How nature nurtures: Amygdala activity decreases as the result of a one-hour walk in nature – Molecular Psychiatry


Participants were recruited from the Castellum database of the Max Planck Institute for Human Development in Berlin, via mailing lists of universities in Berlin, and through the online platform Participants were told that they would take part in an MRI study and that they would go for a walk, but were not informed about the research question of the study. All participants were fluent in the German language, right-handed, and were not diagnosed with any psychological or neurological disorders. A sample size estimation using G*Power resulted in the need of 54 participants to enable a medium effect size. We tested 9 participants more to ensure that potential drop outs would not reduce the sample size below the number we decided on. The final sample consisted of 63 participants (29 females, total mean age = 27.21 years, SD = 6.61, age-range = 18−47 years). The participants were pseudo-randomly assigned either to a nature (32 participants) or an urban walk (31 participants), while controlling that men and women were equally distributed in both environments. During randomization, it was also controlled that the number of afternoon walks were equally distributed between conditions. An overview over the control variables in the two conditions is shown in Supplementary Table 1.

The study was approved by the Local Psychological Ethical Committee at the Center for Psychosocial Medicine at University Medical Center Hamburg-Eppendorf in Hamburg, Germany (LPEK-0054). We obtained written informed consent from all participants and they received monetary compensation for the participation in the study.

Study procedure

The experiment was conducted in late summer/fall 2019 during daylight, between 10:00 a.m. and 5:00 p.m. The flowchart of the study procedure is shown in Fig. 1. Upon arrival, participants signed the informed consent, filled out the questionnaires, and performed a working memory task. Subsequently, the participants underwent an fMRI scanning procedure that included questions on rumination [23], the Fearful Faces Task (FFT) [24], and the Montreal Imaging Stress Task (MIST) [25]. The MIST was administered in order to induce social stress, since the SRT [13] hypothesizes that nature’s restorative potential is most evident when the individual is in a stressed state. The order of the FFT and MIST was counterbalanced between subjects, however, the order was the same within subjects, at pretest and posttest.

Fig. 1: Flowchart of the study procedure.
figure 1

Before the walk participants filled out questionnaires and underwent the fMRI scanning procedure, which included the Fearful Faces Task and the Montreal Imaging Stress Task. Subsequently, each participant was randomly assigned to a 60-min walk, in either a natural or urban environment. After the walk, the participants underwent the fMRI scanning procedure again and filled out the questionnaires.

After the scanning session, participants were randomly assigned to a 60-minute walk in either a natural or urban environment (Fig. 2). Even though the definition and also the dichotomy of ‘natural’ and ‘urban’ environment has been an object of debate [26], the ‘natural environment’ we refer to is an urban forest, the largest green area in the city of Berlin (Grunewald forest; Fig. 2b), whereas the ‘urban environment’ refers to a busy street in one of the city centers in Berlin with shopping malls (Schloßstraße; Fig. 2c). As recommended in the recent review [27], geographic locations of the walk and the landscape features of the environments are reported (see Supplementary information).

Fig. 2: Location of the nature and urban walk.
figure 2

a GPS data of two participants during the walk in the natural environment (Berlin, Grunewald) and the urban environment (Berlin, Schloßstraße) displayed on the OpenStreetMap ( b Sample picture of the walk in the natural environment. c Sample picture of the walk in the urban environment.

The participants were shown the exact walk route on a map (straight path) and they were collected at the lab and brought by taxi to the starting point of the walk. They carried a mobile phone that logged participants’ global positioning system (GPS) data during the walk, to ensure that they walked the intended route (Fig. 2a). During the walk, participants were equipped with an Empatica E4 (Empatica S.r.l, Milan, Italy), a wristband measuring electrodermal activity (EDA), heart rate variability (HRV), and heart rate, as physiological indicators of stress. Participants went on the walk alone and were instructed not to enter shops or use their mobile phones, to avoid potential distraction. They were given a bagged lunch that they could eat during the walk. After 30 min, as an alarm signal was generated by the phone, they turned around and continued the walk back to the starting point. Here they were picked up by a taxi and brought back to the lab.

After the walk the same fMRI scanning procedure was repeated, with one additional stress-inducing task, the Social-Evaluative Threat task (SET) [28], a modified version of the Trier Social Stress Test [29], meant to induce social stress and presented only after the walk, since we reasoned that the participants would not have believed the cover story twice (for detailed SET task procedure see Supplementary information). Additionally, the participants reported the level of restored attention after the walk via a questionnaire. Finally, the participants were debriefed and informed about the aim of the study. Within the scope of this article, we report on the fMRI results on the FFT and the MIST.

Functional imaging paradigms

Fearful Faces Task (FFT)

An adapted version of the Fearful Faces Task (FFT) [24] was used, designed to measure amygdala activity during fearful and neutral facial expressions. While in the MRI scanner, participants were presented with stimuli, consisting of 15 male and 15 female faces, each depicting fearful (Fear condition; Fig. 3 bottom left) or neutral facial expression (Neutral condition; Fig. 3 bottom right). Both fearful and neutral facial expressions were shown either for 1000 ms (unmasked stimuli) or for 17 ms followed by a mask with neutral facial expression presented for 983 ms (masked stimuli). Since the amygdala has been shown to respond to masked stimuli even when most of the participants were not aware of their presence [30,31,32], we used masked stimuli in order to exploratorily examine whether the degree of conscious perception had an effect on amygdala activity. However, we did not have the time to perform a perceptual control test and therefore we have no proof that the masked stimuli were actually processed outside of the participants’ awareness.

Fig. 3: Bilateral amygdala activity during the Fearful Faces Task before and after the walk in the urban and in the natural environment.
figure 3

a a Bilateral amygdala activity while watching fearful faces (Fear condition) decreased after the walk in the natural environment. b Bilateral amygdala activity while watching neutral faces (Neutral condition) decreased after the walk in the natural environment. c Region of interest, the bilateral amygdala as defined in Automated Anatomic Labelling Atlas 2. Bottom: Stimuli in the Fearful Faces Task showing fearful facial expression, within the Fear condition (left) and neutral facial expression within the Neutral condition (right). Note: BOLD stands for Blood-Oxygen Level-Dependent; Significant differences are indicated with asterisks (*P < 0.05; **P < 0.01); error bars represent one standard error of the mean.

We used the set of 60 stimuli from the FACES database by the Max Planck Institute for Human Development in Berlin [33], consisting of face photographs on a gray background, matched on size and luminance. We used the FACES database because it provides a large set of validated high-resolution photographs with natural facial expressions that vary by gender, age, and emotion. The fMRI paradigm consisted of 22 blocks with 6 pictures interleaved with a 200 ms break between pictures. Each block was followed by a white fixation cross presented for 9 s. In order to monitor the participants’ attention, the fixation cross was red on two occasions, and participants were instructed to press the button on the response box as soon as they would see the red cross on the screen. The order of the stimuli was randomized within 10 versions of the FFT, and the task version was introduced in the fMRI data analysis as a covariate. The whole task sequence lasted 8 minutes and 28 s. The task was presented via a projector and mirror system and the participants answered using a response box. The FFT was presented using software Presentation (version: 19.0) and the code for the task used in this study is openly available at

Montreal Imaging Stress Task (MIST)

The Montreal Imaging Stress Task (MIST) [25] is a computerized fMRI-adapted paradigm, based on the Trier Social Stress Test [29], with an aim to induce social stress, in which participants solve mental arithmetic tasks with a time limit designed to be just beyond the participant’s cognitive capacities. The MIST consisted of three different conditions: Experimental, Control, and Rest (Supplementary Fig. 1).

In the Experimental condition, the information about individual performance and a fake-average performance of all participants was graphically presented after each response. This fake-average performance was consistently considerably better than the individual performance in order to induce social stress. In the Control condition, the mental arithmetic tasks had the same level of difficulty as in the Experimental condition, but the participant’s performance as well as the fake-average performance of all participants was not displayed and there was no time limit for solving the task. In the Rest condition, treated as a baseline, no task was displayed and the participants were asked to simply passively look at the screen [25]. For detailed MIST procedure see Supplementary information.

Magnetic Resonance Imaging

Data acquisition

All images were acquired on a Siemens Tim Trio 3 T scanner (Erlangen, Germany) using a 32-channel head coil. The T1-weighted images were obtained using a three-dimensional T1-weighted magnetization prepared gradient-echo sequence (MPRAGE; repetition time (TR) = 2500 ms; echo time (TE) = 4.77 ms; TI = 1100 ms, acquisition matrix = 256 × 256 × 192, flip angle = 7°; 1 x 1 x 1 mm3 voxel size). Whole brain functional images were collected using a T2*-weighted echo-planar imaging (EPI) sequence sensitive to BOLD contrast (TR = 2000 ms, TE = 30 ms, acquisition matrix = 216 × 216 × 129, flip angle = 80°, slice thickness = 3.0 mm, distance factor = 20%, FOV = 216 mm, 3 × 3 × 3 mm3 voxel size, 36 axial slices, using GRAPPA).

Data preprocessing

Functional imaging data were preprocessed and analyzed using Statistical Parametric Mapping software (SPM12; EPIs were corrected for slice timing and head motion and transformed into the stereotactic normalized standard space of the Montreal Neuroimaging Institute (MNI) using the unified segmentation algorithm. Finally, spatial smoothing with a 6-mm full width at half-maximum (FWHM) Gaussian kernel was performed. The voxel size was not changed during preprocessing but kept in the original acquisition dimension (3 × 3 × 3 mm3).

Data analysis

At the first level analysis of the FFT estimates of functional activation during conditions (unmasked Fear, unmasked Neutral, masked Fear, masked Neutral, Response) were obtained using an event-related paradigm. A high-pass filter (cut-off 128 s) was applied. Subsequently, a whole brain analysis was performed, using flexible factorial design with a focus on the interaction of environment (urban vs. natural) and time (before vs. after the walk). Both interaction contrasts were analyzed (Fear > Neutral and Neutral > Fear), using family-wise error (FEW) correction with a threshold at P < 0.05, and no significant clusters survived. Additionally, in order to perform a whole brain analysis with less rigorous threshold, the contrasts were thresholded at P < 0.001, uncorrected while controlling for multiple testing on the cluster level using 3DClutSim in AFNI (Analysis of Functional Neuroimages) [34] and again no significant clusters survived.

We then used a ROI-based approach, based on our a priori hypothesis, focusing on ROI amygdala, ACC (both derived from the Automated Anatomic Labelling atlas 2 [35],, and dlPFC (left and right frontal superior gyrus), derived from the SPM Anatomy Toolbox [36], using WFU PickAtlas ( Volume of the bilateral amygdala was 3744 mm3, dlPFC volume was 79,968 mm3, and ACC volume was 21,704 mm3. We extracted mean BOLD signal from a time window of 4–6 s after stimulus onset across all voxels within each ROI using a Matlab script based on the marsbar toolbox (version 0.44 [37]). We reasoned that the intervention, namely a one-hour walk, would globally affect the stress level and therewith stress-related brain activity, not only when contrasting the Fear > Neutral condition. To test this, we examined activity of each ROI (bilateral amygdala, dlPFC, and ACC) in Fear and Neutral condition separately. Since the results in both conditions were similar, we also examined pooled ROI activity of Fear and Neutral condition. We averaged data from unmasked and masked stimuli, because the results were similar.

We conducted a two-way mixed ANOVA with environment as a between-subject factor (urban vs. natural) and time as a within-subject factor (before vs. after the walk), in the Fear and Neutral condition separately, and also in the ROI pooled activity of Fear and Neutral conditions, while focusing on environment-by-time interaction. Two-tailed post-hoc t-tests were performed within the urban and the natural environment to examine the differences in ROI activity before and after the walk in each environment as well as separately within Fear and Neutral conditions, and the pooled activity of the latter conditions. Additionally, the amygdala subregions (centromedial and laterobasal amygdala) were derived from an atlas of the SPM Anatomy Toolbox [36] and the two-way mixed ANOVA was performed in the same ways as described above.

At the first level analysis of the MIST we obtained estimates of functional activation during the three conditions within a block-design paradigm (Experimental, Control, and Rest) and applied a high-pass filter (cut-off 520 s). We first performed a whole brain analysis, using flexible factorial model and focusing on the interaction of environment (urban vs. natural) and time (before vs. after the walk). Both interaction contrasts (Exp > Cont and Cont > Exp) were analyzed, using family-wise error correction with a threshold at P < 0.05 and no significant clusters survived. Subsequently, to present a more lenient thresholding, the contrasts were thresholded at P < 0.001, uncorrected while controlling for multiple testing on the cluster level using 3DClustSim in AFNI [34]. Significant clusters within the Experimental > Control contrast are shown in the Supplementary Table 2. No significant clusters survived within the Control > Experimental contrast.

To analyze ROI activity within the MIST, we extracted the beta values within each ROI separately for the contrasts Experimental > Rest and Control > Rest, in order to obtain the beta values in the Experimental and Control condition relative to baseline (Rest condition). Subsequently, a 2 x 2 x 2 mixed ANOVA was conducted with condition (Experimental vs. Control) and environment as a between-subject factor (urban vs. natural) and time as a within-subject factor (before vs. after the walk) for the amygdala activity, also focusing on environment-by-time interaction. Additionally, and in accordance with how the FFT data was analyzed, post-hoc t-tests were conducted with pooled amygdala activity of the Experimental and Control condition as a dependent variable in order to examine if the environment-by-time interaction was driven by a change in the amygdala activity after the walk in the urban or in the natural environment.

Behavioural data and Physiological data are reported in the Supplementary information.

Leave a Comment

Your email address will not be published.