Secondary Logo

Journal Logo

Understanding Physical Activity Environmental Correlates: Increased Specificity for Ecological Models

Giles-Corti, Billie1; Timperio, Anna2; Bull, Fiona3; Pikora, Terri1

Exercise and Sport Sciences Reviews: October 2005 - Volume 33 - Issue 4 - p 175-181

Ecological models are now used to understand the complex array of factors that influence physical activity, resulting in a greater emphasis on environmental correlates. This selective review examines whether the predictive capacity of these models could be improved if behavior-specific measures of the environment were used to predict context-specific behaviors.

To improve the predictive capacity of research examining environmental correlates of physical activity, behavior and context-specific constructs are required.

1School of Population Health, The University of Western Australia, Perth, Australia; 2Centre for Physical Activity & Nutrition Research, Deakin University, Melbourne, Australia; and 3Adjunct Appointment School of Population Health, The University of Western Australia, Perth, Australia and BHF National Centre for Physical Activity and Health, School of Sport & Exercise Sciences, Loughborough University, Loughborough

Address for correspondence: Billie Giles-Corti, Ph.D., Associate Professor and NHMRC/NHF Career Development Award Fellow, School of Population Health, The University of Western Australia, M431, 35 Stirling Highway, CRAWLEY 6009 Australia (E-mail:

Accepted for publication: July 12, 2005.

Back to Top | Article Outline


In the past decade, there has been a rapid increase in the number of studies examining the impact of the physical environment on physical activity, particularly walking. The multidisciplinary nature of this research has paved the way for unparalleled opportunities for physical activity researchers to work with sectors outside of health, including transportation, planning, urban design, and environmental protection. Recent funding initiatives, such as those through the Robert Wood Johnson Foundation in the United States, have fueled multidisciplinary, multisector collaborative research activity, by providing incentives for seemingly disparate groups to work together in this challenging area. Although in its embryonic stage, this field of research holds considerable promise (8,9,11) and, given the substantial research endeavor currently underway, our understanding of environmental correlates of physical activity behavior will increase enormously in the next 5–10 yrs.

Motivation to explore environmental correlates of physical activity has been prompted by the modest performance of interventions focused on individual and social environmental factors, and concomitant research that predicted a disappointingly low amount of variance in behavior (9,13). Environmental interventions (such as the design of more walkable neighborhoods or the provision of bicycle paths) are appealing because they have the potential for sustained impact on populations rather than short-term impacts on individuals.

Hence, it has been proposed that ecological models be used to guide physical activity research with the aim of identifying potentially high-leverage physical environmental interventions in relevant behavior settings (13). Within a multilevel ecological framework, complex interactions between the multitude of individual, cultural and social, physical, and policy environmental factors in settings in which people live, work, and play are studied in an attempt to better predict physical activity behavior. Nevertheless, few studies have reported the relative influence of these factors on behavior (4,5). Over the past 5 yrs, the main focus has been on identifying and measuring attributes of the physical environment, a completely new area of focus for the field of physical activity research.

Recent reviews confirm that environmental correlates appear to be important (8,9,11). In developed countries at least, some consistent findings are pointing the way forward for future research activity.

The overall aim of this review is to reflect on early progress and to consider whether the predictive capacity of studies of environmental correlates of physical activity could be further enhanced if specific behaviors were studied within clearly defined environments, (7) and if behavior-specific environmental correlates were considered (9,13). That is, if context-specific behaviors, such as walking for recreation in the neighborhood, were studied, along with specific aspects of the neighborhood environment hypothesized to be specifically associated with that behavior within this context.

Measurement development issues that support the focus on the enhanced specificity are first considered. Then, reflections are presented on how the predictive capacity of models might be improved if specific behaviors and behavior-specific environmental attributes were studied, as well as micro-scale environment-specific attributes. Finally, we propose a research agenda for testing behavior- and context-specific ecological models.

Back to Top | Article Outline


A central theme throughout this review is the need for greater correspondence between behavioral outcome measures and physical environmental independent variables; in particular, we argue for a context-specific behavioral outcome measure and the inclusion of environmental correlates that are specific to the behavioral outcome of interest. This concept is not new. Ajzen and Fishbein (1) provide a step-by-step guide for careful development of measures designed to predict, explain, and influence human behavior. Whereas their approach is rooted in attitudinal research, there is still much to be gained from reflecting on lessons derived from this well-established tradition.

In terms of measuring behavior, Ajzen and Fishbein (1) propose a number of steps: (1) defining the behavioral criterion (i.e., whether it is made up of a single action or a behavioral category); (2) defining the target (i.e., whether a general category, e.g., physical activity) or a specific category (e.g., walking); (3) defining the time frame (e.g., past 7 d, past year, or usual behavior); and (4) defining the context (i.e., where the behavior takes place). In terms of explanatory variables, they also propose developing items that are behavior and context-specific.

Consistent with this approach (1), the physical activity research field has performed well in measuring leisure-time physical activity. It has long-recognized that physical activity is not a single action, but made of a number of behavioral categories or types, e.g., vigorous-intensity activity, moderate-intensity activity, and walking. Moreover, specific targets of behavior have been defined (e.g., specific behaviors such as walking for transport or walking for recreation) as has the time frame in which the behavior has taken place (e.g., the past week or month, or usual behavior). However, until now, one element has not been fully explored—the behavioral context.

In general, the behavioral “context” has been interpreted in terms of the domains of physical activity, e.g., recreational, domestic, and work-related. However, another interpretation of “context” relevant to studies of environmental correlates of behavior relates to the setting in which the behavior takes place. The setting is likely to be important because people behave differently in different settings. For example, the correlates of active transport en route to work are likely to be different to active transport in the local neighborhood. Similarly, the physical and policy environmental correlates of children’s physical activity at school are likely to be different to physical activity performed at home or in the local community.

Back to Top | Article Outline


Defining the Behavior

As evidenced by the extensive literature on the measurement of physical activity, “defining and measuring behavior is not as simple as it first appears” (1). For studies of environmental correlates, defining a behavior for which environmental correlates can be identified is the first task. After reviewing 19 studies of environmental correlates of physical activity, Humpel and colleagues (8) concluded that future research should focus on environmental correlates of specific behaviors such as walking. However, even broad categories of walking need to be broken down, because walking is made up of more than one behavior (e.g., walking for recreation and walking for transport).

Although recent physical activity research has begun focusing on specific behaviors, most research examining environmental correlates uses context-free behavioral outcome measures. For example, when undertaking research on walking, most investigators collect environmental data at the neighborhood level, but the behavior of interest is usually a general measure (i.e., overall walking or overall transport-related walking). This disregards where the behavior took place. Few studies to date have measured context-specific behaviors. Humpel and colleagues (7) found that a general approach to studying environmental correlates may underestimate the association between environmental and behavioral variables. They examined the relationship between perceptions of different aspects of the neighborhood and total min of walking in the neighborhood, total walking (regardless of its context) and total physical activity and found that, particularly for men, perceptions of the neighborhood better predicted walking undertaken in the neighborhood than total walking or total physical activity. This study provides initial evidence of the potential benefit of studying context-specific behaviors (e.g., walking for transport in the neighborhood).

Back to Top | Article Outline

Behavior-Specific Environmental Correlates

The first published study of environmental correlates of physical activity provided a clue about the importance of studying behavior-specific environments (14). Sallis and colleagues found a modest association between achievement of recommended levels of vigorous exercise and the density of user-pay recreational facilities suitable for vigorous exercise (such as tennis clubs, aerobic and dance studios, and health and fitness clubs) within 1–5 km of the respondents’ homes, but not for “free” facilities. However, these latter facilities were more diffuse (e.g., parks, sports fields, public recreation centers, colleges and universities, and public schools) and may or may not be used for vigorous physical activity.

Since this early work, it has become clear that rather than using a general model to predict a general behavior (e.g., physical activity overall), the capacity to predict a behavior is enhanced when there is greater correspondence between a specific behavioral outcome measure and the specific environmental variables hypothesized to be associated with that behavior. For example, Pikora and others proposed the need for separate conceptual frameworks for environmental correlates of walking for transport and walking for recreation (10). Subsequent reviews have supported this view, concluding that environmental correlates vary for different types of walking (9,11).

Our own research findings highlight this point (Table 1). An ecological model was used to compare the relative influence of individual, social environmental, and physical environmental factors associated with overall physical activity (5) and walking (4). Nonspecific objectively measured environmental correlates (Table 1 footnote) were used to predict “overall physical activity,” a complex set of behaviors comprising vigorous activity, moderate activity (including activity at home, gardening, and household chores), and walking. The model for “walking,” however, included objectively measured environmental attributes, but these were specific to recreational and transport-related walking. Not surprisingly, odds ratio for the influence of the physical environment variables was higher relative to individual and social environmental variables in the behavior-specific walking model than the model for “overall physical activity” (Table 1). Nevertheless, the walking model still combined two very different walking behaviors: walking for transport and for recreation, the correlates of which have since been shown to be different (9).



The transportation field is advanced in terms of conceptualizing behavior-specific environmental correlates. This field has successfully used objective measures of urban form to predict walking for transport, principally: land-use mix, the connectivity of street networks, and residential density (3). More “walkable” neighborhoods are those with greater access to shops and destinations (i.e., mixed land uses), gridlike street networks (i.e., good connectivity), and higher density. Handy compared residents of highly and not highly walkable neighborhoods, generally defined using these measures of urban form, and found significant differences in transport-related or utilitarian trips, but no difference in walking for exercise (6). This approach was recently replicated using objective and self-reported measures of physical activity (3,12). Saelens and colleagues found more than a 60-min·wk−1 difference in objectively measured moderate-intensity physical activity and self-reported errands between residents of high- and low-walkable neighborhoods, but an insignificant difference in objectively measured vigorous activity and self-reported leisure time physical activity (12). Nevertheless, the neighborhoods were selected on the basis of their walkability from a transport, rather than recreation perspective.

These findings highlight the need for future studies to examine specific behaviors, undertaken in specific settings and to explore behavior-specific environmental attributes associated with those context-specific behaviors. However, it may be possible that the study of environmental factors might be further enhanced if greater emphasis were also placed on better specifying the environments being studied. Thus, the sections that follow highlight some areas that could be further explored in future studies.

Back to Top | Article Outline

Ensuring Variability in the Environmental Variables

Recent studies have highlighted the benefit of developing environment-specific sampling frames with a view to maximizing variability in the environments being studied (3,12). A central hypothesis of research examining environmental correlates of behavior is that exposure to a supportive physical environment enhances physical activity behavior. Thus, studies of environmental correlates need to include respondents who live in both supportive and nonsupportive environments. In attitudinal research, items that fail to discriminate between respondents are eliminated from scales (1). Without variation in the explanatory variables, they will fail to discriminate between respondents with different behaviors (1), and it would not be possible to test the hypothesis.

Moreover, environmental correlates could independently contribute to behavior or could be a moderator of behavior given other individual and social-environmental influences such as self-efficacy and social support (1). The ability to study the effect modification of the environment or how modifying the environment changes behavior will only be possible if there is variation in the environmental conditions to which respondents are exposed.

Thus, given the dominance of low-density suburban sprawl that characterizes urban form in Australia and the United States, randomly selecting subjects sampled from heterogeneous environments is likely to be more informative than selecting a random sample from the general population. Future studies might benefit from developing sampling frames designed to maximize the variability in environmental exposure measures.

Back to Top | Article Outline

What Is a Neighborhood?

Another area that needs further conceptual development to better contextualize behaviors of interest is the scale of the environment to be studied. For example, in studies that include neighborhood environmental correlates, a clearer definition of “the neighborhood” is required in terms of measurement of respondent perceptions and objective measures of the environment. Surveys include measures of respondent perceptions of “the neighborhood” (9), walking “near home” (10), and walking undertaken in the “neighborhood” (7). However, with few exceptions (12), there appears to be little attempt to define for the respondent what constitutes a “neighborhood.”

Similarly, studies that collect objective area-level data vary enormously in terms of the scale of the environment being measured or the definitions of the neighborhood boundary. Measures include: the quality of streetscapes within the subject’s own street (4); urban form attributes within 400 m (10) or 1 km (3) of the subject’s home or the metropolitan area or county level (2); spatial access to every facility within a study area with distance weighted by a distance of decay parameter (4,5); road network distance to a specific destination (e.g., trails) (15); or access to facilities for vigorous-intensity activity within buffer distances of the subject’s home (i.e., 1–5 km) (13). As evidenced, there appears to be little agreement about which boundary or scale to use, and how this might impact on the association between predictor and outcome variables. Moreover, the boundary to be used may differ for different target groups (i.e., children, adults, and older adults).

The extent to which the effect of distance varies for different physical activity behaviors and destinations also is largely unexplored. Undoubtedly, the likelihood of using a destination is influenced by distance. However, the influence of distance may vary depending upon the behavior being studied (e.g., walking, cycling, vigorous activity), the destination to be accessed and the attractiveness of the destination to the user. For example, in terms of walking for transport, the neighborhood attributes closer to a respondent’s home will be more important (3,10) than those further way. However, for cycling for transport, the physical environment in close proximity to home may be less important and, thus, the geographical scale of the environment to be studied may be considerably larger. In addition, the type of destination and its attractiveness to the user may also be affected by distance thereby impacting on the scale of the environment that is relevant to be studied. We found that the use of parks, for example, is more sensitive to distance than the use of other facilities such as a gym or recreation center (5).

For studies examining differences between respondent perceptions of the environment and objective measures of the environment, there also needs to be greater correspondence between the boundary within which environmental data are collected and the environment to which the perceptions refer. For example, the research team’s definition of the neighborhood within which to collect area-level data may be substantially different to the definition used by respondents to answer questions. This lack of specificity weakens the correlation between respondent perceptions and objective measures of the environment.

So what does this mean in practice for future research? We have argued that the lack of specificity may contribute to weakening the relationship between outcome and predictor variables. However, it also limits comparison across studies. In the interest of greater specificity of environmental, behavioral, and perceptual measures there needs to be better definition of the scale of the environment to be studied and the choice of scale should be behavior-specific. Given current guidelines, encouraging moderate-intensity activity for at least 30 min on most days of the week, it would seem appropriate that for studies of walking locally, to conceptualize the “neighborhood environment” as somewhere between 10 and 15 min from the subject’s home (12). A destination within 15 min of the subject’s home could be incorporated in a 30-min circuit for recreational walks, and most transport-related destinations would need to be within at least 15 min to be used regularly. For the purpose of collecting environmental data, this could further be conservatively defined as being up to 1.6 km from a subject’s home, based on the distance likely to be covered by the average person walking briskly for 15 min, (i.e., 4 miles or 6.4 km·h−1)—the speed of walking suggested by the U.S. Surgeon General as being required to achieve “moderate” intensity physical activity. For studies of other behaviors (e.g., cycling) and different target groups (i.e., children, adults, and older people) the definition of the environment and its scale would need to vary.

Back to Top | Article Outline

The Quality of Environmental Attributes

To date, most studies have focused on the presence of a destination (i.e., a shop) or neighborhood attribute (e.g., connectivity of street network) rather than its quality, such as whether the streetscapes en route are pleasant or the quality attributes of the shop or park. Handy has suggested it may be important to assess the quality of destinations and how this affects use (6). For example, studies repeatedly find that perceiving one’s neighborhood to be pleasant is associated with walking for exercise, walking for transport, and total walking or walking at recommended levels (9). However, at this stage few published studies have collected objective data on aspects of neighborhood quality. There is some evidence, for example, that in terms of predicting higher levels of walking (180 min of walking per week), having access to large attractive parks is more important than simply having a park nearby (5). Similarly, is the presence of one shop enough to encourage more walking for transport, or is it the mix and quality of existing shops that is important?

The collection of micro-level data on quality requires the conduct of environmental audits. Although the collection of these data is time-consuming, research is required to determine whether some micro-level data on the quality of environmental attributes are warranted to provide a more comprehensive picture of the role of the environment on specific physical activity behaviors.

Back to Top | Article Outline

The Need for Behavior-Specific and Environment-Specific Ecological Models

Throughout this review we have argued the need for studies of environmental correlates to include context-specific measures of behavior and behavior-specific environmental measures. However, given that comprehensive ecological models include individual and social-environmental factors (along with biological, cultural and policy factors), we also hypothesize that the predictive capacity of these models may be even further enhanced if the individual (e.g., attitudes, knowledge, beliefs, self-efficacy) and social-environmental (e.g., social support, social norms) variables were also behavior-specific and contextualized. For example, rather than measuring self-efficacy generally, one could measure respondents’ self-efficacy to walk for transport or recreation in their neighborhood.

This suggests the need for ecological models of specific behaviors. Because the inclusion of behavior-specific measures is likely to increase respondent burden, we propose that correlates of specific behaviors be studied rather than correlates of a number of behavioral outcomes. The development of specific models would seem warranted for some behaviors and also would assist in the design of future interventions. For example, the environmental and motivational factors that encourage the use of active modes of travel to work, choosing sedentary behaviors, playing sports, walking for transport, walking for recreation, and using a gym or formal facility would all appear to be different and worthy of study in their own right.

Given the high prevalence of walking in both developed and developing countries, the health benefits associated with moderate-intensity physical activity, the promising results already demonstrated (9), and the fact that the general models still appear to predict only a small amount of variance in behavior (2,3) it would seem particularly justified to develop comprehensive ecological models of different types of walking. For example, the transportation field has excelled in conceptualizing and measuring environmental correlates of walking for transport, but appears to have given little attention to psychosocial correlates of behavior. Ecological models of transport-related walking could assist in the development of both communication and environmental interventions. With this in mind, Figure 1 provides some examples of behavior and context-specific variables that could be included along with broader constructs from ecological models including demographic, biological, and cultural variables. The relationship between variables could be explored using structural equation modeling with a view to examining the interrelationships between dependent and independent variables.

Figure 1.

Figure 1.

Back to Top | Article Outline


In this new and evolving field of research, our progress is being propelled by “standing on the shoulders of the (many) giants who have gone before us” (as suggested by Isaac Newton 1642–1727). What now seems self-evident has resulted from thoughtful conceptual work by leading researchers in this field and the iterative process of inquiry that has followed and continues through the work of investigators globally.

Nevertheless, in this article we have argued that to take additional steps forward, we need to perhaps step back and stand on the shoulders of even earlier giants in the area of attitudinal research (1) with a view to introducing greater specificity to models that seek to explain the impact of the environment on behavior. This would involve the development of behavior-specific ecological models of context-specific behaviors.

As demonstrated by this selective review, the predictive capacity of models appear to improve when environmental measures more closely match the behavior of interest and the setting in which the behavior takes place. Moreover, as illustrated by recent work by Humpel et al. (7), it is possible that the predictive capacity of models could be further enhanced if greater specificity were introduced by using context-specific measures of behavior. The development of behavior and context-specific explanatory variables is consistent with traditional approaches to item development in attitudinal research (1).

The arguments in favor of behavior-specific models are supported by research to date, including important insights derived from transportation and planning research that focus on transport-related behaviors. Although primarily concerned with traffic management and sustainability, transportation research provides some of the most compelling evidence that the built form has a powerful influence on mode of travel. Nevertheless, when the walkability indices based on this literature are applied, they only appear to predict a small amount of the variance in objectively measured or self-reported overall moderate-intensity physical activity behavior (2,3). It may be possible that the lack of correspondence between outcome and predictor variables and insufficient consideration of the scale of the environment to be studied might all contribute to the lower levels of explanatory power achieved by these models.

In addition, research needs to continue to explore emerging environmental variables that might assist in explaining additional variance in specific physical activity behaviors (3). For example, the inclusion of objective measures of sidewalks, exposure to traffic, and neighborhood esthetics are yet to be fully explored, as is the role of other micro-level data that might affect the real or perceived quality of destinations or environments of interest.

Back to Top | Article Outline

Future Research

Over the past two decades, enormous progress has been made in measuring physical activity. Similarly, it will take time to develop valid measures of environmental correlates of behavior. This review has attempted to reflect on research undertaken to date, to highlight the conclusions drawn by ourselves and others about future directions, and to identify research gaps. In summary, we have suggested that future research might consider: 1) studying specific behaviors; 2) the context within which the behavior is performed; 3) using context-specific behavior measures; 4) including behavior and context-specific individual and physical and social environmental correlates; 5) defining the geographical scale of the environment to be studied according to the behavior of interest and the target group, defining the geographical scale for study subjects, and measuring environmental attributes with the same geographical scale; and 6) testing whether the inclusion of micro-scale environmental attributes that measure aspects of the quality of the environment to which study participants are exposed add value to the explanatory power of models. In addition, prospective studies are required to study causal relationships (2).

Back to Top | Article Outline


This review has a number of limitations. In accordance with journal requirements, it was restricted to citing 15 journal articles, which precluded inclusion of all relevant literature from the broad range of contributors to this new field (1). The articles reviewed were selected specifically with a view to developing the arguments presented. The review also was restricted to studies of adults. Issues related to the study of children, older adults, or other subgroups are likely to be different and have not been considered in any details. Moreover, most of the evidence to date and all of the studies presented here relate to developed countries. It is likely that the issues in developing countries are different. Despite these limitations, we have attempted to build on the work undertaken to date with a view to contributing some thoughts for consideration about how the field may develop even further.

Back to Top | Article Outline


Billie Giles-Corti is supported by a NHMRC/NHF Career Development Award (grant 254688). Anna Timperio is supported by a VicHealth Public Health Research Fellowship (2004 0536). The ideas for this review are based on work undertaken for the RESIDential Environment Project funded by the Western Australian Health Promotion Foundation (Healthway). The other investigators on this project (Matthew Knuiman, Kimberly van Niel, and Max Bulsara) and the study team are gratefully acknowledged. The authors acknowledge the following who, among others, have significantly influenced their thinking on this topic: R. G. Barker, Stuart Biddle, Michael Booth, U. Brofenbrenner, Ross Brownson, Cora Craig, Anne Ellaway, Paul Estabrooks, Lise Gauvin, Christine Hoehner, Abby King, Dawn King, K. A. Kirtland, K. J. Krizek, Rebecca Lee, Sally Macintyre, Gavin McCormack, Kenneth McLeroy, Jo Salmon, John Spence, Dan Stokols, and Frank van Lenthe.

Back to Top | Article Outline


1. Ajzen, I., and M. Fishbein. Understanding Attitudes and Predicting Social Behavior. Englewood Cliffs, NJ: Prentice-Hall Inc., 1980.
2. Ewing, R., T. Schmid, R. Killingsworth, A. Zlot, and S. Raudenbush. Relationship between urban sprawl and physical activity, obesity and morbidity. Am. J. Health Promotion. 18:47–57, 2003.
3. Frank, L., T. Schmid, J.F. Sallis, J. Chapman, and B.E. Saelens. Linking objectively measured physical activity with objectively measured urban form. Findings from SMARTRAQ. Am. J. Prev. Med. 28:117–125, 2005.
4. Giles-Corti, B., and R. Donovan. Increasing walking: the relative influence of individual, social environmental and physical environmental factors. Am. J. Pub. Health. 93:1583–1589, 2003.
5. Giles-Corti, B., and R. Donovan. The relative influence of individual, social and physical environmental determinants of physical activity. Soc. Sci. Med. 54:1793–1812, 2002.
6. Handy, S. Regional versus local accessibility: Neo-traditional development and its implications for non-work travel. Built Environment 18:253–267, 1992.
7. Humpel, N., N. Owen, E. Leslie, A.L. Marshall, A.E. Bauman, and J.F. Sallis. Associations of location and perceived environmental attributes with walking in neighborhoods. Am. J. Health Promotion. 18:239–242, 2004.
8. Humpel, N., N. Owen, and E. Leslie. Environmental factors associated with adults’ participation in physical activity. Am. J. Prev. Med. 22:188–199, 2002.
9. Owen, N., N. Humpel, E. Leslie, A. Bauman, and J.F. Sallis. Understanding environmental influences on walking; Review and research agenda. Am. J. Prev. Med. 27:67–76, 2004.
10. Pikora, T., B. Giles-Corti, F. Bull, K. Jamrozik, and R. Donovan. Developing a framework for assessment of the environmental determinants of walking and cycling. Soc. Sci. Med. 56:1693–1793, 2003.
11. Saelens, B., J. Sallis, and L. Frank. Environmental correlates of walking and cycling: Findings from the transportation, urban design, and planning literatures. Ann. Behav. Med. 25:80–91, 2003.
12. Saelens, B.E., J.F. Sallis, J.B. Black, and D. Chen. Neighborhood-based differences in physical activity: An environment scale evaluation. Am. J. Public Health. 93:1552–1558, 2003.
13. Sallis, J., A. Bauman, and M. Pratt. Environmental and policy interventions to promote physical activity. Am. J. Prev. Med. 15:379–397, 1998.
14. Sallis, J.F., M.F. Hovell, C.R. Hofstetter, J.P. Elder, M. Hackley, C.J. Caspersen, and K.E. Powell. Distance between homes and exercise facilities related to frequency of exercise among San Diego residents. Public Health Reports 105:179–85, 1990.
15. Troped, P., R. Saunders, R. Pate, B. Reininger, J. Ureda, and S. Thompson. Associations between self-reported and objective physical environmental factors and use of a community rail-trail. Prev. Med. 32:191–200, 2001.

built environment; walking; physical activity; ecological models; environment; environmental attributes; review

©2005 The American College of Sports Medicine