A total of n = 130 participants were recruited from Austin, TX through online and flyer advertisements. All participants were healthy males and females between the ages of 25–59 years of age, with English as their primary language, and who had a moderate and regular exercise regimen (defined as exercising one or two times per week for 20 min or more for the past 3 months). Participants were excluded if they currently smoked, had back, hip or knee issues or other preexisting health conditions that made exercise difficult or unsafe. Smokers were excluded as smoking is an independent risk factor for a variety of chronic conditions, and smoking can impair cardiorespiratory functioning, making exercise difficult or uncomfortable (Johannsen et al., 2014). Participants were also excluded if they had a current diagnosis of and/or took medication for psychiatric or neurological conditions including anxiety, depression, bipolar disorder, schizophrenia, or epilepsy. Prior to participation, all participants gave their informed consent. All study documentation and data collection methods were approved by and in compliance with the New York University Committee on Activities Involving Human Subjects. For the final analysis, participants were excluded if they did not complete 12 or more total workouts (i.e., at least one per week). The final analysis was conducted on a total of n = 80 participants; however, some analyses include fewer than n = 80 due to missing data. We note in the results where this is the case.

All participants engaged in 3 months of a physically active experience. Participants were randomly assigned to either maintain their current exercise regimen (control group) or increase their exercise regimen to 4–7 classes per week (experimental group). Control participants were assigned to take either one or two classes based on their previous exercise regimen, which was evaluated via a self-reported, screening questionnaire. On the screening questionnaire, participants were asked during the past 3 months, whether exercise was a routine part of their life, how many days per week they exercised, and the level of intensity of these workouts. They were also asked to write a qualitative description of their workout habits. Based on this information as well as follow-up phone calls, if necessary, research personnel determined whether participants were moderately active. Fitness level was later quantified at fitness testing. Experimental participants were encouraged to engage in a minimum of four sessions per week but could increase their exercise regimen to seven per week. All exercise sessions took place at RIDE Indoor Cycling¹ in Austin, TX. All classes were cycling classes that were 45 min in duration. Participants were given a Mio FUSE², wrist-based heart rate monitor, to wear during all exercise sessions. Heart rate (HR) was captured for each exercise session throughout the course of the 3 months, and participants manually recorded their average HR for each exercise session using MyFitnessPal. Participants self-reported other exercise sessions that occurred outside of RIDE.

Before and after the 3-month period, participants completed a modified cardiorespiratory fitness test as well as a series of neuropsychological tasks and self-reported metrics.

At both the beginning and end of the intervention, a submaximal cycle test was used to assess each participant’s aerobic fitness. Participants placed a heart rate monitor (Polar) around their chest and mounted a stationary bicycle. Because of the remote nature of the study, all fitness assessments were conducted via video conferencing (Skype) and remote access to the computer. During the fitness assessment, participants were instructed to maintain a cycle cadence of 50 rotations per minute (RPM). The first 2 min of the fitness assessment was a warm-up period, with no added resistance on the stationary bike. Following the warm-up, the resistance was increased by 6 watts (equivalent to 0.12 kp or 36 kgm/min) every min, with RPE and heart rate being recorded at the end of every min. In addition, RacerMate was attached to the stationary bicycle and tracked speed [both miles per hour (mph) and RPM], watts, and caloric expenditure. Perceived exertion was also monitored each min of the test using the Borg Rating of Perceived Exertion Scale (RPE). The test was continued until at least 3 HRs were recorded between 110 bpm and 80% of age-predicted maximal HR [206 − (0.67 × age)]. The test was terminated when the participant requested to stop, could not maintain an RPM of 50, or they reached or exceeded 80% of their age-predicted maximal HR. Following the fitness assessment, the participants were given a 2-min cool-down period. The maximal aerobic capacity was then estimated by extrapolating the HR response to each workload to an age-predicted maximal HR.

The fitness test was designed to measure several (at least three) workloads and HR responses between 110 BPM and approximately 80% of age predicted maximal HR, also described in the American College of Sports Medicine’s Guidelines for Exercise Testing and Prescription section on sub-maximal cardiorespiratory fitness testing using cycle-ergometers. The HRs were measured in beats per minute and the workloads were recorded as kgm/min. The correlation of these values was determined using the following equation: ∑ (x−x_)(y−y_)÷∑ (x−x_)2∑ (y−y_)2, where x values are workloads, and y values are HR’s. This value was then multiplied by the quotient of the standard deviations for each set of points (i.e., (∑ (x−x_)2÷(n−1))÷(∑ (y−y_)2÷(n−1)) to calculate the slope of a best fit line. Then the intercept of the best fit line was calculated as follows: x − (calculated slope × y). Thus, with an equation that best describes the HR response to workload, the workload expected to elicit age predicted maximal HR was calculated [(calculated slope × age predicted max HR) + calculated intercept]. To estimate the metabolic cost of this predicted maximal workload, the American College of Sports Medicine’s Guidelines for Exercise Testing and Prescription equation for energy expenditure (ml/kg/min O2) during leg cycling was used: Resting component (3.5) + Horizontal component (3.5) + ((1.8 × calculated max workload) ÷ body mass(kg). This is a standard procedure used when performing submaximal exercise testing (a notable example would be the YMCA submaximal cycle ergometer test; Fitchett, 1985; Beekley et al., 2004).

Before and after the 3-month intervention, participants took a series of neuropsychological tasks and self-reported questionnaires at home, at least 2–4 h but no longer than 7 days after their last exercise session. Neuropsychological tasks were administered via the lab’s website³ designed to present all cognitive tasks as well as store participant data. Self-reported questionnaires were administered via Google Forms. Participants were asked to refrain from drinking alcohol or using other illicit substances both before and during these assessments.

Participants were instructed to complete a series of validated and reliable self-report questionnaires to assess their affective state, eating attitudes, body image, and exercise motivation.

At baseline, there was no difference between groups in age (t₍₇₈₎ = −0.384, p = 0.702), gender (χ²₍₁₎ = 0.450, p = 0.502), or education (χ²₍₃₎ = 6.969, p = 0.073; Table 1).

The experimental group (47.87 ± 2.24) engaged in significantly more total cycling workouts over the course of the intervention than the control group (20.73 ± 0.72; t_(45.76) = −11.554, p < 0.001; Figure 2A). Additionally, the experimental group engaged in significantly more workouts outside of the cycling studio than the intervention group (t_(54.320) = −3.586, p < 0.001). We additionally confirmed that even when including workouts outside of the cycling studio, the control group maintained their previous regimen of 1–2 exercise sessions per week (23.63 (±1.11).

There was no significant difference in average heart rate during class sessions between the experimental (145.38 ± 1.92) and control groups (144.96 ± 2.10; t₍₇₈₎ = −0.147, p = 0.883; Figure 2B).

There was no significant time nor time^*group effect on body mass or BMI. There was a significant time effect on estimated VO₂ max (F_{(1, 57)} = 18.809, p < 0.001), with both groups increasing in fitness over time (Table 2). Of note, the average aerobic capacity of the control group at baseline was 33.46 ml/kg/min oxygen consumption and the increaser group was 30.52 ml/kg/min, thus confirming that both groups were moderately fit at baseline (Graves et al., 2015).

All correlational analyses included all participants, both in the control and experimental groups. In regard to the psychological measures, the total number of cycling workouts was significantly correlated with reduced anxiety (BAI r = −0.236, p = 0.035; STAI r = −0.237, p = 0.035), general negative affect (r = −0.280, p = 0.012), fear (r = −0.301, p = 0.007), sadness (r = −0.222, p = 0.048), hostility (r = −0.286, p = 0.010), rumination (r = −0.242, p = 0.030), and disordered eating as measured by the EDE-Q (r = −0.324, p = 0.003) as well as improved body image (r = −0.372, p = 0.001; Figure 4).

Regarding cognitive measures, the total number of cycling workouts was significantly correlated with the improvement in average seek duration (r = −0.321, p = 0.007) as well as the improvement in total time (r = −0.242, p = 0.045) in the Spatial Navigation Task (Figure 4). There were no significant correlations between the total number of cycling workouts and MST, Stroop, Eriksen Flanker, or N-back measures.

Regarding psychological measures, increased estimated VO₂ max was significantly correlated with decreased stress (r = −0.269, p = 0.039) and disordered eating (EDE-Q r = −0.327, p = 0.011; EAT r = −0.278, p = 0.033) as well as increased self-determination for exercise (r = 0.260, p = 0.047; Figure 5).

Regarding cognitive measures, increased estimated VO₂ max was significantly correlated with improvements in the Item Score (r = 0.401, p = 0.004), Order Score (r = 0.284, p = 0.043), Association Score (r = 0.491, p < 0.001), and Episodic Memory Score (r = 0.449, p < 0.001) in the Spatial Navigation Task (Figure 5). Additionally, increased estimated VO₂ max was correlated with decreased percent correct on N3 trials of the N-back task (r = −0.277, p = 0.049). There were no significant correlations between changes in estimated VO₂ max and changes in MST, Stroop, or Eriksen Flanker measures.

Following 12 weeks of aerobic exercise training, marked decreases in several negative mood indicators were seen including decreases in hostility, guilt, sadness, fear, and general negative affect, with the experimental group showing greater decreases than controls. A large body of work has shown that both acute and chronic exercise is beneficial at improving mood, including both increases in positive affect and decreases in negative affect, especially anxiety and depression (Cramer et al., 1991; Arent et al., 2000; Hoffman and Hoffman, 2008; Basso and Suzuki, 2017; Bonham et al., 2018; Aparicio et al., 2021). Here, we show that increasing aerobic exercise in middle-aged individuals with a previous exercise regimen can confer additional affective state improvements beyond maintaining an exercise regimen. Though the majority of studies have focused on younger or older populations (Reed and Ones, 2006; Chang et al., 2012; Hogan et al., 2013) and newer research has focused on middle-aged populations (Chen et al., 2019; Quinlan et al., 2021), our work is some of the first longitudinal exercise research in middle-aged adults. These beneficial effects on affective state are in part thought to be due to exercise-induced increases in neurotrophins and neuromodulators including dopamine, serotonin, norepinephrine, endocannabinoids, and endogenous opioids (Dietrich and McDaniel, 2004; Fuss and Gass, 2010; Lin and Kuo, 2013; Siebers et al., 2021). Additionally, other research has shown that affective-state changes may result from the fact that exercise produces a stress-resistant brain, having potent effects on the sympathetic-adrenomedullary and hypothalamus-pituitary-adrenal (HPA) axes (Greenwood and Fleshner, 2008; Fleshner et al., 2011). Future research is warranted to determine the structural and physiological brain changes associated with these exercise-induced mood-state changes and how long after exercise cessation they persist.

In addition to the decreased negative mood parameters, we found decreases in negative body image after 12 weeks of training, with the experimental group showing greater decreases than controls. Interestingly, both groups demonstrated decreases in negative body image despite lack of change in either BMI or eating attitudes. Regardless, those who increased their exercise regimen demonstrated significantly greater reductions in negative body image compared to the control group. Other research has shown that exercise interventions improve body image in a range of age groups across the lifespan (Hausenblas and Fallon, 2006; Campbell and Hausenblas, 2009). While BMI was assessed before and after 12 weeks, there were no measures of body composition nor of body circumferences. Thus, there may have been redistributions of weight from fat mass to fat-free mass, or changes in waist circumferences that influenced the improved body attitudes. Future research may seek to disentangle the changes in negative body image that are due to the act of performing exercise and the potential anthropometric and body compositional changes that result from that exercise.

We also found that greater engagement in exercise and improved cardiorespiratory fitness improved mood, eating motivation, exercise motivation, and body image. This is one of the first studies to show exercise frequency-related improvements in affect, motivation, and body image in healthy, active adults.

In line with our findings, previous work has demonstrated that exercise leads to decreased anxiety symptoms for individuals with anxiety disorders as well as healthy individuals (Mochcovitch et al., 2016; Stubbs et al., 2017). Additionally, the association between number of workouts or fitness gains and reduction in rumination, general negative affect, sadness, and hostility is consistent with previous findings indicating that increased exercise reduces the risk of depression (Hassmén et al., 2000; Mammen and Faulkner, 2013). Importantly, our work and others have shown that exercise engagement may reduce negative mood state even in the absence of changes in cardiorespiratory fitness (Olson et al., 2017). Research additionally shows that exercise or cardiorespiratory fitness may be related to improved physiological stress reactivity and reduced psychological stress (Holmes and Roth, 1985; Aldana et al., 1996; Brockmann and Ross, 2020; Allesøe et al., 2021). In a large cross-sectional sample (N = 55,185), Allesøe et al. (2021) found that higher levels of self-reported physical activity and fitness were associated with lower levels of perceived stress. Our current study also found that improved fitness, measured through an objective exercise test, is related to decreased perceived stress. Collectively, these findings suggest that exercise engagement and improved fitness lead to improvements in a range of negative mood measures.

Although some studies have found that engagement in a regular exercise regimen leads to improvements in exercise motivation and body image (Pearson and Hall, 2013), the relationships between these findings and either number of workouts or gains in fitness have not previously been reported. Additionally, some studies have suggested that the effect of exercise on body image may be moderated by exercise motivations indicating that there is a complex relationship among psychological outcomes affected by exercise (Lepage and Crowther, 2010). Although there is limited research regarding the relationship between physical fitness and eating motivations, one study found that adolescents with lower levels of fitness had a greater risk of developing eating disorders (Veses et al., 2014). As much of the previous research in this area has focused on the relationship between compulsive exercise and eating disorders, this is the first interventional study to show a relationship between improvements in fitness and decreased disordered eating in a healthy population.

Additionally, we found that greater engagement with exercise and improvements in cardiorespiratory fitness were significantly associated with improvements in spatial navigation and episodic memory abilities. That is, individuals who exercised more frequently and showed larger gains in fitness were more efficiently able to navigate to previously learned locations as well as remember information presented to them during this experience, tasks that are both dependent largely on the hippocampal formation. This is the first time that exercise has been shown to improve spatial navigation in a healthy adult population using a virtual maze task. A recent pilot study in 14 older adults (aged ≥ 60) found that 2 months of exergaming significantly improved spatial navigation as assessed by the immediate course performance maze time in the Floor Maze Test, a task of exocentric and allocentric navigation (Oliveira et al., 2020). Previous reports have also found cross-sectional evidence in adolescents that enhanced cardiorespiratory fitness levels are associated with increased hippocampal volume, which was subsequently associated with spatial learning on a Virtual Water Morris Maze (Herting and Nagel, 2012; Prathap et al., 2021). Other cross-sectional work has shown significant positive relationships between physical activity level (e.g., total step count and step rate) and episodic memory ability in older adults (Hayes et al., 2015). Additionally, acute exercise has been shown to have a significant positive effect on episodic memory (Sng et al., 2018; Johnson and Loprinzi, 2019; Loprinzi et al., 2019). We add to this literature by showing for the first time that enhancing physical activity and fitness in middle-aged adults improves both spatial and episodic memory abilities. This is consistent with extant research in rodents showing that exercise improves hippocampal-dependent functioning, especially spatial navigation (Voss et al., 2013).

Previous work in both human and animal literature has identified that this effect is due to exercise-induced changes in the hippocampus, a critical structure for learning and memory. A pivotal study in older adults found that a one-year intervention of aerobic walking increased the volume of the anterior hippocampus, which included the subiculum, CA1, and dentate gyrus, the seat of exercise-induced neurogenesis (Erickson et al., 2011). This increase in bilateral hippocampal volume was positively associated with the change in VO₂ max, serum BDNF levels, and memory performance (using the dots fixation task) — though the aerobic exercise intervention did not have a significant effect on spatial memory performance itself. Another hallmark study found that chronic exercise increases dentate gyrus neurogenesis and cerebral blood volume (CBV) in rodents in parallel to dentate gyrus CBV in humans (ranging from 21 to 45 years of age), which correlated to increases in both VO₂ max and learning on the Rey Auditory Verbal Learning Task (Pereira et al., 2007). Additionally, an extant body of rodent research has shown that voluntary wheel running and forced treadmill running improves spatial navigation through tasks such as the Morris Water Maze, Y-maze, T-maze, and radial arm maze as well as other tasks dependent on the hippocampus such as contextual fear conditioning, passive avoidance learning, novel object recognition, and pattern separation (Fordyce and Farrar, 1991; Van Praag et al., 2005; O’Callaghan et al., 2007; Chen et al., 2008; van Praag, 2008; Creer et al., 2010; Falls et al., 2010). The behavioral effects appear dependent on exercise-induced enhancements in hippocampal BDNF levels, neurogenesis, long-term potentiation, and functional integration into the existing hippocampal network (Neeper et al., 1995; van Praag et al., 1999a, b; Kobilo et al., 2011; Vivar et al., 2016; Voss et al., 2019).

This collection of research in both the animal and human literature suggests that exercise-induced improvements in spatial navigation and episodic memory are dependent upon both structural and physiological changes at the level of the hippocampus. We add to this literature in humans by demonstrating that greater engagement with exercise can enhance hippocampal abilities into middle age, which has significant implications for aging as detriment in these cognitive abilities emerge with age-related atrophy of hippocampal structures (Ramanoël et al., 2019).

In the current study, we also found no effects of increased exercise engagement on prefrontal cortex-dependent function as measured by the Stroop, Eriksen Flanker, and N-back tasks. These findings come in contrast to our work that shows an acute bout of aerobic exercise in adults of a similar age range improves prefrontal cortex functioning (Basso et al., 2015). We speculate that while acute exercise, with its numerous neural mechanisms of action (especially at prefrontal cortical sites; Basso and Suzuki, 2017), may acutely improve executive function, this level of chronic exercise is not stringent enough to induce baseline executive function improvements. We further hypothesize that the present null finding may be due to a ceiling effect on task performance as percentage correct in each of these tasks was at or near 100%. Previous work in this area has shown that aerobic exercise training significantly improves prefrontal cortex-dependent functioning, specifically using tasks including the Stroop, Eriksen Flanker, and N-Back Task (Dustman et al., 1984; Colcombe et al., 2004; Hansen et al., 2004; Smiley-Oyen et al., 2008; Stroth et al., 2010; Coetsee and Terblanche, 2017; Ludyga et al., 2018; Amatriain-Fernández et al., 2021). The majority of these studies were conducted in older adults (Dustman et al., 1984; Colcombe et al., 2004; Smiley-Oyen et al., 2008; Coetsee and Terblanche, 2017), while some were conducted in children or adolescents (Ludyga et al., 2018; Amatriain-Fernández et al., 2021). Other work in this area has produced null findings (Madden et al., 1989; Panton et al., 1990; Blumenthal et al., 1991; Hassmén et al., 1992; Hill et al., 1993; Dustman et al., 1994), with this lack of effect being attributed to methodological factors such as baseline levels of high cognitive functioning or short intervention periods that do not lead to fitness changes. Kramer et al. (1999), in fact, proposed a “selective improvement” hypothesis whereby aerobic exercise specifically acts on brain regions and cognitive processes that are sensitive to neurodegeneration, such as the prefrontal cortex and executive function. This hypothesis extends to the idea that aerobic exercise may only provide beneficial effects if: (1) enough cognitive decline is in place (i.e., there is room for cognitive improvement); or (2) the cognitive challenge is sufficient (i.e., tasks are challenging enough to show improvement). In fact, previous work in rodents and humans have indicated that the effects of exercise on cognition may be dependent on task difficulty (Creer et al., 2010; Déry et al., 2013; Heisz et al., 2017; Suwabe et al., 2017b). We suggest that there may be a “sweet spot” to examine the effects of exercise on cognitive function, whereby baseline levels of cognition demonstrate room for improvement (e.g., elderly or other clinical populations) or the task demands are challenging enough such that participants will be able to improve performance. Recently, studies in our lab have demonstrated that task difficulty can be modulated by decreasing the amount of time that participants are given the opportunity to respond to stimuli (e.g., decreasing stimuli presentation time from 1,500 ms to 1,000 ms). Future studies are warranted to systematically test the effects of exercise on prefrontal cortex tasks with various stimuli presentation times.

We acknowledge several limitations of the current study. First, our findings may be biased by the relatively high dropout rate (~38%) in the study. This dropout rate occurred most likely due to the remote nature of the study and despite several recruitment strategies including emails, text messages, and phone calls. Second, though we found extant correlational relationships between exercise sessions and cognitive and psychological measures, we found limited between groups effects on these outcomes—suggesting that our intervention and control groups may not have been distinct enough to show the effects of increased exercise engagement. For example, the experimental group attended on average, 4 weekly exercise sessions, which was on the low end of the assigned range of 4–7 weekly exercise sessions. Additionally, as all participants were confirmed as having a moderate exercise regimen for at least 3 months before the study began, they were all highly motivated to exercise. Most participants, including the control group, engaged in other forms of exercise (e.g., running, yoga, aerobic exercise classes) during the study period, and therefore, this additional amount of exercise may have contributed to mood, motivation, or cognitive effects. Further, the fact that participants were enrolled in an exercise study and assessed at two time points may have motivated them to exercise more. We also did not obtain a good estimate of workout intensity for all workouts completed outside of the cycling studio, which is another limitation. General physical activity levels were also not assessed; future studies may want to consider utilizing activity monitoring throughout the study period or self-reported questionnaires (e.g., International Physical Activity Questionnaire; Global Physical Activity Questionnaire). Alternatively, the lack of causational findings (i.e., lack of time^*group effects) on hippocampal function may indicate that it may take a longer bout of exercise to cause hippocampal changes due to the involvement of stimulating the growth of new hippocampal neurons; 3 months of exercise may be too short of a time period for hippocampal cellular integration and functional improvement. Considering that previous reports of exercise-induced improvements in hippocampal function have been in young adults in response to high-intensity exercise regimens (Déry et al., 2013; Heisz et al., 2017), the age of our study population may contribute to the lack of time^*group effects; moreover, higher exercise intensity may be needed to induce effects. Another limitation comes in the realm of the cardiorespiratory fitness test. Because the study happened remotely, we needed to rely on the participants themselves as well as staff at the fitness facility to help set up and start the equipment. We were not able to conduct a traditional VO₂ max test and as such needed to calculate an estimate of aerobic capacity, and in some instances, technical issues arose with the test.

Considering the behavioral effects, future studies should investigate both the structural and functional changes associated with these improvements through neuroimaging techniques such as MRI and fMRI. We hypothesize that these behavioral improvements will be accompanied by increases in hippocampal volume as well as heightened functional activation during task engagement. Additionally, future research is warranted to investigate cellular and molecular mechanisms underlying this effect, especially for growth factors such as brain-derived neurotrophic factors (BDNF). Incorporating genetic testing for the BDNF gene may be an interesting avenue of research, specifically to determine whether individuals with different genetic variants (e.g., BDNF Val66Met polymorphism) may be more susceptible to these exercise-induced effects.