Journal Logo


Exercise versus Nonexercise Activity

E-diaries Unravel Distinct Effects on Mood


Author Information
Medicine & Science in Sports & Exercise: April 2017 - Volume 49 - Issue 4 - p 763-773
doi: 10.1249/MSS.0000000000001149
  • Free


Physical activity is not only linked to reduced risks for severe somatic diseases (e.g., noncommunicable diseases such as cardiovascular illnesses, type 2 diabetes mellitus or obesity) (30,39), it is also associated with mental well-being in the general population (15,25,27–29) and known to ameliorate psychiatric symptoms in clinical samples (21,31). For instance, physical activity has been shown to significantly reduce levels of depression with effects comparable to cognitive therapy (7,20,22,24). Moreover, following behavioral theories (23), affective responses may drive physical activity. In practice, a person who experiences positive feelings during or after physical activities will likely repeat those activities and thus benefit from health promoting effects of an active life style. Accordingly, how exercise (such as jogging, swimming, dancing, and playing football) and nonexercise activities (NEAs) (e.g., daily routine of climbing stairs, gardening, biking for transport, and catching the train) influence mood has been investigated multiple times (for a discussion, refer to Schlicht et al. [32]).

Unfortunately, most of these studies used between-subject designs, that is, compared sporty and nonsporty participants cross-sectionally or compared participants who took part in a treatment intervention with those who did not. However, it has been shown theoretically and empirically that between-subject differences do not translate to within-subject effects (1,19). In other words, between-subject studies do not clarify whether physical activity and mood wax and wane synchronously over time in persons. However, this is a major question that must be answered to garner a deeper understanding of the associations between physical activity and mood.

Fortunately, progress in methodology, technology, and statistics allows us to answer these questions on a within-subject level, currently. In practice, Ambulatory Assessment (AA) approaches using repeated real-time e-diary mood assessments on smartphones as well as continuous and objective monitoring of physical activity via accelerometer are state of the art techniques used to assess whether physical activities influence mood in everyday life (19). There are several reasons for using this type of approach. First, because the link between physical activity and mood is (most likely) dynamic and time-dependent, AA is appropriate to capture this association using repeated assessments (11). Second, AA allows the assessment of mood in real time, which is important because retrospective assessments are vulnerable to systematic distortions (12,35), such as the mood-congruent memory effect, the peak end rule, and the duration neglect effect (11). Third, assessment in everyday life is important to prevent distortions caused by laboratory settings (6). Fourth, according to Prince et al. (26) and Adamo et al. (2), there is a gap between objectively assessed and subjectively rated physical activity. Accordingly, for most activities in daily life that are performed unconsciously and spontaneously, the objective assessment of physical activity via accelerometers appears to be adequate and more precise than subjective ratings.

In the last decade, a few studies investigated the association between physical activity and mood in healthy participants using AA. Gauvin and colleagues (13) recruited 86 women mainly from fitness classes and asked them to fill out paper-and-pencil diaries on their feeling states before and after acute bouts of physical activity across 6 wk in everyday life. Analyses showed that after acute bouts of physical activity feelings of tranquility, positive effect, revitalization, and positive engagement were significantly increased and, moreover, negative affect was significantly decreased. In a study that included 124 healthy participants, Schwerdtfeger et al. (33) used accelerometers and hand held computers to assess the effects of physical activities on mood over the course of 12 h. In this sample, the higher intensity and/or duration of everyday life activity was significantly accompanied with higher energetic arousal and a more positive activated affect. However, Schwerdtfeger et al. (33) found no association between tension arousal/negative affect and physical activity. Kanning et al. (17) investigated activity-mood associations in a sample of 44 undergraduates using electronic diaries and accelerometry over 24 h. The study revealed both a significant positive effect of physical activity on energetic arousal and a significant negative effect on calmness. However, Kanning et al. (17) detected no association between physical activity and valence. Bossmann et al. (4) investigated a sample that consisted of 77 university students using accelerometers and e-diaries over the course of 24 h. The participants reported significantly positive effects of physical activity on energetic arousal and valence. However, calmness was not significantly affected by physical activity. Kanning (16) showed a positive effect of physical activity on valence and energetic arousal as well as a negative effect on calmness independent of the context in a sample of 87 undergraduates. The participants rated mood on palmtop devices and specified context information. Additionally, 24-h accelerometry was conducted. In 2015, Kanning et al. (18) investigated the associations between physical activity and mood in older adults using activity triggered e-diaries over the course of 3 d. The adults reported a higher energetic arousal and less calmness following physical activity. However, valence was not affected by physical activity. Moreover, Dunton and colleagues (8,9) investigated how contextual influences moderated the effects of physical activity on mood in everyday life. The researchers' ambulatory assessment study that included 116 adults who rated mood on e-diaries and wore accelerometers for four consecutive days showed that positive affect responses to physical activity were particularly heightened when people were together with others. Additionally, Dunton et al. (8) found that physical contexts (e.g., being outdoors) decreased the negative affect during physical activity.

Summarizing the abovementioned within-subject studies, physical activity appears to enhance energetic arousal (4,16–18,33). However, findings regarding the mood dimensions of valence and calmness remain ambiguous. Specifically, some of the abovementioned studies found positive effects of physical activity on valence (4,16); others did not (17,18). Furthermore, some studies reported negative effects on calmness (16–18), whereas others did not find an association between physical activity and tense arousal (4,33); even more confusing, one study found positive effects on tranquility (13). However, meta-analyses based on intervention studies on the between-person level revealed that both acute and regular exercise of at least 30 min with low to moderate intensity appear to enhance positive affect (28,29).

A closer examination of the AA studies noted above does suggest that the conceptual aggregation level between the two constructs of interest do differ tremendously. Specifically, based on a three-dimensional model, mood is usually assessed by separating it into a valence component in an energetic arousal dimension as well as in a tense arousal component (for a discussion, refer to Wilhelm and Schoebi [38]). However, physical activity is not differentiated, although there are distinct types of physical activity which show clear differences regarding physiological and psychological processes (for a detailed discussion, refer to Kanning et al. [19]). In practice, for example, playing soccer with friends every Monday and Thursday evening does clearly differ from spontaneously fetching papers from the basement at work with regard to the context, motive, level of socialization, structure, timing, duration, and level of energy expenditure. Therefore, in contrast to earlier ambulatory assessment studies, we differentiated between exercise and NEA in our analyses. We followed Kanning et al. (19) defining exercise as structured physical activities with high demands of energy expenditure across prolonged periods. Thus, we asked participants to report their sport (such as playing volleyball) and exercise activities (such as jogging) indicating point in time, duration in minutes, and type of activity. In contrast, again following Kanning et al. (19), we defined nonexercise activities to comprising all other daily physical activities (e.g., daily routine of climbing stairs, gardening, biking for transport, and catching the train), which are “often processed automatically and habitually or performed spontaneously” (19); we assessed NEA intensity continuously and objectively via accelerometers. Based on the idea of the Brunswik lens model (5), we assumed that the differentiation of exercise and NEA leads to more consistent findings on the influence of physical activity on mood by using similar conceptual levels of physical activity and mood. Oversimplified, the Brunswik lens model (5) assumes that meaningful associations can only be evidenced if the aggregation levels of predictor and outcome are comparable. When transferred to the association of physical activity and mood, this would imply that the total physical activity over a whole year might not be related to momentary mood, but that momentary subcomponents of mood are specifically correlated with momentary subcomponents of physical activity.

Until currently, there has been a lack of within-subject studies delineating the specific effects of exercise versus NEA on subcomponents of mood (valence, energetic arousal and calmness) in everyday life. To fully exploit the potential of within-subject designs, we assessed mood repeatedly in real time as well as physical activity, using continuous and objective measures in everyday life over 7 d. We expected to observe differential effects of exercise vs. NEA on valence, energetic arousal, and calmness.

Unfortunately, there are insufficient prior studies that investigate the differential effects between exercise and NEA to delineate specific hypotheses. However, the large number of meta-analyses on between-subject associations of exercise and positive affect as well as the consistent findings of ambulatory assessment studies on the relation between physical activity and energetic arousal enabled us to at least specify two distinct hypotheses. Accordingly, we hypothesized a positive effect of exercise on valence (hypothesis I) and a positive effect of NEA on energetic arousal (hypothesis II). To statistically analyze how NEA influences mood, we aggregated 15-min episodes of physical activity intensity (captured with accelerometers) before each e-diary assessment of mood.


Study participants

From December 2014 to September 2015, the participants were recruited within a larger study, URGENCY (Impact of Urbanicity on Genetics, Cerebral Functioning and Structure and Condition in Young People), by the psychiatric-epidemiological center (PEZ) at the Central Institute of Mental Health in Mannheim (CIMH), Germany. The participants were randomly drawn from local population registers. The PEZ selected the participants applying a 2-stage proportionally layered procedure taking into account specific stratifications, such as age, ethnic background, and place of residence (considering urban vs. rural areas). The participants were included in the URGENCY study if they were between 7 and 28 yr of age and lived in the study region, which was composed of the municipalities Mannheim, Ludwigshafen, Heidelberg, Weinheim, and the Rhine-Neckar district that contained parts of the Forest of Odes. The exclusion criteria included chronic endocrine, cardiovascular, immunological, or clinically manifested mental disorders. Participants with acute diseases or moderate to difficult impairment of intelligence and participants of consent or legal incapacity were excluded.

For the current analyses, a sample of adult participants age between 18 and 28 yr who provided information on their exercise habits within the study week was included (n = 106). From this sample, nine participants were excluded because of technical problems with the accelerometer or lost devices; one participant was excluded for reasons of work (having a shifted diurnal rhythm). Moreover, three participants were excluded for compliance reasons (large nonwear time of accelerometers or <30% responses to e-diary prompts).

The final sample consisted of 93 participants (62.4% female) with a mean age of 23.4 yr (SD = 2.7) and a mean BMI of 22.8 kg·m−2 (SD = 3.6; for detailed participant characteristics, refer to Tab. 1). Moreover, 90 participants were unmarried, two married and one divorced; 83.9% (n = 78) participants received at least 12 yr of school education and 33.3% (n = 31) participants had a migration background.

The participants received monetary compensation for their participation in the ongoing PEZ project.

Study procedures

Participants carried a smartphone (Motorola Moto G, Motorola Mobility LLC, Libertyville, IL, and an accelerometer (movisens Move-II, movisens GmbH, Karlsruhe, Baden-Wuerttemberg, Germany, for seven consecutive days during their everyday lives. Before the assessment, participants received an extensive briefing regarding the usage of the smartphone and the accelerometer, including individual testing at PEZ. After the 7-d assessment period, participants returned devices and reported their exercise activities within the study week. To optimize participants' recall, we reconstructed participants’ daily routines with a procedure similar to the Day Reconstruction Method (DRM; 15). Originally, the DRM combined a time-use-study and experience sampling procedure to reduce recall bias. In practice, participants were asked to reconstruct their previous day (i.e., activities and social interactions during this day) and to remember affective experiences in each situation. Kahneman and colleagues (15) supposed that the DRM would enhance recall and showed that the method provides results consistent with established findings from experience sampling studies.

In our study, participants were first given a definition of exercise according to Kanning et al. (19) (for details refer to the introduction section). Accordingly, they were asked to report both sport (such as playing soccer) and exercise activities (such as jogging and swimming) and to name the point in time, duration in minutes, and the type of exercise activity. Therefore, participants were shown their daily motion profiles across the whole study week on a digital map, that is, any locations visited and routes covered within the study week that had been tracked on the smartphone, combining GPS, WLAN, and CELL to reduce energy consumption (36). Thereafter, they were asked to label the places visited (e.g., at home, at work, out with friends, jogging in the park, playing football, and being in the gym). Following Kahnemann et al.’s idea of the DRM, this procedure most likely helped participants recall an accurate picture of every study day and to recall exercise activities.

E-diary sampling strategy

Since episodes of high NEAs are rare, triggering e-diaries according to a standard (e.g., time-based) sampling scheme is likely to miss those episodes. In other words, when querying mood every hour, the assessments may, by chance, not be linked to high physical activities. According to Ebner-Priemer et al. (10), using sophisticated algorithms to capture episodes of interest enables the most suitable investigation of associations between physical activity and mood. To increase the probability of ratings during episodes of physical activity, we used a location-triggered approach. The GPS-based trigger algorithm detected distances larger than 500 m from the last assessment point to trigger new assessments. In addition, timeout triggers, that is, triggering no more often than every 40 min and at least every 100 min, and triggers at fixed times (08:00 and 22:20), were implemented. Accordingly, the smartphone prompted the participants via an acoustic, visual, and vibration signal every 40 to 100 min within the 7:30 to 22:30 period, which resulted in at least nine to a maximum of 22 triggers per day. The participants were offered the chance to postpone an e-diary prompt for a maximum of 15 min. This mixed sampling strategy (time- and location-based) to acquire participants' mood ratings during everyday life and the implementation on the Android phones were programmed using the experience sampling software movisensXS, version 0.6.3658 (movisens GmbH, Karlsruhe, Baden-Wuerttemberg, Germany,


Mood was assessed using a six-item short scale developed and validated by Wilhelm and Schoebi (38), which is suitable to assess fluctuations of mood as a three-dimensional construct on the within-person level over time on e-diaries. Based on the originally German-language Multidimensional Mood Questionnaire (MDMQ, [34]), this short scale assesses three basic affective states using the bipolar items (German translations) (a) unwell to well (unwohl–wohl) and (b) discontent to content (unzufrieden–zufrieden), representing valence; (c) without energy to full energy (energielos–energiegeladen) and (d) tired to awake (müde–wach), representing energetic arousal; and e) tense to relaxed (angespannt–entspannt) and (f) agitated to calm (unruhig–ruhig), representing calmness. Wilhelm and Schoebi (38) reported suitable psychometric properties for this short scale. Specifically, the reliability coefficients were 0.92 (valence) and 0.90 (energetic arousal and calmness) on the between-person level; 0.70 (valence and calmness) and 0.77 (energetic arousal) were reported on the within-person level. We presented the bipolar items in mixed order and reversed polarity, according to Wilhelm and Schoebi (38). Moreover, we used visual analogue scales (0–100).

Nonexercise activity

Participants wore the accelerometer (movisens Move-II, movisens GmbH, Karlsruhe, Baden-Wuerttemberg, Germany, on the right side of their hip during the entire study week but not during sleep. A validation study (3) revealed the Move-II accelerometer to be appropriate in assessing human physical activity; specifically, the Move-II accelerometer showed more accurate estimates of activity energy expenditure than the GT3X accelerometer (ActiGraph Manufacturing Technology Inc., Pensacola, FL, when comparing estimates of both devices with indirect calorimetry. The triaxial acceleration sensor Move-II captured movements of as much as ±8g with a sampling frequency of 64 Hz and a resolution of 12 bits. The raw acceleration intensity was stored on an internal memory card.

We operationalized nonexercise activity as mean movement acceleration intensity across 15-min segments before each e-diary assessment because earlier results from Schwerdtfeger and colleagues (33) revealed especially intervals of physical activity with a duration of 15 min to be high correlated with both acute subjective ratings of mood and daily physical activity levels. Accordingly, nonexercise activity was parameterized as the intensity of physical activities in the unit mg (i.e., [g]/1000) across 15-min episodes in everyday life. We included this parameter in our models as a continuous variable with a natural zero.

Data analysis

First, we calculated Movement Acceleration Intensity from the accelerometer data in 1-min intervals using the software DataAnalyzer, version 1.6.12129 (movisens GmbH, Karlsruhe, Baden-Wuerttemberg, Germany, This parameter represents the vector magnitude of the acceleration in mg assessed with the three sensor axes. This parameter is computed using a high-pass filter (0.25 Hz) eliminating gravitational components and a low-pass filter (11 Hz) excluding artifacts, such as vibrations when cycling on a rough road surface or shocks of the sensor; for details on data processing refer to von Haaren and colleagues (37). Second, we merged the minute-by-minute movement acceleration values and the e-diary entries using the software DataMerger, version 1.6.3868 (movisens GmbH, Karlsruhe, Baden-Wuerttemberg, Germany, Third, we aggregated mean Movement Acceleration Intensity values within the timeframes of 15, 20, 25, 30, 40, 50, 60, 70, 80, 90, 100, 150, 200, 250, and 300 min before each e-diary entry, using the software SPSS (IBM), version Fourth, we developed a factor that cumulatively summed the exercise duration in minutes over the day. Participants who engaged in exercise within the study week (51.6% of our sample) did this on average 2.2 times per week with a mean exercise duration of 188.8 min per week (refer to Table 1). Although these participants surpassed the recommended physical activity level for adults (40), in statistical terms, exercise was rare in our data set; specifically, participants engaged on average 0.45 h in exercise within 15 h of ambulatory assessment per day. Because exercise has been shown to affect mood for a prolonged period (29) and due to the statistical distribution of exercise in our data set we cumulatively summed the duration of exercise (in minutes) within one study day to capture the effects of exercise on dimensions of mood in a time range as broad as possible. Put into practice, this means that if, for example, a participant engaged in exercise from 7:00 to 7:30 and from 18:00 to 20:00 on 1 d, after 7:30 the variable was set to the value 30; after 20:00 the variable was set to the value 150. Thus, each value represents the duration of exercise in minutes that the participant already undertook on this day.

Participants' characteristics.

Statistical analysis

To analyze whether exercise and NEA waxed and waned with mood synchronously over time within our participants and whether there are differential effects of both predictors, we conducted multilevel analyses using the statistical software SPSS (IBM), version We calculated two-level models for each dimension of mood, with level 1 representing repeated measurements nested within participants (level 2).

First, intraclass correlations were estimated using unconditional models. Second, we successively entered the predictors time, time-squared, exercise, and NEA into our models. To standardize the predictors time and time-squared, we subtracted the start time of the study for each day (7:30). Third, we included random effects for each predictor. However, we subsequently deleted nonsignificant random effects, which resulted in different models for the three mood dimensions of valence, energetic arousal, and calmness. The final models are presented below. Specifically, these models were used for analyses on the time courses of effects of NEA on mood as well; however, the aggregation level of the predictor NEA varied (15, 20, 25, 30, 40, 50, 60, 70, 80, 90, 100, 150, 200, 250, and 300 min before each e-diary entry).


Energetic arousal


On level 1, within-subject effects were estimated with participants' e-diary entries (subscript j) at any time of measurement (subscript i). Yij represents the level of valence, energetic arousal and calmness, respectively, in person j at time i. Beta coefficients represent the intercept and the effects of time, time-squared, exercise, and nonexercise activity at level 1, and rij represents the residuals at level 1. We centered exercise and NEA around the subjects' mean to differentiate within-person from between-person effects.

On level 2, between-subject effects were estimated. For all three mood outcomes, we included random effects (i.e., individual variation around the sample mean effect, represented by γ) of intercept (u0j); random slope parameters (u1jto u4j) were included only if significant variation across subjects was observed.

To compare the magnitude of effects between exercise and NEA on the dimensions of mood, we calculated standardized beta coefficients for all predictors (time, time-squared, exercise, and NEA) and outcomes (valence, energetic arousal, and calmness) according to Hox (14) using the equation below (equation 19) where SD represents standard deviations across the entire sample. We set the α-level to 0.05 for all analyses.

Ethical consideration

This study was approved by the Medical Ethics Committee II of the Medical Faculty Mannheim at the Ruprecht-Karls-University in Heidelberg. This study fulfilled the ethical guidelines for medical research according to the declaration of Helsinki. Written and oral information regarding study procedures were presented to all eligible participants before written informed consent was obtained. There was no surrogate consent procedure. All participants were free to withdraw from the study at any time.


Descriptive Statistics

The dataset consisted of 6588 e-diary entries in total, provided by 93 participants over 7 d. On average, participants answered 81.2% of all prompts (SD = 14.3%; refer to Table 1).

The duration of exercise in the subsample of participants engaging in exercise within the study week (51.6% (n = 48) of the whole sample) ranged from 20 to 570 min per participant per week with an average of 188.8 min per participant per week (SD = 138.0 min per participant per week; for details, refer to Table 1). Participants engaged in various types of exercise, for example, in team games, such as soccer and volleyball, in strength and endurance training such as weight lifting and jogging, and in gymnastics and in dancing. In detail, many participants engaged in jogging (n = 18 participants) with an average of 48.9 min per session (range, 20–90; SD, 20.7), in fitness training at the gym (n = 13 participants) with an average of 89.0 min per session (range, 30–162; SD, 32.1), and in ball sports (e.g., soccer, tennis, handball, volleyball; n = 7 participants) with an average of 111.0 min per session (range, 60–154; SD, 33.5). The highest total exercise duration of 570 min per participant per week was obtained by a participant exercising six times within the study week (i.e., engaging in soccer, jogging, martial arts, and rollerblading/skating).

The NEA ranged from 14.3 to 58.6 mg·min−1, with an average of 36.3 mg·min−1 (SD = 9.8 mg·min−1) per participant over 1 wk (see Table 1). Note that according to Anastasopoulou et al.'s findings (3), sedentary behavior such as sitting results in approximately 7 mg, walking (at 3.1 mph gait speed) results in approximately 367 mg, and jogging (at 6.5 mph gait speed) results in approximately 1103 mg.

Participants reported average scores of 71.2 (SD = 10.4), 58.1 (SD = 10.6), and 67.7 (SD = 11.4) for the mood dimensions of valence, energetic arousal and calmness, respectively (for details, refer to Table 1). The intraclass correlation coefficients were ρI = 0.272 for valence, ρI = 0.186 for energetic arousal, and ρI = 0.317 for calmness. This finding reveals the variance due to within-subject variation in the sample, namely, 73%, 81%, and 68% for valence, energetic arousal, and calmness, respectively.

Effects of Exercise Versus Nonexercise Activity on Dimensions of Mood


Table 2 shows the fixed and random effects of time, time-squared, exercise and NEA on valence. As hypothesized (hypothesis I), exercise positively predicted valence (standardized beta coefficient = 0.023; P = 0.035). However, NEA did not significantly predict valence (P = 0.200). Furthermore, both time and time-squared significantly influenced valence (P < 0.001 and P = 0.004) with positive (time; standardized beta coefficient = 0.180) and negative (time-square; standardized beta coefficient = −0.117) effect directions. In practice, valence increased during the day and peaked at approximately 19:00, followed by a subsequent decrease until the end of the day. Within the typical subject, 2 h of exercise increased the mean valence by 2.5 points on average across all e-diary assessments subsequent to the exercise activity at the study day on which the person undertook exercise. In simpler terms, on average, when a participant undertook exercise, for example, from 16:00 to 18:00, her or his valence was increased by 2.5 points on average across all mood queries after 18:00 at this day. The random effects of time showed significance (P = <0.001), which indicates variability between participants (level 2) with regard to their association between time and valence.

Multilevel model analysis to predict mood: fixed and random effects of time (h), time-squared (h2), exercise (min) and nonexercise activity (mg).

Energetic arousal

Energetic arousal was significantly predicted by NEA (P < 0.001), showing a positive association (standardized beta coefficient = 0.135), and thus, confirming hypothesis II (refer to Table 2). However, surprisingly, exercise did not predict energetic arousal (P = 0.172). Again, both time and time-squared affected energetic arousal (P < 0.001 and P < 0.001), with opposing directions of effects (standardized beta coefficient time = 0.819; standardized beta coefficient time-square = −0.918). In practice, energetic arousal increased during the day until approximately 15:00 and decreased thereafter. Furthermore, on average, when a participant was walking instead of sitting over 15 min before an e-diary entry (367 mg instead of 7 mg), her or his mean energetic arousal was increased by 14.8 points within this e-diary entry. Additionally, Table 2 shows significant random slopes for time (P < 0.001), time-squared (P = 0.002), exercise (P = 0.021) and NEA (P = 0.006), revealing that the slopes of these predictors varied between participants.


Both exercise and NEA significantly predicted calmness (P = 0.040 and P = <0.001). However, NEA decreased calmness (standardized beta coefficient = −0.080), whereas exercise increased calmness (standardized beta coefficient = 0.022; refer to Table 2). Time-squared showed significant negative effects on calmness (standardized beta coefficient = 0.125; P = 0.002); however, there were no significant effects of time on calmness (P = 0.191). In practice, calmness increased during the day, following a quadratic relation between time and calmness. Accordingly, the rise of calmness successively increased during the day. Moreover, on average, a mean value of 360 mg over 15 min before a mood assessment (representing walking instead of sitting) decreased calmness by 7.2 points within this mood assessment, whereas 2 h of exercise increased calmness by 2.4 points on average across all mood queries subsequent to the exercise activity at the respective study day. Time and NEA revealed significant random effects (P = 0.001 and P = <0.001), thus showing variation between participants.

Effects of Nonexercise Activity over Time

Following earlier findings from Schwerdtfeger and colleagues (33; for details refer to the methods section “Nonexercise activity”), we used aggregated mg values of 15-min segments before each e-diary prompt to test our hypotheses (refer to the calculations above). However, to our knowledge, the temporal relations between NEA and mood for time frames above 15 min are largely unknown. In other words, it is conceivable to detect influences of NEA on mood within 30 min of mean activity before an e-diary prompt, which remains undiscovered by solely considering 15-min segments of NEA. To investigate how stable the relations between NEA and mood are over varying time frames (segments), we plotted the standardized beta coefficients (refer to Fig. 1) of a variety of different time frames. Specifically, the x-axis represents the temporal aggregation level of NEA in minutes before an e-diary prompt. Accordingly, a value of, for example, 50 indicates that an average value for the NEA (mg) of the last 50 min before the e-diary prompt was entered into the model. A value of 150 represents the mean mg value of the 2.5 h of NEA before each e-diary prompt. Furthermore, the y-axis depicts the standardized beta coefficients for NEA on valence, energetic arousal, and calmness.

Effects of NEA in predicting mood over time: course of standardized beta coefficients according to mean physical activity aggregated over different time frames. The x-axis shows the temporal aggregation level of NEA in minutes before a specific e-diary prompt. Accordingly, e.g., the value 50 represents the mean acceleration aggregated over 50 min before each e-diary prompt. The y-axis depicts the standardized beta coefficient for NEA on valence, energetic arousal and calmness.

As noted above, the underlying multilevel models do solely differ with regard to the duration of the aggregated time frame of NEA from the models used to examine the differences of exercise versus NEA; accordingly, the models that estimate effects of 15-min segments of NEA on mood dimensions within Figure 1 are exactly the same models we used above to compare the effects of exercise and NEA.


Descriptively, the standardized beta coefficients of NEA on valence are close to zero across the entire range of aggregated time frames illustrated by a flat green line in Figure 1. None of the depicted effects are significant. Thus, our finding that NEA did not enhance valence is not only valid for the mean mg values 15 min before each e-diary prompt, but it is confirmed across the entire range of time frames.

Energetic arousal

Figure 1 shows a red line illustrating the standardized beta coefficient of NEA on energetic arousal across the different time frames. The line begins at 0.135, as stated above (Table 2), confirming hypothesis II in which we assumed a positive effect of NEA on energetic arousal. Descriptively, Figure 1 depicts a stable high effect of NEA on energetic arousal to the aggregation level of 40 min, with a small rise from 15 to 40 min. Additionally, the standardized beta coefficients decrease from data point 40 (0.143) to 300 (0.015), with diminishing effects after 100 min, and approach zero. This finding is not surprising because the aggregated time frames average the episodes of activity and inactivity and, therefore, lose the power to predict time-dependent effects; in other words, the effects of NEA on mood are successively diluted across aggregated time frames. Nevertheless, for a broad range of aggregated time frames, namely, from 15 to 200 min, the significant effects of NEA on energetic arousal support hypothesis II.


In Figure 1, the blue line depicts the standardized beta coefficient of NEA on mood starting at −0.080. The effects reveal significance across all time frames from 15 to 250 min. Therefore, this result reinforces our finding that NEA predicted and decreased calmness. Moreover, the negative standardized beta coefficient slightly diminishes from data point 15 to data point 300 (−0.080 vs −0.044). Again, this finding is attributable to the successive diluted effects of NEA on calmness across time frames.


As expected, we found differential effects of exercise versus nonexercise activity on valence, energetic arousal, and calmness. First, exercise did show a significant positive effect on valence, whereas NEA did not enhance valence across the entire time frames investigated. This result may help clarify the ambiguous findings of previous ambulatory assessment studies. Whereas some studies showed positive effects of physical activity on valence (4,16), others were not able to reveal significant associations (17,18). Due to our results and inspired by Bossmann et al.'s (4) assumption that certain activity thresholds must be surpassed to enhance one’s valence, we speculate that valence is affected particularly by activities that demand high energy levels and/or long duration such as exercise. Activities in daily life most likely do not exceed certain thresholds to raise the levels of valence. This assumption is supported by meta-analyses including 263 studies from 1979 to 2008, which shows that, generally, both acute and regular exercise of at least 30 min with low to moderate intensity appear to enhance the positive affect at the between-person level (28,29).

Second, exercise versus NEA also showed differential effects on energetic arousal. As hypothesized, NEA significantly increased energetic arousal. This finding is in accordance with previous studies (4,16–18,33). Furthermore, we were able to show the enhancing effects of NEA on energetic arousal for a broad range of time frames within our sample. Thus, our study reveals consistent findings regardless of specific aggregation levels of NEA. In our study, exercise was not associated with energetic arousal. A close examination of the meta-analyses of between-subject studies that showed positive effects of exercise on mood reveals that these investigations often refer to the positive activated affect, a combination of the mood dimension valence and activation (e.g., 28, 29). Due to our results, it appears necessary to separate these dimensions of mood because there appears to be a differential effect of exercise on these dimensions of mood. Furthermore, we did not consider the differences in the types of exercise, which may have affected the association between exercise and energetic arousal. In practical terms, certain participants may have gathered energy throughout a Yoga session, whereas other participants most likely felt exhausted after their football match. Accordingly, our analyses revealed significant random effects for exercise on energetic arousal (but not for valence and calmness), indicating unsolved variance on the person-level.

Third, again we found differential effects of exercise and NEA on calmness. Although both exercise and NEA predicted calmness significantly, exercise increased calmness, whereas everyday life activities decreased calmness. The effect of NEA on calmness is in accordance with three studies of Kanning et al. (16–18), while other investigations did not find any effects (4,33). On a theoretical level, our finding may be explained by activities of daily life that require high intensities of physical activity that are often related to stressful events (e.g., catching the train, and being busy at work). Kanning et al. (19) requested a time-lagged investigation of the effect of physical activity on calmness to clarify whether a positive relation is time-dependent and occurs after a certain time. In other words, the researchers hypothesized that activities in daily life directly decrease calmness, whereas positive effects arise after a time lag. We checked certain broader time frames in our time course analyses, up to a maximum of 300 min. Although methodologically very different, it may be argued that the hypotheses by Kanning et al. (19) may obtain visibility with this approach. However, we did not find supportive evidence for this hypothesis. In contrast, we found that exercise does affect calmness positively. Therefore, a positive effect of physical activity may be evoked by activities that require higher energy levels, such as exercise rather than by short-term physical activities in everyday life. This might provide an explanation for the opposed findings of former studies, for example, Kanning et al. (16–18) who found acute physical activity to be related to decreased calmness in contrast to Gauvin and colleagues (13) who found increased tranquility after self-reported physical activity bouts of at least 20 min.

There are certain limitations to our study that deserve discussion. First, the study devices (i.e., smartphone and accelerometer) were not water and shock proof and could therefore not be worn during all types of exercise sessions (e.g., while swimming). Thus, we examined the influence of exercise on mood not by using the accelerometer data but by solely including duration of exercise in minutes in our statistical models. Because different types of exercise may influence mood in various ways, future studies should address this differentiation and thus take objectively assessed exercise intensity (via accelerometer) into account. Second, our sample was not equally distributed by sex; the majority of our participants were women (62.4%). However, we tested sex effects but did not find systematic differences within our results between women and men. Third, the age range of our sample was restricted because the reported study was conducted within a larger longitudinal research project with follow-ups every 2 yr using an accelerated longitudinal design. Therefore, we cannot generalize our results to other age groups. Fourth, only 51.6% (n = 48) of the participants in our sample (n = 93 participants) performed exercise within the study week. Therefore, the number of valid data points contributing to the estimation of exercise on mood (i.e., data from 48 participants) was smaller compared with those contributing to the nonexercise activity on mood estimation (i.e., data from 93 participants). Fortunately, questionnaire data on exercise habits did not differ significantly with regard to usual exercise frequency and duration between participants performing or not performing exercise within the study week. Thus, we assume that participants were comparable regarding their usual exercise habits. Moreover, both types of physical activity were significantly associated with two of three mood dimensions, that is, NEA with energetic arousal and calmness and exercise with valence and calmness. Fifth, we cannot exclude the possibility that variations in higher-order cognitive and motivational states may relate to the observed differences in mood dimensions between exercise and NEA. Specifically, the motivation for exercising activity may be more often related to intrinsically orientated and longer-term goals or ambitions with a positive valence (e.g., to maintain weight or reduce high blood pressure), whereas the initiation of NEA may rather be related to the satisfaction of more immediate, emotionally more neutral and external demands (e.g., to fetch a needed item from the basement). Further research is needed to substantiate this proposal. Sixth, lifestyle factors influencing mood, such as social behavior, nutritional behavior, quality of partnership, employment relationship, quantity and quality of sleep, and drug consumption (e.g., alcohol and caffeine), were not included in our models. Since these factors might have moderated the findings, this issue should be broached in the future. Seventhly, we applied a fixed daily study start time at 7:30 both in assessment and analyses, so we did not take into account individual wake-up times of participants. The differences in durations of waking hours before the first daily mood assessment might have influenced our results. However, participants were told not to change their usual daily rhythm and to ignore prompts before getting up (i.e., to mute the smartphone alarm). Moreover, our sample does not include specific groups of persons in which this issue might present a major limitation (e.g., pupils during holidays, unemployed people, or shift workers).


In conclusion, our study demonstrated that exercise and nonexercise activity influenced the three basic dimensions of mood, namely valence, energetic arousal, and calmness within persons in different ways. Whereas exercise was associated with an increase in valence and calmness, nonexercise activity increased energetic arousal and decreased calmness.

Our findings may contribute to refine the understanding of this association, which is important because physical activity is known to raise well-being in the general population, to ameliorate psychiatric symptoms in clinical samples and because affective responses are assumed to drive physical activity therewith promoting active lifestyles for prevention of and recovery from severe somatic and psychiatric diseases. In practice, upcoming evidence on distinct influences of exercise and nonexercise activity on mood might help to specify prevention and treatment of psychiatric disorders. For instance, in an oversimplified scenario, patients suffering from symptoms of fatigue might benefit from increasing the amount of physical activity episodes in everyday life (e.g., climbing stairs instead of using the elevator, walking for transport instead of taking the bus), whereas patients suffering from symptoms of inner tension might benefit from increasing their number of exercise sessions.

Accordingly, to garner a deeper understanding of the associations between physical activity and mood, future studies should consider the distinct components of physical activity, that is, exercise and nonexercise activity, in assessment and analysis.

This study was supported by the Ministry of Science, Research and the Arts of the State of Baden-Wuerttemberg.

A. M.-L. received consultancy fees from: Astra Zeneca, Elsevier, F. Hoffmann-La Roche, Gerson Lehrman Group, Lund-beck foundation, Outcome Europe Sárl, Outcome Sciences, Roche Pharma, Servier International, and Thieme Verlag, and lecture fees—including the travel fees—from: Abbott, Astra Zeneca, Aula Médica Congresos, BASF, Groupo Ferrer International, Janssen-Cilag, Lilly Deutschland, LVR Klinikum Düsseldorf, Servier Deutschland, Otsuka Pharmaceuticals. The content is solely the responsibility of the authors and not influenced by any financial relationship of A. M.-L. with the named organizations. All other authors declare no conflicts of interest. The results of the present study do not constitute endorsement by the American College of Sports Medicine. Furthermore, the results of the present study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.

The authors would like to thank Andreas Hoell, Christina Andras, Alan Schary and Beate Höchemer for recruiting the participants.


1. Aarts E, Verhage M, Veenvliet JV, Dolan CV, van der Sluis S. A solution to dependency: using multilevel analysis to accommodate nested data. Nat Neurosci. 2014;17(4):491–6.
2. Adamo KB, Prince SA, Tricco AC, Connor-Gorber S, Tremblay M. A comparison of indirect versus direct measures for assessing physical activity in the pediatric population: a systematic review. Int J Pediatr Obes. 2009;4(1):2–27.
3. Anastasopoulou P, Tubic M, Schmidt S, Neumann R, Woll A, Haertel S. Validation and comparison of two methods to assess human energy expenditure during free-living activities. Plos One. 2014;9(2):e90606.
4. Bossmann T, Kanning M, Koudela-Hamila S, Hey S, Ebner-Priemer UW. The association between short periods of everyday life activities and affective states: a replication study using ambulatory assessment. Front Psychol. 2013;4:102.
5. Brunswik E. The conceptual framework of psychology. In: Neurath O, Carnap R, Morris CW, editors. International Encyclopedia of Unified Science. Chicago: The University of Chicago Press; 1952. pp. iv–102.
6. Bussmann JB, Ebner-Priemer UW, Fahrenberg J. Ambulatory activity monitoring: progress in measurement of activity, posture, and specific motion patterns in daily life. Eur Psychol. 2009;14(2):142–52.
7. Conn VS. Depressive symptom outcomes of physical activity interventions: meta-analysis findings. Ann Behav Med. 2010;39(2):128–38.
8. Dunton GF, Liao Y, Intille S, Huh J, Leventhal A. Momentary assessment of contextual influences on affective response during physical activity. Health Psychol. 2015;34(12):1145–53.
9. Dunton GF, Liao Y, Kawabata K, Intille S. Momentary assessment of adults' physical activity and sedentary behavior: feasibility and validity. Front Psychol. 2012;3:260.
10. Ebner-Priemer UW, Koudela S, Mutz G, Kanning M. Interactive multimodal ambulatory monitoring to investigate the association between physical activity and affect. Front Psychol. 2013;3:596.
11. Ebner-Priemer UW, Trull TJ. Ambulatory assessment: an innovative and promising approach for clinical psychology. Eur Psychol. 2009;14(2):109–19.
12. Fahrenberg J, Myrtek M, Pawlik K, Perrez M. Ambulatory assessment—monitoring behavior in daily life settings. Eur J Psychol Assess. 2007;23(4):206–13.
13. Gauvin L, Rejeski WJ, Norris JL. A naturalistic study of the impact of acute physical activity on feeling states and affect in women. Health Psychol. 1996;15(5):391–7.
14. Hox JJ. Multilevel Analysis. 2nd ed. New York: Routledge; 2010. p. 392.
15. Kahneman D, Krueger AB, Schkade DA, Schwarz N, Stone AA. A survey method for characterizing daily life experience: the day reconstruction method. Science. 2004;306(5702):1776–80.
16. Kanning M. Using objective, real-time measures to investigate the effect of actual physical activity on affective States in everyday life differentiating the contexts of working and leisure time in a sample with students. Front Psychol. 2013;3:602.
17. Kanning M, Ebner-Priemer U, Brand R. Autonomous regulation mode moderates the effect of actual physical activity on affective states: an ambulant assessment approach to the role of self-determination. J Sport Exerc Psycho. 2012;34(2):260–9.
18. Kanning M, Ebner-Priemer U, Schlicht W. Using activity triggered e-diaries to reveal the associations between physical activity and affective states in older adult's daily living. Int J Behav Nutr Phys Act. 2015;12:111.
19. Kanning MK, Ebner-Priemer UW, Schlicht WM. How to investigate within-subject associations between physical activity and momentary affective states in everyday life: a position statement based on a literature overview. Front Psychol. 2013;4:187.
20. Lawlor DA, Hopker SW. The effectiveness of exercise as an intervention in the management of depression: systematic review and meta-regression analysis of randomised controlled trials. BMJ. 2001;322(7289):763–7.
21. Mata J, Thompson RJ, Jaeggi SM, Buschkuehl M, Jonides J, Gotlib IH. Walk on the bright side: physical activity and affect in major depressive disorder. J Abnorm Psychol. 2012;121(2):297–308.
22. Mead GE, Morley W, Campbell P, Greig CA, McMurdo M, Lawlor DA. Exercise for depression. Cochrane Database Syst Rev. 2008;(4):CD004366.
23. Mellers BA. Choice and the relative pleasure of consequences. Psychol Bull. 2000;126(6):910–24.
24. National Collaborating Centre for Mental Health (UK). Depression: The Treatment and Management of Depression in Adults. Leicester: The British Psychological Society; 2010. p. 705.
25. Netz Y. Type of activity and fitness benefits as moderators of the effect of physical activity on affect in advanced age: a review. Eur Rev Aging Phys Act. 2009;6(19):19–27.
26. Prince SA, Adamo KB, Hamel ME, Hardt J, Connor Gorber S, Tremblay M. A comparison of direct versus self-report measures for assessing physical activity in adults: a systematic review. Int J Behav Nutr Phys Act. 2008;5:56.
27. Puetz TW, O'Connor PJ, Dishman RK. Effects of chronic exercise on feelings of energy and fatigue: a quantitative synthesis. Psychol Bull. 2006;132(6):866–76.
28. Reed J, Buck S. The effect of regular aerobic exercise on positive-activated affect: A meta-analysis. Psychol Sport Exerc. 2009; 10(6):581–94.
29. Reed J, Ones DS. The effect of acute aerobic exercise on positive activated affect: a meta-analysis. Psychol Sport Exerc. 2006;7(5):477–514.
30. Reiner M, Niermann C, Jekauc D, Woll A. Long-term health benefits of physical activity—a systematic review of longitudinal studies. BMC Public Health. 2013;13:813.
31. Rosenbaum S, Tiedemann A, Sherrington C, Curtis J, Ward PB. Physical activity interventions for people with mental illness: a systematic review and meta-analysis. J Clin Psychiatry. 2014; 75(9):964–74.
32. Schlicht W, Reicherz A, Kanning M. Affective reactions to physical activity, exercise and the activities of daily living: A review. In: Mohiyeddini C, Eyseneck M, Bauer S, editors. Handbook of Psychology of Emotions. New York: Nova Science; 2013. pp. 307–24.
33. Schwerdtfeger A, Eberhardt R, Chmitorz A. Is there a correlation between everyday-life physical activity and psychological well-being? A methodological study using ambulatory monitoring. Zeitschrift für Gesundheitspsychologie. 2008;16(1):2–11.
34. Steyer R, Schwenkmezger P, Notz P, Eid M. The Multidimensional Mood Questionnaire (MDMQ). Göttingen: Hogrefe; 1997. p. 33.
35. Stone AA, Shiffman S, Schwartz JE, Broderick JE, Hufford MR. Patient non-compliance with paper diaries. BMJ. 2002;324(7347):1193–4.
36. Stumpp J. Platform for interactive ambulatory psychophysiological assessment. [dissertation]. Karlsruhe; 2014. p. 219.
37. von Haaren B, Ottenbacher J, Muenz J, Neumann R, Boes K, Ebner-Priemer U. Does a 20-week aerobic exercise training programme increase our capabilities to buffer real-life stressors? A randomized, controlled trial using ambulatory assessment. Eur J Appl Physiol. 2016;116(2):383–94.
38. Wilhelm P, Schoebi D. Assessing mood in daily life: Structural validity, sensitivity to change, and reliability of a short-scale to measure three basic dimensions of mood. Eur J Psychol Assess. 2007;23(4):258–67.
39. World Health Organisation. Global Status Report on Noncommunicable Diseases 2014. Geneva: WHO Press; 2014. p. 280.
40. World Health Organisation. Global Recommendations on Physical Activity for Health. Geneva: WHO Press; 2010. p. 58.


© 2017 American College of Sports Medicine