URN

B

M

R

C

I

PAoupiaaaawwca

FHlWSND

tQn

A©

nderstanding Environmental Influences on Walkingeview and Research Agenda

eville Owen, PhD, Nancy Humpel, PhD, Eva Leslie, PhD, Adrian Bauman, PhD, James F. Sallis, PhD

ackground: Understanding how environmental attributes can influence particular physical activitybehaviors is a public health research priority. Walking is the most common physical activitybehavior of adults; environmental innovations may be able to influence rates of participation.

ethod: Review of studies on relationships of objectively assessed and perceived environmentalattributes with walking. Associations with environmental attributes were examined sepa-rately for exercise and recreational walking, walking to get to and from places, and totalwalking.

esults: Eighteen studies were identified. Aesthetic attributes, convenience of facilities for walking(sidewalks, trails); accessibility of destinations (stores, park, beach); and perceptions abouttraffic and busy roads were found to be associated with walking for particular purposes.Attributes associated with walking for exercise were different from those associated withwalking to get to and from places.

onclusions: While few studies have examined specific environment–walking relationships, early evi-dence is promising. Key elements of the research agenda are developing reliable and validmeasures of environmental attributes and walking behaviors, determining whether envi-ronment–behavior relationships are causal, and developing theoretical models thataccount for environmental influences and their interactions with other determinants.(Am J Prev Med 2004;27(1):67–76) © 2004 American Journal of Preventive Medicine

acttpdpiicBiicdtbhegctap

a

ntroductionromoting higher levels of participation by adultsin regular, moderate-intensity physical activity isa public health priority.1,2 Recent evidence from

ustralia suggests that, although public campaigns andther initiatives to increase participation have beennderway for more than 10 years, population levels ofhysical activity have been static and may have declined

n some groups.3 There is a strong case that substantialnd long-lasting environmental and policy initiativesre an important opportunity for making physicallyctive choices easier and more realistic choices.4–6 Ifdvocacy for this public health agenda is to be pursuedith confidence, research is needed to determinehether environmental changes (such as providingycle paths and walkways, or public outdoor recre-tional settings) do increase the likelihood of more

rom the Cancer Prevention Research Centre, School of Populationealth, University of Queensland (Owen, Leslie), Brisbane, Queens-

and, Australia; Health & Productivity Research Centre, University ofollongong (Humpel), Wollongong, New South Wales, Australia;

chool of Public Health, University of Sydney (Bauman), Sydney,ew South Wales, Australia; and Department of Psychology, Saniego State University (Sallis), San Diego, CaliforniaAddress correspondence to: Neville Owen, PhD, Cancer Preven-

ion Research Centre, School of Population Health, The University ofueensland, Herston Road, Herston QLD 4006, Australia. E-mail:

t.owen@sph.uq.edu.au.

m J Prev Med 2004;27(1)2004 American Journal of Preventive Medicine • Published by E

ctive behavioral choices. However, there are signifi-ant conceptual and methodologic challenges in iden-ifying how such physical-environment factors might acto influence such choices.7 Conceptually, there is alausible case that environmental influences can play airect role in shaping habitual behavior patterns. Ex-erimental evidence from several behavioral domains

dentifies circumstances in which direct environmentalnfluence can be a stronger determinant of behavioralhoice than are cognitively mediated influences.8,9

ecause cognitive social theories have been a predom-nant influence on behavioral studies of physical activ-ty,10–13 the field has been shaped by assumptions thathoices to be active or inactive are conscious andeliberate—that is, consequent upon attitudes, inten-ions, self-efficacy, and other cognitive mediators ofehavioral change.11,12 Social cognitive models do,owever, identify a strong role for environmental influ-nces under some circumstances. Bandura14 has ar-ued that when behavior is strongly facilitated oronstrained by attributes of the environment in which itakes place (and plausibly this is often likely for physicalctivity), direct environmental influences would be theredominant class of determinants.Studies of environment–activity relationships, if they

re to be of practical use in public health policy, ought

o focus on the environmental influences that may

670749-3797/04/$–see front matterlsevier Inc. doi:10.1016/j.amepre.2004.03.006

dcmmlbirsrcw

mfevlffrtlphmtamttscpisti

tluaSateafp

M

QlpP

cSmit

R

ESspatmaewFtas(nmsa

Sw

BttDt�Haatwcwniaawmfbi“rm

6

etermine particular behavioral choices.4,8,9,15 In theontext of the public health goal to increase regular,oderate-intensity physical activity, the behavior ofost relevance is walking. The public health policy

iterature has identified walking as the physical activityehavior of adults that should be most amenable to

nfluence.12,16 Walking is also the most commonlyeported physical activity behavior.16,17 Thus, there is atrong conceptual and practical case for public healthesearch on the environmental determinants of physi-al activity to focus on the particular behavior ofalking.Humpel et al.18 reviewed the evidence for environ-ental influences on physical activity generally. They

ound that both perceived and objectively determinednvironmental attributes (particularly aesthetics, con-enience, and access) were associated with an increasedikelihood of physical activity. Adopting a more specificocus, Saelens et al.19 synthesized the findings of studiesrom transportation and urban design and planningesearch on factors related to walking and cycling forransportation purposes. Much of the transportationiterature focuses on vehicular travel. However, human-owered modes of travel such as walking and cyclingave also been examined in many studies. Given thatost nonwork trips are within walking or cycling dis-

ance, findings from this closely related area of researchre helpful in identifying objectively measured environ-ental attributes (particularly mixed land use, residen-

ial density, and intersection density19) that are relevanto the choice to walk. A conceptual model of thepecificity of environmental correlates of walking andycling resulted from this review.19 Environmental andolicy initiatives to increase physical activity4 must be

nformed by such conceptual models and also by atrong body of evidence on the environmental at-ributes that are related to such particular active behav-oral choices.

Here, studies from the public health research litera-ure specifically addressing the environmental corre-ates of walking are reviewed. The term “correlates” wassed advisedly,7,11 given that much of the evidencevailable is from studies using cross-sectional designs.pecifically, the focus was on relationships of perceivednd objectively assessed environmental attributes withhe walking behaviors of adults. The evidence wasvaluated on specific environmental attributes associ-ted with subcategories of walking behavior—walkingor exercise or recreation, walking to get to and fromlaces, and total walking.

ethods

uantitative studies examining environmental attributes re-ated to the walking behavior of adults were identified from arevious literature review,18 from database searches including

sycInfo, Cinahl, Medline, and by using preprints from p

8 American Journal of Preventive Medicine, Volume 27, Num

olleagues of papers that were in press at the time of writing.tudies were included if they used any type of walking as theain outcome variable and if the independent variables

ncluded environmental attributes, whether measured objec-ively or by self-report.

esults

ighteen studies were identified as meeting the criteria.ixteen studies used cross-sectional designs, and twotudies were prospective. Thirteen used measures oferceived environmental attributes, while 12 includedt least one objective measure of environmental at-ributes. Ten studies examined associations of environ-

ental attributes with walking for exercise or recre-tion (including “neighborhood” walking). Ten studiesxamined associations with total walking (includingalking sufficiently to meet public health guidelines).our studies examined associations with walking to geto and from places (including walking for errands, tond from work, during breaks, to and from transittops). One study examined walking simply for pleasure“social” walking, such as going for a stroll after din-er). Table 1 summarizes the environmental attributeseasured, demographic variables for which the analy-

es were statistically adjusted, type of walking outcome,nd major findings and their direction.

tudies Examining Environmental Relationshipsith Walking for Exercise or Recreation

rownson et al.20 evaluated the use of a new walkingrail (Table 1). Among people who reported using therail, 55.2% had increased their amount of walking.istance to the trail was not associated with walking, but

his may be due to 43% of respondents having to travel15 miles to the trail. An early cross-sectional study byovell et al.21 found that neighborhood environmental

ttributes (safety and ease of exercising) were associ-ted with walking for exercise. A subsequent prospec-ive study by Hovell et al.22 examined changes inalking over 2 years and found that the number ofonvenient facilities reported at baseline was associatedith an increase in walking at follow-up, whereas theeighborhood environment was not related to change

n walking. A study using an Australian sample ofdults23 found two categories of local environmentalttributes to be associated with increased likelihood ofalking by residents: an aesthetically pleasing environ-ent (e.g., ratings of pleasant and attractive natural

eatures) and a convenient environment (e.g., storeseing nearby, park or beach nearby). One study exam-

ned the relationship of environmental attributes withwalking for pleasure.”24 Men who perceived the envi-onment as safest for walking were less likely to walkore for pleasure, and women with moderately positive

erceptions of the accessibility of places for walking

ber 1

welaa

owawea

iLewpcupbamatqcnwaltiipt

SG

Cehvwdubwtcmsnft

nTnoww

SW

Ba1hnrbtaA“afawsca

itwftsthf(ormcat

S

Ttwb(dw

ere more likely to walk for pleasure. Walking forxercise and recreation and walking for pleasure areikely highly correlated, as enjoyment is often a keyttribute and purpose for choosing particular recre-tional activities.

Giles-Corti and Donovan25 examined associations ofbjective and perceived environmental attributes withalking. They found perceptions of an attractive, safe,nd interesting neighborhood to be associated withalking for recreation. Sallis et al.26 found having homequipment and convenient facilities not to be associ-ted with walking for exercise.

Humpel et al.27 used participants’ postal code todentify coastal versus noncoastal place of residence.iving in a coastal location (an objectively determinednvironmental attribute) was found to be associatedith a greater likelihood of neighborhood walking (aarticular walking behavior index) for men only. Fourategories measuring environmental perceptions weresed. Neighborhood aesthetics and traffic not being aroblem were associated cross-sectionally with neigh-orhood walking for men. Convenience of facilities andccess to services were associated with walking for bothen and women. The participants in this study were

lso followed up prospectively as part of an interventionrial, and further findings were reported in a subse-uent paper.28 Prospectively, coastal location was asso-iated with men being less likely to increase theireighborhood walking. This may be due to men whoere living in a coastal location being already morective, and thus the potential for more walking wasimited. An improvement in perceived convenience ofhe neighborhood environment was associated with anncrease of �60 minutes of walking (Table 1). Anncrease in the perception that traffic was a minorroblem resulted in men being less likely to increaseheir walking.

tudies Examining Relationships with Walking toet to and from Places

raig et al.29 found that a high, positive, neighborhoodnvironment score (observer rating of 18 neighbor-ood characteristics, such as number of destinations,isual interest) was significantly related to walking toork (Table 1). This association was moderated byegree of urbanization, with higher scores found inrban neighborhoods compared to suburban neigh-orhoods. Saelens et al.30 found that living in a highlyalkable neighborhood (as defined by higher residen-

ial density, more mixed land use, and greater streetonnectivity) was associated with participants spendingore time walking for errands and on breaks at work or

chool, compared to those living in a low walkableeighborhood. There was no association with walking

or exercise or with total walking. Giles-Corti25 found

hat objectively verified access to a beach had a strong h

egative relationship with walking for transportation.his may reflect differences in transportation or desti-ation options in areas adjacent to the coast. Presencef sidewalks, perceptions of traffic safety, and storesithin walking distance were positively associated withalking for transport.

tudies Examining Relationships with Totalalking

errigan and Troiano31 used age of respondent’s homes a proxy measure of an urban form attribute (Table). They proposed that neighborhoods with olderomes are more likely to have denser interconnectedetworks of streets, and to have a mix of business andesidential use. Homes built before 1973 were found toe associated with the occupants walking more than 20imes a month for any reason. Other forms of physicalctivity were not found to be associated with home age.study by Carnegie et al.32 also found aesthetics and a

practical” (similar to “convenient”) environment to bessociated with walking. Giles-Corti and Donovan25

ound access to open spaces and perceived aestheticttributes were associated with an overall index ofalking at levels recommended for health benefits. In a

ubsequent study,33 they found that a higher score on aomposite objective physical environment measure wasssociated with walking at recommended levels.

De Bourdeaudhuij et al.34 developed an extensivenstrument to assess environmental attributes. Al-hough some environmental variables were related toalking, this only explained 4% of variance in walking

or men and 3% for women. Other variables not foundo be correlated with walking included residential den-ity, access to local shopping, safety from crime andraffic, connectivity of streets, worksite environment,ome equipment, and convenience of physical activity

acilities. Ewing et al.35 developed “sprawl” indiceslow-density residential development; rigid separationf homes, shops, and workplaces) to examine theirelationship with walking and physical activity. Bothetropolitan and county “sprawl” were negatively asso-

iated with minutes walked. Two other recent studieslso found associations of environmental attributes withotal walking.36,37

ynthesis of Findings

able 2 shows groupings of the environmental at-ributes with the four categories of walking. These werealking for exercise or recreation (including “neigh-orhood” walking); walking to get to and from placesincluding walking for errands, to and from work,uring breaks, to and from transit stops); and totalalking (including walking sufficiently to meet public

ealth guidelines). Both significant and nonsignificant

Am J Prev Med 2004;27(1) 69

Table 1. Characteristics and main findings of studies examining relationships of environmental attributes with a main outcome of walking

Author (year)refNumber, age,gender Design Walking outcome Environmental attributes

Associations with walkingoutcomes

Statisticaladjustment

Ball (2001)23 3392Adults

CS, p For exercise Aesthetically pleasingenvironment

Significant (�) A, G, E

Men 46% Convenient environments Significant (�)Berrigan (2002)31 14,827

�20 yearsCS, o �20 times per month Age of home Significant for older homes

(�)A, G, E, Eth, I

Men 48%Brownson (2000)20 1269

AdultsCS, o Increased walking since

using trailWalking trail length Significant for users of longer

trails (�)None reported

Men 35% Trail surface Significant for users of asphalttrails (�)

Distance to trail NonsignificantCarnegie (2002)32 1200

40–60 yearsCS, p Total Aesthetic environment Significant (�) A, G, E,

Men 43% Practical environment Significant (�)Craig (2002)29 Canadian Census,

1996CS, o To and from work Composite environment score Significant (�) U, E, I, P

De Bourdeaud huij (2003)34 521Mean�44 years

CS, p, o Total minutes in lastweek

40 Neighborhood items in 12categories

Availability of sidewalkssignificant for men (�)

A, E, Em

Men 51.7% 41 Recreational items in 3categories

Land use mix; ease of walk topublic transport significantfor women (�)

Others not detailed

Ewing (2003)35 Total 206,992 CS, p, o Minutes leisure timewalking in past month

Metropolitan sprawl index More sprawl significant (�) None reported

18–75 years County sprawl index More sprawl significant (�)Eyler (2003)37 1816

18–65�CS, p Total for usual week

categorized as regular,occasional, and neverwalkers

No sidewalks Significant with never walkers:no sidewalks (�); noenjoyable scenery (�); nowalk/jog trails (�)

A, E, Eth

Heavy trafficHillsNo streetlightsUnattended dogsFoul airNo enjoyable sceneryNo walk/jog trailsHigh crime

Giles-Corti (2002)25 1803 CS, p, o For transport 1. Access to open space Significant for 1 (�), 2 (�), 4(�), 5 (�), 6 (�)

A, G, E, I, C, W

18–59 years For recreation 2. Access to beach Significant for 2 (�), 3 (�)Meeting guidelines 3. Neighborhood aesthetics Significant for 1 (�), 3 (�)

4. Traffic, busy roads5. Sidewalks present6. Stores in walking distance

Giles-Corti (2003)33 1803 CS, o Meeting guidelines Composite physicalenvironment score

Significant for high score (�) A, G, C, E, I

(continued on next page)

70A

merican

Journal

ofPreven

tiveM

edicine,

Volum

e27,

Num

ber1

Table 1. (continued)

Author (year)refNumber, age,gender Design Walking outcome Environmental attributes Associations with walking outcomes

Statisticaladjustment

18–59 years Access to open space SignificantMinor traffic, some trees SignificantSidewalk, store Nonsignificant

Hovell (1989)21 194adults

CS, p For exercise Home equipment Nonsignificant A, G

Neighborhood environment Significant (�)Convenient facilities Nonsignificant

Hovell (1992)22 1701 Pros, p,o

Change in walking over 2years

Neighborhood environment Nonsignificant A, G

Mean�48 years Convenient facilities Significant (�)Men 58%

Humpel (2004)24 399 CS, p 1. Neighborhood Coastal location Significant for women (�) A, E, and othercategories ofenvironmentvariables

Mean�60 years 2. For exercise Aesthetics Significant with 1, 2 for men (�)Men 43% 3. For pleasure Accessibility of facilities for

walkingSignificant with 1 for men (�) and

3 for women (�)4. To get to places Safety (personal and traffic) Significant with 3 for men (�)

Weather Significant with 1, 2 for men (�)and 1, 2 for women (�)

Humpel (2004)27 800 CS, p, o Neighborhood walking Coastal location Significant for men (�) A, G, E18–71 Years Aesthetics Significant for men (�)Men 50% Convenience Significant for men and women

(�)Access to services Significant for men (�) and

womenTraffic not a problem Significant for men (�)

Humpel (2004)28 512 Pros, p,o

�60 min increase inneighborhood walking

Coastal location Significant for men (�) A, G, E

18–69 Years Change in aesthetics NonsignificantMen 49% Change in convenience Significant for men and women

(�)Change in access to services NonsignificantChange in traffic not a

problemSignificant for men (�)

King (2003)36 149 Women CS, p, o 1. Pedometer steps in 1week;

Convenience of walking to 11destinations

Significant with 1, biking orwalking trail (�); store(s) (�);park (�); neighborhood rating

None reported

Mean�74 Years 2. Self-report walkingkilocalories per week

Overall quality ofneighborhood for walking

Significant with 2, neighborhoodrating

(continued on next page)

Am

JPrev

Med

2004;27(1)71

aa

mhwtwCseowa

aspaw

D

TmpthsiplTefpupaG(tctewsoefwstaw

able

1.C

har

acte

rist

ics

and

mai

nfi

ndi

ngs

ofst

udie

sex

amin

ing

rela

tion

ship

sof

envi

ron

men

tal

attr

ibut

esw

ith

am

ain

outc

ome

ofw

alki

ng

(con

tinue

d)

utho

r(y

ear)

ref

Num

ber,

age,

gend

erD

esig

nW

alki

ngou

tcom

eE

nvir

onm

enta

lat

trib

utes

Ass

ocia

tion

sw

ith

wal

king

outc

omes

Stat

isti

cal

adju

stm

ent

Sign

ifica

nt

wit

h1,

2,to

tal

num

ber

ofpe

rcei

ved

con

ven

ien

tde

stin

atio

ns

(�)

aele

ns

(200

3)30

110

CS,

oFo

rer

ran

dsou

tsid

eh

ome

Hig

hly

wal

kabl

en

eigh

borh

ood

Sign

ifica

nt

(�)

Un

adju

sted

Mea

n�

48ye

ars

On

wor

k/sc

hoo

lbr

eaks

Sign

ifica

nt

(�)

Men

46.7

%Fo

rex

erci

seN

onsi

gnifi

can

tT

otal

wal

kin

gN

onsi

gnifi

can

tal

lis(1

997)

26

110

CS,

pFo

rex

erci

seH

ome

equi

pmen

tN

onsi

gnifi

can

tA

,G

,E

th,

SES

Mea

n�

21ye

ars

Con

ven

ien

tfa

cilit

ies

Non

sign

ifica

nt

Men

25%

Tot

aln

eigh

borh

ood

Non

sign

ifica

nt

�),

posi

tive

asso

ciat

ion

;(�

),n

egat

ive

asso

ciat

ion

;A,a

ge;C

,num

ber

ofch

ildre

n;C

S,cr

oss-

sect

ion

al;E

,edu

cati

on;E

th,e

thn

icit

y;E

m,e

mpl

oym

ent

stat

us;G

,gen

der;

I,in

com

e;L

,loc

atio

n;o

,bj

ecti

vely

mea

sure

d;p,

perc

epti

ons;

P,po

vert

y;Pr

os,

pros

pect

ive;

SES,

soci

oeco

nom

icst

atus

;U

,ur

ban

izat

ion

;W

,w

ork

stat

us.

72 Am

ssociations with particular environmental attributesre listed for each walking outcome.

Perceptions of the aesthetic nature of the environ-ent have most often been measured. This attribute

as been found to be significantly associated withalking for exercise or recreation in four studies and in

wo studies with total walking, but was not associatedith walking to get to and from places (Table 2).omposite convenience of facilities for walking mea-

ures has been found to be associated with walking forxercise or recreation in four studies. Specific facilitiesr destinations (such as convenience of biking oralking trails, stores in walking distance) were associ-ted with total walking.

Access to beach and public open spaces, and having“highly walkable” neighborhood were found to be

ignificantly related to walking to get to and fromlaces. Perceptions about traffic were found to bessociated with walking for exercise or recreation andith walking to get to and from places.

iscussion

he pattern of findings summarized in Table 2 shows aodest but consistent body of evidence indicating

atterns of positive relationships of environmental at-ributes with particular types of walking. What must beighlighted, however, is the number of studies in whichome of these relationships were not statistically signif-cant. Also, while studies accounted for only smallroportions of variance in physical activity, on a popu-

ation-wide basis these proportions can be substantial.hese findings support the broad conclusions of anarlier review of environment–behavior relationshipsor all types of physical activity.18 The findings of theresent review support the authors’ argument for these of behavior-specific measures that help identify thearticular environmental attributes that might promptnd maintain habitual physical activities.8,11,12,14,19

iven the small number of studies specific to walking18 were identified in this review), it would be prema-ure to state definitive conclusions, although there isonsistency in the patterns of associations found inhese studies (Table 2). Only four studies have thus farxamined relationships of environmental attributesith walking to get to and from places; three of these

tudies have found significant associations25,29,30 andne found no association.24 Most studies have foundnvironmental attributes to be associated with walkingor exercise, recreation, or total walking. While thereas some overlap (“stores in walking distance” being

imilar to “access to services”), the environmental at-ributes found to be associated with walking to get tond from places differed from those associated withalking for exercise or recreation (Table 2).Comparisons of patterns of walking behavior be-

tween neighborhoods that differ on objectively assessedT A S S ( o

erican Journal of Preventive Medicine, Volume 27, Number 1

T

T

W

W

T

a

able 2. Overview of environmental attributes significantly associated with particular types of walking

ype of walking Environmental attributes

Associations reported in studies(Ref nos.)

Significant Nonsignificant

alking for exercise orrecreation (includesneighborhood and pleasurewalking)

Aesthetically pleasant 23, 25, 27 (m) 27 (w)24 (m) 24 (w)

Convenient facilities/environment 23, 22, 27, 28 21, 26Walking trail lengtha 20Trail surface 20Distance to traila 20Access to beacha 25Neighborhood environment 21Coastal locationa 27 (m), 28, 24 (w) 27 (w), 24 (m)Access to services 27Traffic not a problem 27 (m) 27 (w)Home equipment 21, 26Total neighborhood 26Highly walkable neighborhooda 30Neighborhood environment 22Change in aesthetics 28Change in convenience 28Change in access to services 28Change in traffic not a problem 28 (m) 28 (w)Access to public open spacesa 25Perceptions of traffic, busy roads 25Sidewalks presenta 25Stores within walking distance 25Accessibility of facilities 24 (w) 24 (m)Weather 24Safety 24 (m) 24 (w)

alking to get to and fromplaces (includes walking towork)

Composite environmenta 29Access to public open spacea 25Access to beacha 25Perceptions of traffic, busy roads 25Sidewalks presenta 25Stores in walking distance 25High walkable neighborhooda 30Aesthetically pleasing 24, 25Accessibility 24Safety 24Weather 24

otal walking (includes meetingpublic health guidelines)

Age of homea 31Neighborhood/environmental aesthetics 25, 32, 37Practical environment 32Access to public open spacesa 25, 33Composite environment scorea 33Access to beach 25Perceptions of traffic, busy roads 25, 37Sidewalks presenta 37 25Stores in walking distance 25Minor traffic, some treesa 33Sidewalk, storea 37 33High walkable neighborhooda 30County/metropolitan sprawla 35Convenience of biking/walking trails 36 (w)Convenience of department, discount, or hardware store 36 (w)Convenience of a park 36 (w)Convenience of eight other community destinations 36 (w)Overall neighborhood quality 36 (w)Availability of sidewalks 34Land use mix (diversity) 34Easy walk to public transport 3412 other environmental categories 34Hills 37No street lights 37Unattended dogs 37Foul air 37High crime 37

Objectively assessed environmental attributes. m, men; w, women.

Am J Prev Med 2004;27(1) 73

ethwiRdfihiCusb

TB

Wrrpwstttswaafinipcap

UW

Tmemficrmlsesmmf

esiriifl(

gtgfs

ev

RttcmHpfc“p1bhcrlcsttcrhcrmrrwe

VtpfSr

7

nvironmental attributes can be particularly informa-ive. Saelens et al.30 did not find differences betweenigh- and low-walkable neighborhoods in self-reportedalking for exercise, self-reported leisure physical activ-

ty, and objectively measured vigorous physical activity.esidents of the high- and low-walkable neighborhoodsid, however, differ on their walking for errands. Thisnding parallels those of transportation studies thatave found no differences in leisure or exercise walk-

ng, but significant differences in walking for transport.onsistent with earlier studies, low levels of othertilitarian walking forms (walking to and from work orchool and to or from transit stops) were reported inoth neighborhoods.

he Utility of Specific Measures of Walkingehaviors

hile there is an insufficient number of studies toeach definitive conclusions, the pattern of findingseviewed here suggests some specific associations ofarticular environmental attributes with particularalking behaviors. For example, Saelens et al.30 as-

essed the minutes spent walking during the past weeko and from work or school, during breaks or lunch-ime at work or school, as part of errands done outsidehe household, for exercise, and to and from transittops. An index of total (sum of the minutes acrossalking purposes) self-reported walking could be cre-ted from these specific indices. Generally, strongerssociations with environmental attributes were foundor the more particular indices of walking behaviorndices.30 Humpel et al.27 found a greater likelihood ofeighborhood walking (a particular walking behavior

ndex) for those who lived in a coastal location and hadositive perceptions of neighborhood aesthetics andonvenience of and access to places to walk. Less-strongssociations were found for total walking or for totalhysical activity.27

nderstanding Environmental Influences onalking: Research Opportunities

he available research findings that identify environ-ental correlates of walking include approximately

qual numbers of associations for objectively deter-ined and perceived environmental attributes. Fewndings from prospective studies are available. Toonclude that environmental attributes have a causalole, there is a need to go beyond looking at environ-ental attributes on their own and to develop multi-

evel studies that include the strongest individual- andocial-level influences on physical activity, such as self-fficacy and social support, ideally using prospectivetudy designs. Multilevel analyses require statisticalethods that can take into account the role of possibleediators (intervening variables) and moderators (ef-

38,39

ect modifiers) in complex causal modeling. How- s

4 American Journal of Preventive Medicine, Volume 27, Num

ver, examinations of the relative importance personal,ocial, and environmental influences on physical activ-ty33 require the use of well-developed measures of theelevant environmental attributes; these remain to bedentified and refined.6,7,11 It would be premature, fornstance, to conclude that individual- and social-levelactors (domains where measures are more well estab-ished) are more influential than environmental factorsa domain where measures are less well established).

Few of the studies reviewed here reported data onender differences in the relationships of environmen-al attributes with walking. Considering the strongender differences reported in some studies,21,27,28

uture investigations should examine these relation-hips with walking for men and women separately.

Four key elements of the research agenda relating tonvironmental influences on walking are reliability,alidity, causality, and conception.

eliability. Reliable measures of environmental at-ributes are required; this is particularly so for percep-ions of environmental attributes such as aesthetics andonvenience, but also for reports on specific environ-ental features such as presence of sidewalks or shade.umpel et al.28 examined the test–retest reliability oferceptions of the neighborhood environment andound excellent agreement between tests with intraclassorrelations ranging from 0.73 to 0.93 for “aesthetics,”convenience,” “access to services,” and “traffic” as aroblem. Saelens et al.30 found that the majority of-week test–retest values for items used in their Neigh-orhood Environment Walkability Scale to be �0.75, aigh level of consistency. Individual test–retest intra-lass correlations were generally in the 0.60 to 0.80ange for residential density, land use–mix diversity,and use–mix access, street connectivity, walking/cy-ling facilities, aesthetics, pedestrian/traffic safety, andafety from crime. Kirtland et al.40 examined 3-weekest–retest reliability for items measuring perceptions ofhe neighborhood and community supports (access,haracteristics, barriers, social issues). They found thatetest reliability was slightly higher for the neighbor-ood items, ranging from 0.42 to 0.74 overall. In someircumstances, obtaining measures of perceived envi-onmental attributes may be less costly than objectiveeasures. The inclusion of standardized, reliable self-

eport measures in multiple studies would help thisesearch field to advance more rapidly. In particular, itould facilitate comparisons of environmental influ-nces across a variety of locations and populations.34

alidity. Rated and self-reported environmental at-ributes should be objectively verifiable, either by inde-endent observation or by objective indices derivedrom geographic information system (GIS) databases.19

aelens et al.30 assessed the construct validity of envi-onmental attribute measures by comparing subscale

cores across two neighborhoods selected to differ

ber 1

othpbabmtahtsadafPewtpdt(cotwmt

Cbmdoppisswtmpt

CwmcrTutae

dopaatmSeh

C

Ripli(ectartfhoqp

fensctpcmmaapiceppt

astab

bjectively on walkability-related environment charac-eristics. Residents in the highly walkable neighbor-oods perceived greater residential density, closerroximity, easier access to stores and other facilities,etter street connectivity, and better neighborhoodesthetics than did residents in low-walkability neigh-orhoods. Kirtland et al.40 assessed the validity of itemseasuring environmental perceptions by comparing

hem to objective measures using GIS. Overall, lowgreement between measures was found for neighbor-ood and community items (kappa ranged from �0.02

o 0.37). There is a need to improve and refineelf-reported measures of environmental attributes thatre currently in use, and for more studies that allowirect comparisons of self-reported perceptions of suchttributes with objective indices that can be derivedrom GIS databases and other sources. For example,ikora et al.41 developed a framework of potentialnvironmental influences on the specific behaviors ofalking and cycling for recreation and for transporta-

ion. Items based on findings from the health, trans-ortation, and urban planning literature were used toevelop an environmental audit instrument, the Sys-ematic Pedestrian and Cycling Environmental ScanSPACES),42 that collected data via observationalhecklists used by trained observers. If strong patternsf concordance emerge between perceived and objec-ive indices of the same environmental attributes, thisill provide support for the validity of the self-reportedeasures of perceived environmental used in several of

he studies that were reviewed here.

ausality. In order to conclude that physical activityehaviors such as walking are influenced by environ-ental attributes, there is the need to move beyond the

escription of cross-sectional associations, making usef prospective study designs with multiple observationoints as well as intervention study designs.7,43 Whilerospective studies with two observation points28 are

nformative, they are limited in that no firm conclu-ions can be reached on the direction of the relation-hips reported. It is possible that increased levels ofalking might influence participants’ perceptions of

he environment.28 Future prospective studies needeasurements made at a minimum of three time

oints, in order to gain a clearer view of the direction ofhese environment–behavior relationships.38,43

onception. The conceptual models and theories onhich this research draws require considerable refine-ent and development. At present, relatively broad

onceptual models of putative environment–behaviorelationships are being used to guide research.5,13,19

hese models may benefit from considering possiblenderlying mechanisms that might help to improveheir explanatory specificity. For example, Bargh etl.8,9 argue that many of the actions in which people

ngage in everyday life are “automatic.” There is evi- i

ence from a body of experimental studies that featuresf current environments (people, objects, settings inarticular) can drive many habitual behaviors. Theyrgue that behavioral choices can be prompted by theutomatic processing of sets of environmental featureso which people have been repeatedly exposed, without

ediation by conscious reflection or decision making.8,9

uch a perspective adds potential depth to some of thenvironmental influences models of physical activity be-avior that are broadly guiding current research.5,13,19

onclusions

esearch on environmental factors associated with walk-ng shows a promising, although at this stage limited,attern of positive findings. The aesthetic nature of the

ocal environment, the convenience of facilities for walk-ng (footpaths, trails), accessibility of places to walk toshops, beach), level of traffic on roads, and composites ofnvironmental attributes have all been found to be asso-iated with walking for particular purposes. However,hese findings are primarily from cross-sectional studiesnd many of the associations reported are based onespondents’ perceived ratings of environmental at-ributes or subjectively identified specific environmentaleatures. Broadly, the research agenda requires a be-avior-specific approach, paying particular attention tobjectively defined environmental attributes, and re-uires multilevel modeling approaches to identify howotentially relevant determinants might be acting.38,39

From a public health advocacy perspective, the caseor “conception”—the conceptual and theoretical mod-ls that will be most helpful in explaining the determi-ants of walking—is of relevance. It is too easily as-umed, given the focus of social cognitive models ononstructs such as attitudes, self-efficacy, and inten-ions, that conscious individual decision making is therimary determinant of behavioral choice.10,14 In thease of physical activity, we have argued that differentodels are needed that focus primarily on environ-entally cued habitual behavior patterns.4,6,12 In the

dvocacy context, it is probable that political anddministrative decision makers operate from an im-licit framework that sees physical activity as “exercis-

ng,” and thus solely within the domain of individual,onsciously made “lifestyle” choices. In this perspective,nvironmental and policy changes that are needed toromote sustainable increases in physical activity in wholeopulations1,2,4 will need concerted advocacy, if they are

o be taken sufficiently seriously in political and civic life.The central challenge in pursuing public health

dvocacy for physical activity is to build this case onystematic research and theory development to identifyhe most relevant environmental influences on physicalctivity. In doing so, it is imperative that this researche used, if possible, to reframe the predominant,

mplicit model of physical activity as a conscious, dis-

Am J Prev Med 2004;27(1) 75

ctfsr

apoddppa

R

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

4

4

7

retionary lifestyle choice. This is particularly impor-ant in socioeconomic contexts where prolonged, en-orced (or at least strongly reinforced) periods ofedentary behavior in occupational and domestic envi-onments constitute a major public health risk.13,15,44

Understanding environmental influences on physicalctivity is an important and challenging new area ofopulation health research, with many new scientificpportunities.45 Importantly, it is research that is fun-amental to chronic disease prevention, through evi-ence-based environmental, transportation, urbanlanning, and public health policy strategies that willromote walking as a more central component ofdults’ health-enhancing physical activity.46

eferences1. Bauman A, Bellew B, Vita P, Brown W, Owen N. Getting Australia active:

best practice for the promotion of physical activity. Melbourne: NationalPublic Health Partnership, 2002.

2. U.S. Department of Health and Human Services. Physical activity andhealth: a report of the Surgeon General. Atlanta GA: U.S. Department ofHealth and Human Services, Centers for Disease Control and Prevention,National Center for Chronic Disease Prevention and Health Promotion, 1996.

3. Bauman A, Armstrong T, Davies J, et al. Trends in physical activityparticipation and the impact of integrated campaigns among Australianadults, 1997–1999. Aust NZ J Pub Health 2003;27:76–79.

4. Sallis JF, Bauman A, Pratt M. Environmental and policy interventions topromote physical activity. Am J Prev Med 1998;15:379–397.

5. Sallis JF, Owen N. Ecological models of health behavior. In: Glanz K, LewisFM, Rimer BK, eds. Health behavior and health education: theory, re-search, and practice. 3rd ed. San Francisco: Jossey-Bass, 2002;462–84.

6. Bauman A, Sallis JF, Owen N. Environmental and policy measurement inphysical activity research. In: Welk G, ed. Physical activity assessment forhealth: related research. Champaign IL: Human Kinetics, 2002;241–51.

7. Bauman AE, Sallis JF, Dzewaltowski DA, Owen N. Toward a better under-standing of the influences on physical activity: the role of determinants,correlates, causal variables, mediators, moderators, and confounders. Am JPrev Med 2002;23:5–14.

8. Bargh J, Ferguson M. Beyond behaviorism: on the automaticity of mentalprocesses. Psychol Bull 2000;126:925–45.

9. Bargh J, Chartrand T. The unbearable automaticity of being. Am Psychol1999;54:462–79.

0. Godin G. Social-cognitive models. In: Dishman RK, ed. Advances inexercise adherence. Champaign IL: Human Kinetics, 1994;113–36.

1. Trost SG, Owen N, Bauman A, Sallis JF, Brown W. Correlates of adults’participation in physical activity: review and update. Med Sci Sports Exerc2002;34:1996–2001.

2. Sallis JF, Owen N. Physical activity and behavioral medicine. ThousandOaks CA: Sage Publications, 1999.

3. Owen N, Leslie E, Salmon J, Fotheringham MJ. Environmental determi-nants of physical activity and sedentary behavior. Exer Sports Sci Rev2000;28:153–8.

4. Bandura A. Social foundations of thought and action. Englewood Cliffs NJ:Prentice-Hall, 1986.

5. Salmon J, Owen N, Crawford D, Bauman A, Sallis JF. Physical activity andsedentary behavior: a population-based study of barriers, enjoyment, andpreferences. Health Psychol 2003;22:178–88.

6. Australian Bureau of Statistics. Participation in sport and physical activities.Canberra: Australian Bureau of Statistics, 2000.

7. Siegel P, Brackbill R, Heath GW. The epidemiology of walking for exercise:implications for promoting activity among sedentary groups. Am J PublicHealth 1995;85:706–10.

8. Humpel N, Owen N, Leslie E. Environmental factors associated with adults’participation in physical activity: a review. Am J Prev Med 2002;22:188–99.

9. Saelens B, Sallis JF, Frank L. Environmental correlates of walking andcycling: findings from the transportation, urban design, and planningliteratures. Ann Behav Med 2003;25:80–91.

0. Brownson RC, Houseman RA, Brown DR, et al. Promoting physical activityin rural communities: walking trail access, use, and effects. Am J Prev Med2000;18:235–41.

6 American Journal of Preventive Medicine, Volume 27, Num

1. Hovell MF, Sallis JF, Hofstetter CR, Spry VM, Faucher P, Caspersen CJ.Identifying correlates of walking for exercise: an epidemiologic prerequi-site for physical activity promotion. Prev Med 1989;18:856–66.

2. Hovell MF, Hofstetter CR, Sallis JF, Rauh M, Barrington E. Correlates ofchange in walking for exercise: an exploratory analysis. Res Q Exerc Sport1992;63:425–34.

3. Ball K, Bauman A, Leslie E, Owen N. Perceived environmental aestheticsand convenience, and company are associated with walking for exerciseamong Australian adults. Prev Med 2001;33:434–40.

4. Humpel N, Owen N, Iverson D, Leslie E, Bauman A. Perceived environ-ment attributes, residential location and walking for particular purposes.Am J Prev Med 2004;26:119–25.

5. Giles-Corti B, Donovan RJ. Socioeconomic status differences in recre-ational physical activity levels and real and perceived access to a supportivephysical environment. Prev Med 2002;36:601–11.

6. Sallis JF, Johnson MF, Calfas KJ, Caparosa S, Nichols JF. Assessing perceivedphysical environmental variables that may influence physical activity. Res QExerc Sport 1997;68:345–51.

7. Humpel N, Owen N, Leslie E, Marshall A, Bauman A, Sallis JF. Associationsof location and perceived environmental attributes with walking in neigh-borhoods. Am J Health Promot 2004;18:239–42.

8. Humpel N, Marshall A, Leslie E, Bauman A, Owen N. Changes inneighborhood walking are related to changes in perceptions of environ-mental attributes. Ann Behav Med 2004;27:60–7.

9. Craig CL, Brownson RC, Cragg SE, Dunn AL. Exploring the effect of theenvironment on physical activity: a study examining walking to work. Am JPrev Med 2002;23:36–43.

0. Saelens B, Sallis JF, Black J, Chen D. Preliminary evaluation of neighbor-hood-based differences in physical activity: an environmental scale evalua-tion. Am J Public Health 2003;93:1152–8.

1. Berrigan D, Troiano RP. The association between urban form and physicalactivity in U.S. adults. Am J Prev Med 2002;23:74–9.

2. Carnegie MA, Bauman A, Marshall A, Mohsin M, Westley-Wise V, BoothML. Perceptions of the physical environment, stage of change for physical activityand walking among Australian adults. Res Q Exerc Sport 2002;73:146–55.

3. Giles-Corti B, Donovan RJ. The relative influence of individual, socialenvironmental and physical environmental correlates of walking. Am JPublic Health 2003;93:1183–9.

4. De Bourdeaudhuij I, Sallis JF, Saelens B. Environmental correlates ofphysical activity in a sample of Belgian adults. Am J Health Promot2003;18:83–92.

5. Ewing R, Schmid TL, Killingsworth R, Zlot A, Raudenbush S. Relationshipbetween urban sprawl and physical activity, obesity, and morbidity. Am JHealth Promot 2003;18:47–57.

6. King W, Brach J, Belle S, Killingsworth R, Fenton M, Kriska A. Therelationship between convenience of destinations and walking levels inolder women. Am J Health Promot 2003;81:74–82.

7. Eyler AA, Brownson RC, Bacak S, Houseman RA. The epidemiology ofwalking for physical activity in the United States. Med Sci Sports Exerc2003;35:1529–36.

8. Masse LC, Dassa C, Gauvin L, Giles-Corti B, Motl R. Emerging measure-ment and statistical methods in physical activity research. Am J Prev Med2002;23:44–55.

9. Blakely T, Woodward A. Ecological effects in multi-level studies. J Epide-miol Community Health 2002;54:367–74.

0. Kirtland K, Porter D, Addy CL, et al. Environmental measures of physicalactivity supports: perception versus reality. Am J Prev Med 2003;24:323–31.

1. Pikora T, Giles-Corti B, Bull F, Jamrozik K, Donovan RJ. Developing aframework for assessment of the environmental determinants of walkingand cycling. Soc Sci Med 2003;56:1693–1703.

2. Pikora T, Bull F, Jamrozik K, Knuiman M, Giles-Corti B, Donovan RJ.Developing a reliable audit instrument to measure the physical environ-ment for physical activity. Am J Prev Med 2002;23:187–94.

3. King AC, Stokols D, Talen E, Brassington GS. Theoretical approaches tothe promotion of physical activity. Am J Prev Med 2002;23:15–25.

4. Salmon J, Bauman A, Crawford D, Timperio A, Owen N. The associationbetween television viewing and overweight among Australian adults partic-ipating in varying levels of leisure-time physical activity. Int J Obes2000;24:600–6.

5. Spence J, Lee R. Toward a comprehensive model of physical activity.Psychol Sport Exerc 2003;4:7–24.

6. Frank LD, Engelke PO, Schmidt TL. Health and community design: theimpact of the physical environment on physical activity. Washington DC:Island Press, 2003.

ber 1