The impact of exercise on food-related inhibitory control — do calories, time of day, and BMI matter? Evidence from an event-related potential (ERP) study

A growing body of research suggests exercise improves inhibitory control functions. We tested if exercise-related inhibitory control benefits extend to food-related inhibitory control and differ by calorie content, time of day, and weight status. One hundred thirty-eight individuals were pseudo-randomly assigned to a morning or evening group. Each subject participated in two lab sessions where they completed questionnaires (rest session) or walked on a treadmill at 3.8mph (exercise session) for 45 min. After each session, participants completed both a high-calorie and low-calorie go/no-go task while N2 and P3 event-related potentials (ERP), both neural indicators of inhibitory control, were measured. Participants also rated food images for valence and arousal. While N2 and P3 difference amplitudes were larger to high-calorie than low-calorie foods, neither exercise nor time of day affected results. Individuals had faster response times after exercise without decreases in accuracy. Arousal and valence for high-calorie foods were lower after exercise and lower for all foods after morning compared to evening exercise. In a subset of individuals with obesity and normal-weight individuals, individuals with obesity had larger N2 difference amplitudes after morning exercise, while normal-weight individuals had larger P3 difference amplitudes to high-calorie foods after exercise. Results suggest moderate exercise did not affect food-related inhibitory control generally, although morning exercise may be beneficial in improving early recruitment of food-related inhibitory control in individuals with obesity. Moderate exercise, particularly in the morning, may also help manage increased attention allocated to food.

Exercise is a promising intervention for enhancing food-related inhibitory control (Lowe et al., 2016).Exercise enhances catecholamine levels, including norepinephrine, improving cognitive function and altering diet regulation through mechanisms such as inducing satiety (Cooper, 1973;Pontifex et al., 2019;Wellman, 2000).Neuroimaging and psychophysiological data also show exercise improves executive functions specifically linked to regulating eating behavior (Joseph et al., 2011), including inhibitory control (Kao et al., 2017).Taken together, it is increasingly clear that the neurophysiological changes induced by exercise lead to improvements in inhibitory control, which can then be linked to diet regulation; however, studies directly testing how exercise targets neural indices of food-related inhibitory control are lacking.
Our goal was to test if acute aerobic exercise affects neural indices of food-related inhibitory control, taking into consideration critical variables that may influence results-calorie content, time of day, and weight status.We assessed neural indices of inhibitory control using eventrelated potentials (ERPs), which directly capture brain electrical activity in response to stimuli (Luck, 2014), reflecting cognitive processes such as attention and inhibitory control (Folstein & Van Petten, 2008;Hajcak et al., 2010;Krolak-Salmon et al., 2001).Two ERP components that reflect inhibitory control during inhibition-type tasks (e.g., go/no-go tasks) are the N2 and P3 ERP components.The N2 is a frontocentral negative-going amplitude occurring 200-300 ms after stimuli, and the inhibitory P3 is a frontocentral positive-going amplitude occurring 300-600 ms after stimuli (Falkenstein et al., 1999;Folstein & Van Petten, 2008).Both ERPs get larger in amplitude when individuals withhold dominant responses, reflecting recruitment of inhibitory control (Folstein & Van Petten, 2008;Wessel & Aron, 2015).The N2 amplitude is larger when withholding responses to food compared to neutral cues (Watson & Garvey, 2013) and both N2 and P3 amplitudes are larger when withholding from high-calorie compared to low-calorie foods (Carbine et al., 2017a(Carbine et al., , 2017b;;2018a, 2018b, 2021, 2021b), supporting the use of the N2 and P3 ERP components as neural indicators of food-related inhibitory control.
We first tested if the impact of acute aerobic exercise on food-related inhibitory control differs towards high-calorie and low-calorie foods, as measured by N2 and P3 ERP amplitudes.Given increased inhibitory control is needed to withhold from high-calorie compared to low-calorie foods (Carbine et al., 2017a(Carbine et al., , 2017b;;2018a, 2018b, 2021, 2021b) and lower inhibitory control is related to increased consumption of foods high in fat and sugar, leading to poor health outcomes (Appelhans et al., 2011;Hall, 2012), it is critical to know if exercise improves inhibitory control specifically to high-calorie foods.Previous research has found vigorous (i.e., 70% VO 2 max) compared to moderate exercise (i.e., 35% VO 2 max) and seated rest results in larger N2 and P3 amplitudes when withholding to images of high-calorie foods during a high-calorie no-go task (Bailey et al., 2021).Although findings show promise that exercise positively impacts neural indices of food-related inhibitory control, results were not compared to a low-calorie go/no-go task (e.g., withhold responses to low-calorie foods).Therefore, it remains unclear if exercise has a differential effect on inhibitory control to high-calorie compared to low-calorie foods.
Our second aim was to test if the impact of exercise on food-related inhibitory control differs by time of day.Decreases in inhibitory control have been observed at nighttime hours or the longer an individual has been awake (Garcia et al., 2011;Harrison et al., 2007;Hartley & Shirley, 1976;Li et al., 2020;Ramirez et al., 2012), although low levels of inhibitory control have also been observed in the early morning (i.e., before 7am; Valdez, 2019).Individuals also have poorer diet self-regulation and increased consumption of unhealthy nutrients (e.g., saturated fats) in the evening (Bouwman, Reinders, Galama, & Verain, 2021;Khare & Inman, 2006).Interventions aimed at improving self-regulation are more successful at reducing food intake in the evening than the morning (Boland et al., 2013), suggesting interventions may be more effective when targeting lower levels of inhibitory control in the evening when the demand for regulating diet is greater.Exercise itself has a greater impact on cognition at lower levels of inhibitory control, as measured by P3 ERP amplitude (Drollette et al., 2014).
Therefore, exercising in the evening might have a stronger effect on food-related inhibitory control, as both inhibitory control and dietary regulation tend to be lower during this time, and cognitive interventions have proven more effective in reducing food intake in the evening compared to the morning (Boland et al., 2013).Multiple studies have tested if morning or evening exercise differentially affect eating behaviors.Research is mixed on whether morning (Alizadeh et al., 2017) or evening exercise (Arciero et al., 2022) increases satiety.While some studies has shown greater reductions in food intake after morning compared to evening exercise (Alizadeh et al., 2017), the majority have found no differences in food intake (Alizadeh et al., 2015;Maraki et al., 2005;McIver et al., 2019;O'Donoghue et al., 2010) or appetite (Alizadeh et al., 2017;Maraki et al., 2005;McIver et al., 2019;Mode et al., 2023) after morning or evening exercise.Given food intake and appetite tend to increase in the evening (Kant, 2018;Smith & Betts, 2022), the lack of differences after morning or evening exercise provide evidence that evening exercise helped reduce the increased food intake and appetite levels often seen in the evening to the lower levels often observed in the morning.One possible reason evening exercise was more effective was because it had a greater impact on the lower levels of inhibitory control in the evening than the morning, leading to greater improvements in evening diet behaviors (Boland et al., 2013;Drollette et al., 2014, Guerrieri et la., 2007;Li et al., 2020).
Our final aim was to test if the impact of exercise on food-related inhibitory control differs between individuals with obesity and normal-weight individuals.Considering increased inhibitory control is associated with lower risk for future weight gain (Anzman & Birch, 2009;Nederkoorn et al., 2010) and increased weight loss in individuals with obesity (Lavagnino et al., 2016;McCaffery et al., 2009), an intervention that improves inhibitory control in individuals with obesity would have a substantial positive impact on their physical health.Despite occasional inconsistencies (Carbine et al., 2018a(Carbine et al., , 2018b)), data from multiple sources indicate that individuals with obesity generally exhibit lower inhibitory control (Appelhans et al., 2011;Spitoni et al., 2017), highlighting the importance of enhancing inhibitory control in this population.Given that exercise tends to yield greater improvements in those with lower inhibitory control (Drollette et al., 2014), it is plausible that individuals with obesity derive greater neurocognitive benefits from exercise compared to normal-weight individuals.It is important to note that a recent study found no difference in no-go accuracy after rest or exercise in overweight to obese individuals during a high-calorie go/no-go task (Flack et al., 2022).However, with exercise increasing N2 and P3 ERP amplitudes during a high-calorie go/no-go task (Bailey et al., 2021) and increasing P3 amplitude in individuals with lower levels of inhibitory control (Drollette et al., 2014), it is plausible at the neural level that individuals with obesity would demonstrate increased N2 and P3 amplitude after exercise, given their lower levels of inhibitory control.
In sum, we tested if exercise affects neural indices of food-related inhibitory control while taking into consideration calorie content, time of day, and BMI.We first hypothesized that exercise would enhance inhibitory control towards high-calorie foods, particularly in the evening, as indicated by larger N2 and P3 ERP amplitudes compared to lowcalorie foods and morning sessions.Our second hypothesis was that there would be lower levels of food-related inhibitory control in the evening, indicated by smaller N2 and P3 ERP difference amplitudes during evening compared to morning rest sessions.Our third hypothesis was that evening exercise specifically would have a greater impact on food-related inhibitory control, indicated by a larger increase in N2 and P3 ERP difference amplitudes after evening than morning exercise.Our fourth hypothesis was exercise would improve food-related inhibitory control for individuals with obesity, particularly for high-calorie foods in the evening, indicated by larger N2 and P3 ERP difference amplitudes in individuals with obesity compared to normal-weight individuals.

Participants
The study was approved by Brigham Young University's IRB and is accordance with the Declaration of Helsinki.All participants provided written informed consent at the start of their first lab session.Participants were recruited from the university and local community via flyers and the university research participant system and compensated via course credit or monetary payment ($70).Eligibility for study participation was assessed via a pre-enrollment phone call.Inclusion criteria included being between ages 18 and 54 and able to walk comfortably on a treadmill for 45 min at a moderate pace.Potential participants were excluded if they experienced a previous head injury that resulted in a loss of consciousness, were pregnant or lactating, were postmenopausal, had a diagnosed learning disability, exercised more than three times a week for 30 min of vigorous activity, or had a diagnosed chronic, metabolic, or psychiatric condition.The Physical Activity Readiness Questionnaire was also administered to all participants during screening to ensure physical ability to do the exercise manipulation.If individuals answered 'yes' to any questions, they were excluded from participating.
One hundred eighty-one participants reported they met inclusion criteria after screening and were initially enrolled in the study.After enrollment, 14 participants were subsequently excluded: three for not meeting pre-study requirements, two for not completing the exercise session, and one due to a technical error.Of the 161 remaining participants, 14 were excluded during artifact correction processes and 9 were excluded for not having enough ERP trials to obtain sufficient reliability (see Section 2.5).Thus, the final sample consisted of 138 participants (M age = 25.67,SD age = 8.80; 54.35% female [biological]; see Table 1 for additional demographics).Ninety percent (N = 124) of participants selfreported a stable weight ( ± 5 lbs) over the past month.Participants were pseudo-randomly assigned to either the morning or evening group in a counterbalanced manner, ensuring sample sizes were relatively equal between morning and evening conditions.As such, 71 (51.45%) participants were assigned to the morning group and 67 (48.55%) participants were assigned to the evening group.Groups did not significantly differ in number of biological males/females, age, weight, or BMI (see Table 1).

General procedures
All protocols, materials, data, and supplementary material are posted to the Open Science Framework: https://osf.io/stpxq/.A flowchart depicting study procedures can be found in Fig. 1.Participants attended two lab sessions approximately one week apart (M Days = 7.09, SD Days = 0.68), either in the morning (between 7 a.m. and 10 a.m.) or in the evening (between 7 p.m. and 10 p.m.), depending on their group assignment.Before each session, participants were required to get at least 7 h of sleep the night before, stop eating the night before (by 9pm) for morning sessions or 4 h before for evening sessions (water ok), and refrain from vigorous physical activity or consuming caffeine 24 h before (Carbine et al., 2017a(Carbine et al., , 2017b;;2018a, 2018b).The fasting requirement was used to control for hunger between groups and sessions and observe neural activity at a time when participants would be about to consume their next meal (Carbine et al., 2017a(Carbine et al., , 2017b;;2018a, 2018b).
Upon arrival to the lab for both sessions, participants confirmed they had met the sleep, exercise, fasting, and caffeine requirements and consent was obtained at the first session.At both sessions, participants were then given both a wrist and waist accelerometer to wear for the next 24 h.During the first session, participants recorded their food intake using the online ASA24 (Subar et al., 2012).Participants subsequently recorded their food intake for four additional days.Participants then completed their rest or exercise portion of the study (see Section 2.2.1).
After completing the rest or exercise protocol, participants were immediately escorted into the laboratory's EEG room (they were offered water and a towel to wipe any sweat after the exercise session).Participants then completed sleep and hunger VAS while they were fitted with an EEG net and impedances reduced (see Table 2; more details on the VAS in supplementary material: https://osf.io/stpxq/).EEG data were then collected (see Section 2.5) while participants completed a high-calorie and low-calorie food go/no-go task in a counterbalanced order (see Section 2.3).Time from finishing the rest or exercise protocol to starting the go/no-go tasks was approximately 15 min.Following completion of the go/no-go tasks, participants completed a ratings task at the end of each session (see Section 2.4), which concluded their study participation.Participants were then compensated with course credit if recruited through the university research participant system or monetary payment if from the community.

Rest and exercise protocols
Participants completed both a rest and exercise session in random order.During the rest session, height and weight measurements were taken using a Detecto (Webb City, MO) physician scale.Participants then completed a demographics questionnaire followed by the Depression Anxiety Stress Scale, Dutch Eating Behavior Questionnaire, and General Foods Craving Questionnaire-State.As a note, the accelerometer data, ASA24 data, and questionnaire data were not analyzed in the present study but are mentioned for sake of transparency and will be reported elsewhere.For the exercise session, participants walked on a treadmill at 3.8 miles per hour at a 0% incline for 45 min.According to the Compendium of Physical Activities (Ainsworth et al., 2011) a 3.8 mile per hour pace with no incline would have a MET between 4.3 and 5, classifying it as moderate intensity exercise (Bull et al., 2020).A moderate intensity was chosen given moderate exercise seems to have the greatest impact higher-order cognitive functions (Chang & Etnier, 2009) and increase N2 and P3 ERP amplitudes (Kamijo et al., 2004;Kao et al., 2017).For the exercise session, participants were fitted with a chest heart rate monitor and companion watch (Polar RS100, Kempele, Finland) to record max heart rate, average heart rate, and calories burned (see Table 2).

Food go/no-go tasks
Food go/no-go tasks were presented via E-prime 2.0 (Psychology Software Tools, Inc., 2012) and identical to food go/no-go tasks used in previous studies (Carbine et al., 2017a(Carbine et al., , 2017b;;2018a, 2018b).In both tasks, participants were shown images of high-calorie and low-calorie foods, all presented on a white background.During the high-calorie task, participants were instructed to respond with a button press whenever they saw a low-calorie food (go trials) and withhold their response when they saw a high-calorie food (no-go trials).For the low-calorie task, participants were instructed to respond with a button press whenever they saw a high-calorie food (go trials) and withhold their response when they saw a low-calorie food (no-go trials).For each task, participants completed two blocks of 100 trials each.70% of trials were go and 30% were no-go, randomly presented, as is common with go/no-go tasks to establish a prepotent response (e.g., Benikos et al., 2013;Ramos-Loyo et al., 2013).All images were presented on the screen for 500 ms with a 1200-1400 ms interstimulus interval where a black fixation cross was presented on a white screen.Each task took about 7 min to complete.Thirty-eight high-calorie and 38 low-calorie images from Killgore et al. (2003) were used for the tasks (details on images and how they were chosen can be found in supplementary materials: htt ps://osf.io/stpxq/).

Ratings task
Following completion of the EEG protocol for both the rest and exercise sessions, participants completed a ratings task in which they viewed the 38 high-calorie food images, 38 low-calorie food images, and  38 neutral (i.e., non-food) flower images (also from Killgore et al., 2003).Participants were first instructed to indicate if the presented image was a high-calorie food, low-calorie food, or non-food image to confirm they accurately classified the high-and low-calorie foods presented during the go/no-go tasks.Participants then rated each picture for arousal (e.g., engagement, excitement) and valence (e.g., pleasantness) using the SAM.Specifically, using the SAM, valence was rated on a scale from "1" (unpleasant) to "9" (pleasant) and arousal was rated on a scale from "1" (calm) to "9" (excited) (Cuthbert et al., 2000;Lang et al., 1997, pp. 39-58).Arousal and valence are commonly used to measure emotional responses to pictorial cues and the SAM specifically has been validated across multiple populations as a measure of arousal and valence (Bradley & Lang, 1994).The SAM is commonly used in psychophysiology research to measure emotionally relevant stimuli (Amrhein et al., 2004;Cuthbert et al., 2000;Foti et al., 2009).

EEG data acquisition and analyses
EEG was recorded with a 128-channel Electrical Geodesic, Inc. EEG sensor net (equidistant passive Ag/AgCl electrodes) and amplifier system (NA 300; 20K nominal gain, bandpass = 0.01-100 Hz).Online, data were digitized continuously at 250 Hz with electrode impedance maintained at ≤ 50 kΩ, per manufacturer recommendations, and referenced to the vertex electrode (Cz).Offline, EEG data were first digitally high-pass filtered with a first-order 0.1 Hz filter, digitally lowpass filtered at 30 Hz (2 Hz roll off; FIR) and segmented into epochs from 200 ms before to 1000 ms after stimulus onset using NetStation (version 4.5.7).Data was then exported to the ERP PCA Toolkit for the remaining of data reduction (Dien, 2010).Channels where the fast average amplitude exceeded 100 μV (μV) or if the differential average amplitude exceeded 50 μV were identified as bad and corrected via interpolation using the nearest six neighbor electrodes (Dien, 2010).Participant eyeblinks were then corrected using an independent component analysis (Dien, 2010).Components were removed if they correlated at 0.90 or higher with either a template created from previous food go/no-go data or a one provided by the toolkit.Finally, data were re-referenced offline to an average reference.Baseline adjustment was completed using the 200ms window before stimulus onset (Carbine et al., 2017a(Carbine et al., , 2017b;;Carbine, Duraccio, et al., 2018).As decided a priori, N2 amplitude was quantified as the mean amplitude between 200 and 300 ms post-stimulus, averaged over four fronto-central electrodes (6, 7, 106, 129;Carbine et al., 2017aCarbine et al., , 2017b;;2018a, 2018b).Using a collapsed localizer approach (Luck & Gaspelin, 2017), P3 amplitude was quantified as the mean amplitude between 400 and 550 ms post-stimulus, averaged over the same four fronto-central electrodes.
To ensure participants had enough trials to produce a reliable signal, ERP data were processed using the ERP reliability Analysis Toolbox v.0.5.3 (Clayson & Miller, 2017a), which has successfully been used to test the reliability of ERPs to food stimuli (Carbine, Clayson, et al., 2021).It was chosen a priori that each participant would have enough trials to obtain a dependability estimate (the generalizability theory analogue of reliability; Baldwin et al., 2015) of 0.70 or better to be included in analyses (Clayson & Miller, 2017b).Overall dependability estimates were then calculated for each session, task, and trial type combination.Final dependability estimates, 95% credible intervals, minimum required trials, average and range of trials, and noise values for each combination are presented in Table 4 for the N2 component and Table 5 for the P3 component.Briefly, 14 participants were excluded from analyses for erroring out during artifact correction and nine participants for not having enough trials to produce a reliable N2 or P3 signal.Overall dependability estimates were excellent (>0.90) and all trial types had at least 12 trials.

Statistical analyses
First, we tested for any group differences in confounding variables that might affect results, namely sleep quality, hunger levels, maximum heart rate, average heart rate, and calories burned during exercise.Details and results of these analyses are found in supplementary material (https://osf.io/stpxq/).Briefly, there were no differences in sleep Note.Accuracy is the average percentage of correctly identifying a high-and low-calorie food.quality or hunger ratings between groups or sessions.Similarly, there were no differences in maximum heart rate, average heart rate, or calories burned between groups.We also confirmed our tasks worked as expected in eliciting an inhibitory response, with increased error rates, N2 amplitude, and P3 amplitude on no-go than go trials (see supplementary material), which also supported the use of a difference wave (no-go minus go amplitude) in all subsequent ERP analyses (Clayson et al., 2021).There were nine main analyses, decided a priori before analyzing any data, to test our hypotheses and get a complete picture of how exercise and time of day affected food-related inhibitory control. 2 As analyses were decided a priori, no p-value corrections were implemented.To test our first three hypotheses regarding the impact of exercise and time of day on inhibitory control towards high-and low-calorie foods, we conducted two 2-Group (Morning, Evening) by 2-Session (Exercise, Rest) by 2-Task (High-Calorie, Low-Calorie) repeated measures ANOVAs on N2 and P3 ERP difference amplitudes.
To test our fourth hypothesis if exercise or time of day effects depended on BMI status, we tested a sub-sample of our participants and only included individuals if they were individuals with obesity (n = 44; BMI >30 kg/m 2 ) or normal-weight individuals (n = 44; BMI <25 kg/ m 2 ).Participants in the normal-weight and obesity groups were matched for biological sex and age within their time-of-day group.For these analyses, we conducted two 2-BMI (Normal-Weight, Obese) by 2-Group (Morning, Evening) by 2-Session (Rest, Exercise) by 2-Task (High-Calorie, Low-Calorie) repeated measures ANOVA on N2 and P3 ERP difference amplitudes.
We also conducted five additional a priori analyses for ratings and behavioral data: two 2-Group (Morning, Evening) by 2-Session (Exercise, Rest) by 2-Food (High-Calorie, Low-Calorie) repeated measures ANOVA on arousal and valence ratings and three 2-Group (Morning, Evening) by 2-Session (Exercise, Rest) by 2-Task (High-Calorie, Low-Calorie) repeated measures ANOVAs on correct go response times (assessed as the median response time, to not be influenced by outliers), go trial accuracy, and no-go trial accuracy.For all analyses, significant two-way interactions were decomposed using independent and paired samples t-tests.Significant three-way interactions were decomposed using two-way ANVOAs.Partial eta squared (η p 2 ), Cohen's d for between subjects, and Cohen's d z for within subjects are reported for effect sizes.
To know what size of an effect we are powered to detect for each main analysis, sensitivity analyses were conducted using G*Power v3.1.9.6 (University of Dusseldorf; Faul et al., 2009).Full details are reported in the supplementary materials (https://osf.io/stpxq/);all analyses were powered at 80% to detect a medium sized effect (Cohen's f = 0.24-0.31;Selya et al., 2012).

Hypothesis one-the impact of exercise
Means and standard deviations for ERP difference amplitudes can be found in Table 6 . 3See Fig. 2 for waveform images.Both the N2 difference amplitude, F(1,136) = 64.82,p < 0.001, η p 2 = 0.26, and the P3 difference amplitude, F(1,136) = 43.56,p < 0.001, η p 2 = 0.24, showed a main effect of task, with difference amplitudes being larger for the highcalorie task than the low-calorie task.However, the main effect of exercise session was not significant for the N2 difference amplitude nor the P3 difference amplitude.Similarly, the Session by Task interaction was not significant for the N2 difference amplitude nor the P3 difference amplitude.

Hypothesis two-the impact of time of day
The main effect of time of day group was not significant for N2 difference amplitude nor the P3 difference amplitude.Similarly, the Group by Task interaction was not significant for the N2 difference amplitude nor the P3 difference amplitude.

Hypothesis three-the interaction between exercise and time of day
The Group by Session interaction was not significant for N2 difference amplitude nor P3 difference amplitude.Similarly, the Group by Session by Task interaction was not significant for N2 difference amplitude nor P3 difference amplitude.Note.μv = microvolts.
2 Due to technological errors, some participants were missing for analyses: 14 for behavioral data, 14 for ratings, and 16 for heart rate monitor data.Due to protocol error, 17 participants were missing sleep VAS and 14 hunger VAS.
3 10 participants requested to walk slower than 3.8mph after starting the exercise intervention.We removed those 10 participants and ran the N2 and P3 analyses again, but the pattern of results did not change.

N2
Means and standard deviations for ERP difference waves by BMI group can be found in Table 7. See Fig. 3 for waveform images by BMI group.There was a significant main effect of task, F(1,84) = 37.46, p < 0.001, η p 2 = 0.31, with larger N2 difference amplitude for the highcalorie task compared to the low-calorie task.There was also a significant BMI by Group by Session interaction, F(1,84) = 4.51, p = 0.04, η p 2 = 0.05.No other main effects or interactions were significant, Fs(1,84) < 3.71, ps > 0.06, η p 2 s < 0.04.The BMI by Group by Session interaction was followed up by a 2-BMI (Normal-Weight, Obese) by 2-Session (Rest, Exercise) repeated measures ANOVAs conducted separately for the morning and evening groups (collapsed across task) to break down the significant three-way interaction.N2 difference amplitude after exercise compared to rest only in the morning, but not evening.There were no time of day or exercise effects for normal-weight individuals (see Fig. 4).

P3
There was a main effect of task, F(1,84) = 27.40,p < 0.001, η p 2 = 0.25, with larger P3 difference amplitude for the high-calorie task compared to the low-calorie task.There was also a significant BMI by Session by Task interaction, F(1,84) = 4.15, p = 0.05, η p 2 = 0.05.No other main effects or interactions were significant, Fs(1,84) < 3.41, ps > 0.07, η p 2 s < 0.04.The BMI by Session by Task interaction was followed up by a 2-BMI (Normal-Weight, Obese) by 2-Session (Rest, Exercise) repeated measures ANOVAs conducted separately for each task (collapsed across morning and evening groups) to break down the significant three-way interaction.

Low-calorie task.
For the low-calorie task, the BMI by Session ANOVA revealed no significant main effects or interaction, Fs(1,86) < 1.77, ps > 0.19, η p 2 s < 0.02.In sum, normal-weight individuals had a larger P3 difference amplitude after exercise compared to rest only on the high-calorie task, not low-calorie task.There no task or exercise effects in individuals with obesity (see Fig. 5).Fig. 3. BMI group N2 and P3 no-go amplitudes by task and session for A) normal-weight morning group, B) normal-weight evening group, C) obesity morning group, and D) obesity evening group.
No other main effects or interactions were significant, Fs(1,122) < 3.10, ps > 0.08, η p 2 s < 0.03.For the Group by Session interaction, valence ratings to food were lower after exercise compared to rest for the morning group, t(63) = 1.99, p = 0.05, d z = 0.25, but showed the opposite effect in the evening group and were lower after rest compared to exercise, t(59) = 2.53, p = 0.01, d z = 0.33.In addition, valence ratings were higher in the morning compared to the evening group after the rest session, t(129) = 2.06, p = 0.04, d = 0.36, but did not differ after the exercise session, t(128) = 0.28, p = 0.78, d = 0.05.For the Session by Food interaction, high-calorie foods had higher valence ratings than low-calorie foods after the rest session, t(130) = 3.10, p = 0.002, d z = 0.27, but not after the exercise session, t(129) = 1.63, p = 0.11, d z = 0.14.In sum, high-calorie foods elicited higher arousal and valence ratings than low-calorie foods.Arousal and valence to high-calorie foods specifically were lower after exercising.Arousal ratings decreased after exercising for those who exercised in the morning compared to those who exercised in the evening.

Correct go response times
Means and standard deviations for response times and accuracy can be found in Table 8.There was a main effect of exercise, F(1,122) = 4.12, p = 0.05, η p 2 = 0.03, with faster response times after exercise than rest.
There was also a main effect of task, F(1,122) = 39.14, p < 0.001, η p 2 = 0.24, with faster response times on the high-calorie task (when participants respond to low-calorie foods) than the low-calorie task.No other main effects or interactions were significant, Fs(1,122) < 2.49, ps > 0.12, η p 2 s < 0.02.

No-go accuracy
There was a significant main effect of task, F(1,122) = 7.70, p = 0.01, η p 2 = 0.06, with higher accuracy on the high-calorie task (withholding to high-calorie foods) than the low-calorie task.No other main effects or interactions were significant, Fs(1,122) < 3.19, ps > 0.08, η p 2 s < 0.03.In summary, for behavioral data, participants were faster for go trials and more accurate for both go and no-go trials on the high-calorie task than low-calorie task.Individuals were also faster and more accurate (particularly on the high-calorie task) after exercising.There were no significant effects due to time of day.

Discussion
We tested the impact of morning or evening exercise on food-related inhibitory control, specifically to high-calorie foods and in individuals with obesity.Our first three hypotheses that exercise would improve inhibitory control to high-calorie foods, that food-related inhibitory control would be lower in the evenings, and that evening exercise would have a greater impact on food-related inhibitory control were not supported.Neither time of day nor moderate exercise affected food-related inhibitory control as measured by N2 and P3 ERP amplitudes.The only significant N2 and P3 effects were larger difference amplitudes when withholding to high-than to low-calorie foods, replicating previous research (Carbine et al., 2017a(Carbine et al., , 2017b;;2018a, 2018b).Taken with our arousal and valence ratings, which were also elevated when viewing high-compared to low-calorie foods, results support that more inhibitory control is needed to withhold to high-compared to low-calorie foods, as high-calorie foods are more appealing, engaging, and appetizing in nature (Appelhans et al., 2011;Hall, 2012).
Our results surprising, as previous studies suggest moderate exercise, which was used in the current study, is the most beneficial for higher-order cognitive functions (Chang & Etnier, 2009).N2 and P3 amplitudes have also been found to be larger after moderate exercise (defined as 60-70% HRmax or 12-14 on Borge RPE scale) and P3 amplitudes smaller after vigorous exercise (defined as 90% HRmax or 20 on Borge RPE scale; Kao et al., 2017;Kamijo et al., 2004).One reason our results may differ is because we examined inhibitory control to food-related stimuli as opposed to generic inhibitory control.Both N2 and P3 amplitudes are larger when withholding to food compared to non-food stimuli (Carbine, Muir, et al., 2021;Watson & Garvey, 2013), suggesting food requires increased recruitment of inhibitory control at a neural level.Given this larger inhibitory response to food, a larger exercise dose may be required to elicit changes in food-related inhibitory control, potentially in the form of more intense exercise.In support of this, Bailey et al. (2021), found it was vigorous (70% VO 2 max), not moderate (35% VO 2 max), intensity exercise that affected N2 and P3 amplitudes during a high-calorie go/no-go task.N2 amplitude was larger when responding to low-calorie foods (i.e., go trials), withholding to high-calorie foods (i.e., no-go trials), and P3 amplitude was larger when withholding to high-calorie foods (i.e., no-go trials) after vigorous compared to moderate exercise and rest.Importantly, there were no differences in ERP amplitudes between moderate exercise and rest, which our results replicate and extend to a low-calorie go/no-go task.Taken together, results suggest vigorous-intensity exercise plays a role in affecting food-related inhibitory control, among other factors.Research testing how vigorous exercise affects inhibitory control to both high-and low-calorie foods and if this differs by time of day is needed.
Research examining the impact of exercise on catecholamines (i.e., norepinephrine, epinephrine, and dopamine) also supports that higherintensity exercise relative to other less intense forms of exercise are be needed to alter food-related inhibitory control.As mentioned previously, exercise leads to an increase in the concentration, absorption, or regulation of catecholamines, leading to improvements in cognition after exercise (Cooper, 1973;Pontifex et al., 2019) and potentially diet regulation, as norepinephrine activates the alpha(1)-adrenoceptors in the hypothalamic paraventricular nucleus, which induces satiety (Wellman, 2000).It is important to note that the levels of catecholamines rise as the intensity of exercise increases (Zouhal et al., 2008).Therefore, increasing norepinephrine levels, which can be done by exercising at a higher intensity, may lead to decreases in food intake via activation of alpha(1)-adrenoceptors.As such, it seems that higher intensity exercise, not low intensity exercise, is needed to see changes in food-related cognition, as it leads to the necessary increase in catecholamines.
Although increasing exercise intensity is one way to improve foodrelated inhibitory control, it might not always be feasible.In 2020, 46.9% of adults in the United States met the Centers of Disease Control and Prevention guidelines of getting 150-300 min of moderate intensity and 75-150 min of vigorous intensity exercise a week (Elgaddal et al., 2022).These statistics suggest there is a need to increase the participation and motivation to engage in vigorous exercise before we can use it as intervention to improve food-related inhibitory control (Ryan et al., 2009;Teixeira et al., 2012).Further, some individuals due to physical limitations are not be able to complete vigorous intensity exercise.As such, other interventions beyond vigorous exercise should be considered to target and improve food-related inhibitory control.For example, it is possible inhibitory control training could improve food-related inhibitory control (Klerk et al., 2023;Veling et al., 2020); however, some research suggests inhibitory control training does not affect food-related inhibitory control (Carbine & Larson, 2019;2021, 2021b) and instead targets the value of food (Carbine, Muir, et al., 2021;Veling et al., 2022).
Although exercise did not affect ERPs, results from arousal, valence, and behavioral data suggest exercise still had a positive effect on cognition, just not inhibitory control.No-go accuracy (a behavioral indicator of inhibitory control) was not affected by exercise, supporting the ERP results.However, arousal and valence to high-calorie foods decreased after exercise, and exercise in the morning resulted in decreased arousal and valence for high-and low-calorie foods, suggesting moderate exercise decreased motivational attention to food, particularly in the morning.Further, participants had faster go-trial response times on both tasks after exercise in addition to higher gotrial accuracy on the high-calorie task (i.e., responding to low-calorie foods) instead of experiencing a speed-accuracy trade-off.The decrease in response time without a subsequent decrease in accuracy suggests exercise helped improve attention and processing speed during both tasks.Results are supported by previous research that show attention as measured by processing speed and accuracy improve after exercise, regardless of time of day (Maeneja et al., 2022).Together, results suggest moderate exercise positively affects attention, motivation, and processing speed related to food, but not food-related inhibitory control.
Our final hypothesis that exercise, particularly in the evening, would be more beneficial for individuals with obesity was partially supported.Morning, not evening, exercise increased N2 difference amplitude to food in individuals with obesity but not normal-weight individuals.As individuals with obesity exhibit deficits in general inhibitory control (Spitoni et al., 2017), leading to increases in food intake and difficulties losing weight (McCaffery et al., 2009), morning moderate exercise may be an effective way to increase recruitment of inhibitory control and help manage food intake.Our BMI effect also supports that exercise is more beneficial for individuals with lower levels of inhibitory control (Drollette et al., 2014).
Interestingly, individuals with normal-weight, but not individuals with obesity, showed larger P3 difference amplitudes to high-calorie foods after exercise, regardless of time of day.While P3 amplitude on a go/no-go task is believed to reflect inhibitory control processes, the exact aspect of inhibitory control is debated.P3 amplitude on a go/nogo task can reflect the suppression of attention to non-relevant task information (Polich, 2007), the evaluation of an inhibitory response (Huster et al., 2013), or the withholding of a motor response (Gajewski & Falkenstein, 2013).While morning exercise improves initial recruitment of inhibitory control in individuals with obesity, exercise in normal-weight individuals seems to target later inhibitory control processes involved in attention or motor responses.Examining if later inhibitory control processes are as relevant as initial recruitment of inhibitory control resources in regulating diet and weight will be important to fully understand how exercise impacts inhibitory control.
As with any study, there are weaknesses that should be noted.First, although our exercise intervention met the definition of moderate exercise (MET between 4.3 and 5; Ainsworth et al., 2011;Bull et al., 2020) and was consistent across participants, there are multiple ways to define moderate exercise (Pontifex et al., 2019), many of which are individualized to fitness levels (e.g., VO 2 max, heart rate reserve).Importantly, the perception of what constitutes moderate intensity effort can vary among participants due to stress, fatigue, effort, and discomfort.(Pontifex et al., 2019).Although we ensured all participants were not avid exercisers and that heart rate variables were equal across groups, setting a moderate intensity based on individual fitness levels and adding an individually assessed exertion level scale (i.e., the Borg RPE scale; Borg, 1998) would be beneficial to ensure consistency of exercise level among participants.Second, our sensitivity analyses indicated that we were powered to detect only moderate effects, indicating a small effect potentially could have been present and possibly missed, particularly when testing a three-way interaction.However, we feel a moderate effect size for an intervention study such as this is more practical and externally valid, as it would not be feasible to recommend or expect individuals to complete an exercise intervention for only a small effect.Third, not all participants self-reported being weight stable ( ± 5 lbs) over the past month.Although this was a small percentage (10%), weight loss has been associated with changes in inhibitory control (Lavagnino et al., 2016;McCaffery et al., 2009), and this this variable may wish to be controlled for in future research.Finally, we did not assess menstrual cycle phase in our female participants.Although the relationship between menstrual cycle and food-related inhibitory control specifically has not been tested, generic inhibitory control may be lower during the follicular phase of the menstrual cycle (Hidalgo-Lopez & Pletzer, 2019;Colzato, Hertsig, van den Wildenberg, & Hommel, 2010).Increased activation to food images in brain regions associated with inhibitory control (e.g., inferior frontal gyrus, nucleus accumbens, amygdala) have also been observed during the late follicular phase (Alonso-Alonso et al., 2011;Frank et al., 2010).Therefore, it is possible menstrual cycle phase could affect food-related inhibitory control specifically and should be investigated in future research.
Our study also had multiple strengths.Confounding variables (e.g., sleep quality, hunger levels, demographics, heart rate during exercise) were consistent across sessions and groups, making sure any significant results were due to time of day or exercise.For BMI analyses, groups were also matched for key variables (e.g., gender, age, exercise level) to ensure effects were due to BMI.The use of both high-calorie and lowcalorie go/no-go tasks also allowed us to test if effects were specific to certain types of food, which is lacking in food-related cognition research (Carbine et al., 2018a(Carbine et al., , 2018b)).Finally, the use of ERPs was a significant strength, as the ERPs not only demonstrated exceptional reliability, but were an objective measure of food-related cognition (Carbine et al., 2017a(Carbine et al., , 2017b) that is not dependent on blood flow, such as other neuroimaging methods, which can be affected by exercise (Joris et al., 2018;Ogoh & Ainslie, 2009).
In conclusion, findings suggest acute bouts of morning or evening moderate exercise do not affect food-related inhibitory control, as measured by N2 and P3 ERP amplitudes.Vigorous intensity exercise or non-exercise interventions may be needed to affect food-related inhibitory control.Instead, moderate exercise, especially in the morning, is helpful in managing the increased attention allocated towards food, as indicated by valence, arousal, and behavioral data.Finally, while individuals with normal-weight showed increases in later inhibitory control processes after exercise to high-calorie foods, individuals with obesity exhibited increased early recruitment of food-related inhibitory control after morning exercise.This effect has important implications for long-term improvements in physical and cognitive health.
K.A.Carbine et al.

Fig. 2 .
Fig.2.N2 and P3 amplitudes by task and session for A) morning no-go amplitudes, B) evening no-go amplitudes, C) morning difference amplitudes (no-go minus go trials), and D) evening difference amplitudes (no-go minus go trials).

Fig. 4 .
Fig. 4. BMI group x Session N2 mean difference amplitudes (no-go minus go trials) for the A) morning and B) evening groups.*p < 0.05.

Fig. 5 .
Fig. 5. BMI group x Session P3 mean difference amplitudes (no-go minus go trials) for the A) high-calorie and B) low-calorie task.*p < 0.05.

Table 2
VAS and heart rate data.
Note. bmp = beats per minute; VAS = visual analog scale; VAS range from 0 to 10 cm.K.A.Carbine et al.

Table 3
Ratings data.

Table 6
ERP difference waves.

Table 8
Behavioral data.
Note.RT = response time.aMediansandranges are reported for response times.K.A.Carbine et al.