Nature Does Not Always Give You a Helping Hand: Comparing the Prosocial Effects of Nature at Different Resource and Security Levels

Abstract

Humans become more prosocial after nature exposure. We proposed that the prosocial effect pertains to resource (e.g., food, water) and security (e.g., shelter, concealment) features in natural environments. Four studies tested the idea that prosociality changes with variations in environmental resource and security. Study 1 reported that urban greenspace, a resource feature to urban dwellers, predicted more volunteering in low-crime cities, but less so in high-crime cities. Studies 2 and 3 compared prosociality after exposure to natural sceneries in a Resource (high/low) × Security (high/low) design. Participants were more prosocial in the high-resource-high-security and low-resource-low-security conditions. Study 4 compared the four natural environments with two control conditions (urban, shape). It reported that not all natural environments led to higher prosociality, nor did any of them undermine prosociality. The findings supported heterogeneity in nature’s prosocial effect. Implications are discussed in relation to urban greening and the evolutionary basis of nature’s effect.

Keywords

nature prosociality resource security evolution

Exposure to nature, particularly green sceneries, has been reported to increase prosociality. For example, the amount of urban greenspace predicted stronger social ties in communities (Kuo, Sullivan, Coley, & Brunson, 1998; Sullivan, Kuo, & DePooter, 2004), lower crime rates, and fewer incivilities (Kuo, 2003; Kuo & Sullivan, 2001; Troy, Grove, & O’Neil-Dunne, 2012). Experiments showed that people became more helpful after short nature experiences, such as walking in a park (Guéguen & Stefan, 2016) and seeing nature images (Weinstein, Przybylski, & Ryan, 2009; Zelenski, Dopko, & Capaldi, 2015). Higher prosociality was also observed among people who were seated in rooms furnished with plants (Weinstein et al., 2009) and received flowers (Haviland-Jones, Rosario, Wilson, & McGuire, 2005).

Researchers have attributed the prosocial effect to psychological mechanisms such as enhanced mood and autonomy (Weinstein et al., 2009; Zhang, Piff, Iyer, Koleva, & Keltner, 2014). However, it remained unclear as to what natural environmental features caused the prosocial effect. Evolutionary theories shed light on this issue by relating nature experiences to habitat selection in hunter-gatherer societies, a mode of living that had occupied a long period of time in human evolution. During that period, humans were accustomed to assessing environmental information as they constantly moved and looked for good habitats. Among all types of information, resources (e.g., food, water) and security (e.g., shelters, dangerous animals, and competitors) were most relevant to survival and reproductive success (Haviland-Jones et al., 2005; Orians, 2007). Presumably, the sight of lush green vegetation and water would hint at the presence of life-sustaining resources (Hartmann & Apaolaza-Ibáñez, 2010) and the provision of concealment (e.g., trees, shrubs; Appleton, 1996). Feeling pleasant and staying in these areas became adaptive responses, and remained impactful to modern humans (Wilson, 2007).

We speculated that the prosocial effect of nature would pertain to its resource and security features. Research has shown that prosocial helping decreased as its cost increased (Graziano, Habashi, Sheese, & Tobin, 2007). For hunter-gatherers, the cost could be sacrificing resources and security, as in sharing food and helping potential competitors. Supposedly, high-resource and high-security features in natural environments, as in lush green vegetation, could offset the cost and thus promote helping. Although urban dwellers no longer gathered resources and sought concealment from nature, they remained sensitive to these features because of their survival values in hunter-gatherer societies (Andrews & Gatersleben, 2010; Gatersleben & Andrews, 2013; Ng, 2015). Indeed, resource and security statuses could affect a range of social tendencies that hinted at prosocial behavior. In a cross-national study (Gelfand et al., 2011), food deprivation (low resource) and natural disasters (low security) correlated with stronger adherence to social norms. Demanding climates in China, which threatened resource supply and safety, predicted higher social connectedness (Van de Vliert, Yang, Wang, & Ren, 2012). Exposure to high-resource-high-security natural environments led to lower aggression than those that were resource-poor or insecure (Ng & Chow, 2017). In urban settings, more helping was reported among people possessing more social and economic resources (Korndörfer, Egloff, & Schmukle, 2015) and in safe cities with lower crime rates (Bach, Defever, Chopik, & Konrath, 2017; Levine, Martinez, Brase, & Sorenson, 1994). Despite the different definitions and research focuses, these studies suggested that the resource-security analysis could be useful in theorizing how nature influenced helping.

The Heterogeneous Effect of Nature

As resource and security features vary across natural environments, the evolutionary perspective hinted that nature exposure does not always bolster prosocial behavior. Past research that reported the prosocial effect tended to use green landscapes as nature stimuli (e.g., Weinstein et al., 2009; Zelenski et al., 2015), which may overrepresent the high-resource-high-security environment. Little research was done on natural environments with other resource-security profiles. This article aimed to fill the gap with two research approaches.

First, we considered the positive effect of urban greenspace on helping (Kuo et al., 1998; Sullivan et al., 2004). The effect partly aligned with our analysis as vegetation revealed high-resource information to city dwellers. However, vegetation may carry ambiguous information on security. Trees and bushes, for example, could act as concealment to both park users and perpetrators. That is, parks could be a hot spot of crime (Michael, Hull, & Zahm, 2001; Nasar & Fisher, 1993) and could induce fear of crime (Maruthaveeran & van den Bosh, 2015). Research also showed that people perceived green sceneries with dense vegetation and low visibility as stressful, dangerous, and fearful (Andrews & Gatersleben, 2010; Gatersleben & Andrews, 2013).

We proposed that although urban greenspace could contribute to higher resource information, its security information would depend on crime activities in the city. Put otherwise, we expected that crime rate, which was related to security threat (e.g., Chiricos, Padgett, & Gertz, 2000), would moderate the prosocial effect of urban greenspace. In cities with lower crime rates, more greenspace may predict more helping as in a high-resource-high-security setting. However, in cities with higher crime rates, more greenspace may present greater danger to citizens and deter them from helping each other. We thus expected a greenspace-by-crime interaction effect on helping (Hypothesis 1 [H1]).

Second, we considered natural environments that vary in resource and security levels. In Ng’s (2015) study, participants rated 16 natural environments using adjectives related to resource (e.g., fertile, replenishing) and security (e.g., frightful, dark) features. The environments were clustered into four groups based on the two subjective ratings, with cultivated green areas (e.g., garden) as the high-resource-high-security cluster (HRHS), pristine high-density vegetation (e.g., rainforest) as the high-resource-low-security cluster (HRLS), beach and seacoast as the low-resource-high-security cluster (LRHS), and barren lands (e.g., desert) as the low-resource-low-security cluster (LRLS). The study supported the resource-security model in environmental assessment and highlighted the heterogeneity of nature on the two dimensions.

As natural environments vary in resource and security features, their prosocial effects should also be different. We expected that high-resource and high-security features would lead to higher prosociality. Exposure to a HRHS natural condition should thus lead to higher prosociality than its low-resource and/or low-security counterparts. There could be an exception though. Research has reported that extremely unfavorable physical environments could increase social connectedness (Van de Vliert et al., 2012) and helping behavior (Rao et al., 2011), and decrease aggression (Ng & Chow, 2017). From an evolutionary perspective, prosocial behavior in adverse environments was adaptive because it increased overall survival opportunities (Van Vugt & Van Lange, 2006). Therefore, natural environments in the LRLS cluster may as well promote prosociality. We conceptualized these possible outcomes as a resource-by-security interaction effect. While resource and security may independently increase prosociality, they may also interact so that higher prosociality would be observed in the HRHS and LRLS conditions (Hypothesis 2 [H2]).

Psychological Mechanisms

The evolutionary perspective linked environmental features to aesthetic responses (Appleton, 1996; Orians, 2007). It posited that hunter-gatherers would have a better chance to survive and reproduce if they felt pleasant staying in habitats with high survival values. People generally reported pleasurable feelings toward lush green landscapes (HRHS; Hartmann & Apaolaza-Ibáñez, 2010). Research also reported higher prosociality when people saw beautiful (vs. less beautiful) nature images (Zhang et al., 2014). Aesthetic responses is therefore a potential variable that mediates the effects of resource and security on prosociality.

However, the aesthetic mechanism would fail to explain the expected prosocial effect in LRLS environments described above. Deserts, for example, were rated as equally unpleasant as urban sceneries (Hartmann & Apaolaza-Ibáñez, 2010). Alternatively, we proposed that prosociality in adverse natural environments would be driven by people’s trust in others. Research has supported a positive relationship between trust and prosocial behavior (Yamagishi et al., 2015; Yamagishi & Yamagishi, 1994). Trusting others in adverse environments could increase one’s social capital to combat environmental hazards, increasing the collective opportunity to survive (Van de Vliert et al., 2012). Trust may explain why people act prosocially in adverse environments. We therefore included it as another potential mediator of nature’s effect.

Research Overview

This research adopted a multimethod approach to study how resource and security features in natural environments influence prosociality. We conducted four studies to test the above hypotheses and mechanisms. Study 1 examined the effects of park area and crime rate on volunteering in 51 U.S. cities. According to H1, we expected that crime rate would moderate the effect of park area on volunteering. Studies 2 and 3 compared participants’ prosociality after exposing them to four natural environments of different resource-security profiles. According to H2, we expected that the resource and security factors would interact and influence prosociality. Study 4 re-examined H2 with reference to two control conditions (urban sceneries, geometric shapes). The control conditions helped clarify whether certain natural environments could decrease rather than increase prosociality. Studies 2 to 4 also included measures of aesthetic response and trust as mediators of the resource and security effects on prosociality. All test statistics reported below were tested against an alpha level of .05 (two-tailed).

Study 1

This study examined the effects of park area and crime rate on volunteering in 51 U.S. cities. According to H1, crime rate was expected to moderate the effect of park area on volunteering.

Method

Datasets

In searching relevant datasets, we prioritized those released by U.S. Federal Government Agencies and nonprofit organizations. Volunteering data were retrieved from Volunteering and Civic Life in America (VCLA; Corporation for National & Community Service, 2016). The data were collected in a national survey about volunteer activities in 51 cities and their surrounding metropolitan statistical areas (MSAs). We defined city as the largest city in each of the MSAs reported in the VCLA. Park area data were obtained from City Park Facts (CPF) released by the Trust for Public Land (2016), except for Hartford, Providence, and Salt Lake City, where data were retrieved from the city government websites. City crime rates were retrieved from Crime in the United States (CIUS) published by the Federal Bureau of Investigation (2016). The three datasets were dated 2015.

Volunteering

The VCLA surveyed citizens’ engagement in 13 types of volunteer activities, such as fundraising, mentoring, and providing professional services. The responses were combined into three indicators, namely, volunteer hours per capita, volunteer rate, and economic value of volunteer service provided. This study focused on volunteer hours because research usually defined helping as the amount of help offered by people (e.g., time, money). Volunteer rate did not reflect this quantity and failed to distinguish between casual and dedicated volunteers. Economic value was also not suitable because volunteers were probably more aware of their effort rather than the economic value in helping others. Missing data in volunteer rate and economic value also undermined their utility in this study.

Parkland area

Parkland area was divided into designed areas (e.g., playgrounds, neighborhood parks) and undisturbed natural areas (e.g., wetlands, forests). We analyzed total, designed, and natural park areas as percentages of the city area.

Crime rate

Crime rate was operationalized as the number of crimes per 1,000 residents. Both violent crimes (murder, rape, robbery, and aggravated assault) and property crimes (burglary, larceny-theft, motor vehicle theft, and arson) were included in this variable.

Socioeconomic covariates

Population density and cost of living predicted lower helping in previous research (Levine et al., 1994) and were controlled statistically in our analysis. Data were obtained from the U.S. Census Bureau (2011, 2015).

Results

Table 1 presents the data of the surveyed cities.

Table 1.

Data of 51 U.S. Cities in the Year of 2015 (Study 1).

City	City area (acres)	Population per acre	Cost of living^a	Volunteer hours per capita	Park/city ratio (%)			Crimes per 1,000 citizens
City	City area (acres)	Population per acre	Cost of living^a	Volunteer hours per capita	Total	Natural area	Designed area	Crimes per 1,000 citizens
Atlanta, GA	85,248	5.44	95.55	25.1	5.92	1.55	4.38	66.19
Austin, TX	173,952	5.36	95.47	21.9	14.58	9.77	4.81	41.43
Baltimore, MD	51,776	12.01	119.37	28.8	9.56	3.56	6.00	65.16
Birmingham, AL	93,504	2.27	90.80	31.8	3.85^c	1.58	2.27	81.05
Boston, MA	30,912	21.58	132.49	27.4	16.99	8.25	8.74	30.23
Buffalo, NY	25,856	9.98	95.76	27.0	7.52	1.16	6.36	54.49
Charlotte, NC	190,528	4.34	93.21	23.2	6.41	5.62	0.79	44.45
Chicago, IL	145,664	18.68	116.87	25.6	9.20	4.13	5.07	38.50
Cincinnati, OH	49,856	5.99	93.81	24.0	14.14	7.52	6.62	64.35
Cleveland, OH	49,728	7.80	101.05	25.5	6.40	3.21	3.18	79.68
Columbus, OH	139,008	6.12	92.00	27.4	8.95	4.52	4.43	44.81
Dallas, TX	217,920	5.97	91.87	31.9	12.58	7.46	5.12	41.34
Denver, CO	97,920	6.97	103.20	28.7	7.96	1.80	6.16	42.04
Detroit, MI	88,832	7.62	99.38	23.3	6.31	0.04	6.27	58.53
Hartford, CT	11,136	11.14	121.84	22.5	12.33	—^b	—^b	55.31
Houston, TX	383,744	5.98	92.19	20.4	14.29	11.18	3.11	53.64
Indianapolis, IN	231,296	3.69	87.21	23.1	5.07	2.86	2.21	60.79
Jacksonville, FL	478,080	1.82	92.92	18.3	13.82	12.09	1.74	43.21
Kansas City, MO	201,600	2.36	97.78	32.4	9.06	5.64	3.42	58.59
Las Vegas, NV	86,912	7.18	101.86	18.5	5.50	2.28	3.22	39.16
Los Angeles, CA	299,968	13.24	136.37	21.5	13.18	9.76	3.42	29.94
Louisville, KY	208,128	2.96	87.68	21.8	7.31	5.24	2.07	47.98
Memphis, TN	201,664	3.25	88.24	25.8	4.66	3.31	1.36	73.71
Miami, FL	22,976	19.19	106.00	20.6	6.28	3.09	3.19	53.89
Milwaukee, WI	61,504	9.76	101.85	36.4	8.70	4.08	4.62	58.60
Minneapolis, MN	34,560	11.89	110.98	31.0	14.91	1.57	13.34	52.57
Nashville, TN	304,064	2.15	88.93	30.4	10.69	9.36	1.34	49.07
New Orleans, LA	108,416	3.59	—^b	11.0	25.60	24.29	1.31	48.24
New York City, NY	193,664	44.15	185.80	38.8	21.08	11.12	9.96	21.04
Oklahoma City, OK	388,096	1.63	91.72	18.6	6.87	4.55	2.32	47.22
Orlando, FL	65,536	4.13	97.78	27.6	5.46	0.80	4.66	69.56
Philadelphia, PA	85,824	18.26	126.46	30.0	13.06	7.69	5.37	41.76
Phoenix, AZ	330,688	4.73	100.69	34.3	15.03	13.31	1.72	40.85
Pittsburgh, PA	35,456	8.59	91.53	28.3	8.44	5.83	2.61	39.31
Portland, OR	85,376	7.41	111.30	34.9	17.75	13.42	4.34	43.26
Providence, RI	11,776	15.22	123.33	17.8	10.12	—^b	—^b	39.56
Raleigh, NC	91,456	4.93	98.19	18.3	14.24	11.53	2.71	40.12
Richmond, VA	38,272	5.76	90.81	28.6	5.30	0.00	5.30	44.32
Riverside, CA	51,904	6.21	112.47	31.8	7.12	5.55	1.57	37.06
Rochester, NY	22,912	9.16	99.99	33.7	6.32^c	3.19	3.13	48.14
Sacramento, CA	62,656	7.83	116.18	22.7	8.97	5.93	3.04	41.07
Salt Lake City, UT	71,104	2.71	100.65	65.9	1.98	1.61	0.37	101.86
San Antonio, TX	294,976	4.98	95.66	24.9	8.94	6.34	2.60	56.17
San Diego, CA	208,128	6.70	132.26	57.0	22.04	17.73	4.32	24.81
San Francisco, CA	30,016	28.81	164.05	31.9	18.99	10.32	8.67	69.15
San Jose, CA	113,024	9.09	156.14	34.7	14.35	12.52	1.83	27.57
Seattle, WA	53,696	12.75	121.35	32.7	12.49	1.94	10.55	61.21
St. Louis, MO	39,616	7.97	90.41	31.7	9.52	1.45	8.07	81.33
Tampa, FL	72,576	5.09	92.44	39.1	6.87	3.97	2.90	29.27
Virginia Beach, VA	159,360	2.84	—^b	30.1	15.65	13.65	2.00	23.44
Washington, D.C.	39,040	17.22	140.09	38.4	21.88	12.90	8.98	57.19
M	129,642	8.91	107.63	28.6	10.87	6.54	4.32	50.24
SD	111,436	7.60	21.31	9.1	5.27	5.08	2.79	16.40

The cost of living index was dated 2010.

Data not available.

Park data were dated 2011.

Preliminary analysis

Volunteering was regressed on the two socio-economic covariates. Population density marginally predicted lower volunteering, b = −0.51, p = .088, whereas cost of living predicted more volunteering, b = 0.26, p = .015. Of the two covariates, cost of living also predicted more park areas, b = 0.002, p < .001, and lower crime rates, b = −0.26, p = .015. The results necessitated controlling population density and cost of living in subsequent analysis.

Consistent with previous research, park area predicted lower crime rates, r = −.43, b = −1.34, 95% confidence interval (CI) = [−2.15, –0.53], p = .002. For every 10% increase in park area, crime rate would decrease by 13.4 per 1,000 people.

Park-by-crime interaction effect

We regressed volunteering on centered scores of total park area, crime rate, and their interaction. The tolerance values of the predictors were .81, .77, and .95, indicating the absence of multicollinearity. The model explained 32% of variance, F(3, 47) = 7.36, p < .001. As shown in Table 2 (Model 1), the interaction effect was statistically significant, indicating that the effect of park area on volunteering was moderated by crime rate.

Table 2.

Volunteer Hours Per Capita Regressed on Park Area, Crime Rate, and Their Interaction Term (Study 1).

Predictors	b	SE	t	p
Model 1 (N = 51)
Constant	26.80	1.16	23.14	<.001
Total park area	0.16	0.23	0.69	.500
Crime rate	0.04	0.08	0.54	.593
Total Park Area × Crime Rate	−0.05	0.01	−4.41	<.001
Model 2 (N = 49)
Constant	28.96	1.24	23.36	<.001
Designed park area	0.52	0.45	1.15	.256
Crime rate	0.04	0.08	0.59	.559
Designed Park Area × Crime Rate	−0.06	0.02	−2.50	.016
Model 3 (N = 49)
Constant	26.33	1.30	20.27	<.001
Natural park area	−0.19	0.27	−0.70	.488
Crime rate	−0.001	0.08	−0.01	.989
Natural Park Area × Crime Rate	−0.07	0.02	−4.09	<.001
Model 4 (N = 49)
Constant	12.89	9.53	1.35	.183
Population density	−0.36	0.25	−1.44	.157
Cost of living	0.17	0.10	1.59	.120
Total park area	0.29	0.30	0.96	.341
Crime rate	0.11	0.07	1.46	.152
Total Park Area × Crime Rate	−0.04	0.01	−4.03	<.001

We analyzed the conditional effects using the Johnson–Neyman procedure. In cities with below-average crime rates, park area predicted more volunteering. The positive slopes became significant when crime rate was 6.52 below average, b = 0.48, 95% CI = [0.00, 0.95], t(47) = 2.01, p = .05. That is, for every 10% increase in park area, volunteering would increase by 4.8 hr per capita. The cutoff value was less than half a standard deviation of all crime rates (SD = 16.40). Cities with crime rates below the cutoff enjoyed a greater effect of park area, as in Portland (crime rate = −6.98, b = 0.50, 95% CI = [0.02, 0.98], p = .042) and Los Angeles (crime rate = −20.30, b = 1.15, 95% CI = [0.53, 1.77], p < .001).

By contrast, park area predicted lower volunteering in cities with above-average crime rates. The cutoff was 15.68 above average and the effect of park would be –.60, 95% CI = [−1.21, 0.00], t(47) = 2.01, p = .05. Put otherwise, every 10% increase in park area would predict a 6-hr decrease in volunteering. The cutoff value was equivalent to 0.96 of the standard deviation of all crime rates. Examples of cities above the cutoff were Memphis (crime rate = +23.47, b = −0.98, 95% CI = [−1.71, –0.26], p = .009) and St. Louis (crime rate = +31.09, b = −1.35, 95% CI = [−2.22, –0.49], p = .003).

Table 2 also reports the interaction model when parkland was specified as designed (Model 2) or natural (Model 3) areas. The interaction effect was significant for designed areas, b = −.06, 95% CI = [–0.11, –0.01], p = .016, and natural areas, b = −0.07, 95% CI = [–0.10, –0.03], p < .001. Their simple slopes were indistinguishable from those reported using total park area. Parks’ naturalness did not affect the interpretation of the park-by-crime interaction effect.

Alternative explanation

As reported above, cost of living predicted more volunteering, more park areas, and lower crime rates. As such, the effects of park and crime on volunteering may reflect but a city’s socio-economic status. That is, expensive cities may just devote more money to build and manage parks, fight crime, and support volunteer campaigns. Population density may also confound the results as it marginally predicted lower volunteering.

We examined the park-by-crime interaction with cost of living and population density controlled statistically (Table 2, Model 4). The interaction effect remained significant and the conditional effects were consistent with those reported above: Park area predicted more volunteering in low-crime cities (equal to or less than −10.21), but less volunteering in high-crime cities (equal to or greater than +35.36). Moreover, the two covariates no longer predicted volunteering, indicating that the two variables did not explain the data well.

Discussion

The resource-security model predicted that urban greenspace does not always enhance prosociality. Although greenspace contains high-resource features (vegetation) that should promote helping, it also contains features of security and insecurity that can bolster and undermine helping. We proposed that the security feature is tied to crime rate as parks are often perceived as a hot spot of crime. Study 1 supported the argument by showing that park area predicted more helping in low-crime cities, but less helping in high-crime cities. The findings contradict the current view that exposure to greenspace always enhance prosociality.

Study 1 enjoyed much ecological validity because it was based on objective environment variables and real-world data of crime and helping. The findings, however, failed to address individuals’ perception of the resource and security features as proposed. This is problematic because the evolutionary perspective emphasized the connection between environmental features and psychological responses. Study 2 sought to address this issue with participants responding in a laboratory setting. It assessed individual’s perception of resource and security features and explored the mediating role of perceived beauty in nature’s prosocial effect.

Study 2

In this study, we compared participants’ prosociality after having them write about natural environments with different resource-security profiles. A resource (high/low) by security (high/low) interaction effect was inspected (H2).

Method

Participants

A previous research adopting the resource-security model reported a median effect size of $η_{p}^{2}$ = .11 (Ng & Chow, 2017). Power analysis (α = .05, β = .99) demanded a sample of 151 to detect a 2 × 2 interaction effect. We recruited 160 undergraduates (119 women; M_age = 20.06, SD_age = 1.51) from an English-medium university in Hong Kong. Of these, 85% were Chinese, 5.63% Koreans, 5.63% White, and 3.75% Others. Demographic differences did not correlate with major outcome variables and were thus not reported below. Participants signed up the experiment for course credits.

Natural environments

Participants were instructed to write about a natural environment in the 2 × 2 between-participants design (n = 40). The instructions read, “We would like you to think about a natural environment that you consider as [rich/poor] in resources (e.g., food, water) and [high/low] in security (e.g., having shelters, free from natural hazards).” Participants put down the environment and described it briefly to justify its resource and security levels. They then wrote three actions that they must take and three that they must not to survive in the environment. They then estimated the days they could survive on their own without assistance.

Participants then completed the Environmental Resource and Security Scale (ERSS; Ng, 2015). It had seven adjectives assessing their perceived resource level (nurturing, resource-rich, replenishing, fertile, lively, flexible, varied) and 10 adjectives assessing perceived security level (dangerous, frightful, creepy, safe, mysterious, dark, friendly, gentle, rough, degraded) of the environment. Previous research has supported the validity and two-factor structure of the scale (Ng, 2015) and has used it to differentiate natural environments based on the resource-security model (Ng & Chow, 2017). Respondents should report higher resource levels in high-resource conditions and higher security levels in high-security conditions. Finally, they responded to three aesthetic items (beautiful, magnificent, attractive). All items were rated on a 7-point scale (1 = very uncharacteristic, 4 = neutral, 7 = very characteristic). The Cronbach’s alphas of the three scores were .94, .87, and .91 respectively.

Aspiration Index

We defined prosociality as valuing other-focused intrinsic aspirations and devaluing self-focused extrinsic aspirations. The two aspirations formed the motivational basis for prosocial behavior and have been used to demonstrate the prosocial effect of nature exposure in laboratory settings (Weinstein et al., 2009). We expected that participants exposed to natural environments in the Resource × Security matrix would report different levels of intrinsic and extrinsic values.

Participants responded to the 42-item Aspiration Index (Kasser & Ryan, 1996). It measured seven life aspirations, namely, self-acceptance, affiliation, community feeling, physical fitness, financial success, attractive appearance, and social recognition. They rated the importance of each aspiration on a 5-point scale (1 = not at all, 3 = so/so, 5 = very important).

An intrinsic composite was computed by averaging subscale scores of Affiliation (e.g., “You will share your life with someone you love”; α = .75) and Community Feeling (e.g., “You will work for the betterment of society”; α = .87). The two subscales (r = .42, p < .001) emphasized others’ benefits and prosocial values, such as helping and sharing with others. An extrinsic aspiration composite was computed by averaging subscale scores of Financial Success (e.g., “You will have a lot of expensive possessions”; α = .76) and Social Recognition (e.g., “You will be admired by many people”; α = .87). The two subscales (r = .58, p < .001) emphasized self-interests that were built on external rewards (i.e., money, fame).

We computed an intrinsic aspiration index and extrinsic aspiration index by subtracting the grand average of all seven subscales from each composite score. The two indexes reflected the importance of intrinsic and extrinsic aspirations relative to the overall life aspiration profile. Higher intrinsic aspirations and lower extrinsic aspirations would indicate higher prosociality (Weinstein et al., 2009).

Procedure

Experiment sessions were held individually in a quiet cubicle. All instructions and materials were delivered on a tablet computer. In the first part, participants were led to believe that it was a test of their writing skills. They wrote about one natural environment that belonged to the Resource × Security matrix and rated it using the ERSS and beauty items. They then proceeded to an allegedly unrelated second study that was about life values and completed the Aspiration Index. As a suspicion check, they were asked if the two studies were actually related and how. Twelve participants responded “yes,” but their responses were unrelated to the environmental effect on prosociality. Participants were then debriefed and dismissed.

Results

Table 3 summarizes the descriptive statistics of all major variables.

Table 3.

Means and Standard Deviations (in Parentheses) of Major Outcome Variables in Studies 2 to 4.

	LRLS	HRLS	LRHS	HRHS	Urban	Shape
Study 2 (N = 160)
ERSS—Resource	2.96 (1.64)	4.88 (1.22)	3.26 (1.18)	4.69 (1.33)	—	—
ERSS—Security	3.29 (1.00)	3.39 (1.02)	3.88 (0.90)	3.99 (1.40)	—	—
Beauty	3.33 (1.60)	5.19 (1.15)	4.39 (1.75)	5.16 (1.21)	—	—
Intrinsic Aspiration	0.53 (0.26)	0.45 (0.29)	0.25 (0.43)	0.46 (0.38)	—	—
Extrinsic Aspiration	−0.73 (0.41)	−0.53 (0.31)	−0.53 (0.48)	−0.68 (0.45)	—	—
Study 3 (N = 160)
ERSS—Resource	3.05 (1.58)	4.83 (1.09)	3.51 (1.40)	4.93 (0.75)	—	—
ERSS—Security	3.88 (1.14)	3.75 (0.97)	5.29 (1.01)	5.70 (0.65)	—	—
Beauty	4.47 (1.51)	5.08 (1.43)	5.72 (0.98)	5.53 (0.98)	—	—
General Trust	3.63 (0.62)	3.24 (0.66)	3.35 (0.68)	3.54 (0.59)	—	—
SVO	36.37 (10.98)	26.99 (16.21)	25.79 (18.07)	32.91 (12.03)	—	—
Study 4 (N = 360)
ERSS—Resource	3.74 (1.28)	5.44 (0.99)	4.04 (1.46)	5.21 (1.26)	3.99 (1.05)	—
ERSS—Security	4.28 (1.32)	4.66 (0.96)	5.45 (0.69)	5.54 (0.83)	4.50 (.89)	—
Beauty	4.59 (1.33)	5.66 (1.19)	5.38 (1.22)	5.89 (0.84)	4.03 (1.37)	—
Easy tangrams	5.57 (2.71)	4.58 (2.35)	4.90 (2.90)	6.28 (2.82)	4.92 (2.31)	4.90 (2.80)

Note. LRLS = low-resource-low-security; HRLS = high-resource-low-security; LRHS = low-resource-high-security; HRHS = high-resource-high-security; ERSS = Environmental Resource and Security Scale; SVO = social value orientation.

Preliminary analysis

The ERSS resource and security scores were compared in the 2 × 2 ANOVA. The main effect of resource manipulation was significant on the resource score, F(1, 156) = 61.18, p < .001, $η_{p}^{2}$ = .282. High-resource (vs. low) conditions scored higher on the resource dimension. Similarly, the main effect of security manipulation was significant on the security score, F(1, 156) = 11.64, p < .001, $η_{p}^{2}$ = .069. High-security (vs. low) conditions scored higher on the security dimension. Other main and interaction effects on ERSS scores were nonsignificant. The manipulation was successful.

We also examined the estimated days of survival with the 2 × 2 ANOVA. Both the resource and security main effects were significant. Participants expected a longer period of survival in high-resource (vs. low) conditions, F(1, 152) = 4.37, p = .038, and high-security (vs. low) conditions, F(1, 152) = 6.81, p = .01. The interaction effect was not significant (p > .250). The results showed that both environmental resource and security were related to the survival opportunities of people.

Perceived beauty

We entered beauty scores into the 2 × 2 ANOVA. The main effects of resource, F(1, 156) = 32.78, p < .001, $η_{p}^{2}$ = .174, and security, F(1, 156) = 5.00, p = .027, $η_{p}^{2}$ = .031, were both significant. Participants perceived high-resource and high-security conditions as more attractive than their low-level counterparts. A significant interaction effect was also observed, F(1, 156) = 5.67, p = .018, $η_{p}^{2}$ = .035. While resource led to higher perceived beauty, the effect was greater in low-security conditions (g = 0.90, 95% CI = [0.58, 1.23], p < .001) than in high-security conditions (g = 0.37, 95% CI = [0.06, 0.68], p = .019).

Intrinsic aspirations

Intrinsic aspiration index was analyzed in the 2 × 2 ANOVA. The main effect of resource was nonsignificant, F(1, 156) = 1.18, p > .250, $η_{p}^{2}$ = .008, but that of security was significant, F(1, 156) = 5.98, p = .016, $η_{p}^{2}$ = .037. Participants reported higher importance of intrinsic aspirations in low-security conditions (M = 0.49, SD = 0.27) than in high-security conditions (M = 0.35, SD = 0.41).

Moreover, the interaction effect of resource and security on intrinsic aspirations was significant, F(1, 156) = 6.85, p = .010, $η_{p}^{2}$ = .042, supporting the second hypothesis that the effect of resource is qualified by security (Figure 1, left panel). The graph shows that the low-resource-low-security condition recorded the highest intrinsic aspirations, followed by the high-resource-high-security condition. We further examined the simple main effects of resource at different levels of security. In low-security conditions, higher resource led to lower intrinsic aspirations, although the effect was nonsignificant, g = −0.17, 95% CI = [–0.48, 0.14], p > .250. However, in high-security conditions, resource significantly led to higher intrinsic aspirations, g = 0.41, 95% CI = [0.10, 0.73], p = .010.

Figure 1.

Results from Study 2: Mean scores of intrinsic aspirations (left panel) and extrinsic aspirations (right panel) across the resource (high vs. low) by security (high vs. low) design.

Extrinsic aspirations

Similarly, we entered the extrinsic aspiration index into the 2 × 2 ANOVA. The main effects of resource, F(1, 156) = 0.10, p > .250, $η_{p}^{2}$ = .001, and security, F(1, 156) = 0.09, p > .250, $η_{p}^{2}$ = .001, were both nonsignificant.

A significant resource-by-security interaction effect (Figure 1, right panel) was also observed, F(1, 156) = 6.93, p = .009, $η_{p}^{2}$ = .043. In low-security conditions, higher resource led to higher extrinsic aspirations (g = 0.33, 95% CI = [0.02, 0.64], p = .038). In high-security conditions, resource led to lower extrinsic aspirations, although the effect did not reach statistical significance (g = −0.26, 95% CI = [–0.57, 0.05], p = .105).

Moderated mediation

The resource-by-security interaction effect was significant on perceived beauty, and intrinsic and extrinsic aspirations. However, beauty did not correlate with intrinsic (r = .07, p > .250) and extrinsic (r = −.05, p > .250) aspirations, suggesting that it would not mediate the interaction effect on both outcome variables. Confirming this, the moderated mediation by perceived beauty was nonsignificant for intrinsic aspirations (indirect effect = −0.03, SE = 0.03, 95% CI = [–0.10, 0.02]) and extrinsic aspirations (indirect effect = 0.04, SE = 0.04, 95% CI = [–0.01, 0.13]). The moderated mediation model is summarized in Figure 2.

Figure 2.

Moderated mediation model from Study 2 showing the effects of resource, security, and their interaction on intrinsic and extrinsic aspirations, mediated by perceived beauty of the environment.

Discussion

According to H2, we expected that natural environments of different resource-security profiles would lead to different levels of prosociality. Study 2 supported this hypothesis: Participants were more prosocial in the most favorable (HRHS) and unfavorable (LRLS) conditions compared with the LRHS and HRLS conditions. Confirming the evolutionary perspective, high-resource and high-security features led to more positive aesthetic responses, but such responses did not explain the differences in prosocial behavior.

Despite its strength, there were several notable limitations in Study 2. Although the Aspiration Index has been used to demonstrate nature’s effect on prosociality (Weinstein et al., 2009), it was not a behavioral measure and people could help because of both intrinsic and extrinsic motives. A better measurement of prosociality is needed. Second, the writing procedure may not engage participants to visualize the environment, which may lead to the failure of the aesthetic mechanism. The unrestricted writing task may also introduce other confounding variables (e.g., mortality salience) that obscured the aesthetic mechanism. In view of these limitations, we conducted Study 3 to re-examine H2 with modified procedures.

Study 3

Study 3 re-examined H2 with two major modifications. We used slideshows of nature images as stimuli, which should address the limitations of the writing task. We also replaced the Aspiration Index with a social value orientation (SVO) task to measure prosociality. Similar procedures have been used to study nature’s prosocial effect (e.g., Weinstein et al., 2009, Study 3; Zelenski et al., 2015, Study 3).

Method

Participants

We recruited 160 undergraduates (105 females, 55 males, M_age = 20.44, SD_age = 1.56) from an English-medium university in Hong Kong, who participated for course credits. Among them, 73.2% were Chinese, 4.4% Caucasians, 3.1% Koreans, 1.3% Arabs, and 18.1% others. Demographic differences did not correlate with major variables and were not included in the following analysis.

Nature stimuli

Ng (2015) used the ERSS to survey people’s perceived resource and security levels of various natural environments. The ratings yielded four environment clusters, namely, high-resource-high-security (HRHS), high-resource-low-security (HRLS), low-resource-high-security (LRHS), and low-resource-low-security (LRLS). Ng and Chow (2017) used picture stimuli from each cluster to test the effects of environmental resource (high/low) and security (high/low) on aggression. We adopted the same approach in this study.

Based on Ng’s (2015) cluster analysis, we used pictures of gardens (HRHS), rainforests (HRLS), beaches (LRHS), and deserts (LRLS) as stimuli because respondents have consistently reported different resource and security ratings of them. For each condition, we collected 60 pictures from the Internet and selected 20 of them as visual stimuli. We screened out images with animals, humans, artificial structures, ongoing actions (e.g., lightning), and artistic effects, and retained those that had similar clarity, brightness, depth, and field of view. The procedure yielded 20 mundane pictures for each environment (see Figure 3 for samples). A pilot test (n = 10, repeated measures) on students naïve to the study confirmed the manipulation: High-resource pictures received higher ratings on resource features (ΔM = 2.13, F = 18.66, p = .002) and high-security pictures received higher ratings on security features (ΔM = 1.60, F = 9.35,p = .014).

Figure 3.

Sample picture stimuli of the four conditions in Study 3.

Each picture was shown on a computer screen for 6 s. Participants were told to imagine themselves being in the environment and pay attention to various aspects of the place (e.g., colors, smells, sounds). The imagery script was adopted from Weinstein et al. (2009). It was reminded intermittently during the slideshow. Participants then reported the environment they had seen and rated its resource, security, and beauty with the items used in Study 2 (α_resource = .92, α_security = .92, α_beauty = .82).

General Trust Scale

Participants completed the six-item General Trust Scale (Yamagishi & Yamagishi, 1994), which measured their positive bias toward strangers (e.g., “Most people are trustworthy”). Participants responded on a 5-point scale (1 = strongly disagree, 5 = strongly agree). As mentioned previously, it was included as a possible mechanism in nature’s effect. The Cronbach’s alpha in this study was .82.

Prosocial decision making

We used the SVO Slider (Murphy, Ackermann, & Handgraaf, 2011) to measure prosociality. In six trials, participants had to allocate points among themselves and another ostensible participant. The points assigned to each player would become a bonus factor to the assessment of their introductory psychology course. Participants were assured that their responses would be anonymous, and the pair would not know the identity of each other.

An SVO index was calculated using the average payoffs for the self (Ā_S) and others (Ā_O):

SVO = \arctan (\frac{{\bar{A}}_{O} - 50}{{\bar{A}}_{S} - 50}) .

The index reported how much people were willing to benefit others at the expense of self-interests, with higher scores meaning higher prosociality. The range of scores could be divided into four categories: competitiveness that maximized gain differences (−16.26 to −12.04), individualism that maximized self-gain (−12.04 to 22.45), prosociality that minimized gain difference or maximized joint payoffs (22.46 to 57.15), and altruism that maximized others-gain (57.15 to 61.39).

Procedure

The experiment took place individually in an isolated cubicle. Once seated, participants watched a slideshow of one of the four natural environments. They then identified the environment they saw, and completed the ERSS, the three beauty items, and the General Trust Scale. A message was shown on the next screen, apparently connecting them to another student in a different room. They were then instructed to finish the SVO task. No participant reported any suspicion about the instructions. Participants were then debriefed and dismissed.

Results

Table 3 summarizes the descriptive statistics of all major variables.

Manipulation check

All participants correctly identified the environment they saw. ERSS ratings confirmed the intended manipulation: High-resource (vs. low) conditions scored higher on the resource dimension, F(1, 156) = 65.89, p < .001, $η_{p}^{2}$ = .297, whereas high-security (vs. low) conditions scored higher on the security dimension, F(1, 156) = 122.57, p < .001, $η_{p}^{2}$ = .440. The other main and interaction effects were nonsignificant.

Beauty

High-security (vs. low) environments were perceived as more beautiful, F(1, 156) = 18.52, p < .001, $η_{p}^{2}$ = .106. The effect of resource was nonsignificant, F(1, 156) = 1.11, p = .293, $η_{p}^{2}$ = .007, but it was qualified by security in a significant two-way interaction, F(1, 156) = 4.10, p = .045, $η_{p}^{2}$ = .026. In low-security conditions, participants gave higher ratings to high-resource rainforests than low-resource deserts (g = 0.34, 95% CI = [0.03, 0.65], p = .031). The effect of resource, however, was nonsignificant in high-security conditions (g = −0.11, 95% CI = [–0.42, 0.20],p > .250).

General trust

The main effects of resource and security on trust were nonsignificant, F(1, 156) = 1.06 and 0.01 respectively, ps > .250. The Resource × Security interaction effect was significant, F(1, 156) = 8.33, p = .004, $η_{p}^{2}$ = .051 (Figure 4, left panel). The simple main effect of resource in low-security conditions was negative (g = −0.44, 95% CI = [–0.75, –0.12]) and significant (p = .006): Participants reported higher trust after seeing deserts (LRLS) than rainforests (HRLS). In high-security conditions, the simple main effect of resource was positive but nonsignificant (g = 0.21, 95% CI = [–0.10, 0.52], p = .191).

Figure 4.

Results from Study 3: Mean scores of general trust (left panel) and social value orientation (right panel) across the resource (high vs. low) by security (high vs. low) design.

SVO

SVO was entered into the 2 × 2 ANOVA. The main effects of resource and security were nonsignificant, F(1, 156) = 0.24 and 1.02, respectively, ps > .250. Supporting H2, the Resource × Security interaction effect was significant, F(1, 156) = 12.74, p < .001, $η_{p}^{2}$ = .075 (Figure 4, right panel). The simple main effect of resource in low-security conditions was negative (g = −0.45, 95% CI = [–0.77, –0.14], p = .005). Participants were more prosocial after seeing the desert pictures (LRLS) than the rainforest pictures (HRLS). In high-security conditions, the simple main effect of resource was reversed and smaller in size (g = 0.34, 95% CI = [0.03, 0.65], p = .031). Put otherwise, participants were more prosocial after seeing the garden pictures (HRHS) than the beach pictures (LRHS). The findings were consistent with Study 2.

It is worth noting that the SVO mean scores of the four conditions all fell into the range of prosociality (22.45 to 57.15). The 95% CIs of LRLS, [32.97, 39.77], and HRHS, [29.18, 36.64], were within the prosocial range, whereas those of the LRHS, [21.23, 30.36], and HRLS, [22.43, 31.56], overlapped with the range of individualism (−12.04 to 22.45). Taken together, participants in the four conditions were mostly prosocial and focused on maximizing joint gains, but there was no sign of altruistic helping. Besides, participants in the LRHS and HRLS conditions showed hints of self-focused individualism.

Moderated mediation

As reported above, general trust and perceived beauty were influenced by the Resource × Security interaction effect. However, SVO correlated with general trust only (r = .31, p < .001), but not with perceived beauty (r = −.08, p > .250). This suggested that general trust was a better mediator of the two in explaining the interaction effect. We tested the two mechanisms with moderated mediation analysis.

Figure 5 summarizes the model in which trust and beauty were entered as mediators between the effects of resource and security on SVO. Bootstrapped results (10,000 iterations) showed that the Resource × Security interaction effect on SVO was mediated by general trust (indirect effect = 3.36, SE = 1.90, 95% CI = [0.58, 8.20]) but not by beauty (indirect effect = 0.45, SE = 0.84, 95% CI = [–0.74, 2.91]). We then examined the indirect effect of resource on SVO through general trust at the two security levels. The indirect effect was significant in low-security conditions (indirect effect = −2.28, SE = 1.29, 95% CI = [–5.57, –0.39]): High resource led to lower trust among participants, which further predicted lower prosociality in the SVO task. However, in high-security conditions, the indirect effect was nonsignificant (indirect effect = 1.08, SE = 1.00, 95% CI = [–0.29, 3.85]), which was due to the nonsignificant simple effect of resource on trust. Finally, entering trust and beauty into the model reduced the Resource × Security interaction effect from 16.49, 95% CI = [7.36, 25.62], to 12.68, 95% CI = [3.38, 21.97]. The reduced interaction effect remained significant (p = .008), suggesting that substantial variance was left unexplained.

Figure 5.

Moderated mediation model from Study 3 showing the effects of resource, security, and their interaction on social value orientation, mediated by general trust and beauty.

Discussion

Study 3 provided further evidence in support of the heterogeneity in nature’s prosocial effect. Consistent with Study 2, resource led to higher prosociality when security was high, but lower prosociality when security was low. Study 3 further showed that helping in the adverse LRLS condition was driven by people placing more trust in strangers. The trust mechanism, however, did not explain helping in the pleasant HRHS condition. Aesthetic response did not explain the prosocial effect in any conditions.

The evidence so far suggests that different natural environments lead to different levels of prosociality. Does that imply that certain nature experiences can undermine helping? Findings from Study 3 are ambiguous because the SVO mean scores all fall within the prosocial range, but the 95% CIs of the LRHS and HRLS groups also cover the range of individualism. However, it remains unclear whether all four natural environments promote helping to different extents, or some of them actually reduce helping. Study 4 sought to address these alternatives by comparing the four natural environments with two control conditions.

Study 4

Study 4 compared prosociality in the Resource × Security matrix with two control conditions (urban, shapes). If some nature experiences reduce helping, they should record less prosociality than the control. It also sought to extend the findings to another validated measure of prosociality: the Tangram Help/Hurt Task.

Method

Participants

We recruited 360 undergraduates (208 females, 152 males, M_age = 20.80, SD_age = 2.20) from three English-medium universities in Hong Kong. The majority of the sample were Chinese (98.3%). Demographic differences did not correlate with the outcome variables and were not included in the following analysis. Students signed up for course credits or a reward of HK$50 (~US$7).

Environment stimuli

Participants were randomly assigned to one of six conditions (ns = 60), including four natural environments and two control conditions. The natural environments were identical to the 2 × 2 design in Study 3. The control conditions comprised an urban control and a shape control, which exposed participants to urban environments and geometric shapes, respectively.

In each condition, participants saw a slideshow of 20 pictures, each lasting for 6 s. We used the same nature stimuli as in Study 3. For the urban condition, stimuli were adapted from Berman, Jonides, and Kaplan (2008). We discarded urban images with ongoing actions, humans, and green elements (e.g., lawns, trees). Participants in the natural and urban conditions received the imagery script in Study 3. For the shape control, participants saw 20 geometric shapes printed in black and white.

After the slideshow, participants in the natural and urban conditions responded to the ERSS (α_resource = .91; α_security = .88) and the three beauty items used in Study 3 (α = .91). Those in the shape control saw the same items but were not required to respond.

Tangram help/hurt task

The task (Saleem, Anderson, & Barlett, 2015) was used to measure prosociality. Tangrams are puzzles that require players to form outlined shapes using at most seven blocks. Difficult puzzles require more blocks to solve. Participants learned the objective of the tangram and worked on three practice puzzles at easy (2 to 3 blocks), medium (4 to 5 blocks), and hard levels (6 to 7 blocks). They then learned that they were connected to another ostensible participant from a different campus and they had to assign 11 puzzles to the stranger from a list of 30 (10 easy, 10 medium, 10 hard). The stranger would receive a HK$50 (~US$7) voucher if they completed all the puzzles within 10 min. Participants were assured of anonymity and confidentiality of their responses.

After the tangram assignment, participants rated the easy, medium, and hard subsets on a 7-point difficulty scale. Hard puzzles were rated the most difficult (M = 4.36, SD = 1.61), followed by medium (M = 2.96, SD = 1.51) and easy puzzles (M = 1.88, SD = 1.16), F(2, 718) = 387.03, p < .001. Helping was operationalized as the number of easy puzzles selected. Previous research has supported the construct validity of the task (e.g., Saleem et al., 2015).

Procedure

The experiment took place in computer labs that could seat up to eight persons in one session. The seats were separated so that participants could not see or communicate with each other. Participants saw a 2-min slideshow of the assigned condition and rated it using the ERSS and beauty items (except for the shape control condition). They then completed the tangram assignment task. Thirty-four participants reported suspicions related to the true purpose of the study. Discarding their responses did not substantially change the results and were retained. Finally, they were thoroughly debriefed and dismissed.

Results

Table 3 summarizes the descriptive statistics of all major variables.

Manipulation check

We analyzed resource and security ratings in the 2 × 2 ANOVA. High-resource natural environments received higher resource ratings than the low-resource counterparts, F(1, 236) = 78.37, p < .001, $η_{p}^{2}$ = .25. Similarly, high-security natural environments received higher security ratings than low-security environments, F(1, 236) = 65.45, p < .001, $η_{p}^{2}$ = .22. Other main effects and interaction effects were nonsignificant. The manipulation was successful.

We further compared resource and security ratings between the five environmental conditions using one-way ANOVA. Resource and security varied significantly across the five conditions: F_resource(4, 295) = 24.69, p < .001, $η_{p}^{2}$ = .25; F_security(4, 295) = 21.23, p < .001, $η_{p}^{2}$ = .22. Tukey’s test reported two homogeneous subsets for resource ratings, with LRLS, LRHS, and urban in the low-resource subset (Ms = 3.74 to 4.04), and HRHS and HRLS in the high-resource subset (Ms = 5.21 and 5.44). Tukey’s test also reported two subsets for security ratings, with LRLS, HRLS, and urban in the low-security subset (Ms = 4.28 to 4.66), and LRHS and HRHS in the high-security subset (Ms = 5.45 and 5.54). The urban environment was perceived as resource-poor and insecure as the LRLS natural environment (deserts).

Beauty

Beauty ratings were analyzed in the 2 × 2 ANOVA. High-resource and high-security natural environments received higher ratings than their low-level counterparts:F_resource(1, 236) = 27.88, p < .001, $η_{p}^{2}$ = .11; F_security(1, 236) = 11.49, p = .001, $η_{p}^{2}$ = .05. The interaction effect was nonsignificant, F(1, 236) = 3.36, p = .068, $η_{p}^{2}$ = .01. Resource and security features influenced aesthetic responses.

We further compared beauty scores between the five environmental conditions using one-way ANOVA. The effect was significant, F(1, 295) = 24.96, p < .001. Tukey’s test put urban and LRLS environments into the less beautiful subset (Ms = 4.03 to 4.59), and LRHS, HRLS, and HRHS into the more beautiful subset (Ms = 5.38 to 5.89). The urban environment was perceived as unattractive as the LRLS condition. Finally, beauty did not correlate with prosociality (r < .01), suggesting that helping was not explained by aesthetic responses to the environment.

Helping

Figure 6 presents the number of easy puzzles assigned in each condition.

Figure 6.

Results from Study 4: Mean number of easy puzzles assigned in the natural and control conditions.

Nature-control contrasts

One-way ANOVA revealed significant differences between the six conditions, F(5, 354) = 3.31, p = .006, $η_{p}^{2}$ = .05. Contrast tests revealed that the four natural conditions (combined) did not lead to higher helping than the two control conditions (combined), t(354) = 1.43, p = .154. The two control conditions did not differ from each other either, t = 0.03, p > .250. The contrasts indicated that not all natural environments led to more helping compared with the control conditions.

We then tested contrasts comparing each natural condition with the two controls combined. HRHS exposure led to more helping than control, t(354) = 3.27, p = .001. LRLS exposure also recorded more helping, although the effect was nonsignificant, t(354) = 1.57, p = .118. HRLS and LRHS did not change helping significantly: ts(354) = −0.77 and –0.02, respectively, ps > .250. The results indicated that only the HRHS condition significantly enhanced helping, but there was no evidence that the other nature experiences undermined helping.

Resource × Security interaction

We analyzed helping in the four natural conditions using the Resource × Security ANOVA. The main effects of resource and security were nonsignificant, F_resource(1, 236) = 0.33, p > .250, $η_{p}^{2}$ = .001; F_security(1, 236) = 2.19, p = .140, $η_{p}^{2}$ = .009. The interaction effect was significant, F(1, 236) = 11.48, p = .001, $η_{p}^{2}$ = .046. Consistent with findings from Studies 2 and 3, resource decreased helping when security was low (g = −0.36, 95% CI = [–0.72, 0.00], p = .048), but increased helping when security was high (g = 0.51, 95% CI = [0.14, 0.87], p = .006). The interaction pattern remained the same when perceived beauty was statistically controlled, F(1, 235) = 10.70, p = .001, $η_{p}^{2}$ = .044.

Discussion

By introducing two control groups, Study 4 concluded that not all natural environments led to greater prosociality. HRHS and LRLS environments led to highest prosociality, but only the former was significantly different from the controls. This confirmed previous findings that lush green vegetation (HRHS) could improve prosociality when compared with urban environments. Moreover, the LRHS and HRLS groups were least helpful among the four natural conditions, but they were indistinguishable from the controls. There was thus no evidence that natural environments could decrease helping. Taken together, Study 4 supported the heterogeneity of nature’s effect on prosociality (H2). The interaction pattern was consistent with that reported in Studies 2 and 3. Again, controlling perceived beauty did not seem to affect the results.

General Discussion

The prosocial effect of nature is fascinating, but it is unlikely to be uniform because nature is a heterogeneous entity. This article argued that the effect pertained to hunter-gatherers’ resource (food, water) and security (shelter, concealment) concerns in habitat selection. The two features were crucial for survival and reproductive success, and modern humans remained responsive to them. Building on previous findings that resource and security statuses influenced prosociality, we investigated if nature’s prosocial effect was related to its resource and security features. Study 1 showed that urban greenspace, a resource feature to urban dwellers, predicted more helping in low-crime cities but less helping in high-crime cities. Crime as a park-related security factor moderated the effect of resource on helping. Studies 2 and 3 extended the effect to resource and security variations between natural sceneries. People were most prosocial in the supportive, high-resource-high-security condition and unsupportive, low-resource-low-security condition. Study 4 added that compared with control conditions (urban, shape), not all natural environments promoted helping, nor did they undermine helping. Below we discuss the implications of these findings and limitations that deserve attention in future studies.

Urban Greening as Social Intervention

Researchers and public opinions are optimistic about the effects of nature exposure. The World Health Organization (2017) promoted the environmental and psychological benefits of greenspace and made suggestions about how greening should be implemented in urban areas. In a research of urban forestry, McPherson, Simpson, Peper, Maco, and Xiao (2005) estimated the annual economic returns (e.g., energy saving, reduced emission) ranging from 37% to 300% per tree. Psychological research also supported the various psychosocial benefits of nature experiences (Berman et al., 2008; Kuo et al., 1998; Sullivan et al., 2004; Weinstein et al., 2009).

It is therefore tempting to conclude that humans always benefit from nature without considering possible modifiers. Study 1 challenges this view by showing that the prosocial effect of greenspace is susceptible to changes in crime rates. It promotes helping in low-crime cities but backfires as crime increases. Current major theories, which assume a homogeneous nature’s effect, fail to explain such findings. The resource-security model fills the gap by relating crime rates to security concerns in vegetated areas (Michael et al., 2001; Nasar & Fisher, 1993). It posits that greenspace promotes helping through high-resource features (i.e., food, water), but the effect is compromised if the greenspace also evokes insecurity, as is the case when trees and bushes provide concealment for perpetrators (Andrews & Gatersleben, 2010). The findings challenge the homogeneous view of nature’s effect and call for more investigation into structural variables that may modify the effect.

Most parks are built for leisure and recreational purposes, but not for promoting volunteerism. Still, we believe the prosocial effect can be sold as a pleasant outcome of increasing urban greenspace. On the contrary, authorities seeking to monitor and increase volunteerism should be aware of the negative social effect greenspace can bring and consider the issue together with crime issues in town. In high-crime cities, for example, local governments should consider prioritizing crime fighting over urban greening if their main objective is to increase volunteerism.

Psychological Aspects of Resource and Security

Why do people act more prosocially in natural environments? Previous works have associated the effect to the unique psychological connection between humans and nature but left the underlying environmental features relatively unexplored. This research decomposes nature experiences into perceived resource and security. Studies 2 to 4 suggest that people do perceive natural environments differently on the two features and their prosocial response varies systematically within the resource–security matrix. People in high-resource-high-security nature act more prosocially than the resource-deprived or insecure counterparts. The effect is understandable when one considers the cost-benefit trade-off of helping in hunter-gatherer societies. Living together in small groups and helping each other rewarded our ancestors with great evolutionary success, but the prosocial behavior also harmed immediate self-interest. Helping was thus subject to environmental constraints as people sought to minimize the cost of the behavior (Graziano et al., 2007). Apparently, this mentality still applies to present-day people, as more helping is observed among those who are better off and live in secure communities (Bach et al., 2017; Korndörfer et al., 2015). Therefore, it is not surprising that people tend to act more prosocially when the constraints are lessened, as in seeing high-resource-high-security natural environments.

Interestingly, we also observe high prosociality in the low-resource-low-security condition, which is the least favorable environment among the four. The finding deviates from the above argument that people are more helpful in supportive environments. We speculate that in adverse environments, the implied cost of helping is too high (lack of resources and shelter) that the benefit of helping becomes the only limiting factor. As humans are powerless to combat environmental hardships individually, helping each other may have evolved as an adaptive collective response that increases overall survival opportunities (Van Vugt & Van Lange, 2006). In line with this, research showed that extreme ecological threats and natural disasters compelled people to connect with and help each other (Gelfand et al., 2011; Rao et al., 2011; Van de Vliert et al., 2012).

Our studies show that resource and security features enhance aesthetic responses to nature, but people do not help because of seeing beautiful sceneries. Indeed, LRLS environments (deserts) are consistently rated as least beautiful among the stimuli. As such, it is likely that HRHS and LRLS environments increase helping through different mechanisms. Study 3 demonstrates that helping in the LRLS condition is driven by people placing more trust in others. By assuming others as honest and good, people may offer unconditional help despite the constraints they experience in the harsh environment. The trust mechanism, however, does not explain helping in HRHS settings. We speculate that helping in such conditions require more self-regulatory control that suppresses immediate gratification in exchange for greater reward in the future. Research has shown that nature experiences can improve such cognitive abilities (Berman et al., 2008; Beute & de Kort, 2014). Studying the self-regulatory mechanism will inform future research on nature’s prosocial effect.

Limitations and Future Research

There are several notable limitations in this research that warrant further investigation. First, we did not look into physical elements that pertain to resource and security features perceived by people. It is difficult to achieve because there are numerous interrelated elements in natural environments. Based on previous works (e.g., Appleton, 1996; Gatersleben & Andrews, 2013; Van de Vliert et al., 2012; White et al., 2010), we propose some elements that are worth studying, including vegetation of different qualities (e.g., height, density, blockage, fruits), the absence or presence of water, sight of different animals (e.g., predators, prey), and periodic changes in climate. Identifying these effective ingredients will help researchers understand how microscopic changes in the environment alter prosocial behavior.

Second, resource and security certainly do not cover all variations in nature. Other environmental features may as well contribute to the prosocial effect. Awe-inducing nature experiences, for example, can increase prosociality regardless of resource and security features (Piff, Dietze, Feinberg, Stancato, & Keltner, 2015). An awe-inducing volcano eruption may just be as effective as seeing an HRHS environment in inducing helping. These works, though taking different perspectives, call for a more heterogeneous view of nature’s effect in the future. Finally, we did not control participants’ familiarity and knowledge with the environment in Studies 2 to 4. Although the manipulation was successful in all studies, our Hong Kong participants were probably more familiar with the HRHS condition (i.e., gardens) than the others. Future research should take into account such individual differences.

Supplemental Material

Ng_OnlineAppendix – Supplemental material for Nature Does Not Always Give You a Helping Hand: Comparing the Prosocial Effects of Nature at Different Resource and Security Levels

Supplemental material, Ng_OnlineAppendix for Nature Does Not Always Give You a Helping Hand: Comparing the Prosocial Effects of Nature at Different Resource and Security Levels by Henry Kin Shing Ng, Yee-Ling Hong, Tak Sang Chow and Angel Nga Man Leung in Personality and Social Psychology Bulletin

Supplemental Material

Supplementary_Materials_PSPB-17-437_RNR – Supplemental material for Nature Does Not Always Give You a Helping Hand: Comparing the Prosocial Effects of Nature at Different Resource and Security Levels

Supplemental material, Supplementary_Materials_PSPB-17-437_RNR for Nature Does Not Always Give You a Helping Hand: Comparing the Prosocial Effects of Nature at Different Resource and Security Levels by Henry Kin Shing Ng, Yee-Ling Hong, Tak Sang Chow and Angel Nga Man Leung in Personality and Social Psychology Bulletin

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by the Faculty Development Scheme (2017-2018), Research Grants Council, Hong Kong (Project No. UGC/FDS15/H03/16).

Supplemental Material

Supplemental material is available online with this article.

References

Andrews

Gatersleben

(2010). Variations in perceptions of danger, fear, and preferences in a simulated natural environment. Journal of Environmental Psychology, 30, 473-481. doi:10.1016/j.jenvp.2010.04.001

Appleton

(1996). The experience of landscape. New York, NY: John Wiley.

Bach

R. A.

Defever

A. M.

Chopik

W. J.

Konrath

S. H.

(2017). Geographic variation in empathy: A state-level analysis. Journal of Research in Personality, 68, 124-130. doi:10.1016/j.jrp.2016.12.007

Berman

M. G.

Jonides

Kaplan

(2008). The cognitive benefits of interacting with nature. Psychological Science, 19, 1207-1212. doi:10.1111/j.1467-9280.2008.02225.x

Beute

de Kort

Y. A. W.

(2014). Natural resistance: Exposure to nature and self-regulation, mood, and physiology after ego-depletion. Journal of Environmental Psychology, 40, 167-178. doi:10.1016/j.jenvp.2014.06.004

Chiricos

Padgett

Gertz

(2000). Fear, TV news, and the reality of crime. Criminology, 38, 755-786. doi:10.1111/j.1745-9125.2000.tb00905.x

Corporation for National & Community Service. (2016). Volunteering and civic life in America 2015. Available from https://data.nationalservice.gov/

Federal Bureau of Investigation. (2016). Crime in the United States 2015. Available from https://ucr.fbi.gov/

Gatersleben

Andrews

(2013). When walking in nature is not restorative: The role of prospect and refuge. Health & Place, 20, 91-101. doi:10.1016/j.healthplace.2013.01.001

10.

Gelfand

M. J.

Raver

J. L.

Nishii

Leslie

L. M.

Lun

Lim

B. C.

. . . Aycan

(2011). Differences between tight and loose cultures: A 33-nation study. Science, 332, 1100-1104. doi:10.1126/science.1197754

11.

Graziano

W. G.

Habashi

M. M.

Sheese

B. E.

Tobin

R. M.

(2007). Agreeableness, empathy, and helping: A person x situation perspective. Journal of Personality and Social Psychology, 93, 583-599. doi:10.1037/0022-3514.93.4.583

12.

Guéguen

Stefan

(2016). “Green altruism”: Short immersion in natural green environments and helping behavior. Environment and Behavior, 48, 324-342. doi:10.1177/0013916514536576

13.

Hartmann

Apaolaza-Ibáñez

(2010). Beyond savanna: An evolutionary and environmental psychology approach to behavioral effects of nature scenery in green advertising. Journal of Environmental Psychology, 30, 119-128.

14.

Haviland-Jones

Rosario

H. H.

Wilson

McGuire

T. R.

(2005). An environmental approach to positive emotion: Flowers. Evolutionary Psychology, 3, 104-132. doi:10.1177/147470490500300109

15.

Kasser

Ryan

R. M.

(1996). Further examining the American dream: Differential correlates of intrinsic and extrinsic goals. Personality and Social Psychology Bulletin, 22, 280-287. doi:10.1177/0146167296223006

16.

Korndörfer

Egloff

Schmukle

S. C.

(2015). A large scale test of the effect of social class on prosocial behavior. PLoS ONE, 10, e0133193. doi:10.1371/journal.pone.0133193

17.

Kuo

F. E.

(2003). The role of arboriculture in a healthy social ecology. Journal of Arboriculture, 29, 148-155.

18.

Kuo

F. E.

Sullivan

W. C.

(2001). Environment and crime in the inner city: Does vegetation reduce crime? Environment and Behavior, 33, 343-367. doi:10.1177/0013916501333002

19.

Kuo

F. E.

Sullivan

W. C.

Coley

R. L.

Brunson

(1998). Fertile ground for community: Inner-city neighborhood common spaces. American Journal of Community Psychology, 26, 823-851. doi:10.1023/A:1022294028903

20.

Levine

R. V.

Martinez

T. S.

Brase

Sorenson

(1994). Helping in 36 US cities. Journal of Personality and Social Psychology, 67, 69-82. doi:10.1037/0022-3514.67.1.69

21.

Maruthaveeran

van den Bosh

C. K.

(2015). Fear of crime in urban parks—What the residents of Kuala Lumpur have to say? Urban Forestry & Urban Greening, 14, 702-713. doi:10.1016/j.ufug.2015.05.012

22.

McPherson

Simpson

J. R.

Peper

P. J.

Maco

S. E.

Xiao

(2005). Municipal forest benefits and costs in five US cities. Journal of Forestry, 103, 411-416.

23.

Michael

S. E.

Hull

R. B.

Zahm

D. L.

(2001). Environmental factors influencing auto burglary: A case study. Environment and Behavior, 33, 368-388. doi:10.1177/00139160121973034

24.

Murphy

R. O.

Ackermann

K. A.

Handgraaf

(2011). Measuring social value orientation. Judgment and Decision Making, 6, 771-781. doi:10.2139/ssrn.1804189

25.

Nasar

J. L.

Fisher

(1993). “Hot spots” of fear and crime: A multi-method investigation. Journal of Environmental Psychology, 13, 187-206. doi:10.1016/S0272-4944(05)80173-2

26.

H. K. S.

(2015, March). Relating human behaviors to environmental factors: The role of resource and security. Poster session presented at International Convention of Psychological Science, Amsterdam, The Netherlands.

27.

H. K. S.

Chow

T. S.

(2017). The effects of environmental resource and security on aggressive behavior. Aggressive Behavior, 43, 304-314. doi:10.1002/ab.21690

28.

Orians

G. H.

(2007). Human behavioral ecology: 140 years without Darwin is too long. In Penn

D. J.

Mysterud

(Eds.), Evolutionary perspectives on environmental problems (pp. 259-279). New Brunswick, NJ: Transaction.

29.

Piff

P. K.

Dietze

Feinberg

Stancato

D. M.

Keltner

(2015). Awe, the small self, and prosocial behavior. Journal of Personality and Social Psychology, 108, 883-899. doi:10.1037/pspi0000018

30.

Rao

L. L.

Han

Ren

X. P.

Bai

X. W.

Zheng

Liu

. . . Li

(2011). Disadvantage and prosocial behavior: The effects of the Wenchuan earthquake. Evolution and Human Behavior, 32, 63-69. doi:10.1016/j.evolhumbehav.2010.07.002

31.

Saleem

Anderson

C. A.

Barlett

C. P.

(2015). Assessing helping and hurting behaviors through the Tangram help/hurt task. Personality and Social Psychology Bulletin, 41, 1345-1362. doi:10.1177/0146167215594348

32.

Sullivan

W. C.

Kuo

F. E.

DePooter

S. F.

(2004). The fruit of urban nature: Vital neighborhood spaces. Environment and Behavior, 36, 678-700. doi:10.1177/0193841X04264945

33.

Troy

Grove

J. M.

O’Neil-Dunne

(2012). The relationship between tree canopy and crime rates across an urban–rural gradient in the greater Baltimore region. Landscape and Urban Planning, 106, 262-270. doi:10.1016/j.landurbplan.2012.03.010

34.

The Trust for Public Land. (2016). 2015 city park facts (Annual report). San Francisco, CA: Author. Available from https://www.tpl.org/

35.

U.S. Census Bureau. (2011). Cost of living index, annual average 2010. Retrieved from https://www.census.gov/library/publications/2011/compendia/statab/131ed/prices.html

36.

U.S. Census Bureau. (2015). Annual estimates of the resident population for incorporated places of 50,000 or more. Available from https://factfinder.census.gov/

37.

Van de Vliert

Yang

Wang

Ren

X. P.

(2012). Climato-economic imprints on Chinese collectivism. Journal of Cross-Cultural Psychology, 44, 589-605. doi:10.1177/0022022112463605

38.

Van Vugt

Van Lange

P. A

. (2006). The altruism puzzle: Psychological adaptations for prosocial behavior. In Schaller

Simplson

J. A.

Kenrick

D. T.

(Eds.), Evolution and social psychology (pp. 237-261). New York, NY: Psychology Press.

39.

Weinstein

Przybylski

A. K.

Ryan

R. M.

(2009). Can nature make us more caring? Effects of immersion in nature on intrinsic aspirations and generosity. Personality and Social Psychology Bulletin, 35, 1315-1329. doi:10.1177/0146167209341649

40.

White

Smith

Humphryes

Pahl

Snelling

Depledge

(2010). Blue space: The importance of water for preference, affect, and restorativeness ratings of natural and built scenes. Journal of Environmental Psychology, 30, 482-493. doi:10.1016/j.jenvp.2010.04.004

41.

Wilson

E. O.

(2007). Biophilia and the conservation ethic. In Penn

D. J.

Mysterud

(Eds.), Evolutionary perspectives on environmental problems (pp. 249-257). New Brunswick, NJ: Transaction.

42.

World Health Organization. (2017). Urban green spaces: A brief for action. Copenhagen, Denmark: WHO Regional Office for Europe.

43.

Yamagishi

Akutsu

Cho

Inoue

Matsumoto

(2015). Two-component model of general trust: Predicting behavioral trust from attitudinal trust. Social Cognition, 33, 436-458. doi:10.1521/soco.2015.33.5.436

44.

Yamagishi

(1994). Trust and commitment in the United States and Japan. Motivation and Emotion, 18, 129-166. doi:10.1007/BF02249397

45.

Zelenski

J. M.

Dopko

R. L.

Capaldi

C. A.

(2015). Cooperation is in our nature: Nature exposure may promote cooperative and environmentally sustainable behavior. Journal of Environmental Psychology, 42, 24-31. doi:10.1016/j.jenvp.2015.01.005

46.

Zhang

J. W.

Piff

P. K.

Iyer

Koleva

Keltner

(2014). An occasion for unselfing: Beautiful nature leads to prosociality. Journal of Environmental Psychology, 37, 61-72. doi:10.1016/j.jenvp.2013.11.008

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.37 MB

0.11 MB