The research team visited 15 elementary schools in the Yangling Agricultural Hi-Tech Industries Demonstration Zone (Shaanxi, Xianyang, China) in mid-March 2022 to identify study partners. Six principals agreed to distribute recruitment messages within their schools. Class teachers then posted electronic recruitment posters in the student guardians’ information exchange group on the WeChat mobile application to call for participants. Guardians could also share these posters with acquaintances. Students interested in participating could have their parents contact the research assistant using the contact information provided in the poster to acquire specific details about the activities. Considering the practical abilities and comprehension of elementary students, we set the following inclusion criteria: third- to sixth-graders with normal vision and hearing; without allergies to fresh-cut flowers; and without physiological, emotional, or latent heart diseases.

We used G*Power (3.1.9.4) to calculate the sample sizes for paired t-tests, one-way analysis of variance (ANOVA), and repeated-measures analysis with p-values and power values of 0.05 and 0.80, respectively (Buchner et al., 2023). Paired sample t-tests and one-way ANOVA with effect sizes of 0.50 and 0.25 required 34 and 36 participants, respectively. With a partial η ² of 0.06, six groups, and four measures, we determined that 42 participants were required for repeated measures analysis. All analyses used a medium effect size.

The participant recruitment process lasted about 3 weeks, and a total of 55 primary students signed up by the recruitment deadline. Some students withdrew for personal reasons during the experimental phase; ultimately, 48 completed the experiment (27 girls, mean age = 10.42 years, SD = 1.05). The study was approved by the Ethics Committee of Northwest A&F University.

The experiment was conducted in the summer mid-semester (April–June 2022) when elementary students were in an ordinary academic period and their mental state was not negatively affected by pressures from approaching final exams. To faithfully replicate the classroom scenes, we employed a quasi-experimental mode. The experiment site was selected as a multi-purpose classroom at the Children’s Cultural Activity Center in Shaanxi. We controlled for the trial site’s environmental factors considering that views, lighting, and thermal conditions can affect indoor occupants’ cognitive performance (Ko et al., 2020). The classroom doors and windows were closed, curtains were drawn, and the central air conditioning’s dehumidification mode and LED educational lighting were turned on 30 min before the daily activity to control for confounding factors in the environment. The indoor temperature, relative humidity, and illumination were maintained at approximately 25°C, 55%, and 640 lux, respectively.

The activities were conducted as horticultural-themed courses from Monday to Saturday in the afternoons after school, with each day featuring a themed activity. Our study employed a within-subject design, and six elementary students took all courses in each session. Students participated in the daily theme activities in two-person groups. The order of group activities each day followed a rotation. After three courses, the members of each group were reorganized to ensure that each participant in the group was unfamiliar with the other. Horticultural classes were randomized before each session, and the order was not disclosed to participants beforehand. The participants were also advised to refrain from engaging in activities similar to those in this study.

We referred to the horticultural activities classification method proposed by Park et al. (2016), excluding “outdoor gardening,” and selected FA, KC, SATS, PFCM, and DBP as the representative activities for the “flower arranging art,” “live plant crafts,” “indoor planting,” “art or pressed flower crafts,” and “other activities (including food making, art, exhibitions, etc.)” categories. Each activity had unique characteristics and objectives (Table 1): The SATS and KC activities were focused on horticultural techniques, whereas the other activities were more inclined toward artistic creation. Further, the textures and color tones of the plant materials were distinctive, providing unique sensory experiences for every activity. We used free composition writing (CW), a typical classroom activity, as a reference activity. The content of the compositions was based on the knowledge and skills the participants had acquired from the course and their varied preferences for activities.

Four researchers conducted the experiment. One taught the course, while the others interpreted questionnaires; wore, adjusted, and removed physiological devices; maintained order; guided students after each session; and coordinated with the parents. Only the course instructor and one researcher who monitored the devices remained on-site after the experiment began.

In the experiment, two students faced each other, with the instructor 1.10 meters to one side. After the daily horticultural knowledge course, we set up a partition that separated the two students and the instructor to prevent interference. The researcher monitoring the devices behind the partition (Figure 1).

The experiment proceeded in four stages (Figure 2). The instructor started the day by teaching horticultural knowledge and giving activity demonstrations (15–20 min). Students took a brief break after the course when two research assistants helped them put on physiological monitoring devices, made necessary adjustments, and informed them of essential considerations. Students then underwent three-minute baseline monitoring and were given 10 min to complete an arithmetic test during the stress induction stages, during which a stopwatch timer was set for the final 3 min of the test to increase stress. The students completed the Positive and Negative Affect Schedule for Children (PANAS-C) immediately after stress induction. A 20-min activity operation stage followed, and students performed independently without mutual communication. After the activity, the research staff removed the physiological monitoring devices, and students completed the PANAS-C again, as well as the Self-Assessment Manikin (SAM) questionnaire, based on their experiences. After a total of approximately 70 min, students received horticultural work as a souvenir and left the classroom. The students’ basic information (age, grade, sex, only child status, prior horticultural activity experience) and informed consent forms were completed before the initial activity began.

SPSS 26.0 software was used for statistical analysis.

ECG and EEG indicators were set as dependent factors, while five horticulture activities and the reference activity were the independent variables. We used a one-way ANOVA to test baseline-dependent variable differences between activities and conducted paired sample t-tests to examine the effectiveness of stress induction. We compared physiological (Question 1) and affective (Question 2) changes among the students between the stress induction and activity stages using paired sample t-tests. To account for individual differences, we converted raw physiological data into percentage change values, as follows:

where Z = relative change (recovery) values (%); X = mean values during the stress induction stage; Y = Raw values during the activity stage; i1 = HR; i2 = RMSSD; i3 = SI; i4 = LF/HF; i5 = α-EEG; and i6 = β-EEG.

Next, we employed a one-way ANOVA to compare the differences in relative recovery from stress among the six activities. Paired t-tests assessed changes in positive and negative affects before and after activities. The Shapiro–Wilk test showed a non-normality of the SAM data, and Kruskal–Wallis analysis was applied (Question 3).

The activity stage was divided into four time phases (T1–T4) at five-minute intervals. A “model-based” corrected “robust estimator” was chosen as the covariance matrix to ensure the robustness of the model. After testing the model’s goodness of fit with different working correlation matrices, we selected the “exchangeable” matrix with the smallest QIC. After selecting the robust working correlation matrix, we removed non-significant interaction terms and reconstructed the equation. Eventually, we employed maximum likelihood estimation with the smallest QICC and constructed a six (activities) by four (time phases) generalized estimating equation (GEE) model to examine the main effects of activities and time phases on the relative change in HR and EEG levels (Question 4). Furthermore, we included sex and prior horticultural activity experience as factors, with age as a covariate in the GEE. We then examined the interactive effects of “activities × time phases,” “activities × sex,” “time phases × sex,” “activities × prior horticultural activity experience,” and “time phases × prior horticultural activity experience.”

All tests determined significance with a p-value below 0.05.

A one-way ANOVA revealed no significant differences in the baseline data between activities for the HR (p = 0.426), RMSSD (p = 0.320), LF/HF (p = 0.099), SI (p = 0.533), α-EEG (p = 0.458), and β-EEG (p = 0.730). Compared with the baseline, the values of HR, LF/HF, SI, and β-EEG increased after the stress induction stage, whereas RMSSD and α-EEG decreased significantly (see Supplementary Table 1). These findings suggested that the arithmetic tests raised the students’ physiological stress levels effectively.

The HR, RMSSD, SI, and EEG results indicated that 20 min of horticultural-related activities could significantly reduce physiological stress in elementary students. Previous research has also examined the correlation between engagement in horticultural activities and individuals’ physiological stress levels, employing various indicators as measures. Hassan et al. (2018) reported significant decreases in older women’s blood pressure after a 15-min flower planting activity compared with a control group. Another study that used flower appreciation and farming work reported decreased salivary cortisol levels and increased oxytocin and parasympathetic nervous system activity in older adults (Watanabe et al., 2021). The results of our study not only align with previous work focused on older populations but also provide evidence for a younger cohort. Additionally, these results are similar to those from research on stress recovery through various forms of contact with nature (Abd-Alhamid et al., 2020; Yin et al., 2020). Horticultural-related activities serve as a medium for connection with nature, which can help lower physiological stress in elementary students and enable their transition from tension anxiety to a state of calmness and relaxation. An evolutionary theory position may explain these effects: Humans are subconsciously drawn to nature because savannahs and environments where water was present helped them survive (Kaplan and Talbot, 1983). The type of non-taxing attention paid to nature and plants is a key mechanism for people to recover from mental fatigue. This study’s results also support the psycho-evolutionary framework that holds that exposure to non-threatening nature after a stressor leads to more positive emotions and lower physiological arousal (Ulrich et al., 1991). Furthermore, gardening could achieve training effects similar to other activities but with less perceived physiological stress (Tao et al., 2022).

We also found lower physiological stress during the CW activity than in the pre-test. The reason for this result might be that the stress induction had a practical effect, and using a timer more vividly replicated the real-life scenario at school. Notably, although the CW was a learning activity, it was not a test. Huang et al. (2020) demonstrated that even in environments without vegetation, moderate viewing duration and relatively rich scenes have a therapeutic effect.

Increased LF/HF indicates a stressed state. We noted an inconsistency in the LF/HF for the FA and KC activities compared with other ECG indicators, implying that the horticultural activities both improved and exacerbated stress among the students. Differences have also emerged in previous research, possibly owing to horticultural activities triggering more complex physiological responses (Li et al., 2020). Another possible explanation relates to the intricate balance control mechanism of the autonomic nervous system—systematic errors may account for the aforementioned result. Thus, physiological indicators require a multi-angle assessment, and subjective reports should be combined to comprehensively evaluate the recovery effects provided by different activities.

Although the five horticultural activities significantly relieved stress, we found no significant difference in their recovery potential across indicators. Similar results have been found by other studies examining the benefits of different types of horticultural activity for patients with dementia and children with maladjustment issues (Gigliotti et al., 2004; Lee et al., 2018). One possible explanation is the intricate nature of physiological intrinsic feedback. The difficulty levels, metabolic equivalents, and research environments among horticultural activities are similar. Consequently, the effects of activities on physiological indicators may not differ significantly in terms of sympathetic or central nervous system regulation. The relatively short intervention duration may also be a factor. Nonetheless, we noted the following three points. First, horticultural-related activities yielded better recovery than the reference activity. Certain activities recorded significant differences in HR indicators. Thus, the horticultural activities had inherent characteristics that may make them more effective in alleviating students’ stress. Second, previous studies have suggested that KC relaxed older adults more than SATS (Tu et al., 2020). As such, the stress-relief benefits of different horticultural activities cannot be generalized across age populations, given the variations in the difficulty of these activities and the traits of the participants. The indicator trends in phases T2–T3 (Figure 3) were related to activities that involved planting seedlings in a soil ball, twining, and fastening twine, which required hand coordination and may have activated the students’ sympathetic nervous system, complicating their stress recovery. In China, theory is prioritized over practice for contemporary elementary students, weakening their hands-on skills (Qiao and Zhou, 2020). Meanwhile, the operation of SATS is relatively simple. Third, the EEG results showed that the FA activity had the best relative recovery effect, which was inconsistent with the ECG results. This discrepancy could be attributed to the fact that ECG and EEG data reflect feedback from sympathetic nerves in the cardiovascular system and central nerves in the cortex, respectively. The α-EEG indicates occipital and parietal cortex activity involved in visual processing (Hsu and Wang, 2013; Rehman and Al Khalili, 2022), and the bright, pastel colors of the FA activity materials may have boosted children’s brain waves by enhancing their visual perception.

The horticultural activities were related to significantly improved relative recovery among the students compared with the reference activity during all time phases except for the HR in the T2 and T4 phases of the KC activity. Thus, the advantages of horticultural activities for stress relief may not have been limited to one or a few time phases, but rather all stages of the activities.

Trends in stress recovery were inconsistent across the horticultural activities, although the initial 5–10 min of these activities exhibited superior recovery benefits compared with the later phases. As we did not focus on the mechanisms of stress recovery, we could not confirm whether this may be due to the inherent factors of the activity (e.g., complexity and difficulty of steps) eliciting more intricate physiological changes and inducing confounding effects on the recovery outcome in the later stage. A single horticultural activity for children typically lasts 60 to 120 min (Park et al., 2016), with little evidence linking the duration to recovery efficiency. Longer or shorter experiments need to be conducted to identify the optimal duration of activity for restoration.

We also found that girls recovered 3.68% more than boys in terms of their HR. Although we could not assess the underlying mechanism of this result, we could assume the role of biological and social differences between girls and boys in their varied stress responses (Jiang et al., 2014). When stressed by grades or performance, men may feel more pressure than women (Jin et al., 2023). Thus, horticultural activities may provide girls with a greater recovery effect in the same time frame as boys.

In the GEE for α-EEG, the interaction between prior horticultural experience and activities was significant, indicating that horticultural experience potentially impacted the stress recovery of elementary students. For students lacking horticultural experience, the FA, KC, and DBP activities demonstrated higher effectiveness compared with SATS and PFCM activities. Our findings confirmed that elementary students recover better when interacting with actual plants and materials that resemble nature. As previously reported, “fascination” in a restorative environment is a crucial element that elicits involuntary attention (Kaplan and Kaplan, 1989). The delicate colors of the FA materials and the carefree posture of Asparagus setaceus seedlings, which closely resemble the layered structure of natural vegetation, undoubtedly served as engrossing objects for the children. This attraction may have been even stronger for children without prior experience with similar activities, providing a plausible explanation for the aforementioned difference. Through patterns of “living plant crafts” (e.g., the KC activity), indirect contact with nature is made interesting. People typically achieve closeness to nature by playing domestication roles (e.g., gardening and caring for pets), thereby enhancing their “compatibility” with nature (Kaplan, 1995).

Elementary students found the FA and KC activities more effective in improving their positive emotions and providing greater happiness and enjoyment than the other horticultural activities. The petals of the Eustoma russellianum, Dendranthema morifolium cv. “pompon,” and Rosa “Barbie Bubbles” have gentle hues (cool and intermediate) and textures, whereas the Eucalyptus pulverulenta “Baby Blue” has an invigorating aroma that further enhances visual, tactile, and olfactory sensations. These elements may create a cozier, more serene atmosphere and promote a subliminally calming effect in children (Koga and Iwasaki, 2013; Neale et al., 2021), while also evoking positive emotions (Xiao et al., 2021). Meanwhile, students also favored horticultural activities that incorporated live plants, indirectly supporting Wilson’s biological and genetic viewpoint that “humans possess a fundamental, genetically-based desire and inclination toward living and lifelike entities, even with minimal connection to nature” (Wilson, 1986). Playing with mud comes naturally to most children, and this might be why students considered molding soil balls into moss balls intriguing. Dirt, water, fallen leaves, and other materials in the natural world offer children valuable opportunities to play and learn, encouraging them to explore, think critically, and seek out more engaging activities (Nicholson, 1972). This also indicates that elementary students’ emotional and affective recovery varied depending on the content and materials of the horticultural activities.

While our study confirmed the positive effects of horticultural-related activities on the psychophysiological stress of elementary students and the differences in recovery benefits across different activities, some limitations warrant further research. First, although the sample size met the estimates of G*Power, a larger and more diverse sample, including a wider age range, is necessary for more comprehensive conclusions. Second, the gender ratio is close to 1:1, which may affect the research sample’s representativeness and the results’ reliability and generalizability. Therefore, there is a need to control for gender proportions. Third, we did not fully consider demographic factors. Family environments and backgrounds (e.g., parents’ income, single-parent households, and intergenerational care) may also affect the dependent variables, as children’s preferences often mirror those of their parents (Meidenbauer et al., 2019). Finally, all study participants were from urban regions. However, rural children may exhibit differing cognition, stress tolerance, and health behaviors because of their natural growth environment (Hope and Bierman, 1998).