The social, emotional, and mental health benefits associated with gardening have been well documented. Kaplan and others pioneered the idea that gardening contributes to mental health among home and community gardeners (Kaplan, 1973). A comprehensive review of the scientific literature on recreational gardening demonstrated that gardening improves mental health status through reduced levels of stress hormones (K. H. Brown & Jameton, 2000). Milligan and others have documented therapeutic benefits of communal gardening on allotments among older adults (Milligan, Gatrell, & Bingley, 2004). Others have documented the cognitive, social, and community benefits of gardening among school-aged children through changes in children’s confidence, self-esteem, attitudes and learning (Bowker & Tearle, 2007).

Moreover, several studies have demonstrated the role of gardens in shaping health behaviors. For example, Park and others showed that gardening was an effective way for adults to meet the national guidelines for moderate to vigorous physical activity (Park, Shoemaker, & Haub, 2009). In other studies, urban gardeners reported fruit and vegetable intake levels that met national guidelines for fruit and vegetable consumption (Alaimo, Packnett, Miles, & Kruger, 2008; Litt et al., 2011). The increase in fruit and vegetable consumption for all gardeners was almost two times larger than the increase seen across most other published interventions (Ammerman, Lindquist, Lohr, & Hersey, 2002; Thomson & Ravia, 2011).

While there is ample evidence to demonstrate the benefits of gardening, understanding the processes that link garden participation and improvements in health status are not well understood. We posit that gardens impact self-rated health through key ecological, emotional, and social intervening processes, specifically through aesthetics, social involvement, collective efficacy, and neighborhood attachment. Our research draws on multiple theoretical frameworks including ecological system (Bronfenbrenner, 1979; Stokols, 1996), social capital (Carpiano, 2006; Kawachi & Berkman, 2000), social cognitive (Bandura, 1986) theories and the relational nature between people and places (Carolan, 2007; Conradson, 2005; H. Frumkin, 2005; Gesler, 1992; Hartig, 2014; Kaplan & Kaplan, 2005). Together, this multi-theoretical approach recognizes the range of influences on health at intrapersonal, interpersonal, cultural, organizational, and environmental levels.

Aesthetic experiences help us understand the relational unfolding between people and places, such as the relationships between nearby nature (e.g., trees, green space) and neighborhood perceptions (e.g., perceived safety) (F. E. Kuo & Sullivan, 2001; Frances E. Kuo, Bacaicoa, & Sullivan, 1998) and health outcomes (e.g., obesity) (Pereira et al., 2013). These experiences occur at the individual and collective levels, are multisensory and play a key role in the development of environmental knowledge and in everyday engagement in one’s surroundings (Capaldi, 2014; Foster, 2009; Hale et al., 2011; Neves, 2009).

Gardens are spaces that foster a sense of beauty among residents and encourage direct engagement in natural and social processes over time. By providing an alternative to city noise, pollution and heat, for example, urban gardens provide hands-on learning about the biophysical system that supports plants, animals and microorganisms and a range of social and emotional processes associated with growing food in the garden (Hale et al., 2011; Neves, 2009). The tactile experience of gardening helps to embed individuals within these alternative social and natural worlds and, in turn, also helps embody a deeper awareness of others and the local environment (Carolan, 2007, 2011; Conradson, 2005; Milligan et al., 2004).

The garden experience also promotes social and civic learning as gardeners watch each other, ask questions, work together, make decisions, experiment and share results. This is rooted in a shared commitment toward the garden, and this commitment is not static. For example, it is shaped by formal workdays where aesthetics, ethics, and routines are shared, as well as regular garden-related activities and experiences, such as sharing seeds or plant knowledge, all of which are sustained through evolving interpersonal relationships and trust (Glover, Parry, & Shinew, 2005; Hale et al., 2011; Teig et al., 2009).

We posit that the aesthetic experiences in community gardens generate meaning that encourages further engagement in garden-specific and neighborhood level activities, ignites strong emotional bonds to place (e.g., neighborhood attachment) (Arnberger & Eder, 2012; Hale et al., 2011) and develops positive social experiences (de Vries, van Dillen, Groenewegen, & Spreeuwenberg, 2013; Hale et al., 2011; Maas, van Dillen, Verheji, & Groenewegen, 2009; S. Wakefield, Yeudall, Taron, Reynolds, & Skinner, 2007). Such processes relate to attitudinal and behavioral changes (Alaimo et al., 2008; Carolan, 2007; Hale et al., 2011; Lindstrom, Hanson, Wirfalt, & Ostergren, 2001; Litt et al., 2011), overall quality of life (Foster, 2009; Hale et al., 2011; Wood, 2008), and improvements in health conditions (Cohen, Finch, Bower, & Sastry, 2006; Pereira et al., 2013).

Social involvement fosters access to social resources and social learning and helps define and reinforce meaningful social roles (Berkman & Glass, 2000; Hale et al., 2011; Litt et al., 2011). Behavioral settings associated with leisure activities that require active and sustained participation can lead to deeper engagement in civic life (Glover, Shinew, & Parry, 2005; Hemingway, 1999). Studies of community gardening show that the process of direct participation and social engagement promotes individual ownership of and commitment to the garden structure, enhances community engagement, and empowers residents to get involved on a variety of civic issues (Alaimo, Reischl, & Ober-Allen, 2010; Delind, 2002; Litt et al., 2011; McIvor & Hale, 2015; Teig et al., 2009; Travaline & Hunold, 2010). Glover and others found that strong social relationships formed within community gardens facilitated social organization and increased community capacity for accessing resources and social learning (Glover, Parry, et al., 2005). Such involvement seeds the formation of trust and reciprocity, which are necessary ingredients for collective action and exercising informal social control (Anselin, 1995; Subramanian, Kim, & Kawachi, 2002; Teig et al., 2009).

Defined by Sampson and others, collective efficacy is “the link between mutual trust and a shared expectations to intervene for the common good of the neighborhood (R. J. Sampson & Raudenbush, 1999; R. Sampson et al., 1997).” Social cohesion and informal social control are the two major tenets of collective efficacy (R. Sampson et al., 1997). This concept reflects the neighborhood conditions that help actualize social networks to achieve desired outcomes (Rose & Clear, 2001). That is, neighborhood collective efficacy reflects the process by which social resources, such as trust, cohesion, shared norms and values, and informal control, are translated into specific social and health outcomes such as public order, obesity, and self-rated health (Browning & Cagney, 2002; Cohen et al., 2006; Hartig, 2014; R. J. Sampson, Morenoff, & Earls, 1999; R. Sampson et al., 1997). Within the health context, collective efficacy may contribute to positive health outcomes by regulating health-compromising behaviors, increasing access to health-related facilities, managing neighborhood physical hazards and risky conditions, and strengthening social psychological processes such as trust and neighborhood attachment (Browning & Cagney, 2002; Kawachi & Berkman, 2000).

While there is evidence to suggest the positive role collective efficacy plays in shaping health outcomes (Cohen et al., 2006), there is less clarity about the factors that generate collective efficacy. Research has shown that resident involvement in neighborhood organizations and other cooperative experiences that strengthen social networks support the development of generalized trust (Fong & Chang, 2011; Howard Frumkin, 2001; Glover, Parry, et al., 2005; Glover, Shinew, et al., 2005; Hemingway, 1999). Sampson’s early conceptualization of collective efficacy has relevance to understanding the processes by which gardens generate collective efficacy (R. Sampson et al., 1997). Community gardens, as one example of a neighborhood-level social organization, tap into psychological processes that foster norms of trust, solidarity and reciprocity by creating feelings of being part of a community (Glover, Parry, et al., 2005; Sempik & Aldridge, 2005), reciprocity between individual and community (Hale et al., 2011; Teig et al., 2009), sharing history (Hale et al., 2011) and developing bonds through the relationships that develop in the community garden (Comstock et al., 2010). Gardens also help build social networks by facilitating involvement in community life (Alaimo et al., 2010; Glover, Parry, et al., 2005; Teig et al., 2009), deriving community benefits from participation in the community garden, and furthering feelings of connectedness and belonging (Knudsen & Clark, 2013; Wiles, Leibing, Guberman, Reeve, & Allen, 2012). We posit that garden-generated ecological and social processes lead to the generation of collective efficacy.

Understanding environment-health relationships requires consideration of place and one’s emotional connection to social and physical surroundings, also referred to as neighborhood attachment (B. Brown, Perkins, & Brown, 2003, 2004). City planners, geographers, and community development planners have long been interested in attachment because of its role in promoting pro-social and environmental behaviors (Manzo & Perkins, 2006; S. L. Wakefield, Elliot, Cole, & Eyles, 2001) and its relationship with lower perceived incivilities, lower fear of crime, and greater social cohesion and control (e.g., collective efficacy) (B. Brown et al., 2003). Emotional bonds to place can be influenced by individual interactions with physical features and structures within a neighborhood (e.g., community garden) and the informal and formal social ties within a particular place (e.g., gardener to gardener) (S. Wakefield & McMullan, 2005; S. Wakefield et al., 2007). For example, studies have demonstrated that individual perceptions of quality of public green space (Arnberger & Eder, 2012) and garden participation (Comstock et al., 2010) were associated with positive attachments to place. Moreover, others have shown that individual reports of friendly relationships with neighbors were associated with higher neighborhood attachment (Mesch & Manor, 1998). Brown and others showed that individuals who had higher ratings of collective efficacy also felt stronger attachments to place (B. Brown et al., 2003).

While there are different facets of attachment, we focus on neighborhood attachment because of our interest in how neighborhood environments shape health status. Moreover, we are interested in the role of the natural environment in shaping attachment processes (Lewicka, 2011). We postulate that garden participation may strengthen people’s emotional connections to their neighborhoods through positive environmental aesthetic experiences, social involvement (Arnberger & Eder, 2012; Comstock et al., 2010; Conradson, 2005; Lewicka, 2011; Scannell & Gifford, 2010) and collective efficacy (B. Brown et al., 2003).

This research helps to elucidate the more proximal outcomes related to gardening and the direct and indirect pathways by which gardens influence self-rated health. We examined how gardening influences fruit and vegetable intake and self-rated health directly as well as how it influences health through neighborhood aesthetics, social involvement, collective efficacy, and neighborhood attachment. Specifically, we proposed relationships among constructs where we expected neighborhood aesthetics and social involvement to influence self-rated health through collective efficacy (de Vries et al., 2013; Teig et al., 2009) and neighborhood attachment (Arnberger & Eder, 2012); (Comstock et al., 2010). As described above, aesthetics and social involvement represent early psychosocial processes that support feelings of place attachment and social cohesion and informal social control (Arnberger & Eder, 2012; de Vries et al., 2013; Hale et al., 2011; Teig et al., 2009). Finally, in addition to examining the direct relationship between collective efficacy and self-rated health, this model examined the role of neighborhood attachment in mediating the relationship between collective efficacy and self-rated health.

We collected data for this cross-sectional study between October 2006 and November 2007 in Denver, Colorado. We used a multi-frame sampling design, which consisted of an area-based sample of the general population and a list-based census of community gardeners with a recruitment goal of 480 total households. The available sampling frame included 1,454 households. Within the area-based sample, we randomly selected 1,154 households from 40 block groups to participate in the survey. Within the list-based sample, we enumerated all gardens within the study area and randomly selected 13 community gardens to participate in the survey. From there, we selected 300 gardeners associated with these gardens to participate in the survey. One eligible respondent was randomly selected from each household to complete the survey using the “most-recent-birthday” method. Among the total 1,454 housing units within the sample, 1,437 households (99%) were visited by study employees; 799 of these housing units (56%) were physically accessible. Among homes visited, n = 469 households, nested within n = 92 block groups, participated in the study (59% response rate). The Colorado Multiple Institutional Review Board approved this study.

We employed two primary methods of data collection. First, we conducted a 45-minute face-to-face survey at or near the home of sampled English- or Spanish-speaking adults aged 18 years or older. Written consent was obtained from the respondent prior to survey administration. Second, we conducted neighborhood audits to objectively measure physical and social environments for each respondent. Trained data collectors assessed the street environment by auditing both sides of the “face block” between two intersecting streets on which the study participant lived.

The survey instrument was developed using existing questions on aesthetics (Saelens et al., 2003), social cohesion, informal social control and incivilities (Sampson et al., 1997), place attachment (Bonaiuto et al., 1999), and dimensions of health (Centers for Disease Control and Prevention, 2005). The neighborhood audit included measures of respondents’ residential environments. The measures used in our audit were based on measures previously evaluated for validity and reliability (Caughy et al., 2001). For this analysis, we included a measure of objective incivilities as a proxy for neighborhood upkeep. Details about the survey methodology, survey measures, and neighborhood audit instruments are described elsewhere (Comstock et al., 2010; Litt et al., 2011). Study measures are described below.

Analyses were conducted using SAS Version 9.2 for descriptive and correlational analyses and Mplus Version 7.2 (Muthén & Muthén, 1998) for path model estimation. Prior to model estimation, the bivariate correlations among all model constructs were examined. We needed to account for the non-independence of observations because individuals were nested within block groups. Models were therefore estimated using a complex sample function that accounts for non-independence by using a sandwich estimator to adjust standard errors and chi-square (χ²) values. Model relationships on self-reported health were estimated controlling for several individual-level (age, education, and years in neighborhood) and neighborhood-level (percent college educated in neighborhood and observed incivilities in neighborhood) covariates. Path analysis is an extension of multiple regression whereby several regression relationships can be estimated simultaneously (e.g., a variable that is an outcome of one variable may be simultaneously examined as a predictor of another). Models are evaluated in terms of how well the model as a whole fits the data, as well as in terms of the significance of each of the specific proposed relationships. This method provides a way to examine both direct and indirect predictors of self-reported health. Figure 1 depicts the estimated model, where gardening status was estimated as a direct predictor of self-reported health and fruit and vegetable intake and an indirect predictor of self-reported health through aesthetics, social involvement, collective efficacy, and neighborhood attachment. The model was evaluated in terms of the magnitude, direction, and significance of the estimated path coefficients as well as by several standard measures of overall model fit (model χ², comparative fit index (CFI), root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR). There was a small amount of missing data, where the sample size of model constructs ranged from n = 445 (94.9% complete) to n = 469 (100% complete). Missing data were addressed by using full information maximum likelihood, a method that has been shown to be unbiased given the small amount of missing data experienced here (Enders & Bandalos, 2001).

Participant characteristics are shown in Table 1, distinguished by characteristics for n = 469 individuals versus n = 92 neighborhoods. The majority of individuals in the study were female and college-educated. There were also more gardeners (59.3%) than non-gardeners. The average age was 46.1 years and average time in the home was 2.5 years. In terms of neighborhood characteristics, the average percentage of those college educated was just under half, the average percentage of minority residents was around 25%, and the average percentage of those living in the neighborhood 5 or more years was 40.8%.

Table 2 depicts the bivariate correlations among all model variables that formed the basis of the estimated path model. All variables were significantly correlated with one another, with effect sizes ranging from small (r = .16) to moderate (r = .55). The strongest correlates of self-rated health were collective efficacy and aesthetics and the strongest correlates of garden participation were social involvement and aesthetics.

The final model results are shown Figure 1. All path coefficients are standardized for ease of interpretation of the magnitude of the effect. Garden participation significantly predicted greater social involvement (β = .36, p < .001), more positive ratings of aesthetics (β = .35, p < .001), and increased times per day of fruit and vegetable consumption (β = .21, p < .001), but did not directly predict self-rated health over and above other model variables (β = .04, ns.). Social involvement and aesthetics both positively predicted collective efficacy (β= .11, p < .05 and β = .46, p < .001, respectively) and higher levels of collective efficacy, in turn, significantly predicted more positive reports of self-rated health (β = .14, p < .05). Aesthetics (β = .27, p < .001), social involvement (β = .12, p < .001), and collective efficacy (β = .29, p < .001) also predicted neighborhood attachment, but neighborhood attachment did not significantly predict self-rated health (β = .04, ns.). The impact of garden participation on self-rated health thus occurred indirectly by way of social involvement, aesthetics, and collective efficacy. This model fit the data adequately (χ²(8) = 24.48, p < .001; CFI = .985; RMSEA = .066 (95% CIs from .037 to .097); SRMR = .03), and accounted for 22% of the variance in self-rated health and 4% of the variance in fruit and vegetable intake. However, there was variability in the size of the effects, where some significant path coefficients represented smaller effects (e.g., the path from social involvement to collective efficacy) while others were larger (e.g., the path from aesthetics to collective efficacy).