School Gardens and Physical Activity: A Randomized Controlled Trial of Low-Income Elementary Schools

This study examines effects of a school garden intervention on elementary school children's physical activity (PA). Twelve schools in New York were randomly assigned to receive the school garden intervention (n = 6) or to the waitlist control group that later received gardens (n = 6). PA was measured by self-report survey and accelerometry at baseline and follow-up. Direct observation was employed to compare indoor and outdoor PA. Results show promise for school gardens to promote children's PA.

School gardens and physical activity: A randomized controlled trial of
low-income elementary schools
Nancy M. Wellsa,
⁎, Beth M. Myersa, Charles R. Henderson Jr.b
a Department of Design & Environmental Analysis, College of Human Ecology, Cornell University, Ithaca, NY 14853, USA
b Department of Human Development, College of Human Ecology, Cornell University, Ithaca, NY 14853, USA
a b s t r a c ta r t i c l e i n f o
Available online 16 October 2014
Keywords:
Health behavior
Motor activity
Primary schools
Gardening
Child
Objective: This study examines effects of a school garden intervention on elementary school children's
physical activity (PA).
Method: Twelve schools in New York were randomly assigned to receive the school garden intervention
(n = 6) or to the waitlist control group that later received gardens (n = 6). PA was measured by self-report s
vey (Girls Health Enrichment Multi-site Study Activity Questionnaire) (N = 227) and accelerometry (N = 124
schools) at baseline (Fall 2011) and follow-up (Spring 2012, Fall 2012, Spring 2013). Direct observation (N =
117, 4 schools) was employed to compare indoor (classroom) and outdoor (garden) PA. Analysis was by gen
linear mixed models.
Results: Survey data indicate garden intervention children's reports of usual sedentary activity decrease
from pre-garden baseline to post-garden more than the control group children's ( Δ = − .19,p = .001).
Accelerometry data reveal that during the school day, children in the garden intervention showed a greater
increase in percent of time spent in moderate and moderate-to-vigorous PA from baseline to follow-up than
the control group children (Δ = +.58, p = .010; Δ = +1.0, p = .044). Direct observation within-group com-
parison of children at schools with gardens revealed that children move more and sit less during an outdoor
garden-based lesson than during an indoor, classroom-based lesson.
Conclusion: School gardens show some promise to promote children's PA.
Clinical Trials Registration: clinicaltrials.gov # NCT02148315.
© 2014 Elsevier Inc. All rights reserved.
Introduction
Children and youths in the United States are not achieving recom-
mended levels ofphysical activity (PA) (NASPE, 2004; Pate et al.,
2002).Among 11 year olds in the U.S.,only 24% of girls and 30% of
boys achieve the recommended 1 h of moderate-to-vigorous PA
(MVPA) daily (World Health Organization, 2012). In New York State,
34.1% of school-aged children engaged in at least 20 min of vigorous
PA 4–6 days per week,compared to 37.8% of children nationwide.
Additionally, only 24.6% of New York children engaged in vigorous PA
everyday compared to 28.0% of children nationwide (National Survey
of Children's Health).Physical inactivity has been linked both cross-
sectionally and prospectively to obesity (Dietz and Gortmaker, 1985;
Gortmaker et al., 1996; Hancox et al., 2004). Health benefits associated
with PA throughout the life course are well-documented (Blair and
Morris, 2009; Nocon et al., 2008; Woodcock et al., 2011). Strategies to
reduce sedentary behavior and increase PA during childhood may help
to curb the obesity epidemic and set youths on a more active, healthy
life-course trajectory (Elder,1998; Wethington,2005; Wheaton and
Gotlib, 1997).
School gardens have gained prominence as a potential contributor to
public health (Christian et al.,2012; Ozer,2007; Twiss et al.,2003).
Gardens are unique in their potential to affect both sides of the energy
balance equation: dietary intake and physical activity (Hill and Peters,
1998; Wells et al., 2007), and yet relatively few studies have examined
the effects of gardens on children's health or health behaviors. More-
over, the extant research on the topic of gardens and children's health
has focused almost exclusively on the potential for gardens to impact
children's diet-related outcomes such as fruit and vegetable consump-
tion or fruit and vegetable preference (Christian et al.,2012,2014;
Lineberger and Zajicek,2000; Morris and Zidenberg-Cherr,2002;
Morris et al., 2001, 2002, Robinson-O'Brien et al., 2009), while studies
of gardens'effects on children's PA are rare (Hermann et al.,2006;
Phelps et al., 2010). For evidence-based garden interventions to be de-
veloped and implemented, there is a need for a clearer understanding
of the potential for gardens to bolster children's PA and reduce seden-
tary behaviors.
Preventive Medicine 69 (2014) S27–S33
⁎ Corresponding author at: Design & Environmental Analysis, 2429 MVR Hall, Cornell
University, Ithaca, NY 14853, USA.
E-mail address: [e-mail] (N.M. Wells).
http://dx.doi.org/10.1016/j.ypmed.2014.10.012
0091-7435/© 2014 Elsevier Inc. All rights reserved.
Contents lists available at ScienceDirect
Preventive Medicine
j o u r n a lh o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / y p m e d

This study addresses the following three research questions:
1. Is there an effect of a school garden intervention on children's overall
PA and sedentary activity, as measured by self-report survey?
2. Is there an effect of a school garden intervention on children's PA
levels during the school day, as measured with accelerometry?
3. In a within-subjects comparison, does PA, measured by direct obser-
vation, differ during an indoor classroom lesson versus during an
outdoor garden lesson?
Methods
Study design and procedure
In this longitudinal cluster randomized controlled trial, schools were ran-
domly assigned to the garden intervention or to the waitlist control group
that received gardens at the end of the study (Wells et al., in press). Baseline
data were collected in Fall 2011 (wave 1). The garden intervention began in
Spring 2012 and continued through Spring 2013. Three waves of post-garden
implementation data were collected (wave 2: late Spring 2012, wave 3: Fall
2012, wave 4: late Spring 2013). All procedures were approved by the authors'
University's Institutional Review Board.The study was deemed exempt and
therefore did not require child assent or parental consent.
The intervention
The intervention, developed as part of the Healthy Gardens, Healthy Youth
pilot program, consisted of four components. (1) The garden was a 4 ′ × 8′ raised
bed for each class. (2) Access to a curriculum of 20 lessons for children in grades
4–6; 11 lessons for year 1, and 9 for year 2. The curriculum toolkit was created
based on a review of 17 extant garden curricula and focused on nutrition, horti-
culture, and plant science and included activities and snack suggestions. Aside
from the lessons, educators led other activities in the garden such as planting,
weeding, and harvesting. (3) Resources for the school included information
about food safety in the garden and related topics. (4) The garden implementa-
tion guide provided guidance regarding planning, planting and maintaining the
garden throughout the year; gardening during the summer; engaging volun-
teers; building community capacity; and sustaining the program.
Schools and classes
This study targeted low-income schools that did not already have school
gardens used for teaching and learning and had at least 50% of students qualify-
ing for free or reduced price meals (FRPM). A total of 12 schools in 5 regions of
Fig. 1. Flow diagram for school gardens RCT.
S28 N.M. Wells et al. / Preventive Medicine 69 (2014) S27–S33

New York State participated in this study.Fig. 1 shows the flow chart of the
school recruitment procedure and corresponding sample sizes. Schools within
each region (rural, suburban, and urban areas) were randomly assigned (by
the first author, using random number generator) to intervention or waitlist
control such that each region had an equal number of intervention and waitlist
control schools. Nineteen schools were randomized and 7 subsequently left the
study: 4 were lost because the Cooperative Extension staff person resigned;
3 schools dropped out due to personnel changes at the schools. The total num-
ber of classes evaluated within the 12 schools was 21, with most schools having
2 classes that participated in the intervention/evaluation (mean = 1.75). This
study is part of a larger 4-state examination of the effects of school gardens on
fruit and vegetable (FV) intake, FV preference, nutritional knowledge, and relat-
ed outcomes.
Participants
The participants in this study were children in grades 4 –5 (ages 8–12 years)
at the start of the study.
Constructs and measures
The dependent variable, PA, was operationalized in 3 ways. For each of the 3
dependent variable measures, sample size varied. Trained Cooperative Exten-
sion educators and university student research assistant (RAs) carried out the
data collection following standardized procedures to ensure consistent and
unbiased administration.
GEMS Activity Questionnaire (GAQ)
The Girls Health Enrichment Multi-site Study (GEMS) Activity Question-
naire (GAQ) was developed by Treuth et al. (2004) based on the Self-
Administered PhysicalActivity Checklist (SAPAC),that has been validated
with heart rate (r = .57, p b .001) and accelerometry (r = .30, p b .001) with
n = 125 5th graders (Sallis et al., 1996). The GAQ obtains information about
children's PA and sedentary activity “yesterday” and “usually” in 28 sports and
physical activities (e.g. bicycling, volleyball, hiking, gymnastics) and 7 sedentary
activities (e.g., watch TV, videos; computer games; play board games; listen to
music). The GAQ is cost-effective for large numbers of participants, has been val-
idated with 8–9-year-old girls (Treuth et al., 2004), and has been used with girls
(Baranowski et al., 2003) and boys (Jago et al., 2007). The GAQ was adminis-
tered at all 12 schools (n = 227 students at baseline and follow-up), in the class-
rooms. Activity scores1 were weighted by intensity using the appropriate
activity-specific MET values for children for each of the physicalactivities
(Ridley et al., 2008). A MET-weighted mean was computed:
Mw x1; x2; …; xkð Þ ¼
X k
i¼1 wi xið Þ
X k
i¼1 wið Þ
;
where w = METk; x = score; and k is the kth question in the GAQ.
Accelerometry
At each of the 4 waves of data collection, children at 8 schools (4 interven-
tion and 4 control) wore Actigraph GT3X+ or GT1M accelerometers from the
time they arrived at school in the morning until the end of the school day, for
3 days.On average,the children wore accelerometers for 355 min (5 h,
55 min) per day. Accelerometer data from children (collected with two acceler-
ometer units simultaneously) are highly correlated with energy expenditure
(r = .86 and .87), oxygen consumption (r = .86, .87), heart rate (r = .77, .77),
and treadmill speed (r = .90,.89) (Trost et al.,1998; Freedson and Miller,
2000). Trained Cooperative Extension educators or university student RAs dis-
tributed the accelerometers to the children and recorded the belt numbers
and time of day. Children were assisted with attaching the accelerometers to
their waist with an elastic belt and plastic buckle and were instructed to follow
their normal routine. Teachers were instructed on how to ensure that children
properly wore the accelerometers. At the end of each day, classroom teachers
collected the accelerometers and recorded the time collected. Due to a limited
supply of costly accelerometers as wellas the physicaldistance between
sites, accelerometry data were collected from one class at 8 of the 12 schools
(n = 124 at baseline and follow-up). On average, 21 children at each of the
schools participated in accelerometry (range per school 15 –25).
Direct observation
To characterize children's movements and motions during a 1-hour garden
lesson compared to a 1-hour classroom lesson, the Physical Activity Research &
Assessment tool for Garden Observation (PARAGON) (Myers & Wells, in press)
was employed. Direct observation was conducted by trained RAs during waves
2 and 4 (Spring 2012 and Spring 2013) at 4 intervention schools with N = 57
children (N = 117 within-subjects indoor–outdoor paired comparisons).
PARAGON uses momentary time sampling with a trained observer repeatedly
observing a focal child for 15 s and then recording behavior during a
15-second interval. Prior to data collection, each RA completed a five-phase,
~20-hour training protocol to memorize codes and ensure inter-rater reliability
PARAGON's overall test–retest reliability is .94.An Ebel of .97 (and percent
agreement of 88%) indicates strong inter-rater reliability ( Myers & Wells,in
press). The five primary PA codes (lying, sitting, standing, walking, vigorous ac-
tivity) used in PARAGON are based on Behaviors of Eating and Activity for
Children's Health (BEACHES) PA coding and were previously validated using
heart-rate monitors and accelerometers (Rowe et al., 2003; McKenzie et al.,
1991) and by expected convergent validity with accelerometry ( Kelkar et al.,
2011).
Statistical analysis
Accelerometry data were scored using ActiLife6 software.Thirty-second
epochs (Klesges et al., 1995) were converted into minutes and proportions of
time spent in each level of PA: 1) sedentary, 2) light PA (LPA), 3) moderate PA
(MPA), 4) vigorous PA (VPA); and MVPA using child-specific cut-points
(Evenson et al., 2008). Preliminary analyses (i.e. χ2 and t-tests) were conducted
to assess demographic differences between intervention and control at the
school and student levels.
The three PA outcomes have related structures of observations involving as
sessment over multiple waves but differ in whether there are observations on
both controls and treatment, on more than one day, or more than one setting
(i.e., indoors and outdoors). The model for the GAQ survey included treatment
(control versus intervention), sex of child, and wave (1 –4) as fixed classification
factors; the interactions among these factors; and individuals and classrooms a
levels of random classification factors.Ethnicity (white, African American,
Hispanic, and Asian) was included as an additional fixed classification factor.
Multi-racial, “other,”,and unknown-race children were excluded from the anal-
ysis because of small numbers. The model for accelerometry outcomes was the
same as for GAQ but also included day of assessment (i.e., day 1, 2, or 3 of the
measurement days per wave) as an additional fixed classification factor. To tak
into account the variance in time spent in physical education (PE) and time
spent in recess during the school day, we included PE and recess time as covar
iates (with higher-level regressions). We considered models with the depen-
dent variable computed as an average over the 3 days but used the model
with repeated measures on days to examine any differences by days and bette
account for individual variance in testing treatment differences. The model for
direct observation included sex of child, ethnicity, wave (2 and 4), and setting
(indoors versus outdoors) as fixed classification factors; and individuals and
classrooms as levels of random classification factors. In deriving each model,
we examined a number of mixed model specifications involving random classi-
fication factors for schools, classrooms, and children, with various assumptions
about error structure. Models including all 3 of these random factors did not
have stable estimation, despite attempts with a variety of iterative algorithms
for estimation, starting points for the solution, and other tweaking of the esti-
mation. We chose the model described here based on the larger variance asso-
ciated with classrooms than with schools, model fit, and theoretical grounds
that influences on the garden program were more likely to occur at the class-
room level than the school level.
Analysis for all 3 PA outcomes was carried out in general linear mixed
models. An unstructured error assumption was used, and denominator degrees
of freedom were computed by first-order Kenward–Rogers method. The key
test for evaluation of the GAQ and accelerometry outcomes is the test of the
interaction of treatment by wave.We partitioned from this interaction key
pre-specified contrasts of interest—specifically, the test of treatment by wave
1 versus waves 2, 3, and 4 (a 2 × 2 contrast). The tables show the means and
probabilities for these contrasts.
Statistical analyses were carried out using SAS version 9.3.
1 Activity items are scored 0, 1, 10: 0 = none; 1 = b15 min; 10 = 15 min or more. Sed-
entary items are scored 0, .25, .75, 1.5, 2.5: 0 = none; .25 = b30 min; .75 = 30 min–1 h;
1.5 = 1–3 h; 2.5 = more than 3 h.
S29N.M. Wells et al. / Preventive Medicine 69 (2014) S27–S33

Results
Characteristics of the 12 participating schools and their communities
are summarized in Table 1. On average, the 12 schools had 412 students
enrolled (range 280–612 students). The percent of students who were
ethnic minority varied across the 12 schools from 4% to 99% of the school
population; with an average of 58% ethnic minority. All 12 schools were
low-income, with an average of 69% of children participating in FRPM
(range 51% to 97%). Town and city characteristics were derived from
Census 2010 data.2 Five of the schools were urban, 3 suburban, and 4,
rural. On average, across the towns and cities, 16% of families were living
in poverty,and median household income was,on average,$40,882.
Overall, the intervention schools were smaller than the controls (359 v.
465 mean enrollment); had a higher percentage of children qualifying
for FRPM (74% v. 64%) and had a lower percentage of minority children
than the control schools (53% v. 64%); however these differences were
not statistically significant.
Table 2 summarizes the participant characteristics at baseline, Fall
2011 (n = 227). The participating children were 204 4th graders
(89.9%) and 23 5th graders (10.1%); 44% were boys.The average age
of children in the intervention schools was 9.5 years and in the control
schools, 9.0 years. Across all 12 schools, the majority of participating
children were ethnic minorities (30.0% African American; 8.8% Hispan-
ic; and 9.7% Asian); 51.5% were White. Comparing the intervention and
control participants,a higher percentage of participating students at
intervention schools were White (67%) compared to that in control
schools (36%).
Overall physical activity and sedentary activity
Survey (GAQ) results indicate that children in the school garden in-
tervention group showed a greater decrease in usual sedentary activity
from wave 1 to waves 2, 3, and 4 than did children in the control group
(p = .001). See Table 3.
Physical activity during the school day: accelerometry
Table 3 also presents the accelerometry findings.Children in the
garden intervention group showed a greater increase in percentage of
time spent in moderate PA (MPA) (+.45%; +1.6 min) from wave 1
(baseline) to waves 2,3, and 4 than did the control group children
(−.13%; −28 sec (p = .010). Similarly, intervention children showed
a greater increase in percentage oftime spent in MVPA (+1.68%;
+5.96 min) from wave 1 (baseline) to waves 2, 3, and 4 than the control
group children (+.68%; +2.42 min) (p = .044). No significant differ-
ences were found for changes in percentage of time spent in sedentary
(p = .144); LPA (p = .492); or VPA (p = .213).
Physical activity in garden v. classroom: direct observation
Table 4 summarizes the percentage of time children spent across the
7 PA intensities (lying, sitting, kneeling, standing, squatting, walking,
and very active) for both outdoor and indoor lessons. Findings indicate
significant differences for outdoor versus indoor comparisons for 6 of
the 7 PA intensity categories (lying was not significant,p = .108).
During the outdoor garden lessons, children spent significantly more
time walking (14% ~ 8.4/60 min) compared to an indoor lesson
(3% ~1.8/60 min) (p b .0001). On average, during outdoor lessons, chil-
dren sat 14% of their time, compared to 84% (~50.4/60 min) of indoor
class time (p b .0001). The only VPA children obtained was during out-
door garden lessons (2%, ~1.2/60 min) versus indoor lesson time (0%)
(p b .0001). During outdoor garden lessons, children spent the majority
Table 1
Elementary school and town/city characteristics at baseline: school year 2011–2012, New York (school n = 12).
Intervention schools Control schools
1 2a,b 3a,b 4 5a,b 6a,b Mean (sd) 7a 8 9a 10 11a 12a Mean (sd)
Demographic
School type pK-5 pK-12 K-6 K-6 K-6 pK-7 pK-12 K-6 pK-8 K-6 3–5 3–5
# students enrolled 284 286 367 317 280 612 358 (129) 425 436 527 459 496 449 465 (39)
% minority students6% 4% 69% 72% 71% 95% 53% (38%) 26% 94% 82% 60% 99% 23% 64% (33%)
% free and reduced
meals
51% 71% 89% 67% 66% 97% 74% (17%) 55% 61% 82% 68% 63% 56% 64% (10%)
Rural/suburban/
urbanc
R R U U S U R S U U S R
Town and city characteristics
Total Population 1712 617 66,135 66,135 33,506 210,565 63,112 (77,862) 596 32,082 66,135 66,135 11,647 9145 30,957
(29,143)
% families living in
povertyd
13% 15% 18% 18% 13% 28% 18% (6%) 7% 20% 18% 18% 11% 11% 14% (5%)
Median household
incomee
$38,922 $44,038 $37,436 $37,436 $38,922 $30,367 $37,854 ($4403) $41,375 $48,386 $37,436 $37,436 $54,527 $44,306 $43,911
($6680)
a Schools in accelerometry study (n = 8).
b Schools in direct observation (n = 4).
c Classification based on ‘locale codes’ describing school locations as city, suburb, town, or rural (Common Core of Data, National Center for Education Statistics, and Census
d New York state % families living in poverty average (2007–2011) = 14.5% (Census, 2010).
e New York state median household income average (2007–2011) = 56,951 (Census, 2010).
2 Common Core of Data (CCD), National Center for Education Statistics and Census Bu-
reau (nces.ed.gov/ccd/rural_locales.asp).
Table 2
Participant characteristics, New York, 2011 (n = 227).
Intervention
n = 115
Control
n = 112
Total
n = 227
Significant
difference I and C
Mean age (at baseline)9.5 (.7) 9.0 (.5) 9.3 (.7) .000⁎⁎ a
n (%) n (%) n (%)
Gender .967b
Girl 65 (56.5) 63 (56.3) 128 (56.4)
Ethnicityc .000⁎⁎ b
White 77 (67.0) 40 (35.7) 117 (51.5)
African American 25 (21.7) 43 (38.4) 68 (30.0)
Hispanic 10 (8.7) 10 (8.9) 20 (8.8)
Asian 3 (2.6) 19 (17.0) 22 (9.7)
a t-Test analyses were used to compare the differences between intervention and con-
trol groups.
b Chi-square analyses were used to compare the differences between intervention and
control groups.
c Children with missing race (N = 15), “other” (N = 12), and “multi-racial” (N = 1)
not in analyses due to small Ns and lack of information.
⁎⁎ p b .01.
S30 N.M. Wells et al. / Preventive Medicine 69 (2014) S27–S33

of the time standing (53%, ~31.8/60 min); while 9% (~5.4/60 min) of in-
door time was spent standing (p b .0001).
Discussion
This school-based cluster randomized controlled trial examines the
effects of a school garden intervention on children's PA. Results suggest
that school gardens help to promote children's PA and reduce sedentary
activity. Results indicate that compared to children at schools without
gardens,over the course of 2-years,children at garden intervention
schools report a greater reduction in their usual daily sedentary activity.
During the school day specifically,children at schools with gardens
exhibited a greater increase in MPA and MVPA than did children in
the no-garden control schools, as measured by accelerometry.In
addition, within-group comparison revealed that outdoor,garden-
based lessons were associated with more varied postures and move-
ments and less sitting than indoor, classroom-based lessons.Though
there is little prior research examining the effects of school gardens on
PA, our findings are consistent with a pre–post study (n = 43) that
found, based on a 1-item self-report measure of PA, that more children
were “physically active every day” following garden participation
(Hermann et al., 2006).
Interpretation
It is unsurprising that findings from the GAQ did not reveal statisti-
cally significant effects on children's survey responses regarding PA
“usually” and PA “yesterday.” Although we might hope that school gar-
dens would activate children to bicycle, do push-ups, play basketball, go
hiking, or do yoga, it seems unlikely, particularly within the relatively
short two-year time frame of this study.And, although no effect of
school gardens is found on sedentary activities “yesterday,” the signifi-
cant effect of school gardens on sedentary activities “usually” suggests
that school gardens may indeed compete with “screen time” (e.g., TV,
computers,and on-screen games) and other sedentary endeavors to
contribute to a reduction of time spent engaged in sedentary behaviors
even beyond the school environment.Both longitudinal and secular
trends contribute to an increase in sedentary behavior (Nelson et al.,
2006) which has been linked both cross-sectionally and prospectively
to obesity (Gortmaker et al., 1996; Dietz and Gortmaker, 1985;
Hancox et al., 2004). Reduction of sedentary behaviors is an important
objective and is associated with decreases in percentage overweight
and body fat as well as with increased physical fitness (Epstein et al.,
2000).
While the approximately 6-minute increase in MVPA and nearly
2-minute increase in MPA during the school day among garden inter-
vention children were modest, they do contribute to daily MVPA and
they may help to counteract the tendency toward greater inactivity
with age (Sallis et al., 1999; Trost et al., 2002; Trost et al., 1996;
Whitt-Glover et al., 2009). Moreover if schools embrace gardens as a
pedagogical tool and as a health intervention strategy, more time will
be spent gardening and engaging in garden-based lessons, likely yield-
ing a stronger effect. In this study, the intervention varied somewhat
from one class to another, with students, on average, spending only an
hour or less in the garden weekly. Thus, our findings are likely conserva
tive. Changes in other accelerometry-measured levels of PA during the
school day (sedentary, LPA, VPA) were in the predicted direction though
not statistically significant.Given the nature of gardening tasks,the
effect on MVPA and MPA, but not VPA is expected. In fact, the significan
effects on MVPA were primarily driven by the changes in MPA. Schools
have been identified as a promising context for the promotion of youth
PA and other health behaviors (Story et al., 2009; Centers for Disease
Control and Prevention, 2011; Tuckerman, 2013). Thus, if gardens can
be integrated more thoroughly with school curriculum throughout the
Table 3
Physical activity (GAQ & accelerometry) data, by treatment and pre-garden (wave 1) to post-garden (waves 2, 3, & 4); New York 2011–2013.
Intervention Control
Pre (W1) Post (W2–W4) Pre (W1) Post (W2–W4) Mean difference
Mean (SE) Mean (SE) Mean (SE) Mean (SE) (Interv–cont) p-Value
GAQ Survey (n = 227)
Activity — yesterday 2.91 (0.19) 2.48 (0.20) 2.74 (0.17) 2.51 (0.19) −.20 .312
Activity — usually 3.78 (0.18) 3.43 (0.19) 3.61 (0.16) 3.63 (0.18) −.37 .083
Sedentary — yesterday 0.63 (0.04) 0.51 (0.04) 0.57 (0.04) 0.54 (0.04) −.09 .064
Sedentary — usually 0.78 (0.05) 0.68 (0.05) 0.68 (0.04) 0.77 (.05) −.19 .001⁎⁎
Mean % (SE), min Mean % (SE), min Mean % (SE), min Mean % (SE), min (Interv–cont) p-Value
Accelerometry (n = 124)
Sedentary 55.23 (1.71), 196.07 55.00 (1.73), 195.25 54.75 (1.59), 194.36 56.11 (1.60), 199.19 −1.59 .144
Light PA 34.62 (1.00), 122.90 33.17 (1.02), 117.75 35.09 (0.92), 124.57 33.07 (0.93), 117.40 +.57 .492
Moderate PA 5.17 (0.54), 18.35 5.62 (0.54), 19.95 5.41 (0.50), 19.21 5.28 (0.50), 18.74 +.58 .010⁎
Vigorous PA 5.01 (0.58), 17.79 6.24 (0.59), 22.15 4.99 (0.54), 17.71 5.78 (0.54), 20.52 +.44 .213
MVPA 10.14 (1.03), 36.00 11.82 (1.04), 41.96 10.35 (0.95), 36.74 11.03 (0.95), 39.16 +1.00 .044⁎
GAQ survey:
Mean activity: yesterday t(570) = 1.01; usually t(569) = 1.74; mean sedentary: yesterday t(542) = 1.85; usually t(536) = 3.46.
Accelerometry:
% sedentary: t(1169) = 1.46; % light PA: t(1169) = −0.69; % moderate PA: t(1169) = −2.57; % vigorous PA: t(1169) = −1.25; % MVPA: t(1169) = −2.02.
⁎ p b .05.
⁎⁎ p b .01.
Table 4
Percentage of time spent in each physical activity category, during 1-hour indoor v. out-
door, direct observation (PARAGON); New York, 2012–2013 (n = 117 indoor & outdoor
paired observations).
PARAGON activity
category
Outdoor Indoor Mean Difference p-Value
Mean (SE) Mean (SE) (Outdoor − indoor)
Lying 0.73 0.50 0.05 0.50 +0.68 .108
Sitting 14.06 3.68 84.38 3.68 -70.32 .000⁎⁎⁎
Kneeling 9.90 2.28 0.89 2.28 +9.01 .000⁎⁎⁎
Standing 52.80 2.67 9.44 2.67 +43.36 .000⁎⁎⁎
Squatting 6.51 1.46 1.01 1.46 +5.50 .000⁎⁎⁎
Walking 14.09 2.12 3.10 2.12 +10.99 .000⁎⁎⁎
Very active 2.28 0.71 0.11 0.71 +2.17 .000⁎⁎⁎
Direct observation: % lying: t(129) = 1.62; % sitting: t(129) = − 28.93; % kneeling:
t(129) = 5.66; % standing: t(129) = 17.51; % squatting: t(129) = 5.52; % walking:
t(129) = 9.11; % very active: t(129) = 4.00.
⁎⁎⁎ p b .0001.
S31N.M. Wells et al. / Preventive Medicine 69 (2014) S27–S33

day, school gardens may be one component of a school's health promo-
tion intervention strategy,helping to nudge more children toward
achieving the recommended 60 min of daily MVPA, which currently is
only achieved by 42% of children ages 6–11 (Troiano et al., 2008).
In addition to school gardens contributing to a reduction in usual
sedentary activities and nudging children's at-school MVPA a bit higher,
our results suggest that while participating in a garden-based lesson,
children engage in more diverse physical movements and postures
than when participating in a classroom-based lesson. Indoors, children
spend most (84%) of their time being sedentary (sitting) and engage
in few other postures.Outdoors,while gardening,children mostly
stand (53% of time) but also engage in a variety of physical movements,
such as kneeling, squatting, walking, sitting, lying down, and running.
Allowing children more opportunity to move their bodies during the
school day may play a role in children's gross-motor development and
strengthen muscles and bones. The direct observation data provide in-
sight regarding what occurs during garden lessons and activities.
Study strengths
To our knowledge, this is the first randomized controlled trial
to examine the effects of gardening on children's PA.The research
design—including both pre- and post-measures as well as control and
intervention groups—ensures strong internalvalidity and provides
greater understanding of the effects of school gardens on children's
PA. Second, use of multiple measures of PA—all with established reliabil-
ity and validity—is an additional virtue of this study. Third, this study fo-
cuses on the population of children at greatest risk for physical inactivity
and obesity: those from under-resourced communities and predomi-
nantly ethnic minority youth (Centers for Disease Control and
Prevention, 2010; Ogden et al., 2010).
Study limitations
This study is not without its limitations. The focus on New York State
youths in under-resourced communities limits the external validity of
the study, hindering generalizability to other contexts and groups. In
addition, the garden intervention was examined holistically, without
distinguishing the individual components of the intervention.Thus,
findings do not elucidate what aspects of the garden intervention are
particularly potent to increase children PA. Moreover, this article does
not examine the fidelity of the garden intervention, which is likely to
differ from one school to another. With respect to the indoor classroom
and the outdoor garden PA paired comparison, it was not possible to
exactly match the curriculum delivered in the two settings. In addition,
despite randomization of schools to intervention or waitlist control,
there were significant differences in ethnicity and age between the
two groups.
Future research
Future studies might aim to ascertain the dose–response relation be-
tween school gardens and children's PA. Specifically, how many square
feet of garden space per child,how many hours of garden time, and
which and how many garden-based lessons are necessary to increase
children's PA. Also, the garden–PA relation might be examined in tan-
dem with children's learning outcomes and with a focus on how garden
activities affect children's attention once they return to the classroom.
The school garden movement will continue to gain traction if educa-
tional effectiveness is established along with beneficial effects on health
behaviors.Future research comparing PA levels during an indoor
classroom-based and an outdoor garden-based lesson might more pre-
cisely match indoor and outdoor lesson content to enhance comparabil-
ity of the two activities.
A school garden study from a life-course perspective might exam-
ine, over a longer time frame, whether introducing children to
gardening is indeed a turning point that shifts them from a life-
course trajectory of sedentary activities toward a trajectory of gar-
dening and healthy habits.Empirical evidence suggests that life-
long habits and patterns, including those related to PA and diet
(DiNubile, 1993), are established early in life (Elder, 1998;
Wethington, 2005; Wheaton and Gotlib, 1997; Wells and Lekies,
2006). The facts that gardening is one of the most popular home-
based leisure activities in the United States (Ashton-Shaeffer and
Constant, 2006) and the second most common leisure activity, after
walking, among adults over age 65 years (Yusuf et al., 1996) suggest
that, once begun, gardening has great potential as a life-long habit.
Conclusion
This randomized controlled trial of school gardens at under-
resourced New York State schools suggests that school gardens may
contribute to children's levels of PA at school and help to reduce time
spent in sedentary activity. Evidence from this study suggests that gar-
dening programs may merit school districts' allocation of resources.
Funding
This project was funded in part by the Robert Wood Johnson
Foundation through its Active Living Research Program (#69550).
Federal funding was provided by the U.S. Department of Agriculture
(USDA) through the Food & Nutrition Service (FNS) People's Garden
pilot program (Project #CN-CGP-11-0047) and by the Cornell Uni-
versity Agricultural Experiment Station (Hatch funds) (#NYC-327-
465) and Cornell Cooperative Extension (Smith Lever funds) through
the National Institutes for Food and Agriculture (NIFA) USDA. Addi-
tional funding for this study of gardens and physical activity came
from: Cornell University's Atkinson Center for a Sustainable Future
(ACSF); The College of Human Ecology, Cornell University; The
Bronfenbrenner Center for Translational Research (BCTR),Cornell
University; and the Cornell Cooperative Extension Summer Intern
Program.
The contents of this publication do not necessarily reflect the view or
policies of the U.S. Department of Agriculture, nor does mention of trade
names, commercial products, or organizations imply endorsement by
the U.S. Government.
Conflict of interest
The authors declare that there are no conflicts of interest.
Acknowledgments
Our thanks to Elaine Wethington who provided feedback on an
earlier draft of this manuscript.Thanks also go to our collaborators
on the larger four-state Healthy Gardens Healthy Youth study examin-
ing the effects of school gardens on dietary intake and other out-
comes.Thanks to the Extension Educators in New York State who
delivered the garden interventions and assisted with data collection
and to our team of undergraduate research assistants who contribut-
ed greatly to this project.
References
Ashton-Shaeffer, C., Constant, A., 2006. Why do older adults garden? Act. Adapt. Aging 30
(2), 1–18.
Baranowski,T., Baranowski, J.,Cullen,K., et al., 2003. The fun, food,and fitness project
(FFFP): the Baylor GEMS pilot study. Ethn. Dis. 13 (1), S1–S30.
Blair, S.N.,Morris, J.N., 2009. Healthy hearts–—and the universalbenefits of being
physically active: physical activity and health. Ann. Epidemiol. 19, 253–256.
Centers for Disease Control and Prevention, 2010. Nutrition and Physical Activity Program
to Prevent Obesity and other Chronic Diseases.Obesity Among Low–Income
Preschool Children. Centers for Disease Control and Prevention, Atlanta, GA http://
www.cdc.gov/obesity/downloads/PedNSSFactSheet.pdf.
Centers for Disease Control and Prevention, 2011.School health guidelines to promote
healthy eating and physical activity. MMWR.Atlanta GA, CDC & US Department of
Health and Human Services 60 (5).
S32 N.M. Wells et al. / Preventive Medicine 69 (2014) S27–S33

Christian, M., Evans, C., Conner, M., Ransley, J., Cade, J., 2012. Study protocol: can a school
gardening intervention improve children's diets? BMC Public Health 12 (304).
Christian,M.S.,Evans,C.E.,Hancock,N., Cade,J.E.,2014.Evaluation of the impact of a
school gardening intervention on children's fruit and vegetable intake. A randomised
controlled trial. Int. J. Behav. Nutr. Phys. Act. 11 (99).
Dietz, W., Gortmaker, S., 1985. Do we fatten our children at the television set? Obesity and
television viewing in children and adolescents. Pediatrics 75, 807–812.
DiNubile, N., 1993. Youth fitness — problems and solutions. Prev. Med. 22, 589–594.
Elder, G., 1998. The life course and human development. In: Damon, W., Lerner, R. (Eds.),
Handbook of Child PsychologyTheoretical Models of Human Development. vol. 1. J.
Wiley & Sons, Inc., NY.
Epstein, L., Rocco, A., Gordy, C., Dorn, J., 2000. Decreasing sedentary behaviors in treating
pediatric obesity. Arch. Pediatr. Adolesc. Med. 154, 220–226.
Evenson,K., Catellier,D., Gill, K., Ondrak,K., McMurray, R., 2008.Calibration of two
objective measures of physical activity for children. J. Sports Sci. 26 (14), 1557–1565.
Freedson,P., Miller, K., 2000.Objective monitoring of physical activity using motion
sensors and heart rate. Res. Q. Exerc. Sport 71 (2), 21–29.
Gortmaker,S.,Must, A., Sobol,A., Peterson,K., Colditz,G.,Dietz,W., 1996.Television
watching as a cause of increasing obesity among children in the United States,
1986–1990. Arch. Pediatr. Adolesc. Med. 150, 356–362.
Hancox, R., Milne, B., Poulton, R., 2004. Association between child and adolescent
television viewing and adult health: a longitudinal birth cohort study.Lancet 364
(9430), 257–262.
Hermann, J.R., Parker, S.P., Brown, B.J., Siewe, Y.J., Denney, B.A., Walker, S.J., 2006. After-
school gardening improves children's reported vegetable intake and physical activity.
J. Nutr. Educ. Behav. 38 (3), 201–202.
Hill, J., Peters, J., 1998. Environmental contributions to the obesity epidemic. Science 280,
1371–1374.
Jago, R., Baranowski, T., Baranowski, J., Cullen, K., Thompson, D., 2007. Social desirability is
associated with some physical activity, psychosocial variables and sedentary behavior
but not self-reported physical activity among adolescent males. Health Educ. Res. 22
(3), 438–449.
Kelkar,S.,Stella,S.,Boushey,C.,Okos,M., 2011.Developing novel 3D measurement
techniques and prediction method for food density determination.Procedia Food
Sci. 1, 483–491.
Klesges, R.,Klesges,L.,Eck, L.,Shelton,M., 1995.A longitudinal analysis of accelerated
weight gain in preschool children. Pediatrics 95 (1), 126–130.
Lineberger,S.,Zajicek,J., 2000.School gardens: can a hands-on teaching tool affect
students'attitudes and behaviors regarding fruit and vegetables? HortTechnol.10
(3), 593–597.
McKenzie,T., Sallis,J., Nader,P., et al.,1991.BEACHES: an observational system for
assessing children's eating and physical activity behaviors and associated events.J.
Appl. Behav. Anal. 2 (Fall).
Morris, J., Zidenberg-Cherr,S.,2002.Garden-enhanced nutrition curriculum improves
fourth-grade school children's knowledge of nutrition and preference for vegetables.
J. Am. Diet. Assoc. 102 (1), 91–93.
Morris, J., Neustadter, A., Zidenberg-Cherr, S., 2001. First-grade gardeners more likely to
taste vegetables. Calif. Agric. 55 (1), 43–46.
Morris, J., Briggs, M., Zidenberg-Cherr, S., 2002.Development and evaluation of a garden-
enhanced nutrition education curriculum for elementary school children.J. Child
Nutr. Manage. 2 (Fall 2002).
Myers, B.M., Wells, N.M., 2014. Children's physical activity while gardening: development
of a valid and reliable direct observation tool.Journal of Physical Activity & Health
(in press).
NASPE, 2004. Physical Activity for Children: A Statement of Guidelines for Children Ages
5–12. National Association for Sport and Physical Education, Reston, VA.
National Survey of Children's Health.Data query from the Child and Adolescent Health
MeasurementInitiative. Data Resource Center for Child and AdolescentHealth.
Available at: http://www.childhealthdata.org.
Nelson, M., Neumark-Stzainer, D., Hannan, P., Sirard, J., Story, M., 2006. Longitudinal and
secular trends in physicalactivity and sedentary behavior during adolescence.
Pediatrics 118, e1627–e1634.
Nocon, M., Hiemann, T., Müller-Riemenschneider, F., Thalau, F., Roll, S., Willich, S., 2008.
Association of physical activity with all-cause and cardiovascular mortality: a system-
atic review and meta-analysis. Eur. J. Prev. Cardiol. 15, 239–246.
Ogden, C., Carroll, M., Curtin, L., Lamb, M., Flegal, K., 2010. Prevalence of high body mass
index in US children and adolescents,2007–2008.J. Am. Med. Assoc.303 (3),
242–249.
Ozer, E., 2007. The effects of school gardens on students and schools: conceptualization
and considerations for maximizing healthy development.Health Educ.Behav.34,
846–863.
Pate, R., Freedson, P., Sallis, J., et al., 2002. Compliance with physical activity guidelines:
prevalence in a population of children and youth. Ann. Epidemiol. 12, 303–308.
Phelps, J., Hermann, J., Parker, S., Denney, B., 2010. Advantages of gardening as a form o
physical activity in an after-school program. J. Ext. 48 (6), 1–7.
Ridley, K., Ainsworth, B., Olds, T., 2008. Development of a compendium of energy expen-
ditures for youth. Int. J. Behav. Nutr. Phys. Act. 5 (45).
Robinson-O'Brien, R., Story, M.,Heim, S., 2009. Impact of garden-based youth nutrition
intervention programs: a review. J. Am. Diet. Assoc. 109, 273–280.
Rowe, P., Schuldheisz, J., vanderMars, H., 2003. Validation of SOFIT for measuring physic
activity of first- to eighth-grade students. Pediatr. Exerc. Sci. 9 (2), 136–149.
Sallis, J., Strikmiller,P., Harsha,D., et al., 1996.Validation of interviewer- and self-
administered physical activity checklists for fifth grade students.Med. Sci.Sports
Exerc. 28 (7), 840–851.
Sallis,J., Alcaraz,J., McKenzie,T., Hovell, M., 1999.Predictors of change in children's
physical activity over 20 months: variations by gender and level of adiposity.Am.
J. Prev. Med. 16 (3), 222–229.
Story, M., Nanney, M., Schwartz, M., 2009. School and obesity prevention: Creating schoo
environments and policies to promote healthy eating and physical activity. Milbank
Q. 87 (1), 71–100.
Treuth,M., Sherwood,N., Baranowski,T.,et al.,2004.Physical activity self-report and
accelerometry measures from the Girls Enrichment Multi-site Studies.Prev.Med.
38, S43–S49.
Troiano, R., Berrigan, D., Dodd, K., Masse, L., Tilert, T., McDowell, M., 2008. Physical activ
in the United States measured by accelerometer.Med. Sci. Sports Exerc.40 (1),
181–188.
Trost, S., Pate, R., Dowda, M., Saunders, R., Ward, D., Felton, G., 1996.Gender differences in
physical activity and determinants of physical activity in rural fifth grade children. J.
Sch. Health 66 (3), 145–150.
Trost,S.,Ward, D.,Moorehead,S.,Watson,P.,Riner,W., Burke,J., 1998.Validity of the
computer science and application (CSA) activity monitor in children. Med. Sci. Sports
Exerc. 30 (4), 629–633.
Trost, S., Pate, R., Sallis, J., et al., 2002. Age and gender differences in objectively measu
physical activity in youth. Med. Sci. Sports Exerc. 34 (2), 350–355.
Tuckerman,R.,2013.America's childhood obesity crisis and the role of schools.J. Sch.
Health 83 (3), 137–138.
Twiss, J., Dickinson, J., Duma, S., Kleinman, T., Paulsen, H., Rilveria, L., 2003. Community
gardens: lessons learned from California Healthy Cities and Communities.Am.
J. Public Health 93 (9),1435–1438.
Wells, N.M., Lekies, K.S., 2006. Nature and the life course: pathways from childhood na-
ture experiences to adult environmentalism. Environ. Behav. 35 (3), 311–330.
Wells, N.M., Myers, B.M., Henderson, C.R., 2014. Study Protocol: School gardens and phy
ical activity. Archives of Public Health (in press).
Wells, N.M., Ashdown, S., Davies, E., Cowett, F., Yang, Y., 2007. Environment, design and
obesity. Environ. Behav. 39 (1), 6–33.
Wethington, E., 2005. An overview of the life course perspective: implications for health
and nutrition. J. Nutr. Educ. Behav. 37 (3), 115–120.
Wheaton, B., Gotlib, I., 1997. Trajectories and turning points over the life course: concept
and themes.In: Gotlib, I., Wheaton,B. (Eds.), Stress and Adversity Over the Life
Course. Cambridge University Press, Cambridge, England, pp. 1–25.
Whitt-Glover, M., Taylor, W., Floyd, M., Yore, M., Yancey, A., Matthews, C., 2009. Dispar-
ities in physical activity and sedentary behaviors among US children adn adolescents
prevalence,correlates,and intervention implications.J. Public Health Policy 30
(Suppl. 1), S309–S334.
Woodcock, J., Franco, O., Osini, N., Roberts, I., 2011. Non-vigorous physical activity and a
cause mortality:systematic review and meta-analysis ofcohort studies.Int. J.
Epidemiol. 40 (1), 121–138.
World Health Organization, 2012. Social determinants of health and well-being among
young people. In: Currie, C., Zanotti, C., Morgan, A., et al. (Eds.), Health Behaviour in
School-aged Children (HBSC) Study: InternationalReport From the 2009/2010
Survey. Vol Health Policy for Children and Adolescents No. 6 http://www.euro.who.
int/__data/assets/pdf_file/0003/163857/Social-determinants-of-health-and-well-
being-among-young-people.pdf.
Yusuf, H., Croft, J., Giles, W., Anda, R., Casper, M., Caspersen, C., 1996. Leisure-time phys
activity among older adults — United States,1990.Arch. Intern. Med. 156 (2),
1321–1326.
S33N.M. Wells et al. / Preventive Medicine 69 (2014) S27–S33