Trees, grass, or concrete? The effects of different types of environments on stress reduction

Landscape and Urban Planning 193 (2020) 103654

Contents lists available at ScienceDirect

Landscape and Urban Planning journal homepage: www.elsevier.com/locate/landurbplan

Research Paper

Trees, grass, or concrete? The effects of different types of environments on stress reduction

T

Qiuyun Huanga,b, Minyan Yanga, Hao-ann Janea, Shuhua Lia, , Nicole Bauerb ⁎

a b

Department of Landscape Architecture, School of Architecture, Tsinghua University, China Swiss Federal Institute for Forest, Snow and Landscape Research WSL, Switzerland

ARTICLE INFO

ABSTRACT

Keywords: Health and well-being Restorative environment Skin conductance Stress reduction Vegetation types Virtual reality immersion

Despite a number of studies about how natural environments can affect human health and well-being, few have examined the potentially different effects of different types of vegetation. We therefore designed and conducted a randomised controlled experiment to identify the restorative potential of different types of trees and grass in an urban virtual reality (VR) environment. We initially induced stress in participants (n = 89) by asking them to complete a 5-minute math test while listening to noise, after which they were randomly assigned each to one of three VR environments: a courtyard with grass, a courtyard with trees, or a courtyard devoid of any vegetation. Participants were immersed in the VR experience for 10 min, during which we measured skin conductance levels (SCLs). We also assessed participants’ positive and negative affect scores at baseline, before VR immersion, and after VR immersion. Repeated-measures analysis of variance with a general linear model indicated that the grassy environment had the greatest effect on positive affect. SCLs during the second half of VR immersion were significantly lower in both the grass and tree environments than in the concrete-only environment. Our results are consistent with Ulrich’s theory that unlearned factors of evolutionary origin influence affective responses to environments. Our findings provide preliminary practical evidence for landscape planning that can maximise the restorative effects of urban environments.

1. Introduction Amid the recent global trend of rapid and intense urbanisation, exposure to traffic noise, pollution, and stress have been shown to increase the incidence of psychiatric disorders (Verheij, Maas, & Groenewegen, 2008). Moreover, an increasing number of studies have recently confirmed a positive relationship between natural environments and physical and psychological well-being in urban contexts (e.g., de Vries, Verheij, Groenewegen, & Spreeuwenberg, 2003; Maas, Verheij, Groenewegen, de Vries, & Spreeuwenberg, 2006). Landscape planners, designers, and researchers have therefore sought to create urban green spaces with specific characteristics that can support or restore human well-being (Kaplan, 1995; Olmsted, 1865; Ulrich et al., 1991; von Lindern, Bauer, Frick, Hunziker, & Hartig, 2013). Well-being refers to a state of having one’s physical, psychological, and social needs met, regardless of external circumstances (Dodge, Daly, Huyton, & Sanders, 2012), and can be measured as life satisfaction or positive versus negative affect (Headey & Wearing, 1991). Restoration is the process through which individuals renew depleted

physical, psychological, or social resources (Ulrich et al., 1991; Velarde, Fry, & Tveit, 2007; von Lindern et al., 2013) furthermore used ‘restoration’ and ‘stress reduction’ interchangeably to refer to a recharge of energy that has been diminished by psychophysiological responses to a stressor. This process involves numerous positive physiological, psychological, behavioural, and cognitive changes (Ulrich et al., 1991). Ulrich (1986) wrote from the perspective of environmental psychology, and his stress reduction theory (SRT), as well as Kaplan and Kaplan (1989) attention restoration theory (ART), have also proven relevant to the study of landscape design in urban spaces. SRT posits a healing power of nature that can reduce individuals’ stress in an unconscious and automatic way (Ulrich, 1993). ART posits that nature has the capacity to replenish certain types of attention of human beings through unconscious and cognitive processes (Kaplan & Kaplan, 1989). Kaplan (1995) furthermore argued that a natural environment that offered visitors a sense of escape, a high level of fascination, enough extent of space or content to experience and explore, and a high level of compatibility with visitors’ demands could restore well-being that had been diminished by directed attention fatigue, which was caused by the

Corresponding author. E-mail addresses: [email protected] (Q. Huang), [email protected] (M. Yang), [email protected] (H.-a. Jane), [email protected] (S. Li), [email protected] (N. Bauer). ⁎

https://doi.org/10.1016/j.landurbplan.2019.103654 Received 5 January 2019; Received in revised form 28 August 2019; Accepted 1 September 2019 0169-2046/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al.

use of individuals’ energy to inhibit the urge to pay attention to distractions (Kaplan, 1983). Similarly, Appleton (1996) prospect-refuge theory argued that environments with clear field of vision and place to hide can significantly increase individuals’ abilities to restore wellbeing (Gatersleben & Andrews, 2013). Such environments can also increase survival by offering early observation of the surroundings and the ability to achieve shelter (Gatersleben & Andrews, 2013; Grahn & Stigsdotter, 2010). However, natural environments and green spaces exhibit a variety of forms, types, colours, and vegetation (Kaplan & Kaplan, 1989), and while numerous studies have confirmed that natural environments are more restorative to human physiological and psychological well-being than are built environments (e.g., Igarashi, Song, Ikei, & Miyazaki, 2014), the specific elements and attributes that are needed to design spaces that improve human well-being remain unclear. Identifying the features that are most relevant to human well-being, and understanding the extent to which different aspects may contribute, thus remain critical tasks for urban landscape planners and designers (Martens, Gutscher, & Bauer, 2011; Tyrväinen et al., 2014). Researchers have variously suggested relationships between well-being and the amount (Jiang, Chang, & Sullivan, 2014; Ward Thompson et al., 2012), degree (Beil & Hanes, 2013), type (Martens et al., 2011; Van den Berg, Jorgensen, & Wilson, 2014), location (Chiang, Li, & Jane, 2017), leaf colour (Elsadek, Sun, and Fujii, 2017), landscape colour (Akers et al., 2012), or prospect-refuge level (Gatersleben & Andrews, 2013) of a natural environment, as well as whether it has or lacks mountains, parks, forests, or sources of water (Arnberger et al., 2018, Hansmann, Hug, & Seeland, 2007; Tang et al., 2017; Tyrväinen et al., 2014; Ulrich, 1981; Van den Berg et al., 2014). However, questions remain about whether different vegetation types (e.g., grass, trees) affect physiological and psychological well-being and restoration differently. Although a few previous studies have explored the restorative abilities of different types of vegetation in a qualitative method or self-reported method (e.g., grass, bushes, trees; Nordh, 2012; Nordh & Østby, 2013), there is a need for more research that uses randomised controlled experiments to directly analyse the physiological and psychological restorative effects of different types of vegetation. In particular, there have been few comparisons of the effects of grass versus trees on human well-being.

urban square, an urban park, a pine forest, a botanical garden, and a peri-urban area. However, neither study used any objective measures of physiological restoration. Although a few studies have used SCLs to measure restoration in different natural environments, few have focused on how different vegetation types—for example, grass and trees—might affect SCLs. For instance, Jiang et al. (2014) reported that SCLs decreased to various extents depending on the amount of greenery when male participants were exposed to a 6-minute natural treatment, and de Kort, Meijinders, Sponselee, and Ijsselsteijn (2006) found that immersion in natural environments had a restorative effect as measured by SCLs, but neither compared the effects of different types of vegetation. CamachoCervantes, Schondube, Castillo, and MacGregor-Fors (2014) posited that trees, which provide shade and may improve scenic views, create a safer, more relaxing space than other types of vegetation. In line with this, we hypothesised that environments with trees would elicit a stronger restorative effect, as measured by SCLs, than environments with only grass. 1.2. Restorative psychological effects of different types of vegetation Researchers have increasingly sought to understand the effects of different types of natural environments on psychological well-being. The primary components of psychological well-being are positive affect (PA) and negative affect (NA; Diener & Suh, 1997), while life satisfaction is a more subjective component of long-term cognitive well-being (cf., Dodge et al., 2012). Due to our experiment’s design, we were only able to measure acute PA and NA responses rather than long-term cognitive measures of life satisfaction. Martens et al. (2011) found significant differences between the restorative effects of wild versus tended forests on psychological wellbeing. Similarly, Chiang et al. (2017) reported that individuals’ positive moods increased significantly inside a forest than outside it or at its periphery, whereas their negative moods increased outside the forest. Van den Berg et al. (2014) likewise observed a significant pairwise difference in recovery from NA after exposure to a city street versus a park, while Gatersleben and Andrews (2013) reported that positive moods increased significantly more for people who walked through high-prospect, low-refuge environments than low-prospect, high-refuge ones. These studies suggest that the restoration of psychological wellbeing is associated with various visual perceptions of different natural settings (Ulrich, 1983). Ulrich (1983) also emphasised that individuals’ cultural experiences can influence their relationships with different environments. We conducted the present study in a Chinese cultural context. Given that Chinese people typically resist stepping on grass in public green spaces in order to avoid damaging it (Zhou & Gu, 2013), we hypothesised that the primary effects of natural environments for Chinese people stem from the environment’s visual features, rather than its tactile or audial ones. Since trees are more visually complex in form and thus more likely to attract people’s interest than meadows, we hypothesised that environments with trees would achieve greater restoration than ones with only grass (Ulrich, 1983). This is also consistent with attention restoration theory, which posits that the attention-grabbing qualities of a natural environment are central to its restorative potential (Hansmann et al., 2007; Kaplan, 1995).

1.1. Restorative physiological effects of different types of vegetation Previous research has sought to compare the physiologically restorative effects of natural versus urban environments. Ulrich et al. (1991) continuously monitored a series of physiological measures, including 10-minute changes in skin conductance response (SCR), in six different natural and urban environments; they found that participants who watched a stress-inducing film exhibited faster and more complete physiological restoration when later exposed to a natural environment than to an urban one. SCR is a localized change in the tonic Electrodermal Activity (EDA) signal. Tonic EDA is also referred as skin conductance level (SCL; Braithwaite, Watson, Jones, & Rowe, 2013). SCLs are an unbiased measure of sympathetic activity via the electric impulses on the skin’s surface and sweat glands, which are innervated only by the sympathetic nervous system (SNS Alvarsson, Wiens, & Nilsson, 2010; Cummings, El-Sheikh, Kouros, & Keller, 2007), whose major function is to stimulate the body’s fight-or-flight response (Ulrich, 1983). A number of studies have therefore used SCLs as a measure of physiological wellbeing. For example, Alvarsson et al. (2010) found that nature sounds elicited faster physiological restoration than noise of lower, similar or higher sound pressure level over a 4-minute period, as measured by SCLs. A few studies have attempted to compare the restorative effects of settings with different degrees and combinations of vegetation. For example, Van den Berg et al. (2014) compared the restorative effects of parkland settings with either tended woodlands or more complex and wild woodlands, while Carrus et al. (2013) compared the effects of different amounts or combinations of vegetation types, including an

1.3. Mechanisms of physiological and psychological restoration Ulrich (1983) suggested that physiological arousal closely relates to neurophysiological activity, affect, thought, and action. Physiological arousal is generated by numerous bodily systems that mobilized people for dealing with the challenges, which uses energy and contributes to fatigue, if prolonged. Restoration can be construed as the recovery from excessive physiological arousal (Ulrich et al., 1991). Ulrich’s theory (1983) argues that affect-related functions and consequences in various natural environments are essential to understanding why different 2

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al.

natural environments can generate different affective responses, and proposes a psychoevolutionary framework that describes the interactions between physiological arousal and emotional responses. Thus, for example, Ulrich et al. (1991) found that an increase in positive feelings was associated with a decline in autonomic physiological responses. Researchers have also explored both physiological and psychological responses to different natural environments. de Kort et al. (2006) found that more immersive natural environments had stronger physiological restorative effects, as assessed by SCLs (a measure of physiological arousal), but had similar psychological restorative effects, as assessed by measures of affect, compared to less immersive environments. However, to date, although a few studies have explored the restorative abilities of different types of vegetation in a qualitative method or self-reported method (e.g., grass, bushes, trees; Nordh, 2012; Nordh & Østby, 2013), more randomised controlled experiments need to be conducted to compared the restorative effects of different types of vegetation and interpreted the results using this psychoevolutionary framework. The theorised relationship between physiological arousal and psychological well-being in different natural environments has also not been fully substantiated. We therefore sought to analyse the interactions of arousal and affect in different environments with various vegetation types to elucidate how various natural environments foster stress restoration and well-being in humans. In particular, we sought to pinpoint the restorative effects of different types of vegetation on stress in humans by answering three research questions:

test was used to represent their SCL under stress. Participants next completed a second PANAS assessment, after which each participant was randomly assigned to one of three immersive three-dimensional VR environments in which the participant wore an Oculus SDK2 headset for 10 min. We calculated each participant’s mean SCL for each minute during each of the first 9 min of the immersion process (first minute, second minute, … , ninth minute); we excluded data from minute 10 of the VR immersion because the experimenters’ help with detaching the headset could have elicited introduced bias. Following VR immersion, participants completed the third and final PANAS assessment, after which they removed the SCL measurement equipment and moved to another room, where we debriefed them and asked about any problems encountered during the experiment. 2.3. Participants We initially set a target of 30 participants for each group, for a total sample size of 90 participants. This was determined to meet or exceed the minimum sample size needed according to the central limit theorem: as the sample size increases above 30, the sampling distribution of the mean approaches a normal distribution; since the underlying assumption for Student’s t-test is that samples should be obtained from a normally distributed population, the sample size should thus be at least 30. We further calculated the required sample size needed for a repeated-measures design to identify within-between interaction effects. Calculations were performed using G*Power software version 3.1.9.4 (Prajapati, Dunne, & Armstrong, 2010). We set the required parameters as follows: partial η2 = 0.06, p = 0.05, power = 0.8, number of groups = 3, and number of measurements = 3. Since there were too few similar studies to identify the effect size, we chose a medium effect size (partial η2 = 0.06) for estimating the sample size. We also selected the option ‘effect size specification as in SPSS’ in G*Power to calculate the sample size, taking into account the correlation among repeated factors. According to these calculations, the total sample size needed for this study was a total of 99 participants, with 33 for each group. Although we had to exclude some participants due to the incomplete data or technical problems, we consider our sample size to be appropriate for this study. Ninety participants were voluntarily recruited from the campus libraries and dormitory halls of Tsinghua University, as well as via group messages using the WeChat mobile application. All of the participants were healthy students and were native Chinese speakers. All potential participants were informed about the experiment procedure, associated risks, and confidentiality issues, and all participants signed an informed consent form before beginning the experiment. The study was conducted in compliance with the Declaration of Helsinki. Data from one participant were excluded due to a computer system failure during the experiment, so our sample ultimately consisted of 89 participants. The participants included 44 men and 45 women; their average age was 23.0 years old (SD = 2.8). We randomly assigned each participant to one of three immersive VR environments: a courtyard without nature (n = 29, 14 men), a courtyard with trees (n = 30, 15 men), and a courtyard with grass (n = 30, 15 men). We did not inform participants about what they would experience before they donned the VR headset. Each participant was immersed in only one VR environment. Upon completion of the experiment, each participant received a potted plant as compensation for their participation.

1. Do different types of vegetation (i.e., grass and trees) influence physiological restoration differently? 2. Do different types of vegetation (i.e., grass and trees) influence psychological well-being differently? 3. How do physiological arousal and psychological well-being interact in the restorative processes caused by environments with different vegetation? 2. Methods 2.1. Overview We designed a randomised controlled experiment to explore the psychophysiological restorative effects of various built environments with different types of vegetation in virtual reality (VR) scenarios that we created. Participants (n = 89) first took a mathematics test in a noisy environment to induce mental stress, after which each participant was randomly assigned to one of three VR environments: concrete-only (concrete), trees, or grass. Participants were immersed in the VR environment for 10 min. To assess and compare their physiological stress responses, we continuously measured their SCLs during the experiment. We also compared their positive and negative affective states using the positive and negative affect schedule (PANAS; Watson, Clark, & Tellegen, 1988) assessment at three different timepoints during the experiment (Fig. 1). 2.2. Procedure The experimental procedure and the timing of the PANAS measurements are shown in Fig. 1. First, we briefly explained the experiment to potential participants and obtained their written informed consent to participate. Participants then completed a short demographic questionnaire, after which they were brought to a laboratory where we attached electrodes to their skin to measure their SCLs; we monitored and measured their SCLs continuously throughout the rest of the experiment. Next, participants were asked to calm themselves for 5 min; their mean SCL during this time was used as their baseline SCL. They then completed an initial PANAS assessment, the results of which served as the baseline measure of their affect. They were then tasked with completing a revised version of the Markus and Peters arithmetic (MPA) test (de Kort et al., 2006) to stimulate stress, and their mean SCL during the

2.4. Stressor task To stimulate stress, we asked participants to complete a 5-minute abbreviated version of the MPA test (Peters et al., 1998), which previous research has shown can effectively induce measurable stress, in terms of both SCL and NA (de Kort et al., 2006), in a noisy environment. We told participants that the experiment’s purpose was to assess their performance on the MPA and that we would score and rank their performance. The test consisted of one test run and five scoring sections 3

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al.

Fig. 1. Experimental procedure.

that each contained three mathematics questions written in English from the hard question set of the Graduate Record Examinations (GRE; Warner, 2016), which is a standardised test that is required for admission to most graduate schools in the United States. Participants were allowed a maximum of 1 min to complete each of the five sections. Participants were told that each subsequent section would be accompanied by noise only if they correctly answered all of the questions on the previous section; however, in reality, every section was accompanied by noise, which we revealed to participants after they had completed the full experiment.

In the environments with vegetation, we controlled the number of natural elements to be approximately equivalent using Jiang et al. (2014) method to accurately determine the quantity of visible natural elements. Accordingly, we used Adobe Photoshop CS6 to create crossview screenshots for each scenario, from which we identified the numbers of pixels that were associated or unassociated with vegetation. To prevent bias from any perceived differences in the quantity of visible natural elements in the VR environments versus the two-dimensional screenshots, we used cross-view (Figs. 2–4) screenshots instead of direct views to calculate the visible natural elements. We thus established the visible greenery ratio (VGR) as the number of pixels associated with vegetation divided by the number of pixels in the entire cross-view, multiplied by 100 to achieve a percentage value. Accordingly, the VGRs of the environments without nature, with grass, and with trees were 0%, 30.44%, and 27.19%, respectively. We considered the VGRs of the environments with grass and trees to be approximately equivalent. In the post-experiment interviews, we asked participants to evaluate the quality of the VR environment, and we received a total of 114 comments on this topic. Most (63.16%) were positive, such as ‘nothing uncomfortable’. Some (20.18%) reported minor problems, such as slightly blurred vision or mild nausea. A few comments related to more significant problems, such as ‘eyestrain’ (1.75%) or ‘the VR setting was not as real as I expected’ (1.75%).

2.5. VR environments To examine the psychological and physiological effects of nature in built environments, we designed and constructed three different VR environments: an enclosed courtyard without nature (Fig. 2), an enclosed courtyard with grass (Fig. 3), and an enclosed courtyard with trees (Fig. 4). All three environments were created using the VR modelling software Unreal Engine 4. The only difference between the VR environments was the type of vegetation, if any, in the courtyard. The courtyard without nature consisted of a hollow space with a ground surface composed of gravel and surrounded by walls; the courtyard with grass was the same courtyard but with a grass lawn; and the courtyard with trees contained a row of tall trees against the far wall of the otherwise barren courtyard, at and above eye level. Since the spaces were shaped and defined by their vegetation, in the environment with trees we set the tree array along the far edge of the courtyard to make the central space appear as large as in the other two environments. Although the proportion of sky in the environments appears different in the still images (Figs. 2–4), the Oculus headset displayed the entire sky, which occupied the largest portion of each virtual environment and was approximately equivalent in each. The visual characteristics of all other elements in each environment (e.g., sunlight, walls, ground surfaces, solid constructions) were identical.

2.6. Measurements 2.6.1. Physiological measures We used a biofeedback device in a Biopac MP150 workstation, and its accompanying AcqKnowledge software, to monitor, measure, and record SCL parameters. SCL is the most common measure of general tonic-level electrodermal activity, and variations in SCL reflect general changes in arousal (Braithwaite et al., 2013). Previous research has demonstrated that SCL is an effective indicator of physiological stress,

Fig. 2. VR environment. Courtyard without nature (i.e., concrete group) and 0 pixels (0%) of visible vegetation. 4

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al.

Fig. 3. VR environment. Courtyard with grass (i.e., grass group) and 1,135,295 pixels (30.44%) of visible vegetation.

and it is often used in research about stress recovery (Alvarsson et al., 2010; de Kort et al., 2006; Jiang et al., 2014; Li & Sullivan, 2016). We directly measured SCL using the constant voltage technique, in which the Electrodermal Activity Amplifier Module in the Biopac workstation measured absolute SCL every 1 ms (1000 samples/s). To collect SCL data, we applied non-irritating electrode pads followed by Silver-silver chloride (Ag-AgCl) electrodes to the first phalange of the index and middle fingers on participants’ nondominant hands. We also attached an electrodermal activity transmitter from the Biopac workstation to participants’ nondominant wrists. To filter out unrelated noise from body movements, we applied continuous decomposition analysis to the raw electrodermal activity data using the default parametric on Ledalab (Benedek & Kaernbach, 2010). 2.6.2. Psychological measures As previously noted, subjective affect was measured using the PANAS assessment (Watson et al., 1988), which contains 10 words representing PA and 10 words representing NA. Participants indicated the extent to which they felt that specific words described their mood by moving a sliding bar to a position indicating a value from 1 to 100. We analysed both the positive and negative dimensions of PANAS. The measure had good reliability, with Cronbach’s α scores ranging from 0.84 to 0.88. We administered PANAS assessments both before and after the MPA test, as well as after VR immersion.

stressor with a paired sample t-test. To compare the effects of the stressor and the restorative effects of different VR environments, we used repeated-measures ANOVAs (REMANOVAs) in a general linear model. When a violation of the assumption of sphericity was indicated, we used the Greenhouse-Geisser correction. To test whether the type of vegetation in the three environments influenced physiological restoration (Question 1), we calculated a mixedmodel 3 × 3 REMANOVA for SCL. The environment (i.e., concrete, grass, and trees) served as the between-subjects variable, and time (T0, T1, and T2) was the within-subjects variable. To further identify the phases and groups in which significant interactions between environment and time occurred, we examined the interactions between each environment groups (i.e., concrete vs. grass, grass vs. trees, and concrete vs, trees) at each period (i.e., T0-T1 and T1-T2), respectively. To test whether the type of vegetation in the three environments influenced psychological restoration (Question 2), we used REMANOVAs to analyse how the different VR environments affected PA and NA, with PA and NA, respectively, as dependent variables and environment (i.e., concrete, grass, and trees) and time (i.e., post-stressor and post-VR exposure) as the independent variables. Finally, to determine whether each VR environment influenced PA and NA, we analysed PA and NA scores before and after immersion (i.e., poststressor and post-VR exposure) with a paired sample t-test. We excluded all incomplete data.

2.7. Data analysis

3. Results

We used a one-way analysis of variance (ANOVA) to compare baseline and post-stressor SCLs and PA and NA scores among the three groups. We did not expect any significant differences among the groups regarding the likelihood of returning to baseline psychological and physiological levels after experiencing the stress test. However, we expected significant differences in SCL, PA, and NA within each group before and after inducing stress. We examined the effectiveness of the

We present the experiment results in three sections. In the first Section 3.1, we compare the baseline stress levels of the three groups and note any differences among them before the experiment. We also examine the effects of the stressor and compare the groups’ stress levels following exposure to the stressor. In the next two sections, we compare the effects of the different VR environments on stress recovery in terms of SCL (Section 3.2) and PA and NA (Section 3.3).

Fig. 4. VR environment. Courtyard with trees (i.e., tree group) with 1,014,028 pixels (27.19%) of visible vegetation. 5

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al.

indicated no significant differences in mean SCL (p > .32), PA (p > .98), or NA (p > .37) values among the three groups after exposure to the stressor. We were therefore able to attribute later differences in physiological and psychological states to the different VR environments (Table 1).

Table 1 ANOVA results and mean values in skin conductance level (SCL), positive affect (PA), and negative affect (NA) among the three experimental groups, before and immediately after exposure to the stressor. Partial η2

Concrete

Grass

Trees

F

p

SCL Pre-stressor (baseline) M SD n Post-stressor M SD n

3.27 2.13 29 8.41 4.14 29

4.04 2.28 30 9.51 5.62 30

3.48 2.11 30 7.82 2.94 30

0.973 .382 0.022

PA

Pre-stressor (baseline) M SD n Post-stressor M SD n

268.28 174.01 29 209.04 162.83 28

353.69 202.83 29 216.07 163.16 30

333.73 1.626 .203 0.037 188.70 30 218.23 0.023 .977 0.001 180.79 30

NA

Pre-stressor (baseline) M SD n Post-stressor M SD n

81.21 92.61 29 214.79 253.85 28

50.17 65.42 29 157.53 124.08 30

44.17 2.247 .112 0.050 51.69 30 153.47 1.008 .369 0.023 149.96 30

3.2. Effects of different VR environments on SCLs As previously noted, we expected to find significant differences in physiological restoration, as measured by SCL, between individuals in the three different VR environments. To fully capture minute-to-minute changes in SCLs during the experiment, we calculated the mean SCL at baseline, after exposure to the stressor, and at each minute during VR immersion (minutes 1–9). As shown in Fig. 5, although SCLs initially decreased across all three groups, minute 5 was a turning point for participants in the concrete group, after which their SCLs began to increase, while the SCLs of participants in the grass and tree groups continued to decline. We therefore identified minute 5 as the critical dividing point of the VR immersion experiences; we labelled it T1, between T0 (immediately post-stressor) and T2 (minute 9), as shown in Fig. 6. To exclude the influence of variability in baseline SCLs, we calculated the standardised SCL mean as a change in SCL from baseline according to the following equation:

1.145 .323 0.026

Notes. *p < .05. **p < .01. ***p < .001. M = mean value; SD = standard deviation.

MStandardised SCL = MTn

MBaseline (n = 0, 1, 2)

To answer Question 1, we calculated a mixed-model 3 × 3 REMANOVA for SCL, in which environment (i.e., concrete, grass, and trees) was the between-subjects variable and time (i.e., T0, T1, and T2) was the within-subjects variable. The interaction effect of environment × time was significant at the 5% level (F(4, 86) = 2.89, p = .03, η2(Partial) = 0.063), as was the main effect of time alone (F(2, 86) = 72.79, p = .000, η2(Partial) = 0.458). The effect size of η2(Partial) for the significant interaction effect is 0.063, which could be considered as medium based on Cohen’s convention (1988). This indicates that 6.3% proportion of variance in the standardised SCL means is explained by the interaction of time and environment. Similarly, the effect size of the main effect of time is 0.458 that can be considered as medium. However, the main effect of environment alone was not significant (F(2, 86) = 0.06, p = .940, η2(Partial) = 0.001). The significant interaction effects thus suggest significant differences in SCLs among the concrete, grass, and tree groups (Fig. 6; means are presented in Table 3).

3.1. Baseline checks and effectiveness of the stressor Since participants were randomly assigned to the three experimental VR treatments, we expected no significant differences in the physiological and psychological states of the groups prior to stress exposure. ANOVA demonstrated that, as expected, there were no significant differences in the mean SCL (p > .38), PA (p > .20), or NA (p > .11) values among the concrete, grass, and tree groups at baseline (Table 1). As shown in Table 2, the paired sample t-test demonstrated, also as expected, a significant (p < .05) difference in SCL values (t = 13.62, df = 88, p < .001, Cohen’s d = 1.44) between the baseline (mean = 3.60, SD = 2.18) and post-stressor measurements (mean = 8.58, SD = 4.38). Results from the PANAS assessments were consistent with the physiological findings: participants reported higher NA (t = 6.91, df = 86, p < .001, Cohen’s d = 0.741) and lower PA (t = −6.29, df = 86, p < .001, Cohen’s d = 0.674) after exposure to the stressor than at baseline. The effect sizes suggest that there is a large difference between the baseline and post-stressor means for SCL, NA and PA according to Cohen’s convention (1988), since the parameters of Cohen’s d are all close to 0.80. The baseline and post-stress means differed by 1.44 standard deviation for SCL, by 0.741 standard for NA and by 0.674 standard for PA. The stressor therefore clearly produced a stress response in participants, as intended. We also performed an ANOVA to identify any differences in SCLs, PA, or NA across the three groups after stressor. As expected, the results Table 2 Results of paired t-test analysis and mean values in skin conductance level (SCL), positive affect (PA), and negative affect (NA) before and after stress exposure for all participants. Variable SCL PA NA

Pre-stressor Post-stressor Pre-stressor Post-stressor Pre-stressor Post-stressor

n

M

SD

Cohen’s d

t

p

89 89 87 87 87 87

3.60 8.58 315.95 216.17 56.43 175.97

2.18 4.38 189.52 167.66 70.88 183.19

1.44

13.62

.000***

0.674

–6.29

.000***

0.741

6.91

.000***

Notes. *p < .05. **p < .01. ***p < .001. M = mean value; SD = standard deviation.

Fig. 5. Changes in mean skin conductance level (SCL) during the experiment. 6

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al.

3.3. Effects of different VR environments on PA and NA 3.3.1. Positive affect We next used REMANOVA to compare the effects of the different VR environments on PA, with PA as the dependent variable and environment (i.e., concrete, grass, and trees) and time (i.e., post-stressor and post-VR) as the independent variables. With significance set at 5%, the interaction of time and environment was significant with a large effect size (F(2, 84) = 3.60, p = .032, η2(Partial) = 0.079), as was the main effect of time alone with a medium effect size((F(1, 84) = 4.28, p = .042, η2(Partial) = 0.048). However, the main effect of environment alone was not significant (F(2, 84) = 1.28, p = .284, η2(Partial) = 0.030). The difference in PA recovery across the various VR environments before and after VR immersion was therefore significant (Fig. 7). To identify which environments had a significantly different capacity to influence PA restoration, we performed a series of 2 × 2 REMANOVAs between each two-way environment group comparison (i.e., concrete vs. grass, concrete vs. trees, and trees vs. grass) and time. Only the 2 × 2 REMANOVA examining the interaction effects of the concrete and grass environments and time (both post-stressor and post-VR) was significant with a large effect size (F(1, 55) = 7.15, p = .010, η2(Partial) = 0.115). Neither the main effect of environment alone (F(1, 55) = 2.83, p = .098, η2(Partial) = 0.049) nor time alone (F(1, 55) = 2.920, p = .093, η2(Partial) = 0.050) was significant (p < .05). The grass group’s PA recovered far better (mean value increased from 221.79 to 311.79, p = .001) than did that of the concrete group (mean value decreased from 209.04 to 189.21, p = .594), with the tree group in the middle (mean value increased from 218.23 to 250.09, p = .21). To determine whether any of the three environments significantly influenced PA, we next analysed PA scores from before and after VR immersion (i.e., post-stressor and post-VR) with a paired sample t-test. A significant (p < .05) difference was apparent in the grass group, whose mean value increased from 221.79 to 311.79 (t(28) = 3.58, p = .001), but not in the tree group (p = .253) or concrete group (p = .549). These results (Table 4) suggest that only the grass environment significantly influenced PA recovery, although there was no significant difference in PA recovery between the grass and tree groups. We therefore again combined the grass and tree groups into a single green group and compared it with the concrete group using a mixedmodel 2 × 2 REMANOVA analysis, in which environment (i.e., concrete and green) was the between-subjects variable and time (i.e., poststressor and post-VR) was the within-subjects variable. The interaction effect of environment × time was significant at the 5% level (F(1, 85) = 5.12, p = .026, η2(Partial) = 0.057). Neither time alone nor environment alone had a significant effect. This indicates that the environments with natural components—either trees or grass—had a better restorative effect on PA compared to the concrete environment.

Fig. 6. Changes in standardised mean skin conductance levels (SCLs) during the Virtual Reality (VR) immersion experience.

To identify during which phase and between which groups significant interaction occurred, we examined the interactions between environment and time in each phase (i.e., T0–T1 and T1–T2) for each two environments phase. We only found two significant (p < .05) interaction effects: one between the concrete and grass environments at T1 and T2 with a large effect size, (F(1, 57) = 8.23, p = .006, η2(Partial) = 0.126) and another between the concrete and tree environments at T1 and T2 with a medium effect size (F(1, 57) = 5.52, p = .022, η2(Partial) = 0.088). This suggests that although there were no significant differences in the level of decline of the SCLs during the first 5 min of immersion, significant differences emerged during the second half of the VR immersion experiences, both between the concrete and grass groups as well as between the concrete and tree groups. Notably, however, there were no significant differences between the grass and tree groups, although the effect size of interaction between time and environment for comparison of grass and concrete environments is larger than that of tree and concrete environments. We then combined the grass and tree groups into a single ‘green’ group to examine whether the changes in SCLs in the second half of the VR immersion period differed significantly between the green group and the concrete group. We therefore conducted a mixed-model 2 × 2 REMANOVA analysis, in which environment (i.e., concrete and green) was the between-subjects variable and time (i.e., T1 and T2) was the within-subjects variable. The interaction effect of environment × time was significant at the 5% level (F(1, 87) = 9.02, p = .003, η2(Partial) = 0.094). The large effect size (η2(Partial) = 0.094) means that 9.4% proportion of variance of the SCLs is explained by the integration effect of environment and time for the green and concrete environment from T1 to T2. Neither time alone nor environment alone had significant effects. This indicates that the green environments—containing either trees or grass—had a stronger restorative effect on SCLs than did the concrete environment.

3.3.2. Negative affect Finally, we also used REMANOVA to compare the effects of the different VR environments on NA, with environment (i.e., concrete, grass, and trees) as the between-subjects variable and time (i.e., poststressor and post-VR) as the within-subject variable. Although time alone had a significant (p = .05) main effect with a medium effect size (F(1, 84) = 50.334, p = .000, η2(Partial) = 0.375), environment alone did

Table 3 Standardised skin conductance levels (M and SD) during exposure to the stressor and during virtual reality (VR) immersion. Concrete (n = 29)

Standardised skin conductance level

M SD

Grass (n = 30)

Trees (n = 30)

T0

T1

T2

T0

T1

T2

T0

T1

T2

5.14 3.56

0.89 2.64

1.75 3.16

5.47 4.13

1.35 3.46

0.24 4.21

4.34 2.47

1.95 2.70

1.00 3.31

Notes. T0 = stressor (pre-VR); T1 = minute 5 of VR immersion; T2 = minute 9 of VR immersion. M = mean value; SD = standard deviation. 7

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al.

Fig. 8. Negative affect before and after the virtual reality (VR) immersion.

Fig. 7. Positive affect scores before and after the virtual reality (VR) immersion.

environments with natural elements exert greater physiological restorative impacts than do built environments without such elements (Jiang et al., 2014; Li & Sullivan, 2016; Ulrich et al., 1991; Wang et al., 2016). However, contrary to our expectations, we found no difference between the mean level of SCL recovery of participants in the grass and tree groups. However, they did show different trends in the 5–9-minute phase of VR immersion, and future research should conduct experiments with longer immersion periods to assess whether the SCL restoration capacity of grass and trees might differ over a longer period of time. In addition, during the initial 5-minute phase of VR immersion, we observed no significant differences in SCL reduction levels across the three groups. This may be due to the fact that all three environments were likely better than the stress test. Prior research has found that even a concrete environment, given moderate depth and complexity, may elicit some degree of therapeutic influence (Evans, 2003; Frumkin, 2001; Ulrich, 1983), especially when compared to crowded and noisy built environments (Tyrväinen et al., 2014; Ulrich et al., 1991) or monotonous ones, such as built environments with stone walls (Hartig & Staats, 2004; Ulrich, 1984).

not (F(2, 84) = 1.845, p = .164, η2(Partial) = 0.042), and the ordinal interaction between environment and time was also not significant (F(2, 84) = 0.062, p = .940, η2(Partial) = 0.001). These results indicate that there was no significant difference in the changes of NA scores across the three groups (Table 4). The NA scores in all groups changed significantly before and after VR immersion (concrete group: p = .007; grass group: p = .000; tree group: p = .000; Fig. 8). 4. Discussion We sought to examine the influence of different types of vegetation on the restorative potential of urban environments in order to better understand the general effects of urban and natural environments. Our results confirm previous findings that natural environments exert more positive effects on human well-being than do urban ones, and also provide a novel comparison of the differential effects of grass versus trees in urban environments. 4.1. Effects of different vegetation on physiological stress

4.2. Effects of different vegetation on psychological well-being

We observed significantly different levels of SCL recovery between the concrete environment and the environments with either form of vegetation in the 5–9-minute phase of VR immersion, suggesting that both grass and trees are important for the restoration of physiological well-being (Fig. 5). This is consistent with previous findings that

We found that the environment with grass had a stronger effect on PA scores than did the other environments. By contrast, NA scores did not differ significantly across the three experimental conditions. This is consistent with a meta-analysis by McMahan and Estes (2015), which

Table 4 Means of positive affect (PA) and negative affect (NA) after stress exposure and after virtual reality immersion in a paired sample t-test. Variable Group

Group

Time

M

SD

n

t

p

Cohen’s d

PA

Concrete

Post-stressor Post-VR Post-stressor Post-VR Post-stressor Post-VR

209.04 189.21 221.79 311.79 218.23 250.90

162.83 147.08 162.95 202.97 180.79 199.18

28

–0.607

.594

0.11

29

3.575

0.66

30

1.165

.001 *** .253

Post-stressor Post-VR Post-stressor Post-VR Post-stressor Post-VR

214.79 83.11 162.62 40.10 153.47 36.70

253.85 120.07 123.05 41.47 149.96 61.76

28

2.921

0.55

29

6.602

30

5.430

.007 ** .000 *** .000 ***

Grass Trees NA

Concrete Grass Trees

Notes. *p < .05. **p < .01. ***p < .001. M = mean value; SD = standard deviation. 8

0.21

1.23 0.99

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al.

found that PA improved differently under different conditions while NA did not. The stronger increase of PA in the environment with grass than in those with trees or without vegetation is also consistent with Nordh and Østby (2013) study, in which participants indicated that a natural environment with a lot of grass had a high likelihood of restoration. It is also in line with findings by Nordh (2012) that found that participants who spent longer looking at grass were more likely to rate it as having a higher likelihood of restoration. Our results also indicate that different natural settings may hold different capacities to impact positive emotions. This is consistent with studies by Chiang et al. (2017), Jiang et al. (2014), Martens et al. (2011), and Tyrväinen et al. (2014) that found that different natural settings differently affected the recovery of positive emotions. However, in contrast with previous findings that environments with trees are preferential to ones with grass (Kenwick, Shammin, & Sullivan, 2009; Purcell, Lamb, Mainardi Peron, & Falchero, 1994), we observed that the environment with grass more strongly restored PA than either the one with trees or the one without any vegetation. These conflicting results may in fact support the findings of Martens et al. (2011), which found no strong association between preference for urban forests and affective responses. There are several possible explanations for why the grass environment had a stronger restorative effect on PA. First, the ground texture of an environment is important to the perception of depth, which could influence affective reactions according to affordance theory (Gibson, 1986; Menatti & Casado da Rocha, 2016) and the concept of extent in Kaplan (1995) theory. Regarding the latter, the VR environment with grass may thus have elicited greater PA due to its greater extent of restoration, given its more spacious green area in which to move and experience the greenery. Second, individuals might experience vegetation in closer proximity in environments with grass than ones with trees (Fig. 2) and the environment with grass might have therefore elicited stronger restoration; such an explanation would be in line with the biophilia theory (Kellert & Wilson, 1995), which maintains that humans experience an innate urge to associate with other forms of life. Third, the environment with grass might have suggested more pleasant activities to the participants. According to affordance theory (Menatti & Casado da Rocha, 2016), an environment’s physical properties suggest different behavioural options to individuals. In our interviews with participants at the end of the experiment, participants described the environment with grass as being conducive to walking and playing golf, even if Chinese people do not tend to step on the grass in public green spaces, whereas no participant mentioned an intention to interact with the trees. Such findings are consistent with Ulrich (1983) suggestion that unlearned factors of evolutionary origin influence reactions to natural and urban components in environments, in addition to Ulrich (1979) empirical results that indicated cross-cultural similarities in psychological effects from environments with natural components. We observed no significant differences in the changes of NA across the three groups, which concurs with a study by Van den Berg et al. (2014). This may be because all three environmental conditions were considered more peaceful than the stressor, leading to universal decreases in NA (Fredrickson, Mancuso, Branigan, & Tugade, 2000).

In the first phase of our experiment (i.e., the stress phase), physiological stress, as measured by SCLs, was significantly higher than prior to stress exposure, as were NA scores, while PA scores were lower. Upon encountering the stressor task with noise, participants initially exhibited avoidance or dislike (Ulrich, 1983), which activated their SNS (Alvarsson et al., 2010) ‘fight or flight’ responses (Fink, 2017) and, in turn, induced higher SCLs. Because electrodermal activity measures electric impulses on the skin’s surface and sweat glands that are solely innervated by the SNS (sympathetic nervous system; Cummings et al., 2007; Ulrich, 1983), we considered SCLs to constitute an unbiased index of sympathetic activity. Therefore, for simplicity without overgeneralisation, we regarded the physiological and psychological states after stress exposure as the initial affective state, or what Ulrich (1983) called ‘Arousal1’. In the second phase (i.e., the initial 5-minute VR immersion phase), participants’ SCLs decreased in all three VR environments. A possible explanation is that escaping the stressor was tantamount to escaping a threat; compared to the stressor, the perceptibly quieter VR environments were preferable and thus evoked more positive reactions than the stressor (Ulrich, 1983). Accordingly, the improved PA would have relaxed the SNS and gradually stopped mobilising the body’s fight-or-flight response, which, in turn, would have decreased participants’ SCLs (‘Arousal2’; Ulrich, 1983). Ultimately, therefore, all three environments gave participants the opportunity to restore their emotional states, consistent with the results of Alvarsson et al. (2010) and Ulrich et al. (1991). In the third phase (i.e., the second 5-minute VR immersion phase), there were significant differences in the changes of participant SCLs and PA scores across the three groups, and each group’s changes in SCLs were consistent with the changes in PA scores. This is consistent with earlier findings that reported an association between PA and decreasing physiological arousal (Alvarsson et al., 2010; Ulrich et al., 1991). Our finding that the environment with grass caused PA to increase also echoes Ulrich (1983) model. In this phase, individuals were able to cognitively evaluate the scenery. According to affordance theory (Gibson, 1986; Menatti & Casado da Rocha, 2016; Roe & Aspinall, 2011), sentient organisms can directly perceive the meaning and value of environments and react to them (Kyttä, 2002). The environment with grass would thus have been perceived as being more positive, since a grassy lawn naturally provides opportunities for walking, playing, or lying down (Menatti & Casado da Rocha, 2016). The positive appraisal of that environment would thus have generated a corresponding post-cognitive effect and elicited post-cognitive arousal (‘Arousal3’; Ulrich, 1983). Higher PA in turn reduces the urge to mobilise the cognitive resources that are necessary to counter a threat (Russell & Mehrabian, 1978). This would explain why the SCLs of participants in the grass group continued to decline even in the final minutes of VR immersion. 4.4. Strengths, limitations, and directions for future research This study was designed as a non-crossover controlled experiment. In order to avoid an imbalance in some covariates, it was necessary to include a larger sample size than would be needed for a crossover design study. Although less economically efficient, this design helped us avoid any ‘order effects’ or ‘carry-over effects’. We conducted this study with virtual environments instead of real ones, and we should therefore be cautious about extending the results to real landscape designs without additional in situ testing. However, previous research by de Kort et al. (2006) provides support for the argument that virtual natural environments have similar restorative effects to real ones. We did not test immersion effect or VR quality in our experiment, but Valtchanov and Ellard (2010) also demonstrated that the specific content of a VR experience, and especially the presence of natural elements, rather than the VR experience itself, is likely to be responsible for any restorative effect. Although only 1.75% of the comments received in the post-experiment interview were critical of the reality of our VR immersion experience, further research is needed regarding both different VR experiences and potential differences between VR and real environments.

4.3. Interpreting the interaction of physiological arousal and psychological well-being The changes in the trends of physiological arousal, in terms of SCL, and psychological well-being, in terms of PA and NA, can be divided into three phases: the stress phase, the initial 5-minute VR immersion phase, and the second 5-minute VR immersion phase. Ultimately, our results are in line with the model of affective response to nature proposed by Ulrich (1983) and later supported by additional empirical results (Hartig, Evans, Jamner, Davis, & Gärling, 2003; Ulrich et al., 1991). They therefore support Ulrich’s hypothesis that natural environments improve both physiological and psychological reactions to stress more than do environments without nature. 9

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al.

Another limitation of our study is that we excluded confounding factors; for example, we controlled the leaves of trees and blades of grass so that they did not sway as they might in real life, which could have affected our results. Future research should therefore compare the restorative effects of grass and trees in different types of weather. Furthermore, we compared only two types of vegetation, and future research should examine a greater variety vegetation types with different textures and forms. Moreover, we compared vegetation in a single courtyard setting, and future research should explore the effects of vegetation in other settings—for example, open fields. In addition, we observed SCL responses for only 9 min of VR immersion, but the SCL trends in the grass and tree groups showed different trend in the final minute during the second half of the VR immersion period. In the tree group, SCLs continued to decline during the entire second half, while in the grass group, although SCLs declined from minutes 5–8, they increased during the final minute. Future research should therefore explore SCL trends over a longer period of time. Finally, this is a single study. To generate more generalisable findings, additional studies that build on this and other similar research should be conducted with either VR or real environments to explore the effects of different vegetation types on the restoration of physiological and psychological well-being.

urban site-results from a field experiment. International Journal of Environmental Research and Public Health, 15(12), https://doi.org/10.3390/ijerph15122647. Beil, K., & Hanes, D. (2013). The influence of urban natural and built environments on physiological and psychological measures of stress–a pilot study. International Journal of Environmental Research and Public Health, 10(4), 1250–1267. https://doi.org/10. 3390/ijerph10041250. Benedek, M., & Kaernbach, C. (2010). A continuous measure of phasic electrodermal activity. Journal of Neuroscience Methods, 190(1), 80–91. https://doi.org/10.1016/j. jneumeth.2010.04.028. Braithwaite, J. J., Watson, D. G., Jones, R., & Rowe, M. (2013). A guide for analysing electrodermal activity (EDA) & skin conductance responses (SCRs) for psychological experiments. Psychophysiology, 49, 1017–1034. Camacho-Cervantes, M., Schondube, J. E., Castillo, A., & MacGregor-Fors, I. (2014). How do people perceive urban trees? Assessing likes and dislikes in relation to the trees of a city. Urban Ecosystems, 17(3), 761–773. https://doi.org/10.1007/s11252-0140343-6. Carrus, G., Lafortezza, R., Colangelo, G., Dentamaro, I., Scopelliti, M., & Sanesi, G. (2013). Relations between naturalness and perceived restorativeness of different urban green spaces. Psyecology, 4(3), 227–244. https://doi.org/10.1174/217119713807749869. Chiang, Y.-C., Li, D., & Jane, H.-A. (2017). Wild or tended nature? The effects of landscape location and vegetation density on physiological and psychological responses. Landscape and Urban Planning, 167, 72–83. https://doi.org/10.1016/j.landurbplan. 2017.06.001. Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, N.J.: Lawrence Erlbaum. Cummings, E. M., El-Sheikh, M., Kouros, C. D., & Keller, P. S. (2007). Children's skin conductance reactivity as a mechanism of risk in the context of parental depressive symptoms. Journal of Child Psychology and Psychiatry, 48(5), 436–445. https://doi. org/10.1111/j.1469-7610.2006.01713.x. de Kort, Y. A. W., Meijnders, A. L., Sponselee, A. A. G., & IJsselsteijn, W. A. (2006). What’s wrong with virtual trees? Restoring from stress in a mediated environment. Journal of Environmental Psychology, 26(4), 309–320. https://doi.org/10.1016/j.jenvp.2006.09. 001. de Vries, S., Verheij, R. A., Groenewegen, P. P., & Spreeuwenberg, P. (2003). Natural Environments—Healthy Environments? An Exploratory Analysis of the Relationship between Greenspace and Health. Environment & Planning A, 35(10), 1717–1731. https://doi.org/10.1068/a35111. Diener, E., & Suh, E. (1997). Measuring quality of life: Economic, social, and subjective indicators. Social Indicators Research, 40(1), 189–216. https://doi.org/10.1023/ A:1006859511756. Dodge, R., Daly, A. P., Huyton, J., & Sanders, L. D. (2012). The challenge of defining wellbeing. International Journal of Wellbeing, 2(3). Elsadek, M., Sun, M., & Fujii, E. (2016). Psycho-physiological responses to plant variegation as measured through eye movement, self-reported emotion and cerebral activity. Indoor and Built Environment. https://doi.org/10.1177/1420326X16638711 1420326X16638711. Evans, G. W. (2003). The built environment and mental health. Journal of Urban Health: Bulletin of the New York Academy of Medicine, 80(4), 536–555. https://doi.org/10. 1093/jurban/jtg063. Fink, G. (2017). Stress: Neuroendocrinology and Neurobiology. Academic Press/Elsevier. Fredrickson, B. L., Mancuso, R. A., Branigan, C., & Tugade, M. M. (2000). The undoing effect of positive emotions. Motivation and Emotion, 24(4), 237–258. Frumkin, H. (2001). Beyond toxicity: Human health and the natural environment. American Journal of Preventive Medicine, 20(3), 234–240. Gatersleben, B., & Andrews, M. (2013). When walking in nature is not restorative-the role of prospect and refuge. Health & Place, 20, 91–101. https://doi.org/10.1016/j. healthplace.2013.01.001. Gibson, J. J. (1986). The ecological approach to visual perception. Psychology Press. Grahn, P., & Stigsdotter, U. K. (2010). The relation between perceived sensory dimensions of urban green space and stress restoration. Landscape and Urban Planning, 94(3–4), 264–275. https://doi.org/10.1016/j.landurbplan.2009.10.012. Hansmann, R., Hug, S.-M., & Seeland, K. (2007). Restoration and stress relief through physical activities in forests and parks. Urban Forestry & Urban Greening, 6(4), 213–225. https://doi.org/10.1016/j.ufug.2007.08.004. Hartig, T., Evans, G. W., Jamner, L. D., Davis, D. S., & Gärling, T. (2003). Tracking restoration in natural and urban field settings. Journal of Environmental Psychology, 23(2), 109–123. https://doi.org/10.1016/j.ufug.2007.08.004. Hartig, T., & Staats, H. (2004). Restorative environments. Encyclopedia of Applied Psychology, 3, 273–279. Headey, B., & Wearing, A. (1991). Subjective well-being: A stocks and flows framework. In F. Strack, M. Argyle, & N. Schwarz (Eds.). International series in experimental social psychology, Vol. 21. Subjective well-being: An interdisciplinary perspective (pp. 49–73). Elmsford, NY, US: Pergamon Press. Igarashi, M., Song, C., Ikei, H., & Miyazaki, Y. (2014). Effects of olfactory stimulation with perilla essential oil on prefrontal cortex activity. The Journal of Alternative and Complementary Medicine, 20(7), 545–549. https://doi.org/10.1089/acm.2014.0100. Jiang, B., Chang, C.-Y., & Sullivan, W. C. (2014). A dose of nature: Tree cover, stress reduction, and gender differences. Landscape and Urban Planning, 132(Supplement C), 26–36. https://doi.org/10.1016/j.landurbplan.2014.08.005. Kaplan, R. (1983). The role of nature in the urban context. Behavior and the natural environment (pp. 127–161). Boston, MA: Springer. Kaplan, R., & Kaplan, S. (1989). The experience of nature: A psychological perspective. New York, NY, US: Cambridge University Press. Kaplan, S. (1995). The restorative benefits of nature: Toward an integrative framework. Journal of Environmental Psychology, 15(3), 169–182. https://doi.org/10.1016/02724944(95)90001-2.

5. Conclusion We conducted a randomised controlled experiment to examine physiological and psychological responses to different types of vegetation in built environments. Our study provides evidence that natural environments have a stronger restorative effect than do concrete environments without vegetation, as assessed by reductions in SCLs (a proxy for physiological stress levels) and improved PA. However, the different environments had no significantly different effects on NA, which suggests that carefully designed built environments may be able to reduce stress levels to at least a certain extent. Notably, we found that the grassy VR environment had the best ability to support PA restoration. However, our use of VR environments means that landscape planners should be cautious about generalising these results to real settings. In reality, trees may also provide other advantages, such as shade, that were not taken into account in the present study. Thus, future studies should be conducted with real environmental stimuli and real vegetation in order to explore the effects of different vegetations on individuals’ capacities to restore physiological and psychological wellbeing after a stressful event. Ultimately, however, our results support a preliminary conclusion that environments with grass, and to a lesser extent those with trees, have the potential to support physiological and psychological well-being in urban environments. Acknowledgments We gratefully acknowledge financial support from the National Natural Science Foundation of China (project no. 51678327) and the Chinese Scholarship Council (project no. 201706210099). We wish to thank participants and volunteers and express special thanks to Huaxin Qin from Hefei Expected Shine Info Technology for helping with the technical aspects of creating virtual environments used in our study. References Akers, A., Barton, J., Cossey, R., Gainsford, P., Griffin, M., & Micklewright, D. (2012). Visual color perception in green exercise: Positive effects on mood and perceived exertion. Environmental Science & Technology, 46(16), 8661–8666. https://doi.org/10. 1021/es301685g. Alvarsson, J. J., Wiens, S., & Nilsson, M. E. (2010). Stress recovery during exposure to nature sound and environmental noise. International Journal of Environmental Research and Public Health, 7(3), 1036–1046. https://doi.org/10.3390/ijerph7031036. Appleton, J. (1996). The experience of landscape. Chichester: Wiley66–67. Arnberger, A., Eder, R., Allex, B., Ebenberger, M., Hutter, H.-P., Wallner, P., ... Frank, T. (2018). Health-related effects of short stays at mountain meadows, a river and an

10

Landscape and Urban Planning 193 (2020) 103654

Q. Huang, et al. Kellert, S. R., & Wilson, E. O. (1995). The Biophilia Hypothesis. Island Press. Kenwick, R. A., Shammin, M. R., & Sullivan, W. C. (2009). Preferences for riparian buffers. Landscape and Urban Planning, 91(2), 88–96. https://doi.org/10.1016/j. landurbplan.2008.12.005. Kyttä, M. (2002). Affordances of Children’s Environments in the Context of Cities, Small Towns, Suburbs and Rural Villages in Finland and Belarus. Journal of Environmental Psychology, 22(1), 109–123. https://doi.org/10.1006/jevp.2001.0249. Li, D., & Sullivan, W. C. (2016). Impact of views to school landscapes on recovery from stress and mental fatigue. Landscape and Urban Planning, 148, 149–158. https://doi. org/10.1016/j.landurbplan.2015.12.015. Maas, J., Verheij, R. A., Groenewegen, P. P., de Vries, S., & Spreeuwenberg, P. (2006). Green space, urbanity, and health: How strong is the relation? Journal of Epidemiology and Community Health, 60(7), 587–592. https://doi.org/10.1136/jech.2005.043125. Martens, D., Gutscher, H., & Bauer, N. (2011). Walking in “wild” and “tended” urban forests: The impact on psychological well-being. Journal of Environmental Psychology, 31(1), 36–44. https://doi.org/10.1016/j.jenvp.2010.11.001. McMahan, E. A., & Estes, D. (2015). The effect of contact with natural environments on positive and negative affect: A meta-analysis. The Journal of Positive Psychology, 10(6), 507–519. https://doi.org/10.1080/17439760.2014.994224. Menatti, L., & Casado da Rocha, A. (2016). Landscape and health: Connecting psychology, aesthetics, and philosophy through the concept of affordance. Frontiers in Psychology, 7, 571. https://doi.org/10.3389/fpsyg.2016.00571. Nordh, H. (2012). Quantitative methods of measuring restorative components in urban public parks. Journal of Landscape Architecture, 7(1), 46–53. https://doi.org/10.1080/ 18626033.2012.693780. Nordh, H., & Østby, K. (2013). Pocket parks for people–A study of park design and use. Urban forestry & urban greening, 12(1), 12–17. Olmsted, F. (1865). The Yosemite Valley and the Mariposa big trees: A preliminary report, 1865. Landscape Architecture, 43(1), 12–25. Peters, M. L., Godaert, G. L., Ballieux, R. E., van Vliet, M., Willemsen, J. J., Sweep, F. C., & Heijnen, C. J. (1998). Cardiovascular and endocrine responses to experimental stress: Effects of mental effort and controllability. Psychoneuroendocrinology, 23(1), 1–17. Prajapati, B., Dunne, M., & Armstrong, R. (2010). Sample size estimation and statistical power analyses. Purcell, A. T., Lamb, R. J., Mainardi Peron, E., & Falchero, S. (1994). Preference or preferences for landscape? Journal of Environmental Psychology, 14(3), 195–209. https://doi.org/10.1016/S0272-4944(94)80056-1. Roe, J., & Aspinall, P. (2011). The Emotional affordances of forest settings: An investigation in boys with extreme behavioural problems. Landscape Research, 36(5), 535–552. https://doi.org/10.1080/01426397.2010.543670. Russell, J. A., & Mehrabian, A. (1978). Approach-avoidance and affiliation as functions of the emotion-eliciting quality of an environment. Environment and Behavior, 10(3), 355–387. https://doi.org/10.1177/0013916578103005. Tang, I.-C., Tsai, Y.-P., Lin, Y.-J., Chen, J.-H., Hsieh, C.-H., Hung, S.-H., ... Chang, C.-Y. (2017). Using functional Magnetic Resonance Imaging (fMRI) to analyze brain region activity when viewing landscapes. Landscape and Urban Planning, 162(Supplement C), 137–144. https://doi.org/10.1016/j.landurbplan.2017.02.007. Tyrväinen, L., Ojala, A., Korpela, K., Lanki, T., Tsunetsugu, Y., & Kagawa, T. (2014). The influence of urban green environments on stress relief measures: A field experiment. Journal of Environmental Psychology, 38, 1–9. https://doi.org/10.1016/j.jenvp.2013.

12.005. Ulrich, R. S. (1979). Visual landscapes and psychological well-being. Landscape Research, 4(1), 17–23. https://doi.org/10.1080/01426397908705892. Ulrich, R. S. (1981). Natural versus urban scenes: Some psychophysiological effects. Environment and Behavior, 13(5), 523–556. https://doi.org/10.1177/ 0013916581135001. Ulrich, R. S. (1983). Aesthetic and affective response to natural environment. In I. Altman, & J. F. Wohlwill (Eds.). Behavior and the Natural Environment (pp. 85–125). US: Springer. https://doi.org/10.1007/978-1-4613-3539-9_4. Ulrich, R. S. (1984). View through a window may influence recovery from surgery. Science, 224(4647), 420–421. https://www.ncbi.nlm.nih.gov/pubmed/6143402. Ulrich, R. S. (1986). Human responses to vegetation and landscapes. Landscape and Urban Planning, 13, 29–44. https://doi.org/10.1016/0169-2046(86)90005-8. Ulrich, R. S. (1993). Biophilia, biophobia, and natural landscapes. The Biophilia Hypothesis, 7, 73–137. Ulrich, R. S., Simons, R. F., Losito, B. D., Fiorito, E., Miles, M. A., & Zelson, M. (1991). Stress recovery during exposure to natural and urban environments. Journal of Environmental Psychology, 11(3), 201–230. https://doi.org/10.1016/S0272-4944(05) 80184-7. Valtchanov, D., & Ellard, C. G. (2010). Physiological and affective responses to immersion in virtual reality: Effects of nature in urban settings. Journal of Cybertherapy & Rehabilitation. 3(4), 359–373. Van den Berg, A. E., Jorgensen, A., & Wilson, E. R. (2014). Evaluating restoration in urban green spaces: Does setting type make a difference? Landscape and Urban Planning, 127, 173–181. https://doi.org/10.1016/j.landurbplan.2014.04.012. Velarde, M. D., Fry, G., & Tveit, M. (2007). Health effects of viewing landscapes–Landscape types in environmental psychology. Urban Forestry & Urban Greening, 6(4), 199–212. https://doi.org/10.1016/j.ufug.2007.07.001. Verheij, R. A., Maas, J., & Groenewegen, P. P. (2008). Urban—rural health differences and the availability of green space. European Urban and Regional Studies, 15(4), 307–316. https://doi.org/10.1177/0969776408095107. von Lindern, E., Bauer, N., Frick, J., Hunziker, M., & Hartig, T. (2013). Occupational engagement as a constraint on restoration during leisure time in forest settings. Landscape and Urban Planning, 118, 90–97. https://doi.org/10.1016/j.landurbplan. 2013.03.001. Wang, D., Lau, K. K.-L., Yu, R. H. Y., Wong, S. Y. S., Kwok, T. C. Y., & Woo, J. (2016). Neighbouring green space and all-cause mortality in elderly people in Hong Kong: A retrospective cohort study. The Lancet, 388, S82. https://doi.org/10.1016/S01406736(16)32009-8. Ward Thompson, C., Roe, J., Aspinall, P., Mitchell, R., Clow, A., & Miller, D. (2012). More green space is linked to less stress in deprived communities: Evidence from salivary cortisol patterns. Landscape and Urban Planning, 105(3), 221–229. https://doi.org/10. 1016/j.landurbplan.2011.12.015. Warner, S. (2016). 320 GRE Math Problems Arranged by Topic and Difficulty Level: 160 GRE Questions with Solutions, 160 Additional Questions with Answers; 2016 Edition. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54(6), 1063–1070. Zhou, Q., & Gu, B. (2013). A study of lawn trample based on public behaviour pattern. Heilongjiang Agricultural Sciences, 2013(2), 85–89.

11