 7R), and TARGET TYPE (2 levels: sequential, non-sequential).Intercepts for each participant and trials within each block were added as a random effect.We ensured normality and homogeneity of the variance of residuals through visual inspection of quantile-quantile and residual plots.Statistical inference was made based on the lmerTest package (Kuznetsova et al., 2017) that provides p-values from linear mixed effect models using the Satterthwaite's degree of freedom method.If significant interaction effects were found, we proceeded with pairwise comparisons using multcomp R-package (Hothorn et al., 2008) and adjusted for multiple comparisons using the Bonferroni method.As detailed in the results section, we observed a significant interaction between GROUP and BLOCK, but not between GROUP, BLOCK, and TARGET TYPE.Therefore, we excluded a separate analysis of specific performance changes in sequential and non-sequential blocks between groups to avoid conducting unsupported pairwise comparisons.More specifically, this means that performance scores on sequential and non-sequential blocks were considered collectively when comparing the effect of different exercise protocols or rest on online (IR vs. baseline), offline (7R vs. IR), and total learning (7R vs. baseline).However, we did contrast performance scores in sequential and non-sequential blocks across groups (i.e., considering the pooled data from all groups collectively) to gain insights into the respective roles of sequence-specific and general skill learning components, as these across-group comparisons were supported by a significant main effect of TARGET TYPE and a BLOCK × TARGET TYPE interaction.Finally, we used one-way ANOVAs to assess between-group differences in participant characteristics and sleep reports.Model estimates derived from the statistical analysis are presented with their respective standard errors and 95 % confidence intervals when appropriate.Raw data in figures and tables are presented as means with standard deviations.Statistical significance was established if p < 0.05 and marked by the asterisk (*) for within-group comparisons or square (#) for between and across-group comparisons in the presented figures and tables.

Results
Sixty-seven participants were recruited for the present study.Of those, one participant did not meet the inclusion criteria (psychiatric disorder), and two did not complete all experimental sessions.Hence, the results are based on data from 64 participants divided into four groups of 16.

Participant characteristics
Table 1 displays participant characteristics and sleep reports, while Table 2 provides an overview of heart rates measured during the different exercise regimes.

Motor skill learning and effects of acute exercise
The linear mixed effects model revealed a significant GROUP × BLOCK (F (6,9230) = 4.15, p < 0.001) interaction, suggesting that the administered exercise conditions had an impact on motor learning.Notably, there was no indication of differences between the sequential or non-sequential blocks in terms of the group, as indicated by the absence of GROUP × TARGET TYPE (F (3,9230) = 0.18, p = 0.913) and GROUP × BLOCK × TARGET TYPE (F (6,9230) = 1.06, p = 0.385) interactions.These findings suggest that the effects of exercise did not exhibit specific interactions with sequential or non-sequential performance improvements but instead influenced general aspects of motor learning spanning across both target types.Given the lack of a significant three-way interaction, we analyzed the sequential and nonsequential blocks collectively when conducting the between-group comparisons based on the GROUP × BLOCK interaction.However, we did discover a main effect of TARGET TYPE (F (1,9230) = 5.30, p = 0.021) and a significant BLOCK × TARGET TYPE (F (2,9230) = 7.38, p < 0.001) interaction irrespective of the group.These findings suggest the presence of sequence-specific performance improvements that were not affected by the exercise conditions.As a result, data from all groups were combined to evaluate sequence-specific and general motor learning.

Online learning
The results of the study are shown in Fig. 2A, where performance scores are presented as mean values for ToT at baseline, motor practice, IR, and 7R.Our linear mixed effect model found no significant differences between the groups during these time points (p-values ranged from 0.686 to 1).However, all groups exhibited significant improvements from baseline to IR (CON: 10.93 ± 0.71 % ToT; PRE MO : 8.69 ± 0.71 %; POST HI : 9.04 ± 0.71 %; PRE MO + POST HI : 10.09 ± 0.71 %; all pvalues < 0.001).There were no significant differences in performance changes between the groups from baseline to IR (p-values ranged from 0.116 to 1), as shown in Fig. 2C: Online learning.Therefore, none of the exercise conditions had a significant impact on motor memory encoding compared to rest.

Offline learning
No significant performance changes were found among the three exercise groups from the IR to 7R test (PRE MO : − 0.10 ± 0.71 %; POST HI : − 0.76 ± 0.71 %; PRE MO + POST HI : 0.22 ± 0.71 %; range of p-values: 0.859-1), as shown in Fig. 2C: Offline learning.This shows that improvements in performance seen after practice were stabilized and maintained over time.On the contrary, a significant drop in performance was observed in the resting CON group (− 3.84 ± 0.71 %, p < 0.001).This indicates that all exercise conditionsbut not restimproved offline learning by preventing performance decay between sessions.Pairwise comparisons between groups further confirmed this: the drop in performance from IR to 7R was significantly greater in the CON compared to PRE MO (− 3.74 ± 1.00 %, 95 % CI: [− 6.32; − 1.16], p = 0.001), POST HI (− 3.08 ± 1.00 %, 95 % CI: [− 5.67; − 0.51], p =
Taken together, all exercise conditions improved offline learning by preventing performance decay between experimental sessions.Additionally, the PRE MO + POST HI groupbut not the remaining exercise groupsshowed greater total learning compared to rest, supporting an enhanced effect on total learning, as indexed from changes from baseline to 7R, of pairing different exercise regimes.

Sequence-specific and general motor learning
The results showed no significant difference in performance scores between sequential and non-sequential blocks at baseline (− 0.34 ± 0.50 %, p = 0.495), indicating that both types of targets were equally demanding before practice.Significant performance improvements were observed for both sequential and non-sequential blocks from baseline to IR (sequential: 10.96 ± 0.50 %, p < 0.0001; non-sequential: 8.40 ± 0.50 %, p < 0.001).Notably, there were significant contrasts between target types at IR, which favored performance scores in sequential blocks (2.22 ± 0.50 %, p < 0.001).These results demonstrate signs of both general and motor sequence learning, as evidenced by the overall improvements across both target types, but with a higher performance score found in sequential blocks at IR.The sequence-specific improvements were more prone to decay.This was confirmed by a significant drop in performance in sequential blocks (-2.16 ± 0.50 %, p < 0.0001) compared to the stable change in performance in nonsequential blocks (-0.07 ± 0.50 %, p = 0.910).Furthermore, there were no significant contrasts between sequential and non-sequential blocks at 7R (0.128 ± 0.50 %, p = 0.7992).Overall, the results suggest that both sequence-specific and general skill learning played a part in improving performance during practice, leading to significant online learning across all groups.However, the significant performance drop in sequential blocks seven days later and the absence of significant contrast between target types at the 7R indicate that only the general components remained stable over time.Finally, it is important to note that we did not observe a significant three-way interaction effect of GROUP × BLOCK × TARGET TYPE.As a result, the beneficial impact of acute exercise on long-term retention was not limited to either sequential or non-sequential blocks.This indicates that exercise contributed to the stabilization of generalizable skill learning components.

Discussion
Effects of acute exercise on motor learning are thought to be timingand intensity-dependent.Here, we investigated the effects of different exercise protocols that we hypothesized would target motor memory encoding and consolidation processes.Corroborating results from previous studies, we found that all exercise conditions positively affected offline learning compared to rest, suggesting a positive effect of exercise on memory consolidation.In addition, positive effects on total learning were seen when exercise was performed before and after motor practice.The latter strengthens the notion that additional motor learning benefits can be achieved by performing multiple exercise bouts in close temporal proximity to motor practice.However, contrary to our initial hypothesis, moderate-intensity exercise performed before practice did not improve online learning.Thus, exercise did not seem to affect the encodingrelated processes of memory formation.Additionally, we found that the effects of exercise on long-term retention were not specific to the sequential element practiced during the training blocks, suggesting that exercise interacted with motor memory processes that manifested regardless of the target positions and orders.Taken together, our main findings support that (1) acute exercise promotes offline learning by counteracting performance decay between sessions, (2) effects on motor learning can be enhanced when exercise is performed before and after motor practice, (3) exercise promotes motor memory processes that generalize across sequential and non-sequential motor patterns.

Acute exercise enhances offline contributions to motor learning
We initially hypothesized that exercising at moderate intensity prior to motor practice would enhance memory encoding, expressed as improved online learning, whereas high-intensity exercise after practice would promote consolidation, expressed as improved offline learning.
These hypotheses were based on results from previous studies demonstrating improved online learning when motor practice was preceded by moderate-intensity exercise (Moriarty et al., 2022;Perini et al., 2016;Statton et al., 2015) and improved offline learning when motor practice was followed by high-intensity exercise (Roig et al., 2012;Thomas, Beck, et al., 2016;Thomas, Johnsen, et al., 2016).In contrast to our initial hypothesis, the present study did not reveal any positive effects of exercise on online learning.However, we did observe beneficial effects across all exercise conditions compared to rest on offline learning.Specifically, exercise counteracted the between-session performance decay seen in resting controls leading to retained performances after seven days.
Studies applying high-intensity exercise before or after motor practice have consistently reported positive effects on offline learning (Wanner et al., 2020).However, high-intensity exercise does not seem to improve online learning when placed before practice (Mang et al., 2016;Roig et al., 2012).This indicates that the physiological influence of preceding exercise may still be present after the termination of practice and impact the processes of memory consolidation (Skriver et al., 2014) but without affecting memory encoding (Roig et al., 2012).Interestingly, a study by Statton et al. (2016) found that online learning of a sequential visuomotor pinch task was enhanced when moderateintensity exercise was performed before motor practice.This could suggest that exercise at lower intensities can effectively promote encoding processes.Recently, this notion was further supported by the finding that piano-playing performance was improved when preceded by moderate but not high-intensity exercise (Moriarty et al., 2022).It is worth noting that this study also discovered a negative relationship between heart rates recorded during high-intensity exercise and subsequent piano-playing performance, indicating that undertaking highintensity exercise before practice may have negative effects.Contrary to these studies, however, the results of the present study did not find a positive effect of moderate-intensity exercise on online learning.These ambiguous findings may be related to differences between the behavioral paradigms that were used to investigate the effects of exercise on motor learning.The effects of acute exercise on online learning have previously been proposed to be attributed to an effect on cognitive rather than motor-related aspects of motor learning (Baird et al., 2018;Neva et al., 2019).As further discussed later, the performance improvements observed across groups in the present study were mostly general, likely reflecting skill learning related to enhanced speed and/or accuracy of the motor output.Therefore, it is possible that the cognitive elements were less likely to influence performance improvements in the present study, causing a lack of interaction between moderate-intensity exercise and online learning.
Another reason for lowering the exercise intensity in the pre-practice exercise conditions was to specifically target the online learning phase and encoding processes (Loprinzi et al., 2021;Roig et al., 2016).In doing so, we tried to ensure that the transient physiological response elicited by the exercise bout would return to baseline before consolidation processes started to evolve to avoid 'carry-over' effects (Skriver et al., 2014).However, in contrast to our assumptions, moderate-intensity exercise prior to motor practice led to similar improvements in offline learning as when practice was followed by high-intensity exercise.Thus, it is possible that the moderate-intensity exercise regime could still influence the early stages of memory consolidation, improving long-term retention to the same degree as the post-practice exercise condition.Since moderate-intensity exercise has been shown to enhance consolidation processes (Holman & Staines, 2021) -although to a lesser extent (Thomas, Johnsen, et al., 2016) -this explanation seems plausible.Alternatively, the preceding exercise may have affected encoding mechanisms without causingor potentially maskingimmediate performance improvements.Indeed, beneficial effects on memory encoding may not be reliably assessed during or immediately after motor practice, where factors such as fatigue and fluctuating levels of attention may contaminate performance measures (Kantak & Winstein, 2012).However, previous studies have shown that the positive effects of acute exercise on retention are not manifested after one hour (Lundbye-Jensen et al., 2017;Roig et al., 2012;Skriver et al., 2014;Thomas et al., 2017).This suggests that the effects of acute exercise on offline learning develop gradually over time and interact through the ongoing processes of memory consolidation that take place in the hours after motor practice has ended (Roig et al., 2016).We cannot ascertain to which extent the preceding exercise bouts promoted motor memory encoding or consolidation.Nevertheless, the temporal placement of the post-practice exercise interventions only allowed for interaction with memory processes taking place after the termination of practice.Thus, it is plausible that the positive impact observed on offline learning can be attributed to the interaction between acute exercise and memory consolidation processes.
We chose an interval of seven days between the main experiment and long-term retention since this has been shown to be effective in capturing the effects of acute exercise on offline learning (Roig et al., 2012).Contrary to previous studies, however, we did not include retention tests between these time points to avoid the potential contribution of re-exposure to the task and subsequent (re-)consolidation, which may further modify the memory trace (Censor et al., 2010).In the meta-analysis by Wanner et al. (2020), all included studies using 7-day retention intervals also assessed motor performance in between via 24hour retention tests.Thus, our results provide evidence to support that the effects of acute exercise on long-term retention are not reliant on intermediate reactivation of the memory engram.

Acute exercise before and after motor practice enhances the motor learning benefits
An important finding of this study was that the double-bout exercise intervention was the only one to produce significant effects on total learning (i.e., 7R vs. baseline) compared to the resting controls.This result is in line with our initial hypothesis, i.e., that exercise before and after practice would lead to the greatest increase in performance from baseline to long-term retention.However, as we did not observe an accumulated effect of exercising before and after motor practice on online and offline learning, respectively, the interpretation of this result is not straightforward.Previous studies investigating the effect of acute exercise on motor learning have analyzed online and offline learning separately (Wanner et al., 2020), but neither provides a comprehensive measurement of the overall improvement in skilled performance over time.Our results indicate that all exercise protocols applied in this study augment offline learning.We did, however, only find evidence for improved total learning when exercise was performed before and after the practice session.This suggests that the positive effects of acute exercise on skill retention can be enhanced when exercise is performed both prior to and following motor practice.
To our knowledge, this is the first study to demonstrate such an effect for a task that engages the motor memory system.In the declarative memory domain, the results of a recent study support these findings.Incorporating acute exercise into multiple learning phases, Loprinzi et al. (2021) showed that exercise improved memory regardless of the temporal placement of the exercise bout.In line with our findings, they also found that the improvement of long-term retention was most effective when acute exercise was administered both before and after information exposure (Loprinzi et al., 2021).This could suggest that exercising before and after a learning task leads to synergistic effects on memory retention.However, since neither this nor the present study controlled for the increased exercise duration caused by performing a second exercise bout, the larger exercise volume may have contributed to the observed effects.Slutsky-Ganesh et al. (2020) accounted for this problem by comparing the effects of 10 min of exercise applied before and after auditory-verbal memory encoding to the effect of exercising 20 min either before or after memory encoding.Pairing shorter exercise conditions targeting both encoding and consolidation processes produced the greatest impact on long-term retention (Slutsky-Ganesh et al., 2020).This may indicate that it is the temporal pairing of exercise before and after practice that acts in synergy to promote memory rather than being exposed to greater exercise volume.However, as we did not control for the potential dosage effect in the present study, further studies are needed to determine whether the enhanced total learning found by exercising both before and after motor practice is related to synergistic or dose-dependent effects.

Acute exercise contributes to the stabilization of generalizable components of motor skill learning
The SVAT task required participants to reduce their transition times between target locations via rapid modulation of force and to sustain hold phases within targets, demanding a high level of discrete precision control to improve performance scores.The SVAT task also contained repeated target sequences, allowing sequence-specific skill learning.Therefore, the observed effects of exercise on long-term retention could be mediated via processes interacting with at least two distinct components of motor skill learning: improved abilities to deliver fast and accurate motor execution (i.e., general learning) and the ability to organize movements into sequences (i.e., motor sequence learning).Improvements in the first component may lead to better performance in both sequential and non-sequential blocks, while the sequence-specific skill component is likely to only apply to the practiced motor sequence (Krakauer et al., 2019).
After practicing the SVAT task for six consecutive training blocks non-sequential motor performance showed significant improvements, indicating general motor learning.Additionally, there was sequencespecific learning, with sequential blocks showing significantly better performance compared to non-sequential blocks immediately after practice.This shows that different skill components contributed to the observed improvements in SVAT task performance during practice.Despite this, the target type did not influence the interaction between acute exercise and motor learning.Consequently, the exercise-induced effects did not appear to favor one target type over the other.This suggests that exercise influenced general aspects of motor memory, such as sustained changes in rapid and precise motor execution, ultimately leading to improvements in both sequential and non-sequential blocks.At first sight, our results contrast with previous findings showing that acute exercise selectively promotes sequential elements of motor skill learning (Mang et al., 2016(Mang et al., , 2014)).It should, however, be noted that these discrepancies may be related to differences in methods used to evaluate the effects of exercise on motor learning.For example, Mang et al. (2016) found improved sequence relearning at the long-term retention test rather than actual retention per se.Additionally, in the present study, we noted that the improvements in sequential blocks during online learning were more susceptible to decay compared to those in non-sequential blocks.Ultimately, no differences in performance were observed between sequential and non-sequential blocks at the long-term retention test, indicating that the sequence-specific improvement decreased over time while the general improvements proved robust.Therefore, it is also possible that the long-term retention test could not capture potential interactions between exercise and sequence-specific memory due to a lack of sensitivity.Nevertheless, our results add to the existing literature by showing that acute exercise can enhance the stabilization of motor skill components that improve both sequential and non-sequential motor performance.Based on these results, we propose that acute exercise promotes general aspects of motor memory, resulting in lasting improvements in the execution of the generalizable components of the learned skill.

Conclusion
Taken together, our results provide behavioral evidence that acute exercise before or after motor practice improves offline learning, leading L. Jespersen et al. to a resistance to performance decay over days.We further demonstrate a positive effect of performing exercise both before and after practice resulting in increased long-term motor learning benefits.These results have implications for the use of acute exercise as an intervention to improve motor learning in various settings, including rehabilitation and sports.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1. (A) Schematic illustration of the study design.Participants visited the laboratory on three separate occasions involving a screening session (session 1), the main experiment (session 2), and a long-term retention test (session 3).During the main experiment, participants practiced the SVAT task.Exercise intervals prior to motor practice were performed at moderate intensity (45 % of W peak ).Exercise intervals following motor practice were performed at a high intensity (90 % of W peak ).Rest conditions administered before or after motor practice consisted of seated rest on the bicycle.(B) Illustration of SVAT task.Participants controlled a red cursor in the vertical direction by increasing or decreasing the force applied to a load cell with their thumb and index finger.By moving the red cursor up and down, the participants were instructed to track rectangular target boxes, which continuously appeared on the screen one at a time for two seconds.At baseline, immediately after practice, and at the 7-day retention, the participants performed two test blocks containing different target types.One followed a predefined sequential order (S), depicted by the blue arrows, and the other a pseudorandom non-sequential order (N).The order of sequential and non-sequential blocks was counterbalanced between subjects.During the six training blocks, participants only practiced sequential trials.Motor performance was quantified as the percentage of the time spent inside the targets.Online learning was used as a marker of memory encoding, offline learning was used as a marker of memory consolidation, and online + offline learning was used to quantify total learning.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 0.012), and PRE MO + POST HI (− 4.06 ± 1.00 %, 95 % CI: [− 6.64; − 1.5], p = <0.001)(Fig. 2C: Offline).No significant differences were observed between exercise conditions (all p-values = 1).

Fig. 2 .
Fig. 2. Effects of exercise and rest on motor skill learning.(A) Mean time on target scores ± standard deviations at baseline, motor practice (B1-B6), immediate retention (IR), and 7-day retention (7R) for all groups.(B) Differences in performance scores between sequential and non-sequential blocks pooled across groups (C) Delta values displaying differences in online learning [IR vs. Baseline], offline learning [7R vs. IR], and total learning [7R vs. Baseline]  between groups.* Significant within-group performance changes (p < 0.05).# Significant between-group differences in change of performance or significant across-group differences in performance scores between sequential and non-sequential blocks (p < 0.05).

Table 1
Participant characteristics and sleep reports.

Table 2
Heart rate data.Mean heart rate data for all participants allocated to the exercise interventions (n = 48).Data are divided into those performing moderate-intensity exercise (PRE MO and PRE MO + POST HI , n = 32) and high-intensity exercise (POST HI and PRE MO + POST HI , n = 32).Data are reported as means ± standard deviations.HR mean ; average heart rates for intervals 1-3, HR intensity ; exercise intensity in percentage based on the heart rate reserve (HR mean − HR rest )/(HR max − HR rest ).