Attention Restoration Theory: A Systematic Review of the Attention Restoration Potential of Exposure to Natural Environments

Abstract

Attention Restoration Theory (ART) suggests the ability to concentrate may be restored by exposure to natural environments. Although widely cited, it is unclear as to the quantity of empirical evidence that supports this. A systematic review regarding the impact of exposure to natural environments on attention was conducted. Seven electronic databases were searched. Studies were included if (1) they were natural experiments, randomized investigations, or recorded ?before and after? measurements; (2) compared natural and nonnatural/other settings; and (3) used objective measures of attention. Screening of articles for inclusion, data extraction, and quality appraisal were performed by one reviewer and checked by another. Where possible, random effects meta-analysis was used to pool effect sizes. Thirty-one studies were included. Meta-analyses provided some support for ART, with significant positive effects of exposure to natural environments for three measures (Digit Span Forward, Digit Span Backward, and Trail Making Test B). The remaining 10 meta-analyses did not show marked beneficial effects. Meta-analysis was limited by small numbers of investigations, small samples, heterogeneity in reporting of study quality indicators, and heterogeneity of outcomes. This review highlights the diversity of evidence around ART in terms of populations, study design, and outcomes. There is uncertainty regarding which aspects of attention may be affected by exposure to natural environments.

Full text

Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=uteb20
Download by: [51.7.36.64] Date: 27 September 2016, At: 01:57
Journal of Toxicology and Environmental Health, Part B
Critical Reviews
ISSN: 1093-7404 (Print) 1521-6950 (Online) Journal homepage: http://www.tandfonline.com/loi/uteb20
Attention Restoration Theory: A Systematic Review
of the Attention Restoration Potential of Exposure
to Natural Environments
Heather Ohly, Mathew P. White, Benedict W. Wheeler, Alison Bethel, Obioha
C. Ukoumunne, Vasilis Nikolaou & Ruth Garside
To cite this article: Heather Ohly, Mathew P. White, Benedict W. Wheeler, Alison Bethel,
Obioha C. Ukoumunne, Vasilis Nikolaou & Ruth Garside (2016): Attention Restoration Theory: A
Systematic Review of the Attention Restoration Potential of Exposure to Natural Environments,
Journal of Toxicology and Environmental Health, Part B, DOI: 10.1080/10937404.2016.1196155
To link to this article: http://dx.doi.org/10.1080/10937404.2016.1196155
Published with license by Taylor & Francis
Group, LLC© 2016 Heather Ohly, Mathew P.
White, Benedict W. Wheeler, Alison Bethel,
Obioha C. Ukoumunne, Vasilis Nikolaou, and
Ruth Garside
Published online: 26 Sep 2016.
Submit your article to this journal
View related articles
View Crossmark data

Attention Restoration Theory: A Systematic Review of the Attention Restoration
Potential of Exposure to Natural Environments
Heather Ohly
a
, Mathew P. White
a
, Benedict W. Wheeler
a
, Alison Bethel
b
, Obioha C. Ukoumunne
b
,
Vasilis Nikolaou
b
, and Ruth Garside
a
a
European Centre for Environment and Human Health, University of Exeter Medical School, Truro Campus, and Knowledge Spa, Royal
Cornwall Hospital, Truro, Cornwall, United Kingdom;
b
NIHR CLAHRC South West Peninsula, University of Exeter Medical School, South
Cloisters, St Luke’s Campus, Exeter, Devon, United Kingdom
ABSTRACT
Attention Restoration Theory (ART) suggests the ability to concentrate may be restored by
exposure to natural environments. Although widely cited, it is unclear as to the quantity of
empirical evidence that supports this. A systematic review regarding the impact of exposure to
natural environments on attention was conducted. Seven electronic databases were searched.
Studies were included if (1) they were natural experiments, randomized investigations, or
recorded “before and after”measurements; (2) compared natural and nonnatural/other settings;
and (3) used objective measures of attention. Screening of articles for inclusion, data extraction,
and quality appraisal were performed by one reviewer and checked by another. Where possible,
random effects meta-analysis was used to pool effect sizes. Thirty-one studies were included.
Meta-analyses provided some support for ART, with significant positive effects of exposure to
natural environments for three measures (Digit Span Forward, Digit Span Backward, and Trail
Making Test B). The remaining 10 meta-analyses did not show marked beneficial effects. Meta-
analysis was limited by small numbers of investigations, small samples, heterogeneity in reporting
of study quality indicators, and heterogeneity of outcomes. This review highlights the diversity of
evidence around ART in terms of populations, study design, and outcomes. There is uncertainty
regarding which aspects of attention may be affected by exposure to natural environments.
There is increasing practice and policy interest in
the potential for natural environments to provide
positive human health and well-being benefits.
Attention Restoration Theory (ART) is commonly
referenced to explain how this benefit might
accrue; however, it is unclear how strong the
empirical evidence is that exists for this proposed
mechanism. In cognitive psychology, the ability to
focus on a task that requires effort is known as
directed or voluntary attention (Kaplan and
Kaplan 1989). This ability is finite and may
become fatigued. Attention fatigue may occur
when there is a need to focus on a specific stimu-
lus or task with little or no intrinsically motiva-
tional draw, while suppressing distractions that
may be inherently more interesting, with an exam-
ple being filling in a tax return while your children
are playing in the yard (Kaplan 1995; Kaplan and
Berman 2010). Attentional fatigue is important,
not least because it is associated with poorer deci-
sion making and lower levels of self-control, which
in turn have been linked to a variety of health-
related issues such as obesity via increasingly
understood neural and behavioral pathways (Fan
and Jin 2013; Hare, Camerer, and Rangel 2009;
Vohs et al. 2008).
More than half the world’s population lives in
urban areas. From a psychological perspective,
urban lifestyles impose increasing demands on
our cognitive resources (Kaplan and Berman
2010). According to ART these enhanced demands
on directed attention may be linked to attention
fatigue (Kaplan 1995; Kaplan and Kaplan 1989).
The antidote, the theory claims, is to take time out
from attention-demanding tasks associated with
modern life, and spend time in natural
CONTACT Ruth Garside R.Garside@exeter.ac.uk European Centre for Environment and Human Health, University of Exeter Medical School, Truro
Campus, Knowledge Spa, Royal Cornwall Hospital, Truro, Cornwall, TR1 3HD, United Kingdom.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/UTEB.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B
http://dx.doi.org/10.1080/10937404.2016.1196155
Published withlicense by Taylor & Francis Group, LLC © 2016Heather Ohly, Mathew P. White, BenedictW. Wheeler, Alison Bethel, ObiohaC. Ukoumunne, Vasilis Nikolaou, and Ruth Garside
This is an Open Accessarticle. Non-commercialre-use, distribution,and reproduction in any medium, provided the original work is properly attributed, cited,and is not altered, transformed,
or built upon in any way, is permitted. The moral rights of the named author(s) have been asserted.

environments that demand less of our cognitive
resources and enable us to recover our attentional
capacities.
ART proposes that individuals benefit from the
chance to (1) “be away”from everyday stresses, (2)
experience expansive spaces and contexts (“extent”),
(3) engage in activities that are “compatible”with
our intrinsic motivations, and (4) critically experi-
ence stimuli that are “softly fascinating”(Kaplan
1995). This combination of factors encourages
“involuntary”or “indirect attention”and enables
our “voluntary”or “directed”attention capacities to
recover and restore (Kaplan 1995; Staats 2012).
Relaxing settings (such as places of worship) and
activities (such as sleep) may provide restorative
opportunities, but ART argues that nature may be
particularly useful because it has an “aesthetic advan-
tage”(Herzog et al. 2010;KaplanandBerman2010;
Kaplan and Kaplan 1989). It is suggested that send-
ing time in the natural world allows individuals the
opportunity for “reflection”and consideration of
unresolved issues (Herzog et al. 1997;Kaplanand
Berman 2010).
The original development of ART was largely
descriptive, based on observations of human–nature
interactions and analysis of qualitative data (Kaplan
and Frey Talbot 1983). Kaplan (1995) subsequently
linked ART more broadly to attention theory, for
instance, associating directed attention fatigue to
problems in selection and problem solving, inhibi-
tion of competing stimuli, and feelings of irritability.
More general psychological research provides beha-
vioral and neural evidence of a distinction between
“top-down”directed attention and “bottom-up”
involuntary attention (Fan et al. 2005). Specifically,
it has been postulated that directed attention is more
linked to higher order mental functions because of a
greater load on working memory produced by,
among other things, the need to suppress distracting
stimuli or alternative attentional cues. The simpler
attentional processes of alerting (becoming aware of
something; Jonides et al. 2008; Fan et al. 2002)and
orienting (taking actions to focus on a stimulus)
should be less affected by the trials of modern living
because they demand relatively few cognitive
resources. These processes are thus less likely to
need recovery in the same way as the executive
functions of attention, such as working memory,
which not only needs to hold and replay visual and
auditory stimuli, but may also have to manipulate
them according to rules stored in short-term mem-
ory (Jonides et al. 2008).
This discussion is important because it suggests
that only relatively demanding attentional tasks
should show improvements following exposure to
nature. An example is the Backwards Digit Span
test, which requires participants to both remember
and manipulate (reverse) a series of numbers. In con-
trast, tasks that require relatively few cognitive
resources should be less demanding and so less
affected by exposure to nature. An example would
be the Sustained Attention to Response Task (SART),
whichsimplyrequiresparticipantstoreacttothe
presentation of digits from 1 to 9 on a computer
screen, for example, by pressing the “space bar,”
except when the number 3 appears. The task thus
involves “inhibiting”themostfrequentresponse
(pressing a space bar) when a rare event occurs and
reflects the need to “sustain attention”rather than
drift into overgeneralizing a behavioral response.
Taylor, Kuo, and Sullivan (2002) explored
whether children living in apartments with greater
surrounding green space showed higher levels of
“self-discipline,”potentially due to fewer demands
on attention resources. Echoing this, Kaplan and
Berman (2010) and Baumeister, Vohs, and Tice
(2007) proposed that directed attention fatigue may
be linked to a loss in self-regulation, such as the
ability to resist temptation, suggesting these two
processes may share a common resource. If true,
relative depletion on tasks measuring self-regulation
and the ability to inhibit actions, rather than just
directed attention, might also be seen following
exposure to urban versus natural environments.
Kaplan and Berman (2010) presented a number of
studies to support these claims. However, this review
did not aim to systematically collect all relevant
papers in the field and thus may have missed those
that came to a different conclusion.
In contrast, a recent systematic review of the evi-
dence for general health-related benefits of exposure
to nature used a more systematic search strategy
(Bowler et al. 2010). Consequently, it included a
broader evidence base and also included a formal
appraisal of study quality in an attempt to weigh the
relative importance of different investigations. From
the eight studies using cognitive measures of assess-
ment, a meta-analysis of five, focusing on
2H. OHLY ET AL.

postexposure measures, supported ART. However,
the three better designed studies, measuring cognitive
performance both before and after exposure, demon-
stratednomarkedevidenceofimprovementsfollow-
ing exposure to nature.
A number of features of the Bowler et al. (2010)
systematic review are important. First, the investiga-
tions included for the cognitive aspect of the review
are different from those reviewed by Kaplan and
Berman (2010), suggesting the need for a systematic
review of all appropriate studies. Second, investiga-
tions using subjective measures of directed attention,
such as parental reports of children’s ability to con-
centrate, were included. A more robust test of the
theory would focus on those studies using objective
measures. Third, and partly because of the limited
number of investigations reviewed in both papers,
there was little opportunity to examine whether
some attention measures were more sensitive to
exposure to natural environments than others. All
of the measures may be considered as trying to
measure the same underlying construct: directed
attention capacity. As with any measurement tool,
each is subject to measurement error and might be
more or less effective. Moreover, since each measure
may be tapping into a slightly different aspect of
directed attention capacity (such as alerting, orien-
tating, or executive functions), considering data on
the effects of nature on different types of measure
may shed light on the precise mechanisms by which
nature may restore attentional processes (Jonides
et al. 2008).
The aim of the current systematic review was to
identify, select, appraise, and synthesize the evi-
dence for ART among studies that used experi-
mental and quasi-experimental approaches and
objective measures of attention. A larger sample
of studies was included (n= 31) than was the case
in either of the previous reviews. The investiga-
tions were also critically appraised and meta-ana-
lyzed where possible. Such meta-analysis is
consistent with Gifford’s(2014) call for better evi-
dence synthesis in environmental psychology.
Methods
This systematic review followed the general principles
p
ublishedbytheCentreforReviewsand
Dissemination (2009). The predefined protocol is
available on PROSPERO (Reference CRD420130
05008).
Review Question
What is the relative attention restoration potential
of natural settings compared to other settings?
Literature Search
A search strategy was devised by the research team,
led by our Information Specialist (Alison Bethel),
and captured concepts of attention restoration, cog-
nitive function, and natural versus other settings. No
suitable MeSH (Medical Subject Headings) terms
were identified. No methods filters were used. The
master search strategy (Table 1) was adapted and run
in the following electronic databases in July 2013:
PsycINFO, MEDLINE, and EMBASE (using OVID);
AMED, SPORT Discus, and Environment Complete
(using EBSCOHost); and Web of Knowledge (on
Thomson). Reference lists of included studies were
scrutinized for relevant investigations. Forward cita-
tion searches were undertaken on included studies.
All searches were conducted from 1989, when semi-
nal investigations on ART were published. Citation
searches were also performed in Web of Science
using these key references: Kaplan (1995)and
Kaplan and Kaplan (1989).
Inclusion Criteria
Studies were eligible for inclusion in the review if
they met the following criteria:
Population: Any.
Intervention and comparators: Studies reporting a
comparison of the effects of exposure to natural set-
tings and other, nonnatural settings. The definition of
“natural”included real settings (such as parks, forests,
wilderness areas) and virtual settings (images or
videos of similar settings). The definition of “nonna-
tural settings”included real settings (such as city
centers, residential areas, parking lots) and virtual
ones (images or videos of similar settings). Types of
engagement with these settings included active (such
as walking or running) and passive (such as looking at
the view from a window).
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 3

Outcomes: Objective measures of attention capa-
city, for example the Digit Span Forward or
Backward.
Study design: All experimental designs including
randomized controlled trials, quasi-experimental stu-
dies (nonrandomized controlled trials; randomized or
nonrandomized crossover trials) and natural experi-
ments. With the exception of natural experiments,
nonrandomized studies were included if they
recorded measures of attention before and after expo-
sure to nature/non nature settings. Investigations
were excluded if baseline measures were taken after
exposure had commenced because it was necessary to
establish baseline attentional abilities.
Other: Conference proceedings or dissertations
were included if there were sufficient data to assess
the risk of bias. No language restrictions were
applied.
Study Selection
All references identified through the search strategy
were uploaded into ENDNOTE (X7, Thomson
Reuters) and duplicates were removed. Reference
titles and abstracts, where available, were indepen-
dently double screened against the inclusion criteria
(by Heather Ohly and Ruth Garside/Alison Bethel).
Studies appearing to meet these were retrieved in full
text. One article was published in Chinese and this
was professionally translated. Full text screening was
completed independently by two reviewers (Heather
Ohly and Ruth Garside) using the same criteria.
Discrepancies were resolved through discussion
between reviewers.
Data Extraction
A standardized, piloted data extraction sheet was
developed in Excel to ensure consistency between
studies and reviewers. Data extracted for each study
included study design, sample characteristics, setting
characteristics (natural and nonnatural), type of
exposure and engagement, duration of exposure,
measures of attention, and duration of follow-up.
Authors were contacted to clarify or supply missing
data where necessary. Data were independently
extracted by one reviewer (Heather Ohly) and
checked by a second (Ruth Garside/Ben Wheeler/
Mathew White). Disagreements were resolved by
discussion, including the full team where necessary.
Quality Appraisal
The overall quality of the included studies was assessed
using a combination of resources and guidelines: qual-
ity indicators from the Centre for Reviews and
Dissemination (2009); critical appraisal checklists
from the Critical Appraisal Skills Programme (2013);
and quality assessment tool for quantitative studies
from the Effective Public Health Practice Project
(2013). These tools aim to assist reviewers in identify-
ingpotentialsourcesofbiasandmakeaconsidered
judgment how robust the evidence may be.
Table 1. Master search strategy as used in OVID Medline.
1 attention restorat*.tw.
2 (theory or hypothesis).tw.
3 (attention restorat* adj1 (theory or hypothesis)).tw.
4 natur*.tw.
5 outdoor*.tw.
6 green*.tw.
7 forest*.tw.
8 condition*.tw.
9 setting*.tw.
10 4 or 5 or 6 or 7 or 8 or 9
11 environ*.tw.
12 (environ* adj2 (natur* or outdoor* or green* or
forest* or condition* or setting*)).tw.
13 4 or 5 or 6 or 7
14 (setting* adj2 (natur* or outdoor* or green* or
forest*)).tw.
15 12 or 14
16 restorat*.tw.
17 15 and 16
18 attent*.tw.
19 cognitive function*.tw.
20 concentrat*.tw.
21 18 or 19 or 20
22 ((attent* or cognitive function* or concentrat*) adj3
((environ* adj2 (natur* or outdoor* or green* or
forest* or condition* or setting*)) or (setting* adj2
(natur* or outdoor* or green* or forest*)))).tw.
23 17 or 22
24 urban setting*.tw.
25 everyday setting*.tw.
26 garden*.tw.
27 mental fatigue*.tw.
28 4 or 5 or 6 or 7 or 24 or 25 or 26 or 27
29 16 or 18 or 19 or 20
30 ((restorat* or attent* or cognitive function* or
concentrat*) adj3 (natur* or outdoor* or green* or
forest* or urban setting* or everyday setting* or
garden* or mental fatigue*)).tw. (4872)
31 23 or 30
32 3 or 31
33 limit 32 to yr = “1989 -Current”
4H. OHLY ET AL.

A bespoke quality appraisal rating system was
developed using 20 standard indicators of robust
study conduct considered relevant in a review
context. The rationale and method of applying
some of these indicators is provided in Table 2.
Each quality indicator used the following ratings:
yes = 2; partial = 1; no = 0; unclear = 0. To
accommodate the fact that not all questions were
applicable for all study designs, for example, ques-
tions about appropriate randomization, all studies
were rated overall using a percentage score based
only on “applicable”criteria. Overall quality
assessment was given as low (0–33%), moderate
(34–66%), or high (67–100%).
Given changing standards relating to methods
reporting and limited word counts for many journals,
first authors of included papers were contacted where
an indicator had initially been scored “unclear”
(n= 24). They were asked to provide more informa-
tion regarding the study, and these “unclear”indica-
tors in particular, to aid a more informed assessment
of the papers. Of the 24 requests, 9 responses with
authors were received providing further details of
study design. Whether or not authors responded to
our request is recorded in Table 4 (shown later).
Data Synthesis
Random effects meta-analysis models were fitted
using standard methods, in which inverse variance
is used to weight individual study results, to pool
the effect estimates across investigations. The
meta-analyses compared attention outcomes at
follow-up between groups exposed to natural set-
tings (intervention) and groups exposed to non-
natural settings (control) (Sutton et al. 2000).
Summary data, the mean and standard deviation
of the outcome, and sample size in each group for
each study were used in the meta-analysis, with
pooled results reported as mean differences (inter-
vention minus control). Data were pooled from
investigations that used the same measures of
attention and use the same outcome (some atten-
tion measures have multiple associated outcomes).
Care was taken not to double count participants,
for example, in crossover trials when participants
completed the same walk twice, alone and with a
friend (Johansson, Hartig, and Staats 2011).
Onearticleincludedthreeseparatestudieseach
with independent samples (Berto 2005). The first
investigation compared two groups, one viewing nat-
ural images and the other viewing urban images. The
second study had only one group, viewing geometric
images, which were compared with data from the first
investigation. As the urban group from the first study
and the geometric group from the second were inde-
pendent, their results were combined before being
pooled with the third investigation in the meta-
analysis.
None of the studies that measured outcomes at
baseline reported using the correct method to adjust
for baseline imbalance, analysis of covariance
(ANCOVA) (Vickers and Altman 2001).
Consequently, all meta-analyses were carried out
using data at follow-up only. The main meta-ana-
lyses included all studies regardless of the level of
baseline imbalance. Investigations for which the
mean difference between the groups in outcome
score at baseline was greater than one-tenth of the
standard deviation in the control arm (i.e., “effect
size”of 0.1) were considered to have imbalance at
baseline. Cohen (1992), in his summary of effect
sizes in behavioral science studies, used a threshold
of 0.2 to define a small effect size. Sensitivity analyses
were also conducted in which only those investiga-
tions that had low levels of baseline imbalance were
included, to illustrate the potential impact of imbal-
anced experiments on results of the meta-analyses.
Data analysis was carried out using Stata 13 and
Review Manager (RevMan 5.2).
Where studies could not be meta-analyzed, as
they contained outcomes not shared with other
investigations, or where insufficient data were sup-
plied, results are described narratively.
Results
Search Results
Searches identified 10,979 unique records.
Twenty-four articles met the inclusion criteria
(Figure 1). Some of these articles reported results
from more than one investigation, so 31 separate
studies were included. Given the complexity of the
tables and figures in this section, references are
presented using only the first author’s name and
the date, to simplify data presentation.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 5

Table 2. Quality appraisal rating system for selected indicators.
Quality indicators Rationale Yes Partial No Unclear
Random allocation to
groups/condition order
Based on the scientific definition
of randomization for the purpose
of allocation to environment
groups.
Random allocation to groups (or
condition order in studies with
within-subject designs).
NA Nonrandom allocation to groups.
In natural experiments, where
participants had previously been
“randomly”allocated to housing,
this was not classed as random
allocation to environment groups
because it was not within the
experiment’s control.
Method of allocation not clearly
stated.
Randomization procedure
appropriate
Failure to properly randomize
has been shown to be associated
with increased levels of effect in
controlled clinical trials (Jüni,
Altman et al. 2001).
Appropriate method/tools of
randomization used.
NA Inappropriate methods/tools of
randomization used.
Method/tools of randomization
not clearly stated.
Groups balanced at
baseline (attention
scores)
Effect sizes were calculated to
show whether groups were
balanced or imbalanced at
baseline in terms of attention
scores (or in some cases study
authors provided this info).
An effect size <0.1 implies that
groups were balanced at
baseline (i.e., similar attention
scores). Therefore both groups
have similar potential to improve
their attention scores.
Multiple attention outcomes
reported, of which some were
balanced and others were
imbalanced at baseline.
An effect size >0.1 implies that
groups were imbalanced at
baseline (i.e., dissimilar attention
scores).
Data needed to calculate effect
size (baseline means and/or
standard deviations) not
reported.
Participants blind to
research question
If participants were aware of the
purpose of the study or
hypothesis, this may have been a
source of bias in favor of the
intervention group through
attracting a self-selected sample
of those for whom the question
had resonance or influencing
their experience of intervention
and control exposures.
Participants were not aware of
the research question, and/or
may have been given vague
information about the nature of
the study.
NA Participants were aware of the
research question and specifically
its focus on attention outcomes.
Awareness or blindness to
research question not clearly
stated.
(Continued )
6H. OHLY ET AL.

Table 2. (Continued).
Quality indicators Rationale Yes Partial No Unclear
Demonstrated need for
attention restoration
Normal or high attention scores
at baseline would limit the
potential for an exposure to
nature to improve attention (i.e.,
ceiling effect).
Study populations with
diagnosed medical conditions
that affect attention were
considered to have
demonstrated need for attention
restoration, for example, children
with ADHD.
Some studies used a mental
loading task to induce attention
fatigue prior to exposure to
natural/other settings. This was
only considered to have fully
demonstrated need for attention
restoration if measures of
attention were taken before and
after the task (i.e., T1—loading—
T2—exposure—T3), thereby
showing the extent of attention
depletion from baseline.
Studies that measured attention
after the mental loading task but
not before (i.e., loading—T1—
exposure—T2) were assigned
“partial”because, although this
sequence does not show the
effectiveness of the mental
loading task, it does show the
impact of the intervention.
Studies that measured attention
before the mental loading task
but not after (i.e., T1—loading—
exposure—T2) were assigned
“no”because this sequence does
not show the effectiveness of the
mental loading task and the
impact of the intervention on
depleted attention is unclear.
Study populations may or may
not have depleted levels of
attention (e.g., schizophrenia
patients) and this was not
clearly stated. Some study
authors may have considered it
plausible that subjects were
depleted but this was not
demonstrated.
Consistency of intervention
(within and between
groups)
Between-group consistency is
important because some of the
experimental studies involved
participants completing the
same activity in two or more
different settings, i.e., “active
control”rather than “placebo
control.”Any differences
between groups (other than the
setting) may induce performance
bias and lead to false results.
For studies that involved viewing
images of natural and other
settings, aspects assessed for
consistency included number of
images, duration of exposure,
location and conditions.
Consistency of intervention
clearly described. Taking natural
vs. urban walking studies as
example, within-group
consistency might include the
walking route, day of the week
and time of day, so that traffic
volume is similar for each
participant completing the urban
walk. Between-group consistency
might include the distance,
speed, number of walkers,
talking and other conditions of
the walks. Some studies used a
guide to ensure that a defined
route and pace were adhered to
and that conversation/
distractions were kept to a
minimum.
Some attempt to ensure
consistency. For example, some
natural vs. urban walking studies
attempted to ensure compliance
by sending participants out with
maps or global positioning
system (GPS); however, in some
cases it was not clear whether
compliance was actually
checked.
Lack of consistency within and/or
between groups. For example,
some interventions were home-
based and self-directed, where
participants chose from a range
of possible activities. In the
natural experiment studies,
participants may have lived in
their current home for varying
periods of time, and may have
had different levels of exposure
to nature outside the home.
Consistency of intervention not
clearly described.
(Continued )
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 7

Table 2. (Continued).
Quality indicators Rationale Yes Partial No Unclear
Outcome assessors blind to
group allocation
It has been repeatedly shown in
controlled trials that lack of
assessor blinding may lead to
biased assessment and inflated
effect sizes (Schulz et al. 1995;
Jüni, Altman et al. 2001).
However, this kind of detection
bias may be less problematic
where objective outcome
measures (such as attention
tests) are used.
Assessor blindness clearly stated. NA Assessors were clearly not blind
to group allocation, for example,
if one researcher led the group
activities and administered the
attention tests.
Assessor blindness not clearly
stated.
Consistency of data
collection
This includes consistency within
groups and between groups.
Consistency of data collection
clearly described. Within-group
consistency might include the
pre–post assessment interval,
particularly in studies where
participants were recruited at
different times across a wider
study period and self-managed
the intervention phase at home
(such as cancer patients or
pregnant women). Between-
group consistency might include
the location, conditions, and
timing of data collection. All of
these factors may influence the
outcome of the attention tests
and the extent to which any
changes over time (and
differences between groups) are
observed.
Some attempt to ensure
consistency. For example, there
may have been some variation in
the pre–post assessment interval,
but the testing was done in the
same room with all other factors
consistent.
Lack of consistency within and/or
between groups.
Consistency of data collection
not clearly described. Lack of
detail about data collection
methods generally.
Statistical analysis methods
appropriate for study
design
In order to demonstrate the
effect of nature on attention
capacity, statistical comparison
of the mean change in attention
capacity between groups was
expected (i.e., ANOVA or
ANCOVA time × environment).
For natural experiments, baseline
measures were not collected and
this type of analysis would not
be possible; therefore
comparison between groups
after the exposure (e.g., t-test)
was considered appropriate.
Appropriate statistical analysis
methods used.
Some doubt as to the
appropriateness of the statistical
analysis methods used, such as
correct choice of test but
incorrect use of data (e.g.,
collapsing three time points).
Inappropriate statistical analysis
methods used. However, we
recognise that accepted norms of
statistical analysis may have
changed since the time of
publication.
Statistical analysis methods not
clearly stated.
(Continued )
8H. OHLY ET AL.

Table 2. (Continued).
Quality indicators Rationale Yes Partial No Unclear
Sample representative of
target population
Studies were assessed for
selection bias (or sampling bias)
including self-selection,
prescreening or recruitment from
groups likely to show a better
response. If study participants
are not representative of the
target population, this may
introduce bias in favor of the
intervention group.
Sample considered
representative of target
population; recruitment strategy
included measures to reduce
selection bias.
NA Sample not considered
representative of target
population. For example,
students enrolled on
environmental courses such as
forestry would be expected to
show preference for natural
environments, which may
increase the likelihood of
experiencing restoration effects
in natural environments relative
to other types of environments.
Insufficient information about
the sample and how it was
recruited to make a judgment
about representativeness.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 9

Study Characteristics
The 31 included studies were from Europe, the
United States, and Asia, and varied in terms of
experimental design, sample characteristics, and
study duration, as well as the type and extent of
exposure to nature (Tables 3A to 3D).
Study designs included 16 randomized controlled
trials (RCT) from 12 articles (Berto 2005; Chen, Lai,
and Wu 2011; Cimprich and Ronis 2003;Hartigetal.
1996;2003; Hartig, Mang, and Evans 1991;
Laumann, Gärling, and Stormark 2003; Mayer et al.
2009; Perkins, Searight, and Ratwik 2011;Rich2008;
Stark 2003; van den Berg, Koole, and van der Wulp
2003); 7 randomized crossover trials from 6 articles
(Berman et al. 2008;2012; Bodin and Hartig 2003;
Johansson, Hartig, and Staats 2011;Shinetal.2011;
Taylor and Kuo 2009); 3 natural experiments (Kuo
2001; Taylor, Kuo, and Sullivan 2002; Tennessen and
Cimprich 1995); 3 nonrandomized controlled trials
(Berto 2005; Hartig, Mang, and Evans 1991;Wuetal.
2008); and 2 nonrandomized crossover trials
(Ottosson and Grahn 2005; van den Berg and van
den Berg 2011).
Study populations included children, “students”
and adults. Some samples were of individuals with
Records identified through database
searching (n=15443)
Records screened by title and abstract
(n=10979)
Full text articles screened for eligibility
(n=41):
Identified from electronic search (n=37)
Identified from hand searching (n=4)
Records excluded based on title and
abstract (n=10941)
Articles included (n=24)
Studies or experiments included
within articles (n=31)
Articles excluded based on full text (n=17)
with reasons:
No objective measure of attention (n=6)
No nature/other setting comparison (n=5)
No repeated measure of attention and
participants not randomised (n=3)
Dual publication of same data (n=2)
Baseline measures taken after exposure to
nature (n=1)
Duplicate records identified (n=4464)
Full text unobtainable (n=1)
Records included based on title and
abstract (n=38)
Figure 1. PRISMA flow chart.
10 H. OHLY ET AL.

Table 3A. Characteristics of included studies—RCT design; actual exposures (or mixed actual/virtual).
Author, year ()
a
Study
design
Country and
setting n
Sample characteristics: gender, mean age, population,
ethnicity, and s/e status if reported
Intervention characteristics: activity, setting (each
group) and duration of exposure Attention measures (objective)
Cimprich, 2003 RCT United States
University
medical center
185 100% female
53.8 years
Patients with newly diagnosed breast cancer (with
surgery as primary treatment plan)
86% White
Home-based, patient-led program of nature
activities
Control group logged relaxation time
120 min per week (nature activities)
Baseline—follow-up period approx. 36 days (pre-
and postsurgery)
Digit Span Backward
Digit Span Forward
Necker Cube Pattern Control
Trail Making Tests A
Trail Making Test B
Hartig, 2003 RCT United States
University and
local area
112 50% male
20.8 years
Students
Sitting, natural view; then walking, natural
(nature reserve)
Sitting, no view; then walking, urban (city streets)
1 hour (10 min passive; 50 min active)
Necker Cube Pattern Control
Search and Memory Test
Hartig, 1991 (2) RCT United States
University and
local area
102 50% male
20 years
Students
Walking, natural (regional park)
Walking, urban (city centre)
Reading magazines, comfortable laboratory
setting
40 min
Proofreading Task
Mayer, 2009 (1) RCT United States
University
76 29% male
Mean age not reported
Students
Walking, natural (woods/creek)
Walking, urban (downtown)
10 min
Memory Loaded Search Task (similar to
Search and Memory Task)
Mayer, 2009 (2) RCT United States
University
92 30% male
Mean age not reported
Students
Walking, natural (woods)
Watching video, natural (woods)
Watching video, urban (busy streets)
10 min
Memory Loaded Search Task (similar to
Search and Memory Task)
Perkins, 2011 RCT United States
University
26 27% male
Age range 19–24 years
Mean age not reported
Students
Walking, natural (woods)
Walking, urban (residential/business)
Walking, urban (parking lot)
20 min
Digit Span Backward
Digit Span Forward
Logical Memory
Stark, 2003 Cluster
RCT
United States
Prenatal classes
57 100% female
29.1 years
Pregnant women in the third trimester
94.7% White
Outdoor “restorative”activities
Alternative session on the discomfort of
pregnancy
120 min per week (outdoor activities)
Baseline—follow up period varied 13–64 days
Category Matching
Digit Span Backward
Digit Span Forward
Errors Scale
Trail Making Tests A
Trail Making Test B
Note. Outcome measures listed in alphabetical order, not the order in which they were administered.
a
Experiment number in parentheses; each experiment has distinct sample; s/e = socioeconomic.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 11

Table 3B. Characteristics of included studies—RCT design; virtual exposures.
Author, year ()
a
Study design Country and setting n
Sample
characteristics:
gender, mean
age, population
Intervention characteristics: activity, setting
(each group) and duration of exposure Attention measures (objective)
Berto, 2005 (1) RCT Italy
University
32 50% male
23 years
Students
Viewing images, natural
Viewing images, urban
25 images × 15 sec each
Sustained Attention to Response Test
Berto, 2005 (3) RCT Italy
University
32 50% male
22 years
Students
Viewing images, natural
Viewing images, urban
25 images × duration of their choice
Sustained Attention to Response Test
Chen, 2011 (1) RCT China
Senior secondary
school
48 42% male
Mean age not
reported
Students
Viewing images, natural
Viewing images, city
Viewing images, urban nightscape
Viewing images, sports
10 images × 15 sec each
Colored number pictures
Hartig, 1996 (1) RCT Sweden
University and high
schools
102 38% male
21.4 years
Students
Watching simulated walk, natural (trees)
Watching simulated walk, urban (city)
No simulated walk (control)
80 slides × 10 sec each (13.5 min)
Search and Memory Task
Hartig, 1996 (2) RCT Sweden
University
18 50% male
27.4 years
Students
Watching simulated walk, natural (trees)
Watching simulated walk, urban (city)
80 slides manually (12 min)
Search and Memory Task
Laumann, 2003 RCT Norway
University
28 100% female
Age range
18–24 years
Mean age not
reported
Students
Watching video, natural (island waterside)
Watching video, urban (city streets)
80 scenes × 15 sec each
Posner’s Attention Orienting Task (note:
no raw data and comparisons focus on
different types of stimuli; therefore not
reported in Table 5)
Rich, 2008 (1) RCT United States
University
145 17% male
Mean age not
reported
Students
Looking at view, natural (forest)
Looking at view, urban (buildings)
No view
1 min
Vigilance Task
Stroop Colour-Word Test
Rich, 2008 (2) RCT United States
University
36 42% male
Age range
18–21 years
Mean age not
reported
Students
Reading magazines, room with plants
Reading magazines, room with other objects
10 min
Digit Span Backward
van den Berg, 2003 RCT The Netherlands
University
114 32% male (after
exclusions for
n= 106)
21.9 years
Students
Watching simulated walk, natural (forest with or
without water)
Watching simulated walk, urban (city with or
without water)
7 min
D2 mental concentration test
Note. Outcome measures listed in alphabetical order, not the order in which they were administered.
a
Experiment number in parentheses; each experiment has distinct sample; s/e = socioeconomic
12 H. OHLY ET AL.

Table 3C. Characteristics of included studies—Other designs; actual exposures.
Author, year ()
a
Study design Country and setting n
Sample characteristics: gender, mean age,
population, ethnicity, and s/e status if reported
Intervention characteristics: activity, setting (each group),
duration of exposure, and washout period (crossover only)
Attention measures
(objective)
Berman, 2008
(1)
Randomized
crossover trial
United States
University
38 39% male
22.6 years
Students
Walking, natural (park)
Walking, urban (downtown)
50–55 min
Two walks, 1 week apart
Digit Span Backward
Berman, 2008
(2)
Randomized
crossover trial
United States
University
12 33% male
24.3 years
Students
Viewing images, natural (Nova Scotia)
Viewing images, urban (downtown)
50 images in 10 min
Two sessions, 1 week apart
Attention Network Test
Digit Span Backward
Berman, 2012 Randomized
crossover trial
United States
University and local
area
20 40% male
26 years
Adults diagnosed with major depressive
disorder (MDD)
Walking, natural (park)
Walking, urban (downtown)
50–55 min
Two walks, 1 week apart
Digit Span Backward
Bodin, 2003 Randomized
crossover trial
Sweden
Running club
12 50% male
39.7 years (males)
37.0 years (females)
Runners
Running, natural (park)
Running, urban (city streets)
60 min
Two runs, 1 week apart
Combined Digit Span
Backward and Forward
Symbol Digit
Modalities Test
Johansson, 2011 Randomized
crossover trial
Sweden
University
20 50% male
24.2 years (males)
22.4 years (females)
Students
Walking, natural (park)
Walking, urban (streets)
40 min
Four walks, 1 week apart (natural with friend; urban with
friend; natural alone; urban alone)
Symbol Substitution
Test
Shin, 2011 Randomized
crossover trial
South Korea
University
60 58% male
23.3 years
Students
Walking, natural (park)
Walking, urban (city streets)
50–55 min
Two walks, 1 week apart
Trail Making Test B
Taylor, 2009 Randomized
crossover trial
United States 25 88% male (after exclusions for n = 17)
9.2 years
Children diagnosed with ADHD
Walking, natural (urban park)
Walking, urban (downtown)
Walking, urban (neighborhood)
20 min
Three walks, 1 week apart
Digit Span Backward
Stroop Colour-Word
Test
Symbol Digit
Modalities Test
Vigilance Task
(Note: Only DSB
reported)
Hartig, 1991 (1) Nonrandomized
controlled trial
United States
Trailheads and local
clubs
68 62% male
35.9 years (G1)
29.2 years (G2)
31.6 years (G3)
Experienced backpackers
Wilderness backpacking vacation
Nonwilderness vacation
No vacation
4–7 days (vacation groups)
Proofreading Task
Wu, 2008 (1) Nonrandomized
controlled trial
Taiwan
Public psychiatric
centre
23 72% male
Mean age not reported
Schizophrenia patients
Horticulture activities (indoors and outdoors)
Regular hospital activities like watching movies, singing,
drawing, cooking (indoors)
90 min per week × 15 classes
Chu’s Attention Test
(Continued )
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 13

Table 3C. (Continued).
Author, year ()
a
Study design Country and setting n
Sample characteristics: gender, mean age,
population, ethnicity, and s/e status if reported
Intervention characteristics: activity, setting (each group),
duration of exposure, and washout period (crossover only)
Attention measures
(objective)
Ottosson, 2005 Nonrandomized
crossover trial
Sweden
Residential care
home
17 87% female (after exclusions for n= 15)
86 years
Elderly residents of the care home
Leisure time outside (terrace and gardens)
Leisure time inside (own room and shared space)
1h
Two sessions, 14 days apart
Digit Span Backward
Digit Span Forward
Necker Cube Pattern
Control
Symbol Digit
Modalities Test
van den Berg,
2011
Nonrandomized
crossover trial
The Netherlands
Two care farms
12 83% boys
12.8 years
Children diagnosed with ADHD
Building a cabin, natural (woodland)
Walking “expedition,”urban (quiet neighborhood)
1 hour
Two activities, 1 day apart
Test of Everyday
Attention for Children
Kuo, 2001 Natural
experiment
United States
Inner city community
145 100% female
34 years
Heads of household; African American residents
of inner city housing development
Living near high levels of vegetation (“green”)
Living near low levels of vegetation (“barren”)
Digit Span Backward
Taylor, 2002 Natural
experiment
United States
Inner city community
169 54% boys
9.6 years
Children; African American residents of inner
city housing development
High level of near-home nature (“green”view from
apartment)
Low level of near-home nature (“barren”view from
apartment)
At least 1 year living in current location
Alphabet Backward
Category Matching
Delayed Gratification
Task
Digit Span Backward
Matching Familiar
Figures Test
Necker Cube Pattern
Control
Symbol Digit
Modalities Test
Stroop Colour-Word
Test
Tennessen, 1995 Natural
experiment
United States
University
72 42% male
20 years
Students
All natural view from dormitory
Mostly natural view from dormitory
Mostly built view from dormitory
All built view from dormitory
Digit Span Backward
Digit Span Forward
Necker Cube Pattern
Control
Symbol Digit
Modalities Test
Note. Outcome measures listed in alphabetical order, not the order in which they were administered.
a
Experiment number in parentheses; each experiment has distinct sample; s/e = socioeconomic.
14 H. OHLY ET AL.

psychological conditions such as attention deficit
hyperactivity disorder (ADHD), depression, or schi-
zophrenia (Berman et al. 2012; Taylor and Kuo 2009;
vandenBergandvandenBerg2011;Wuetal.2008),
lower income groups such as African American resi-
dents of an inner city housing development (Kuo
2001; Taylor, Kuo, and Sullivan 2002); or partici-
pants experiencing other circumstances that, it is
suggested, might influence their attention capacity,
such as pregnancy or breast cancer (Cimprich and
Ronis 2003;Stark2003). The remainder of the sam-
ples had “normal”cognitive function.
Study duration and intensity varied, from less
than an hour of exposure in controlled conditions,
to multiple days or weeks of exposure in real-life
settings. The longest exposures were seen in the
natural experiments, where participants had been
exposed to their surroundings for months or years
(Kuo 2001; Taylor, Kuo, and Sullivan 2002;
Tennessen and Cimprich 1995). Investigations used
various cognitive tests.
Some studies involved actual exposure to nature:
either through active engagement (walking, running,
or other activities)(Berman et al. 2008;2012;Bodin
and Hartig 2003; Cimprich and Ronis 2003;Hartig
et al. 1996; Hartig, Mang, and Evans 1991;
Johansson, Hartig, and Staats 2011; Mayer et al.
2009; Perkins, Searight, and Ratwik 2011;Shinetal.
2011;Stark2003; Taylor and Kuo 2009; van den Berg
and van den Berg 2011;Wuetal.2008)orpassive
engagement (resting outside or living with a view)
(Kuo 2001; Ottosson and Grahn 2005;Rich2008;
Taylor, Kuo, and Sullivan 2002; Tennessen and
Cimprich 1995). Other investigations involved vir-
tual exposure to nature; this was exclusively passive
engagement (watching video or viewing images)
(Berman et al. 2008;Berto2005; Chen, Lai, and Wu
2011;Hartigetal.1996; Laumann, Gärling, and
Stormark 2003;Rich2008; van den Berg, Koole,
and van der Wulp 2003). Most studies had a com-
parison group that involved equivalent exposure to a
nonnatural (urban or indoor) setting. Four studies
used a placebo control setting involving relaxation
time or usual activities (Cimprich and Ronis 2003;
Hartig et al. 1996; Hartig, Mang, and Evans 1991;
Rich 2008).
Quality scores varied from 22.5 to 75% (Table 4).
Seven of the 31 included studies were classified as
“high”quality (scoring 67–100%), while 22 were clas-
sified as “moderate”(scoring 34–66%) and 2 were
classified as “low”quality (scoring 0–33%). The qual-
ity indicators reflected overall experimental quality
and also how well the study answered our review
question, which may not have been the main focus
of the individual studies. Indicators that few investi-
gations reported clearly were power calculation, ran-
domization procedure, whether participants were
blind to the research question, demonstrated need
for attention restoration, whether outcome assessors
were blind to group allocation, and whether the sam-
ple was representative of the target population.
Evidence for Effects of Nature on Attention
Capacity
Some investigations reported results for sub-
groups, such as men/women (Bodin and Hartig
2003), task/no task (Hartig et al. 1996;2003), and
alone/with friend (Johansson, Hartig, and Staats
2011). One study compared one natural group
(walking in woods) and two nonnatural groups
(walking in neighborhood and walking in parking
lot), but only presented attention scores for the
sample as a whole (Perkins, Searight, and Ratwik
2011).
Table 3D. Characteristics of included studies—Other designs; virtual exposures
Author, year ()
a
Study design
Country and
setting n
Sample characteristics: gender,
mean age, population,
ethnicity, and s/e status if
reported
Intervention characteristics: activity,
setting (each group), duration of
exposure, and washout period
(crossover only)
Attention
measures
(objective)
Berto, 2005 (2) Nonrandomized
controlled trial
Italy
University
64 50% male
23 years
Students
Viewing images, natural
Viewing images, urban
Viewing geometric images
Note: Two groups from Berto 2005 (1)
25 images × 15 sec each
Sustained
Attention
to
Response
Test
Note. Outcome measures listed in alphabetical order, not the order in which they were administered.
a
Experiment number in parentheses; each experiment has distinct sample; s/e = socioeconomic.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 15

Table 4. Indicators of quality of included studies.
RCT design; real exposures RCT design; virtual exposures
Quality indicators
Cimprich,
2003
Hartig,
2003
Hartig,
1991 (2)
Mayer,
2009 (1)
Mayer,
2009 (2)
Perkins,
2011
Stark
2003
Berto
2005 (1)
Berto
2005 (3)
Chen,
2011 (1)
Hartig,
1996 (1)
Hartig,
1996 (2)
Laumann,
2003
Rich
2008
(1)
Rich
2008
(2)
van den
Berg, 2003
Study design
Power calculation reported No No No No No No No No No No No No No No No No
Inclusion criteria reported Yes Yes Pa. No No No Yes No No No Yes No No No No No
Individual level allocation Yes Yes Yes Yes Pa. Yes No Yes Yes Yes No Yes Yes Yes Yes Yes
Random allocation to groups/
condition order
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Randomization procedure
appropriate
Un. Yes Yes Un. Un. Yes Un. Un. Un. Yes Un. Un. Un. Un. Un. Un.
Confounders
Groups similar
(sociodemographic)*
Pa. Yes Yes Un. Un. Yes Yes Un. Un. Un. Yes Yes Un. Un. Un. Un.
Groups balanced at baseline
(attention scores)
No Yes Pa. Un. Un. No Pa. Pa. Pa. No No No Un. Un. Un. Un.
Participants blind to research
question
No Yes Yes Un. Yes Yes No Un. Un. Un. Yes Un. Un. Un. Un. Un.
Intervention integrity
Demonstrated need for attention
restoration
No Pa.Un. No No No No No No Yes Un. Un. Pa. No No No
Clear description of intervention
and control
Yes Yes Yes Yes Yes Pa. Yes Yes Yes Yes Yes Yes Yes Pa. Yes Yes
Consistency of intervention
(within and between groups)
No Yes Pa. Pa. Yes Pa. No Yes Yes Yes Yes Pa. Yes Un. Yes Yes
Data collection methods
Outcome assessors blind to
group allocation
Un. No No Un. Un. No Un. Un. Un. Un. No No Un. No No Un.
Baseline attention measures
taken before the exposure
Yes Yes Yes No No Yes Yes Yes Yes Yes No No Yes No No No
Consistency of data collection Pa. Yes Pa. Un. Un. Yes Yes Yes Yes Yes Yes Yes Yes Un. Yes Yes
Analyses
All attention outcomes reported
(means and SD/SE)
Yes Pa.
†
Pa. Yes Yes No No
†
Yes Yes Yes Yes Yes No No No No
All participants accounted for
(i.e., losses/exclusions)
Yes Yes Yes Un. Yes Yes Yes Yes Yes Yes No Yes Yes Un. Un. Yes
ITT analysis conducted (all data
included after allocation)
No No No Un. No No Yes Yes Yes No No Yes No Un. Un. No
Individual level analysis Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Statistical analysis methods
appropriate for study design
Yes Yes No Yes Yes Pa. Yes No No Pa. Yes Un. Yes Yes Yes Yes
External validity
Sample representative of target
population
No Un. Un. Un. Un. Un. No Un. Un. Yes Un. Un. Un. Un. Un. Un.
(Continued )
16 H. OHLY ET AL.

Table 4. (Continued).
RCT design; real exposures RCT design; virtual exposures
Quality indicators
Cimprich,
2003
Hartig,
2003
Hartig,
1991 (2)
Mayer,
2009 (1)
Mayer,
2009 (2)
Perkins,
2011
Stark
2003
Berto
2005 (1)
Berto
2005 (3)
Chen,
2011 (1)
Hartig,
1996 (1)
Hartig,
1996 (2)
Laumann,
2003
Rich
2008
(1)
Rich
2008
(2)
van den
Berg, 2003
Overall quality score
Total number of points (out of
possible 40)
20 30 23 13 17 21 21 21 21 23 20 19 19 9 14 16
Quality rating as percent 50 75 57.5 32.5 42.5 52.5 52.5 52.5 52.5 57.5 50 47.5 47.5 22.5 35 40
Responded to query about
“uncertain”ratings
No Yes Yes No No Yes No No No No Yes Yes No No No No
Other designs; real exposures Other; virtual
Quality indicators
Berman,
2008 (1)
Berman,
2012
Bodin,
2003
Johansson,
2011
Shin,
2011
Taylor,
2009
Hartig,
1991 (1)
Wu,
2008
(1)
Ottosson,
2005
van den
Berg, 2011
Kuo
2001
Taylor,
2002
Tennessen,
1995
Berman,
2008 (2)
Berto
2005
(2)
Study design
Power calculation reported No Yes No No No No No No No No No No No No No
Inclusion criteria reported No Yes Yes Yes No Yes Yes Un. Yes Yes Yes Yes Yes No No
Individual level allocation Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes
Random allocation to groups/condition
order
Yes Yes Yes Yes Yes Yes No No No No NA NA NA Yes No
Randomization procedure appropriate Yes Yes Yes Un. Un. Yes NA NA NA NA NA NA NA Yes. NA
Confounders
Groups similar (sociodemographic)* Yes Yes Yes Yes Un. Yes Yes Yes Un. Yes Un. Un. Pa. Yes Un.
Groups balanced at baseline (attention
scores)
Yes No Pa. No Yes Un. Yes Un. Un. Un. NA NA NA Pa. Pa.
Participants blind to research question Yes Yes Yes Yes Un. Yes Un. Un. Un. Yes Yes Yes Un. Yes Un.
Intervention integrity
Demonstrated need for attention
restoration
No No Un. No No Yes No Un. No Yes NA NA NA No No
Clear description of intervention and
control
Yes Yes Yes Yes Yes Pa. Yes Yes Yes Yes Yes Yes Yes Yes Yes
Consistency of intervention (within and
between groups)
Yes Yes Yes Pa. Pa. Yes No Yes Pa. No No No No Yes Yes
Data collection methods
Outcome assessors blind to group
allocation
Un. Un. No No Un. Yes No Un. Un. No Un. Yes Un. Un. Un.
Baseline attention measures taken before
the exposure
Yes Yes Yes Yes Yes No Yes Yes Yes Yes NA NA NA Yes Yes
Consistency of data collection Yes Yes Yes Yes Yes Yes Pa. Un. Pa. Un. Pa. Pa. Pa. Yes Yes
Analyses
All attention outcomes reported (means
and SD/SE)
Yes Yes Yes Yes Yes No Yes No No No Yes No Yes Yes Yes
All participants accounted for (i.e., losses/
exclusions)
Yes Yes Yes Yes Un. Yes Un. Yes Yes Un. Un. Un. Yes Yes Yes
(Continued )
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 17

Table 4. (Continued).
Other designs; real exposures Other; virtual
Quality indicators
Berman,
2008 (1)
Berman,
2012
Bodin,
2003
Johansson,
2011
Shin,
2011
Taylor,
2009
Hartig,
1991 (1)
Wu,
2008
(1)
Ottosson,
2005
van den
Berg, 2011
Kuo
2001
Taylor,
2002
Tennessen,
1995
Berman,
2008 (2)
Berto
2005
(2)
ITT analysis conducted (all data included
after allocation)
Yes
†
No Yes Yes Un. No Un. No No Un. Un. Un. Yes Yes Yes
Individual level analysis Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Statistical analysis methods appropriate
for study design
Yes Yes Pa. Yes No Yes Yes Un. Yes Un. Yes Yes Yes Yes No
External validity
Sample representative of target
population
Un. Un. Un. Un. No Un. Un. Un. No Un. Yes Yes Un. Un. Un.
Overall quality score
Total number of points (out of possible
40, or fewer where criteria are NA)
30 30 30 27 17 27 21/38 14/38 16/38 14/38 17/30 17/30 18/30 29 19/38
Quality rating as % 75 75 75 67.5 42.5 67.5 55.3 36.8 42.1 36.8 56.7 56.7 60 72.5 50
Responded to query about “uncertain”
ratings
Yes Yes Yes Yes No Yes Yes No No No No Yes No Yes No
Note. Yes = 2; Partial (Pa.) = 1; No = 0; Unclear (Un.) = 0; NA = criterion not applicable to this study design. Results in boldface reflect changes to “unclear”categorization after further information was
provided by authors.
*Significance level for differences between groups was p< .05.
†
Additional data were provided by the author for our meta-analyses. In some cases additional information was provided by study authors for the quality appraisal, and therefore this table would not be
entirely replicable by other reviewers. Any changes made after consultations are highlighted in boldface.
18 H. OHLY ET AL.

Various cognitive tests were used to measure
attention capacity. Evidence for the effects of nat-
ure on attention capacity is presented below for
each attention measure separately although, as
noted in the preceding text, it is not clear which
of these measures is the most appropriate in the
context of ART. Where data could be pooled,
forest plots were produced and are reproduced
here if three or more studies were included.
Meta-analyses pooling only two studies are
described in the text, and forest plots are presented
in the Supporting Information. Data relating to
two outcome measures could not be pooled and
these data are reported narratively (Proof Reading
Task and Symbol Substitution Test). Measures of
attention unique to a single study are described
narratively. Full details of all study outcomes are
presented in Tables 5A to 5L.
Digit Span
Participants are presented with a series of digits
(e.g., “8, 3, 4”) and need to immediately repeat
them back. If this is done successfully, they are
given a longer list of digits (e.g., “9, 2, 4, 0”). The
length of the list is increased until the participant
fails to accurately recall a list of that length on two
subsequent occasions. The length of the longest list
a subject can remember is that subject’s digit span.
In the Digit Span Forward (DSF), participants
have to recall the digits in the same order they
are presented. In the Digit Span Backward (DSB),
participants have to reverse the order with which
they are presented.
Digit Span Forward (DSF)
Five studies reported DSF scores (Table 5A)
(Cimprich and Ronis 2003; Ottosson and Grahn
2005;Perkins,Searight,andRatwik2011;Stark2003;
Tennessen and Cimprich 1995). The meta-analysis
included data from three experiments, none of
which were balanced at baseline (Cimprich and
Ronis 2003;Stark2003; Tennessen and Cimprich
1995). The natural exposure groups performed signif-
icantly better than controls (Figure 2).
Digit Span Backward (DSB)
Eleven studies, reported in 10 articles, reported
DSB scores (Table 5B) (Berman et al. 2008,2012;
Cimprich and Ronis 2003; Kuo 2001; Ottosson
and Grahn 2005; Perkins, Searight, and Ratwik
2011; Rich 2008; Stark 2003; Taylor and Kuo
2009; Tennessen and Cimprich 1995). The meta-
analysis included data from eight investigations
reported in seven articles(Berman et al. 2008;
2012; Cimprich and Ronis 2003; Kuo 2001; Stark
2003; Taylor and Kuo 2009; Tennessen and
Cimprich 1995). The natural exposure groups per-
formed significantly better than controls
(Figure 3). Sensitivity analysis, including only the
two studies that were balanced at baseline (Berman
et al. 2008,2012), also indicated better DSB per-
formance in the intervention groups (natural).
Combined Digit Span Backward/Forward (DSB/
DSF)
One study reported combined DSB/DSF scores,
obtained by summing the two scores (Bodin and
Hartig 2003)(Table 5C). The meta-analysis
included data from two independent groups
(men and women) from one experiment, which
were not balanced at baseline (Bodin and Hartig
2003). There was little evidence of a marked dif-
ference between groups at follow-up (Figure 4).
Proofreading Task (PR)
The participant is asked to find simple misspell-
ings, typographical errors, and grammatical errors
in a five-page passage of text. The score is percent
of errors detected from the total present at the
point in the text reached after 10 min. Higher
scores indicate better performance. Due to the
length of the task, it measures attentional vigi-
lance, a key aspect of which is the inhibition of
distractions.
One article, containing two studies, reported
proofreading scores (Table 5D)(Hartig,Mang,
and Evans 1991). The first study found that
proofreading scores improved in the natural
group (wilderness backpacking) and declined in
the two nonnatural groups (nonwilderness vaca-
tion and no vacation); the difference in change
between groups was not statistically significant.
The second experiment reported proofreading
scores at follow-up only, which were signifi-
cantly higher in the natural group (nature
walk)comparedtothetwononnaturalgroups
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 19

Table 5A. Results of included studies—Digit Span Forward (DSF): Length of sequence repeated correctly.
Study design
Author, year, and
sample (n)
Natural settings Nonnatural settings
Difference between
groups at follow-up
Difference in change
between groups
Baseline mean
(SD) and n
Follow-up mean
(SD) and n
Baseline mean
(SD) and n
Follow-up mean
(SD) and n
Baseline mean
(SD) and n
Follow-up mean
(SD) and n
RCTs Cimprich, 2003
n= 185
6.71 (SE 0.14) 6.98 (SE 0.15) 6.54 (SE 0.13) 6.53 (SE 0.16) –– p= .04 NR
Perkins, 2011
n=26
Woods
NR
Woods
NR
Neighborhood
NR
Neighborhood
NR
Parking lot
NR
Parking lot
NR
NR No significant
difference
Stark, 2003
n=57
6.8 (1.2) 7.1 (1.4) 6.6 (1.0) 6.7 (1.1) –– NR NR
Nonrandomized
crossover trial
Ottosson, 2005
n=17
NR NR NR NR –– NR p< .001
Study design
Author, year,
and sample (n)
Natural settings Nonnatural settings
Difference between groups at follow-up
Follow-up mean
(SD) and n
Follow-up mean
(SD) and n
Follow-up mean
(SD) and n
Follow-up mean
(SD) and n
Natural experiment Tennessen, 1995
n=72
All natural
7.50 (1.08)
Mostly natural
7.40 (1.35)
Mostly built
7.38 (0.90)
All built
6.96 (1.11)
No significant difference
Note. SD, standard deviation; SE, standard error; NR, not reported. There were no baseline values for this study; therefore, no pre–post exposure (baseline to follow-up) change values.
20 H. OHLY ET AL.

Table 5B. Results of included studies—Digit Span Backward (DSB): Length of sequence reversed correctly.
Study design
Author, year ()a
and sample (n)
Natural settings Nonnatural settings
Difference between groups
at follow-up
Difference in change
between groups
Baseline mean
(SD) and n
Follow-up
mean
(SD) and n
Baseline
mean
(SD) and n
Follow-up
mean
(SD) and n
Baseline
mean
(SD) and
n
Follow-up
mean
(SD) and n
RCTs Cimprich, 2003
n= 185
4.99 (SE 0.15) 5.20 (SE 0.14) 4.51 (SE 0.13) 4.58 (SE 0.14) —— p= .002 NR
Perkins, 2011
n=26
NR NR NR NR —— NR No significant difference
Rich, 2008 (2)
n=36
NR NR NR NR ——F(1, 35) = 2.43; p=ns NR
Stark, 2003
n=57
5.1 (1.3) 5.4 (1.6) 4.7 (1.1) 5.0 (1.6) —— NR NR
Randomized
crossover trials
Berman, 2008 (1)
n=38
7.90 (SE 0.37) 9.4 (SE 0.41) 7.90 (SE 0.30) 8.4 (SE 0.33) —— NR F(1, 36) = 6.055; Prep = 0.95
Berman, 2008 (2)
n=12
7.92 (SE 0.96) 9.33 (SE 0.86) 7.83 (SE 1.04) 8.83 (SE 0.90) —— NR F(1, 10) = 0.486; Prep = 0.68
Berman, 2012
n=20
7.42 (3.00) 8.63 (2.87) 8.26 (2.51) 7.84 (2.24) —— NR F(1, 18) = 20.5; p< .001
Taylor, 2009
n=25
NR Urban park
4.41 (1.18)
NR Downtown
3.82 (1.07)
NR Neighborhood
3.71 (1.21)
F(2,16) = 4.72; p= .02 NR
Nonrandomized
crossover trial
Ottosson, 2005
n=17
NR NR NR NR —— NR p< .001
a
Experiment number in parentheses; each experiment has distinct sample.
Study design Author, year, and sample (n)
Natural settings Nonnatural settings
Difference between groups at follow-up
Follow-up mean
(SD) and n
Follow-up mean
(SD) and n
Follow-up mean
(SD) and n
Follow-up mean
(SD) and n
Natural experiments Kuo, 2001
n= 145
Green
4.96 (1.0)
—Barren
4.64 (1.2)
—t=−1.74; p= .05
Tennessen, 1995
n=72
All natural
5.60 (1.35)
Mostly natural
5.11 (1.62)
Mostly built
5.04 (1.18)
All built
5.00 (1.13)
No significant difference
Note. There were no baseline values for these studies; therefore, no pre–post exposure (baseline to follow-up) change values.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 21

Table 5C. Results of included studies—Combined Digit Span Backward and Forward (DSB/DSF): Sum of scores.
Study design
Author, year, and
sample (n)
Subgroups within
sample
Natural settings Nonnatural settings
Difference between
groups at follow-up
Difference in change
between groups
Baseline mean
(SD) and n
Follow-up mean
(SD) and n
Baseline mean
(SD) and n
Follow-up mean
(SD) and n
Randomized crossover
trial
Bodin, 2003
n=12
Men 10.83 (2.32) 11.92 (2.31) 11.92 (2.76) 12.17 (2.89) NR Men and women
F(1, 10) = 0.92; p= .36Women 11.67 (3.82) 10.58 (3.68) 10.67 ± 2.89 11.25 (3.22) NR
Table 5D. Results of included studies—Proofreading Task (PR): Percent of errors detected.
Study design
Author, year
()a,
and sample
(n)
Natural settings Nonnatural settings Difference
between
groups at
follow-up
Difference in
change
between
groups
Baseline mean (SD)
andn
Follow-up mean (SD)
and n
Baseline mean (SD)
and n
Follow-up mean (SD)
and n
Baseline mean (SD)
and n
Follow-up mean (SD)
and n
RCT Hartig, 1991
(2)
n= 102
Natural
NR
Natural
56.5
Urban
NR
Urban
49.5
Relaxation
NR
Relaxation
48.0
t(94) = 2.45;
p< .01
NR
Nonrandomized
controlled
trial
Hartig, 1991
(1)
n=68
Wilderness
51.12
Wilderness
53.92
Nonwilderness
56.77
Nonwilderness
53.78
No vacation
60.16
No vacation
56.52
NR F(2, 65) = 2.39;
p< .09
a
Experiment number in parentheses; each experiment has distinct sample.
22 H. OHLY ET AL.

Table 5E. Results of included studies—Necker Cube Pattern Control (NCPC) (three separate outcomes).
Study design
Author, year,
and
sample (n)
Attention outcome;
subgroups within sample
Natural settings Nonnatural settings
Difference between groups at
follow-up
Difference in change
between groups
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
RCTs Cimprich,
2003
n = 185
Percent reduction in reversals (30 sec
period)
10.29 (SE
6.61)
19.48 (SE
4.15)
17.87 (SE
7.73)
13.87 (SE
7.31)
No significant difference NR
Hartig, 2003
n = 112
Number of reversals (average 2 x 30 s
periods);
Task
T1 Baseline
4.37 (1.92)
T2 During
walk
3.96 (1.93)
T3 Follow-
up
3.98 (2.44)
T1 3.87
(1.85)
T2 4.76
(2.19)
T3 4.67
(2.63)
NR Task and no task
T1-T2
F(1,98) = 13.15; p < 0.001
T1-T3
F(1,100) = 5.59; p = 0.02
Number of reversals (average 2 × 30 sec
periods):
No task
T1 Baseline
4.17 (1.92)
T2 During
walk
4.06 (1.90)
T3 Follow-
up
3.88 (2.21)
T1 3.70
(1.35)
T2 4.43
(1.67)
T3 4.09
(1.72)
NR
Nonrandomized
crossover trial
Ottosson,
2005
n=17
Difference between baseline and controlled
(60 s period)
NR NR NR NR NR p < 0.05
Note. T1, T2, and T3 used where attention measures were taken at three time points (as described in nature group cells). Underscore: inverse outcome, therefore lower score indicates better performance.
Table 5E. Continued—natural experiment.
Study design
Author, year,
and
sample (n) Attention outcome
Natural settings Nonnatural settings
Difference between groups at
follow-up
Follow-up mean (SD)
and n
Follow-up mean (SD)
and n
Follow-up mean (SD)
and n
Follow-up mean (SD)
and n
Natural
experiment
Tennessen,
1995
n=72
Percent reduction in reversals
(30 sec period)
All natural
60.53 (13.61)
Mostly natural
64.18 (22.48)
Mostly built
35.38 (38.14)
All built
36.35 (39.78)
No significant difference
Note. There were no baseline values for this study, and therefore no pre–post exposure (baseline to follow-up) change values.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 23

Table 5F. Search and Memory Task/Memory Loaded Search Task (SMT).
Study
design
Author,
year ()a
sample (n)
Attention outcome;
subgroups within sample
Natural settings Nonnatural settings
Difference between groups
at follow-up
Baselinemean
(SD) and n
Follow-up
mean
(SD) and n
Baseline
mean
(SD) and
n
Follow-up
mean
(SD) and n
Baseline
mean
(SD) and
n
Follow-up mean
(SD) and n
Baseline
mean
(SD) and
n
Follow-up
mean
(SD) and n
RCTs Hartig,
1996 (1)
n= 102
Percent error (accuracy)
Task
NR Natural slides
Block 1
34.0 (18.3)
Block 2
43.4 (16.0)
––NR Urban slides
Block 1
33.0 (15.4)
Block 2
44.7 (14.6)
NR No slides
Block 1
33.6 (15.1)
Block 2
42.5 (15.5)
No significant
difference
% error (accuracy)
No task
NR Natural slides
Block 1
40.6 (18.6)
Block 2
37.5 (17.2)
––NR Urban slides
Block 1
33.5 (16.3)
Block 2
39.9 (18.7)
NR No slides
Block 1
33.4 (14.5)
Block 2
47.1 (18.0)
No significant
difference
Number of letters (speed)
Task
NR Natural slides
Block 1
611.4 (156.9)
Block 2
679.5 (196.5)
––NR Urban slides
Block 1
688.2 (135.5)
Block 2
772.5 (218.2)
NR No slides
Block 1
697.6
(252.0)
Block 2
774.7
(222.2)
No significant
difference
Number of letters (speed)
No task
NR Natural slides
Block 1
547.2 (155.4)
Block 2
573.7 (153.3)
––NR Urban slides
Block 1
591.4 (161.5)
Block 2
652.4 (200.2)
NR No slides
Block 1
589.1
(180.4)
Block 2
622.1
(207.2)
No significant
difference
Hartig,
1996 (2)
n=18
Percent error (accuracy) NR Block 1
33.18 (14.20)
Block 2
40.21 (13.56)
––NR Block 1
32.30 (15.55)
Block 2
42.97 (15.55)
––No significant
difference
Number of letters (speed) NR Block 1
620.90
(139.76)
Block 2
656.99
(145.58)
––NR Block 1
658.39 (177.77)
Block 2
718.22 (243.01)
––No significant
difference
Hartig,
2003
n= 112
Accuracy x speed NR NR ––NR NR –– NR
Mayer,
2009 (1)
n=76
Number of errors per line
completed
NR 1.18 (0.47) ––NR 1.60 (0.74) ––F(1,67) = 8.49;
p < 0.01
Mayer,
2009 (2)
n=92
Number of errors per line
completed
NR Actual nature
1.00 (0.47)
NR Virtual nature
1.15 (0.59)
NR Virtual urban
1.41 (0.49)
––F(1, 56) = 1.31;
p = 0.28
Note. There were no baseline values for these studies, therefore no pre–post exposure (baseline to follow-up) change values. The exception was Hartig (2003), which did measure SMT at baseline but did
not report this datum; there was no significant difference pre–post exposure.
a
Experiment number in parentheses; each experiment has distinct sample. Underscore: inverse outcome, therefore lower score indicates better performance.
24 H. OHLY ET AL.

Table 5G. Results of included studies—Sustained Attention to Response Test (SART) (four separate outcomes).
Study design
Author, year ()
a, and
sample (n) Attention outcome
Natural settings Nonnatural settings
Difference between groups at
follow-up
Difference in change
between groups
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
RCTs Berto, 2005 (1)
n=32
Reaction time (msec) Natural
313.71
(38.36)
Natural
267.38
(73.78)
Urban
319.59
(70.98)
Urban
299.61
(41.43)
——t(30) = −2.19; p= .03 NR
Berto, 2005 (1)
n=32
Number of correct
responses
Natural
11.68
(5.28)
Natural
13.62
(5.37)
Urban
13.25
(5.09)
Urban
13.00 (5.4)
——t(30) = 0.32; p= .74 NR
Berto, 2005 (1)
n=32
Number of incorrect
responses
Natural
1.81 (3.83)
Natural
2.06 (4.79)
Urban
3.25 (6.22)
Urban
1.62 (4.96)
——t(30) = 0.25; p= .80 NR
Berto, 2005 (1)
n=32
d-prime or sensitivity Natural
1.40 (0.71)
Natural
1.86 (0.89)
Urban
1.97 (0.96)
Urban
2.00 (0.95)
——t(30) = −0.40; p= .68 NR
Berto, 2005 (3)
n=32
Reaction time (msec) Natural
311.27
(35.6)
Natural
302.22
(32.09)
Urban
306.21
(78.79)
Urban
297.91
(52.98)
——No significant difference NR
Berto, 2005 (3)
n=32
Number of correct
responses
Natural
14.81
(5.55)
Natural
17.12
(4.09)
Urban
12.5 (6.36)
Urban
14.56
(5.95)
——No significant difference NR
Berto, 2005 (3)
n=32
Number of incorrect
responses
Natural
1.62 (2.5)
Natural
0.81 (1.42)
Urban
1.68 (2.62)
Urban
0.75 (1.52)
——No significant difference NR
Berto, 2005 (3)
n=32
d-prime or sensitivity Natural
2.12 (1.08)
Natural
2.47 (1.04)
Urban
1.79 (1.24)
Urban
2.03 (0.93)
——No significant difference NR
Nonrandomized
controlled trial
Berto, 2005 (2)
n=64
Reaction time (msec) No new data; used data
from Berto 2005 (1)
No new data; used data
from Berto 2005 (1)
Geometric
310.24
(43.89)
Geometric
289.46
(55.2)
F(2, 61) = 5.57; p= .00;
rsq = 0.01
NR
Berto, 2005 (2)
n=64
Number of correct
responses
No new data; used data
from Berto 2005 (1)
No new data; used data
from Berto 2005 (1)
Geometric
13.90
(5.15)
Geometric
13.59 (5.66)
F(2, 61) = 4.60; p= .01;
rsq = 0.10
NR
Berto, 2005 (2)
n=64
Number of incorrect
responses
No new data; used data
from Berto 2005 (1)
No new data; used data
from Berto 2005 (1)
Geometric
1.5 (2.59)
Geometric
1.71 (3.12)
No significant difference NR
Berto, 2005 (2)
n=64
d-prime or sensitivity No new data; used data
from Berto 2005 (1)
No new data; used data
from Berto 2005 (1)
Geometric
1.98 (0.92)
Geometric
1.95 (1.01)
No significant difference NR
a
Experiment number in (); each experiment has distinct sample. Inverse outcome therefore lower score indicates better performance.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 25

Table 5H. Results of included studies—Symbol Digit Modalities Test (SDMT): Number of correct symbol/digit pairs (90 sec period).
Study design
Author,
Year
Sample
(n)
Subgroups
within
sample
Natural settings Nonnatural settings
Difference between
groups at follow-up
Difference in change
between groups
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
Randomized
crossover trial
Bodin,
2003
n=12
Men 51.17 (7.17) 48.83 (4.22) 51.50 (3.67) 49.17 (5.38) NR Men and women
F(1, 10) = 0.02; p= .90
Women 56.33 (11.33) 52.50 (10.04) 56.00 (9.36) 52.83 (9.56) NR
Nonrandomized
crossover trial
Ottosson,
2005
n=17
Not
applicable
NR NR NR NR NR p< .001
Table 5H. Continued—natural experiment.
Study design
Author, year, and
sample (n)
Natural settings Nonnatural settings
Difference between
groups at follow-up
Follow-up mean
(SD) and
n
Follow-up mean
(SD) and
n
Follow-up mean
(SD) and
n
Follow-up mean
(SD) and
n
Natural
experiment
Tennessen, 1995
n=72
All natural
74.00 (9.87)
Mostly natural
64.40 (10.76)
Mostly built
61.50 (10.36)
All built
63.08 (8.74)
F(3, 67) = 3.78;
p< .05
Note. There were no baseline values for this study, and therefore no pre–post exposure (baseline to follow-up) change values.
Table 5I. Results of included studies—Symbol Substitution Test (SST); number of correct assignments (60 sec period).
Study design
Author,
year, and
sample (n)
Subgroups
within sample
Natural settings Nonnatural settings
Difference between
groups at follow-up
Difference in change
between groups
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
Randomized
crossover
trial
Johansson,
2011
n=20
Alone 38.65 (5.28) 37.85 (5.10) 37.85 (5.21) 37.80 (4.87) NR Alone and with friend
F(1, 18) = 5.99;
p= .025, n
2
= 0.250
With friend 40.00 (6.78) 36.35 (5.09) 37.70 (4.78) 36.85 (4.79) NR
Table 5J. Results of included studies—Trail Making Test A (TMTA); completion time (seconds).
Study
design
Author, year
sample (n)
Natural settings Nonnatural settings
Difference between
groups at follow-up
Difference in change
between groups
Baseline
mean
(SD) and n
Follow-up
mean
(SD) and n
Baseline
mean
(SD) and
n
Follow-up
mean
(SD) and n
RCTs Cimprich, 2003 n= 185 30.23 (SE 1.3) 25.65 (SE 1.03) 37.08 (SE 2.3) 31.21 (SE 1.34) p= .001 NR
Stark, 2003 n= 57 22.06 (7.09) 19.84 (4.79) 21.67 (5.46) 19.60 (5.33) NR NR
Note. Inverse outcome, therefore lower score indicates better performance.
Table 5K. Results of included studies—Trail Making Test B (TMTB); completion time (seconds).
Study design
Author,
year, and
sample
(n)
Subgroups
within sample
Natural settings Nonnatural settings
Difference
between groups at
follow-up
Difference in
change between
groups
Baseline
mean (SD)
and
n
Follow-up
mean (SD) and
n
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
RCTs Cimprich,
2003
n= 185
Not applicable 65.01 (SE 3.4) 56.51 (SE 2.92) 77.09 (SE 4.9) 68.01 (SE 4.06) p= .02 NR
Stark,
2003
n=57
Not applicable 46.55 (11.92) 38.89 (10.75) 48.15 (10.42) 40.40 (9.30) NR NR
Randomized
crossover
trial
Shin,
2011
n=60
First walk 37.03 (6.81) 29.48 (6.82) 37.03 (6.81) 39.24 (21.23) NR NR
Second walk 37.04 (6.90) 29.45 (6.72) 37.04 (6.90) 39.17 (21.23) NR NR
Note. Inverse outcome, therefore lower score indicates better performance.
26 H. OHLY ET AL.

Table 5L. Results of included studies—Other measures of attention (each used only in one study).
Study
design
Author, year ()*,
and sample (n) Attention test and outcome
Natural settings Nonnatural settings
Difference between
groups at follow-up
Difference in change
between groups
Baseline
mean (SD)
and
n
Follow-
up
mean
(SD) and
n
Baseline
mean (SD)
and
n
Follow-up
mean (SD)
and
n
Baseline
mean
(SD) and
n
Follow-
up
mean
(SD) and
n
RCTs Chen, 2011 (1)
n=48
Coloured Number Pictures:
reaction time (msec)
Nature
T1 Baseline
587.0 (9.5)
T2 After
loading task
777.7 (6.9)
Nature
T3
Follow-
up
557.2
(7.7)
City
T1 652.8
(11.2)
T2 836.9
(12.3)
City
T3 749.8
(10.2)
Urban
night
T1 615.9
(10.3)
T2 709.9
(11.1)
Urban
night
T3 534.8
(7.3)
p< .001 T1–T2–T3
F= 8.27; p< .001
Sports
T1 600.5
(10.4)
T2 808.8
(11.0)
Sports
T3 605.1
(9.4)
Cimprich, 2003
n= 185
Total attention: combined scores (DSB,
DSF, TMA, TMB, NCPC)
0.56 (SE
0.34)
1.75 (SE
0.27)
−0.56 (SE
0.37)
0.04 (SE 0.34) —— p< .001 p< .001
Perkins, 2011
n=26
Logical Memory: number
of segments correctly recalled
Woods
NR
Woods
NR
Neighborhood
NR
Neighborhood
NR
Parking
lot
NR
Parking
lot
NR
NR No significant
difference
Rich, 2008 (1)
n= 145
Stroop Colour-Word: % errors NR Forest
view
1.05
(0.976)
NR Buildings
3.53 (0.937)
NR No view
1.43
(0.950)
F(2, 129) = 2.61;
p= .08
NR
Vigilance task: outcome unclear NR NR NR NR NR NR No significant difference NR
Stark, 2003
n=57
Category Matching: number of correct
pairs circled
41.48 (9.13) NR 43.82 (9.97) NR —— NR NR
Errors Scale:
sum of uncorrected errors (TMA, TMB,
CM)
2.6 (2.8) NR 3.3 (2.3) NR —— NR Standardized
beta = −0.68;
t=−2.98; p= .001
Study design
Author, year ()*,
and sample (n) Attention test and outcome
Natural settings Nonnatural settings
Difference between
groups at follow-up
Difference in change
between groups
Baseline
mean
(SD) and
n
Follow-up
mean (SD) and
n
Baseline
mean
(SD) and
n
Follow-up
mean (SD) and
n
RCT van den Berg, 2003
n= 114
D2 mental concentration test: concentration
index
NR 399.84 NR 379.30 F(1, 102) = 2.79; p= .098 NR
D2 mental concentration test: speed index NR 421.18 NR 399.92 F(1, 102) = 2.74; p= .10 NR
D2 mental concentration test: Accuracy
index
NR NR NR NR No significant difference
(p= .86)
NR
(Continued )
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 27

Table 5L. (Continued).
Study design
Author, year ()*,
and sample (n) Attention test and outcome
Natural settings Nonnatural settings
Difference between
groups at follow-up
Difference in change
between groups
Baseline
mean
(SD) and
n
Follow-up
mean (SD) and
n
Baseline
mean
(SD) and
n
Follow-up
mean (SD) and
n
Randomized
crossover trial
Berman, 2008 (2)
n=12
Attention Network Test: Executive: Response
time (msec)
86 (SE
11.30)
67 (SE 8.45) 81 (SE
15.50)
93 (SE 17.96) NR F(1, 10) = 17.089,
Prep = 0.99
Attention Network Test: Orienting: Response
time (msec)
47 (SE
6.46)
55 (SE 7.33) 46 (SE
10.01)
43 (SE 4.73) NR NR
Attention Network Test: Alerting Response
time (msec)
32 (SE
6.86)
31 (SE 5.23) 36 (SE
6.52)
46 (SE 5.63) NR NR
Nonrandomized
crossover trial
van den Berg, 2011
n=12
Test of Everyday Attention: Difference in
time between two tests (sec)
NR 3.20 (1.39) NR 3.82 (2.47) eta squared = 0.21;
p= .07
NR
Nonrandomized
controlled trial
Wu, 2008
n=23
Chu’s Attention Test: number of correct
answers
45.78 52.89 53.11 47.66 No significant difference NR
Chu’s Attention Test: number of errors 7.18 6.38 6.29 2.83 Significant difference (no
pvalue)
NR
Note. Experiment number in parentheses; each experiment has distinct sample. T1, T2, and T3 used where attention measures were taken at three time points (as described in nature group cells).
Underscore: inverse outcome, and therefore lower score indicates better performance.
28 H. OHLY ET AL.

Table 5L. Continued—natural experiment.
Study design
Author,
year, and
sample (n) Attention test and outcome
Subgroups
within sample
Natural
settings
Nonnatural
settings
Difference between
groups at follow-up
Follow-up
mean (SD)
and
n
Follow-up
mean (SD)
and
n
Natural
experiment
Taylor,
2002
n= 169
Concentration: combined z-scores
(SDMT, DSB, AB, NCPC)
Girls Green
view
NR
Barren view
NR
F(1, 76) = 10.9;
B= 0.23; p< .01
Boys Green
view
NR
Barren view
NR
No significant difference
Impulse inhibition: combined z-scores
(MFF, SCW, CM)
Girls Green
view
NR
Barren view
NR
F(1, 76) = 3.8;
B= 0.17; p= .05
Boys Green
view
NR
Barren view
NR
F(?, ?) = 2.3;
B= 0.12; p= .13
Delay of gratification score Girls Green
view
NR
Barren view
NR
F(1, 76) = 12.7; B= 0.42;
p< .001
Boys Green
view
NR
Barren view
NR
No significant difference
Self-discipline score: average of the
above three scores
Girls Green
view
NR
Barren view
NR
F(1, 76) = 19.4; B= 0.27;
p< .001
Boys Green
view
NR
Barren view
NR
No significant difference
Note. There were no baseline values for this study, therefore no pre–post exposure (baseline to follow-up) change values,
Study or Subgroup
Cimprich 2003
Stark 2003
Tennessen 1995
Total (95% CI)
Heterogeneity: Tau² = 0.00; Chi² = 0.22, df = 2 (P = 0.90); I² = 0%
Test for overall effect: Z = 2.50 (P = 0.01)
Mean
6.98
7.1
7.45
SD
1.37
1.4
1.16
Total
83
29
20
132
Mean
6.53
6.7
7.17
SD
1.38
1.1
1
Total
74
25
52
151
Weight
50.6%
21.1%
28.3%
100.0%
IV, Random, 95% CI
0.45 [0.02, 0.88]
0.40 [-0.27, 1.07]
0.28 [-0.30, 0.86]
0.39 [0.08, 0.70]
Experimental Control Mean Di fference Mean Difference
IV, Random, 95% CI
-2-1012
Favours [control] Favours [experimental]
Figure 2. Forest plot showing meta-analysis for DSF.
Study or Subgroup
Berman 2012
Berman(1) 2008
Berman(2) 2008
Cimprich 2003
Kuo 2001
Stark 2003
Taylor 2009
Tennessen 1995
Total (95% CI)
Heterogeneity: Tau² = 0.00; Chi² = 2.80, df = 7 (P = 0.90); I² = 0%
Test for overall effect: Z = 4.47 (P < 0.00001)
Mean
8.63
9.4
9.33
5.2
4.96
5.4
4.41
5.35
SD
2.87
2.49
2.98
1.27
1
1.6
1.18
1.47
Total
19
37
12
83
76
29
17
20
293
Mean
7.84
8.4
8.83
4.58
4.64
5
3.76
5.02
SD
2.24
2
3.12
1.2
1.2
1.6
1.13
1.15
Total
19
37
12
74
69
25
34
52
322
Weight
1.7%
4.4%
0.8%
31.4%
35.9%
6.4%
10.2%
9.1%
100.0%
IV, Random, 95% CI
0.79 [-0.85, 2.43]
1.00 [-0.03, 2.03]
0.50 [-1.94, 2.94]
0.62 [0.23, 1.01]
0.32 [-0.04, 0.68]
0.40 [-0.46, 1.26]
0.65 [-0.03, 1.33]
0.33 [-0.39, 1.05]
0.49 [0.28, 0.71]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-2 -1 0 1 2
Favours [control] Favours [experimental]
Figure 3. Forest plot showing meta-analysis for DSB.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 29

(urban walk, and passive relaxation indoors).
This was an RCT but it was not clear whether
the groups were balanced at baseline. Lack of
data precluded meta-analysis.
Necker Cube Pattern Control (NCPC)
An image of a three-dimensional cube is presented,
which may be perceived from alternative perspec-
tives resulting from reversal of the foreground and
background. The participant needs to indicate the
number of times the cube appears to “flip”or change
perspectives in a short, timed period. The test is
performed twice: first with the participant just obser-
ving the cube (baseline) and the second time
attempting to hold one perspective (controlled).
The score may be calculated in various ways, includ-
ing percent reduction in reversals between the two
tests (higher scores indicate better performance), the
difference in reversals between the two tests (higher
scores are better), or the number of reversals in the
controlled test (lower scores are better).
Four studies reported NCPC scores (Table 5E)
(Cimprich and Ronis 2003;Hartigetal.2003;
Ottosson and Grahn 2005; Tennessen and Cimprich
1995). Meta-analysis was conducted separately for
different calculations of the NCPC score.
For percentage reduction in reversals, two stu-
dies were included in the meta-analysis. Study
populations were either not balanced at baseline
(Cimprich and Ronis 2003) or balance was
unknown (Tennessen and Cimprich 1995). There
was little evidence of a difference between groups
at follow-up. Wide confidence intervals indicate
substantial uncertainty in the pooled effect esti-
mate (Figure 5).
For number of reversals, data from two inde-
pendent groups, reported in the same study, were
included in the meta-analysis (Hartig et al. 2003).
One group completed a mental loading task prior
to the environmental exposure; the other group
did not. The groups were not balanced at baseline.
Pooled results indicated little evidence of a signifi-
cant difference between groups at follow-up
(Figure 6).
Study or Subgroup
Cimprich 2003
Tennessen 1995
Total (95% CI)
Heterogeneity: Tau² = 160.88; Chi² = 3.82, df = 1 (P = 0.05); I² = 74%
Test for overall effect: Z = 1.60 (P = 0.11)
Mean
19.48
62.35
SD
37.8
18.1
Total
83
20
103
Mean
13.87
35.86
SD
62.88
37.6
Total
74
52
126
Weight
46.9%
53.1%
100.0%
IV, Random, 95% CI
5.61 [-10.86, 22.08]
26.49 [13.55, 39.43]
16.70 [-3.72, 37.12]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-50 -25 025 50
Favours [control] Favours [experimental]
Figure 5. Forest plot showing meta-analysis for NCPC: percentage reduction in reversals.
Study or Subgroup
Hartig(1) 2003
Hartig(2) 2003
Total (95% CI)
Heterogeneity: Tau² = 0.00; Chi² = 0.29, df = 1 (P = 0.59); I² = 0%
Test for overall effect: Z = 0.91 (P = 0.36)
Mean
3.98
3.88
SD
2.44
2.21
Total
26
26
52
Mean
4.67
4.09
SD
2.63
1.72
Total
27
27
54
Weight
38.0%
62.0%
100.0%
IV, Random, 95% CI
-0.69 [-2.06, 0.68]
-0.21 [-1.28, 0.86]
-0.39 [-1.23, 0.45]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-2 -1 0 1 2
Favours [experimental] Favours [control]
Figure 6. Forest plot showing meta-analysis for NCPC: number of reversals.
Study or Subgroup
Bodin(1) 2003
Bodin(2) 2003
Total (95% CI)
Heterogeneity: Tau² = 0.00; Chi² = 0.06, df = 1 (P = 0.81); I² = 0%
Test for overall effect: Z = 0.47 (P = 0.64)
Mean
11.92
10.58
SD
2.31
3.68
Total
12
12
24
Mean
12.17
11.25
SD
2.89
3.22
Total
12
12
24
Weight
63.6%
36.4%
100.0%
IV, Random, 95% CI
-0.25 [-2.34, 1.84]
-0.67 [-3.44, 2.10]
-0.40 [-2.07, 1.27]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-4 -2 0 2 4
Favours [control] Favours [experimental ]
Figure 4. Forest plot showing meta-analysis for combined DSF/DSB.
30 H. OHLY ET AL.

Search and Memory Task (SMT)
The participant memorizes 5 target letters, subse-
quently searches lines of 59 letters, and crosses off
any target letters found. Subjects need to complete as
many lines, and find as many target letters, as possi-
ble in 10 min. The number of target letters per line,
and in total, varies between studies. The score may
be calculated in various ways, including the percent
missed targets (accuracy: lower scores indicate better
performance), the number of errors per line (accu-
racy: lower scores are better), the number of letters
searched in a given time (speed: higher scores are
better), or accuracy multiplied by speed (higher
scores are better). This task combines elements of
vigilance and working memory capacity. As targets
are essentially random, it might be argued that this is
a more demanding task than proofreading.
Five studies, reported in three articles, reported
SMT scores (Table 5F) (Hartig et al. 1996;2003;
Mayer et al. 2009). Meta-analysis was conducted
separately for different methods of calculating the
SMT score.
For percentage error (accuracy), data from three
studies, reported in the same article, were pooled
(Hartig et al. 1996). Baseline data were not reported,
so balance between groups is unknown. The first study
reported data for two independent groups, one of
which completed a mental loading task prior to envir-
onmental exposure. These data appear as separate
investigations in the forest plot. For both studies,
data were reported in two blocks because data
collection was conducted in two halves, completed
back-to-back, in order to assess change in accuracy
and speed over the course of the task (Table 5F)
(Hartig et al. 1996). Data from both blocks were aver-
aged for meta-analysis. There was no evidence of a
significant difference between groups at follow-up.
The confidence interval indicates substantial uncer-
tainty in the pooled effect estimate (Figure 7).
For number of letters searched (speed), as
already described, data from three studies,
reported in the same article, were pooled (Hartig
et al. 1996). The control groups (nonnatural) per-
formed significantly better than the intervention
groups. (Figure 8).
Sustained Attention to Response Test (SART)
One digit (1–9) is assigned as the target. Digits are
presented to the participant on a computer screen
in quick succession. Individuals need to press the
space bar every time a nontarget digit is seen, and
avoid pressing the space bar when viewing the
target. The one paper that used this test focused
on four separate scores: reaction times (lower
scores indicate better performance), number of
incorrect responses (lower scores are better),
quantity of correct responses (higher scores are
better), amount of incorrect responses (lower
scores are better), and sensitivity (or d-prime),
which takes into account correct and incorrect
responses simultaneously (higher scores are
Study or Subgroup
Hartig (1_1) 1996
Hartig (1_2) 1996
Hartig (2) 1996
Total (95% CI)
Heterogeneity: Tau² = 0.00; Chi² = 0.03, df = 2 (P = 0.98); I² = 0%
Test for overall effect: Z = 0.04 (P = 0.97)
Mean
38.7
39.05
36.69
SD
17.15
17.9
13.88
Total
17
17
9
43
Mean
38.45
38.47
37.63
SD
14.9
16.7
15.55
Total
34
34
9
77
Weight
42.1%
37.1%
20.8%
100.0%
IV, Random, 95% CI
0.25 [-9.32, 9.82]
0.58 [-9.61, 10.77]
-0.94 [-14.56, 12.68]
0.13 [-6.08, 6.33]
Experimental Control Mean Diffe rence Me an Difference
IV, Random, 95% CI
-20 -10 010 20
Favours [experimental] Favours [control]
Figure 7. Forest plot showing meta-analysis for SMT: percentage error (accuracy).
Study or Subgroup
Hartig (1_1) 1996
Hartig (1_2) 1996
Hartig (2) 1996
Total (95% CI)
Heterogeneity: Tau² = 0.00; Chi² = 0.18, df = 2 (P = 0.91); I² = 0%
Test for overall effect: Z = 2.58 (P = 0.010)
Mean
645.45
560.45
638.94
SD
176.7
154.35
142.67
Total
17
17
9
43
Mean
733.25
670.87
688.3
SD
205.98
223.28
447.99
Total
34
34
9
77
Weight
45.4%
48.9%
5.7%
100.0%
IV, Random, 95% CI
-87.80 [-196.65, 21.05]
-110.42 [-215.38, -5.46]
-49.36 [-356.53, 257.81]
-96.66 [-170.03, -23.29]
Experimental Control Mean Di fference Mean Difference
IV, Random, 95% CI
-200 -100 0100 200
Favours [Control] Favours [ex perimental]
Figure 8. Forest plot showing meta-analysis for SMT: number of letters searched (speed).
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 31

better). The SART was designed to measure ability
to withhold responses to infrequent and unpre-
dictable stimuli during a period of rapid and
rhythmic response to frequent stimuli and so
may be viewed as a vigilance task (Robertson
et al. 1997).
Three studies within one article reported SART
scores (Table 5G) (Berto 2005). Meta-analysis was
conducted separately for the four approaches to
measuring SART outcomes. However, there are
some concerns that outcomes reported in these
experiments are not the same as those intended
by the developers of SART (Robertson et al. 1997).
For reaction time (milliseconds), the three stu-
dies were balanced at baseline (Berto 2005). There
was no evidence of a significant difference between
groups at follow-up (Figure 9).
For number of correct responses, the three stu-
dies were not balanced at baseline (Berto 2005).
There was no evidence of a significant difference
between groups at follow-up (Figure 10).
For number of incorrect responses, of the three
meta-analyzed studies, only one (study 3) was
balanced at baseline (Berto 2005). There was little
evidence of a significant difference between groups
at follow-up (Figure 11). This was also true for the
balanced study alone (Berto 2005)(Figure 11).
For sensitivity (or d-prime), the three investiga-
tions were not balanced at baseline (Berto 2005).
Study or Subgroup
Berto(1_2) 2005
Berto(3) 2005
Total (95% CI)
Heterogeneity: Tau² = 126.26; Chi² = 1.40, df = 1 (P = 0.24); I² = 28%
Test for overall effect: Z = 0.54 (P = 0.59)
Mean
267.38
302.22
SD
73.78
32.09
Total
16
16
32
Mean
292.84
297.91
SD
50.78
52.98
Total
48
16
64
Weight
41.3%
58.7%
100.0%
IV, Random, 95% CI
-25.46 [-64.36, 13.44]
4.31 [-26.04, 34.66]
-7.99 [-36.72, 20.74]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-50 -25 0 25 50
Favours [experimental] Favours [control]
Figure 9. Forest plot showing meta-analysis for SART: reaction time (msec).
Study or Subgroup
Berto(1_2) 2005
Berto(3) 2005
Total (95% CI)
Heterogeneity: Tau² = 0.00; Chi² = 0.96, df = 1 (P = 0.33); I² = 0%
Test for overall effect: Z = 1.04 (P = 0.30)
Mean
13.62
17.12
SD
5.37
4.09
Total
16
16
32
Mean
13.39
14.56
SD
5.47
5.95
Total
48
16
64
Weight
57.3%
42.7%
100.0%
IV, Random, 95% CI
0.23 [-2.82, 3.28]
2.56 [-0.98, 6.10]
1.22 [-1.09, 3.54]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-4 -2 0 2 4
Favours [control] Favours [experimental]
Figure 10. Forest plot showing meta-analysis for SART: number of correct responses.
Study or Subgroup
Berto(1_2) 2005
Berto(3) 2005
Total (95% CI)
Heterogeneity: Tau² = 0.00; Chi² = 0.05, df = 1 (P = 0.82); I² = 0%
Test for overall effect: Z = 0.21 (P = 0.83)
Mean
2.06
0.81
SD
4.79
1.42
Total
16
16
32
Mean
1.68
0.75
SD
3.74
1.52
Total
48
16
64
Weight
13.5%
86.5%
100.0%
IV, Random, 95% CI
0.38 [-2.19, 2.95]
0.06 [-0.96, 1.08]
0.10 [-0.84, 1.05]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-4 -2 0 2 4
Favours [experimental] Favours [control]
Figure 11. Forest plot showing meta-analysis for SART: number of incorrect responses.
Study or Subgroup
Berto(1_2) 2005
Berto(3) 2005
Total (95% CI)
Heterogeneity: Tau² = 0.06; Chi² = 1.59, df = 1 (P = 0.21); I² = 37%
Test for overall effect: Z = 0.43 (P = 0.67)
Mean
1.86
2.47
SD
0.89
1.04
Total
16
16
32
Mean
1.97
2.03
SD
0.97
0.93
Total
48
16
64
Weight
58.7%
41.3%
100.0%
IV, Random, 95% CI
-0.11 [-0.63, 0.41]
0.44 [-0.24, 1.12]
0.12 [-0.41, 0.65]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-1 -0.5 00.5 1
Favours [control] Favours [experimental]
Figure 12. Forest plot showing meta-analysis for SART: sensitivity (or d-prime).
32 H. OHLY ET AL.

There was little evidence of a significant difference
between groups at follow-up (Figure 12).
Symbol Digit Modalities Test (SDMT)
The participant is given nine pairs of symbols and
digits (e.g., 1#, 2X, 3$, . . . 9%). After practicing writing
the correct number under the corresponding symbol
on a test sheet, the participant is given a blank copy of
the test and asked to write the correct number for each
symbol in 90 sec. This is repeated orally. The numbers
of correct symbol/digit pairs for the written and oral
tests are combined, with higher scores indicating bet-
ter performance.
Given the complexity of the task, it probably
reflects several perceptual, attentional, and executive
function processes. Pfeffer et al. (1981) suggested
that it is one of the best tools to distinguish between
early signs of dementia and depression, indicating
that while many of the attentional tests may be
affected by mood, the SDMT is tapping into cogni-
tive function over and above any mood effects
(p. 524).
Three studies reported SDMT scores as pre-
sented in Table 5H (Bodin and Hartig 2003;
Ottosson and Grahn 2005; Tennessen and
Cimprich 1995). The meta-analysis included data
from two studies (Bodin and Hartig 2003;
Tennessen and Cimprich 1995). Only the latter
was balanced at baseline (Bodin and Hartig
2003). This study reported data for men and
women as subgroups and these appear as separate
rows in the forest plot. There was little evidence of
a significant difference between groups at follow-
up (Figure 13). This was also the case for the
balanced study alone (Figure 13).
Symbol Substitution Test (SST)
AsfortheSDMT,participantsaregivenpairsofnine
symbols and digits. For the SST, they are asked to
assign the correct digits to a series of blanks, each
paired with a symbol. After a practice trial, the parti-
cipant is given 60 sec to fill in as many of the 110
available blanks as possible. The score is the number
of correct assignments completed, and higher scores
indicate better performance. Comments already given
about the processes for SDMT measures also apply to
SST, where speed is an even more important consid-
eration. It is unclear why this test was renamed and
administered for a shorter test period compared to the
SDMT.
Only one investigation reported SST scores
(Table 5I) (Johansson, Hartig, and Staats 2011).
This study was a crossover design in which parti-
cipants walked four times: alone and with a friend,
in natural and nonnatural settings. SST scores
declined for all four conditions, with the greatest
Study or Subgroup
Bodin(1) 2003
Bodin(2) 2003
Tennessen 1995
Total (95% CI)
Heterogeneity: Tau² = 10.80; Chi² = 4.72, df = 2 (P = 0.09); I² = 58%
Test for overall effect: Z = 0.83 (P = 0.40)
Mean
48.83
52.5
69.2
SD
4.22
10.04
11.18
Total
12
12
20
44
Mean
49.17
52.83
62.29
SD
5.38
9.56
9.52
Total
12
12
52
76
Weight
42.9%
23.5%
33.6%
100.0%
IV, Random, 95% CI
-0.34 [-4.21, 3.53]
-0.33 [-8.17, 7.51]
6.91 [1.37, 12.45]
2.10 [-2.83, 7.02]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-10 -5 0 5 10
Favours [control] Favours [experimental]
Figure 13. Forest plot showing meta-analysis for SDMT.
Study or Subgroup
Cimprich 2003
Stark 2003
Total (95% CI)
Heterogeneity: Tau² = 14.43; Chi² = 7.03, df = 1 (P = 0.008); I² = 86%
Test for overall effect: Z = 0.89 (P = 0.37)
Mean
25.65
19.84
SD
9.38
4.79
Total
83
29
112
Mean
31.21
19.6
SD
11.53
5.33
Total
74
25
99
Weight
48.6%
51.4%
100.0%
IV, Random, 95% CI
-5.56 [-8.87, -2.25]
0.24 [-2.48, 2.96]
-2.58 [-8.26, 3.10]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-10 -5 0 5 10
Favours [experimental] Favours [control]
Figure 14. Forest plot showing meta-analysis for TMTA.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 33

decline observed in the natural setting walked with
a friend. Differences in change between the natural
and nonnatural groups, regardless of social con-
text, were significant. However, SST performance
differences between groups were greater at base-
line and gaps closed during the intervention,
which may reflect regression to the mean. Meta-
analysis was not appropriate for this outcome
measure because the two environment groups
were not independent (Johansson, Hartig, and
Staats 2011). Although this test was virtually the
same as the SDMT, the difference in test periods
(60 vs. 90 sec) indicated that it was not possible to
meta-analyze SST/SDMT scores together.
Trail Making Test A (TMTA)
On paper or computer, participants must connect
25 numeric targets (1, 2, 3, 4, 5, etc.) in the correct
ascending order as quickly as possible. The score is
completion time, so lower scores indicate better
performance. Unlike most, the TMTA has a strong
motor component, since the respondent has to
coordinate actions as well as attention.
Two studies reported TMTA scores and both were
included in the meta-analysis (Table 5J)(Cimprich
and Ronis 2003;Stark2003). Only one was balanced
at baseline (Stark 2003). There was no evidence of a
marked difference between groups at follow-up
(Figure 14). The one balanced experiment on its
own also provided no evidence of a significant differ-
ence between groups at follow-up (Stark 2003)
(Figure 14).
Trail Making Test B (TMTB)
This test follows the same procedures as Trail Making
Test A, but targets alternate numbers and letters (1, A,
2, B, 3, C, etc.). It places more demands on working
memory as participants need to know whether they
are shifting from numbers to letters, or vice versa.
All three studies reporting TMTB scores were
included in the meta-analysis (Table 5K)
(Cimprich and Ronis 2003; Shin et al. 2011; Stark
2003). Only one study was balanced at baseline
(Shin et al. 2011). The intervention groups (nat-
ural) performed significantly better than controls
(Figure 15). This finding was supported by the
balanced study on its own (Shin et al. 2011)
(Figure 15).
Other Measures of Attention
Ten studies uniquely reported other measures of
attention (Table 5L). Three investigations showed
that attention performance improved in both (or
all) groups, with a significantly greater improve-
ment in the natural group (Chen, Lai, and Wu
2011; Cimprich and Ronis 2003; Stark 2003). The
other studies reported incomplete data (Laumann,
Gärling, and Stormark 2003; Perkins, Searight, and
Ratwik 2011; Taylor, Kuo, and Sullivan 2002; van
den Berg and van den Berg 2011) and/or found no
significant differences in change between the nat-
ural and nonnatural groups (Berman et al. 2008;
van den Berg and van den Berg 2011; Wu et al.
2008).
Discussion
Key Findings
This systematic review is based on 31 studies with
a variety of study designs, reported in 24 articles,
and found some empirical evidence to support
ART for three measures such as DSF, DSB, and
TMTB; the meta-analyses demonstrated significant
evidence that participants exposed to natural set-
tings displayed better postexposure attention
Study or Subgroup
Cimprich 2003
Shin 2011
Stark 2003
Total (95% CI)
Heterogeneity: Tau² = 19.69; Chi² = 4.67, df = 2 (P = 0.10); I² = 57%
Test for overall effect: Z = 1.98 (P = 0.05)
Mean
56.51
29.48
38.89
SD
26.6
6.82
10.75
Total
83
30
29
142
Mean
68.01
39.24
40.4
SD
34.92
21.23
9.3
Total
74
30
25
129
Weight
25.8%
31.8%
42.5%
100.0%
IV, Random, 95% CI
-11.50 [-21.30, -1.70]
-9.76 [-17.74, -1.78]
-1.51 [-6.86, 3.84]
-6.71 [-13.36, -0.05]
Experimental Control Mean Difference Mean Difference
IV, Random, 95% CI
-20 -10 010 20
Favours [experimental] Favours [control]
Figure 15. Forest plot showing meta-analysis for TMTB.
34 H. OHLY ET AL.

scores than those exposed to nonnatural settings.
However, meta-analyses for 10 other attention
outcomes (using 7 different attention measures)
did not show any marked differences between set-
tings. Further, meta-analysis for one attention out-
come, SMT, indicated that participants exposed to
nonnatural settings displayed significantly better
postexposure attention scores than participants
exposed to natural settings.
Several measures demonstrating significant
effects, such as ANT, have thus far only been
employed in a single published study, precluding
synthesis. ANT is the only measure that attempts
to delineate which of the attention processes
(alerting, orientating, or executive processes;
Jonides et al. 2008)) may be restored through
exposure to natural environments. As noted ear-
lier, more agreement about the most appropriate
measures of attention restoration is needed in the
field. Further trials using these agreed-on mea-
sures would help future appraisals of the theory
because more studies may then be included in
fewer meta-analyses, resulting in greater power.
It was not always clear how meaningful
improvements were. The pooled data from the
DSF test represents a mean increase of 0.39 digits
recalled by those exposed to natural compared to
nonnatural settings. Despite being significant,
practical significance in real-world settings is
unclear.
A full critique of the outcomes used for ART is
beyond the scope of this review, but one can make
some observations. Some tasks, including proof-
reading and SART, appear to measure vigilance
processes, with relatively limited demands on
executive functions such as working memory.
These require that attention is inhibited from
shifting to more interesting stimuli than the task,
but require few things to be remembered or cog-
nitively manipulated. Other tasks, such as the DSF,
DSB, SMT, SDMT, and SST, involve more obvious
demands on working memory, in terms of either
the amount of information remembered or the
need to manipulate it. Further, these measures
encompass graded demands on cognitive pro-
cesses; for instance, DSB involves more working
memory and executive function than DSF. Berman
et al. (2008) suggested that those tasks concerned
with working memory may be most likely to be
affected by natural exposure and so most relevant
for measuring its impact on attention. However,
not all the review analyses support this. Some tasks
that imposed higher levels of demand on working
memory failed to show significant effects for expo-
sure to nature (SDMT). Conversely, DSF and DSB
displayed similar significant effects for exposure to
nature compared to controls, although DSB places
greater demands on working memory than DSF. It
not known whether these anomalies are related to
lack of study power and limited numbers of stu-
dies contributing to each meta-analysis, low-mod-
erate quality investigations, or inappropriate
outcome measures. Again, better understandings
of the mechanisms for attention restoration, and
the best ways to measure them, are needed.
Only two studies measured attention during the
exposure to nature, as well as before and after the
exposure (Hartig et al. 2003; Wu et al. 2008). This
is potentially important because any positive
effects on attention during exposure may also be
considered beneficial, even if they do not persist
beyond the exposure. Future experiments might
consider including “during exposure”measures
to determine whether effects of nature are short-
lived or longer lasting.
Review Strengths
This is the first systematic review regarding atten-
tion restoration potential of natural compared to
other settings to focus on objective measures and
use systematic methods to identify, select,
appraise, and synthesize relevant experimental stu-
dies. By focusing on attention, it was possible to
include many more relevant studies than Bowler
et al. (2010), and also to conduct meta-analysis on
several attention measures that had been utilized
across several studies, allowing greater confidence
in results (e.g., in the effect of nature to impact
DSB scores). Nonetheless, a key outcome from the
review process as a whole demonstrates that the
field has yet to arrive at a clear consensus regard-
ing exactly how to best operationally define “direc-
ted attention”as conceptualized by ART. By
clearly identifying those attention-related tasks
that are most affected by nature, future research
may therefore help refine ART by increasing our
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 35

understanding of which precise attentional pro-
cesses may be the most relevant.
Review Limitations
The studies included in this systematic review
were heterogeneous in terms of study design,
experimental population, sample size, attention
capacity at baseline, type of natural setting, type
of exposure to and engagement with nature, dura-
tion of exposure, measures of attention, outcomes,
statistical methods used, and data reported. This
limited the scope for meta-analysis and it was not
possible to determine which groups of individuals,
settings, and exposures might result in the greatest
attention restoration. Many meta-analyses con-
tained few investigations, reducing their power
and preventing statistical determination of publi-
cation bias. It is recommended that future studies
consider using standardized approaches, to enable
subsequent systematic reviews to explore potential
differential effects more completely.
The quality of most (22 of 31) of the included
studies was rated “moderate,”with two “low”and
only seven rated “good.”This was partly because
some aspects of experimental design known to be
particularly important (such as blinding and rando-
mization procedures) were rarely reported in the
included investigations (Schulz et al. 1995). It is also
important to acknowledge that some of these stu-
dies were published more than 20 years ago, and
since then, reporting standards have progressed
considerably. As research in this discipline has not
previously been judged against systematic review
quality appraisal systems, some of the apparent
limitations in studies where the authors did not
reply to our request for further information may
be explained by lack of reporting, rather than by
deficits in conduct. Only 9 of 24 investigators
replied to our request for further information, and
it is possible that this may have introduced an
element of bias into our rating process. In addition,
it may be hard for subsequent reviewers to fully
replicate the current evidence synthesis. In medical
and health services research there is greater con-
sensus around best practice reporting standards
than in this field (Schulz, Altman, and Moher
2010). There is thus a need for researchers, journal
editors, and reviewers to come to an agreement
regarding key elements of study conduct required
to be reported in experiments in this field, in order
for quality to be judged fairly. All of the quality
indicators were given equal weighting in this system
of appraisal. Therefore, although our quality apprai-
sal system necessarily simplifies a complex issue, it
is postulated that this provides much-needed
impetus and guidance for better research practice
and reporting in future studies.
Ordinarily, quality appraisal in a systematic
review would include an appraisal of the validity
and reliability of outcome measures reported in
the included papers. However feedback from peer
reviewers led us to reconsider this. It was suggested
that “directed attention,”“voluntary attention,”and
“top-down attention”are synonymous, and thus
from this perspective any measure considered to
be an appropriate way of operationalizing either of
these concepts is, in theory, appropriate for either
concept. However, it is recognized that any given
task may be associated with demands on other
resources over and above directed attention.
Moreover, since nearly all measures used in the
ART literature are examples of widely used attention
measures with reliable psychometric properties, vir-
tually all measures might be considered valid and
reliable. However, it was also suggested that “direc-
ted attention,”as defined by ART, is not clearly
elucidated, making it unclear how validity should
be determined in the context of measures employed
to appraise the attention restoration value of differ-
ent environments. In this case, it is difficult for
researchers to know whether they have adopted an
appropriately valid measure. The debate examined
was broad and beyond the scope of our review, and
subsequently removed this criterion form the qual-
ity appraisal tool. It is postulated that the ART
community needs to attempt to address this short-
coming in the future such that there is clearer agree-
ment regarding which tools are deemed valid and
reliable measures of “directed attention”as defined
by ART in future studies.
In order to reduce bias in the meta-analyses,
only the follow-up data were used from studies
that had not employed appropriate analysis of
covariance methods to adjust for baseline imbal-
ance. A potential limitation of discarding base-
line data is that information is lost, making it
less likely that analyses might detect differences
36 H. OHLY ET AL.

between the trial arms or estimate differences
precisely. Where studies use multiple measures,
it is possible that performance may be influenced
by the order in which they were administered,
with previous tests impacting on performance in
subsequent ones. It was not possible to account
for this possibility in the meta-analyses. There
are complexities around establishing the need
for restoration in a studied population, which
may be influenced by the characteristics of the
participants, the nature and length of any load-
ing task and the nature of the measurement task
itself, and their interactions. Clear reasons were
provided for our assessment of the need for
restoration in studies included in this review,
but the true picture may be more complicated
and requires further investigation.
Finally, the current review was not designed to
examine attention outcomes alongside the range of
other outcomes claimed to be associated with expo-
sure to natural environments, such as recovery
from physiological stress, improvements in mood,
encouragement to exercise, facilitating social con-
tact, encouraging optimal development in children,
providing opportunities for personal development,
and a sense of purpose, despite many of our
reviewed studies also including one or more of
these outcomes (Mayer et al. 2009). Nonetheless,
as the causal mechanisms for attention restoration
may be co-related with other restorative effects,
further synthesis, building on theoretical develop-
ments on how nature may restore individuals
through multiple and interacting pathways, is
needed.
Research Recommendations
This systematic review highlighted a number of
issues for the future research agenda: (1) It is
unclear how validity and reliability of measures
of directed attention should be assessed, and
which are likely to be most sensitive to nature
exposure. More needs to be done to articulate
which specific characteristics of a task might be
important and thus gain a better understanding
of exactly which underlying attentional pro-
cesses nature may influence the most. (2)
Meta-analysis would be facilitated if the ART
community could articulate more clearly which
measures of attention are likely to measure the
impact of restoration most appropriately, and
then use these measures in a consistent way
across multiple studies. (3) Researchers and
journal editors should encourage complete
reporting of experimental outcomes, including
publishing negative findings, so that accurate
assessments can be made of the attention
restoration potential of natural settings. (4)
Investigators and journal editors should work
together to agree the key elements of research
reporting and experimental conduct to allow an
accurate and fair appraisal of study quality to
be made by readers and reviewers. (5) Future
studies could usefully assess the impact of
employing multiple measures, and the order in
which they are administered, on the outcomes
of attention themselves.
Funding
We thank the authors of previous studies who responded
to our queries relating to this review; Rachel Wigglesworth
for her help with double data extraction; and Shanker
Venkatasubramanian for his advice about cognitive tests.
We thank the peer reviewers for their insightful comments
on a previous version of this article. The European Centre
for Environment and Human The authors of this review
are supported by the National Institute for Health Research
(NIHR) Collaboration for Leadership in Applied Health
Research and Care (CLAHRC) for the South West
Peninsula (AB, OU, VN, RG) and by the European
Regional Development Fund and the European Social
Fund Convergence Programme for Cornwall and the Isles
of Scilly (HO, MW, BW, RG). The views expressed in this
publication are those of the authors and not necessarily
those of the NHS, the NIHR, the Department of Health in
England, or the European Union.
Notes on contributor
All authors contributed to the design of this review, critically
revised the article, and approved the final versions. HO con-
tributed to all stages of the systematic review (searching,
screening, data extraction, quality appraisal and synthesis)
and drafted the article. MW and BW contributed to double
data extraction and preparation of the article. AB devised the
search strategy, ran the literature searches, carried out cita-
tion searching, and contributed to double screening. OU and
VN provided statistical advice and designed and conducted
the meta-analyses. RG conceived the idea for the review,
contributed to double screening, double data extraction,
quality appraisal, and preparation of the article, and is the
guarantor.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 37

ORCID
Ruth Garside http://orcid.org/0000-0003-1649-4773
References
Baumeister,R.F.,K.D.Vohs,andD.M.Tice.2007.Thestrength
model of self-control. Current Directions in Psychological
Science 16:351–55. doi:10.1111/cdir.2007.16.issue-6.
Berman, M. G., J. Jonides, and S. Kaplan. 2008. The cognitive
benefits of interacting with nature. Psychological Science
19:1207–12. doi:10.1111/psci.2008.19.issue-12.
Berman, M. G., E. Kross, K. M. Krpan, M. K. Askren, A.
Burson, P. J. Deldin, S. Kaplan, L. Sherdell, I. H. Gotlib,
and J. Jonides. 2012. Interacting with na ture improves
cognition and affect for individuals with depression.
Journal of Affective Disorders 140:300–05. doi:10.1016/j.
jad.2012.03.012.
Berto, R. 2005. Exposure to restorative environments helps
restore attentional capacity. Journal of Environmental
Psychology 25:249–59. doi:10.1016/j.jenvp.2005.07.001.
Bodin, M., and T. Hartig. 2003. Does the outdoor environ-
ment matter for psychological restoration gained through
running? Psychology of Sport and Exercise 4:141–53.
doi:10.1016/S1469-0292(01)00038-3.
Bowler, D. E., L. M. Buyung-Ali, T. M. Knight, and A. S.
Pullin. 2010. A systematic review of evidence for the added
benefits to health of exposure to natural environments.
BMC Public Health 10. doi:10.1186/1471-2458-10-456.
Centre for Reviews and Dissemination. 2009. Systematic
reviews: CRD’s guidance for undertaking reviews in health
care. York, UK.
Chen, C., Y.-H. Lai, and J.-P. Wu. 2011. Restorative affections
about directed attention recovery and reflection in different
environments. Chinese Mental Health Journal 25:681–85.
Cimprich, B., and D. L. Ronis. 2003. An environmental
intervention to restore attention in women with newly
diagnosed breast cancer. Cancer Nursing 26:284–92.
doi:10.1097/00002820-200308000-00005.
Cohen, J.. 2011. A power primer. Psychological Bulletin 1992.
112:155–159.
Critical Appraisal Skills Programme. 2013. CASP UK critical
appraisal checklists. Critical Appraisal Skills Programme
2013.http://www.casp-uk.net (accessed September 2013).
Effective Public Health Practice Project. 2013. Quality assess-
ment tool for quantitative studies.http://www.ephpp.ca/
tools.html (accessed September, 2013).
Fan,J.,B.D.McCandliss,J.Fossella,J.I.Flombaum,andM.I.
Posner. 2005. The activation of attentional networks.
Neuroimage 26:471–79. doi:10.1016/j.neuroimage.2005.02.004.
Fan, J., B. D. McCandliss, T. Sommer, A. Raz, and M. I.
Posner. 2002. Testing the efficiency and independence of
attentional networks. Journal of Cognitive Neuroscience
14:340–47. doi:10.1162/089892902317361886.
Fan, M., and Y. Jin. 2013. Obesity and self-control: Food
consumption, physical activity, and weight-loss intention.
Applied Economic Perspectives and Policy 36:125–45.
doi:10.1093/aepp/ppt034.
Gifford, R.. 2014. Environmental psychology matters. Annual
Review of Psychology 65:541–579.
Hare, T. A., C. F. Camerer, and A. Rangel. 2009. Self-control in
decision-making involves modulation of the vmPFC valuation
system. Science 324:646–48. doi:10.1126/science.1168450.
Hartig, T., A. Book, J. Garvill, T. Olsson, and T. Garling.
1996. Environmental influences on psychological restora-
tion. Scandinavian Journal of Psychology 37:378–93.
doi:10.1111/j.1467-9450.1996.tb00670.x.
Hartig, T., G. W. Evans, L. D. Jamner, D. S. Davis, and T.
Gärling. 2003. Tracking restoration in natural and urban
field settings. Journal of Environmental Psychology 23:109–
23. doi:10.1016/S0272-4944(02)00109-3.
Hartig, T., M. Mang, and G. W. Evans. 1991. Restorative
effects of natural environment experiences. Environment
and Behavior 23:3–26. doi:10.1177/0013916591231001.
Herzog, T. R., A. M. Black, K. A. Fountaine, and D. J. Knotts.
1997. Reflection and attentional recovery as distinctive ben-
efits of restorative environments. Journal of Environmental
Psychology 17:165–70. doi:10.1006/jevp.1997.0051.
Herzog, T. R., P. Ouellette, J. R. Rolens, and A. M. Koenigs. 2010.
Houses of worship as restorative environments. Environment
and Behavior 42:395–419. doi:10.1177/0013916508328610.
Johansson, M., T. Hartig, and H. Staats. 2011. Psychological
benefits of walking: Moderation by company and outdoor
environment. Applied Psychology: Health and Well-Being
3:261–80. doi:10.1111/aphw.2011.3.issue-3.
Jonides, J., R. L. Lewis, D. E. Nee, C. Lustig, M. G. Berman,
and K. S. Moore. 2008. The mind and brain of short-term
memory. Annual Review of Psychology 59:193–224.
doi:10.1146/annurev.psych.59.103006.093615.
Jüni, P., D.G. Altman, and M. Egger. 2001. Restorative affections
about directed attention recovery and reflection in different
environments. BMJ 323.
Kaplan, R., and S. Kaplan. 1989. The experience of nature: A
psychological perspective. New York, NY: Cambridge
University Press.
Kaplan, S. 1995. The restorative benefits of nature: Toward
an integrative framework. Journal of Environmental
Psychology 15:169–82. doi:10.1016/0272-4944(95)90001-2.
Kaplan, S., and M. G. Berman. 2010. Directed attention as a
common resource for executive functioning and self-reg-
ulation. Perspectives on Psychological Science 5:43–57.
doi:10.1177/1745691609356784.
Kaplan, S., and J. Frey Talbot. 1983. Psychological benefits of
a wilderness experience. In Behavior and the natural envir-
onment, eds. I. Altman and J. Wohlwill, 163–203. Boston:
q6: Springer.
Kuo, F. E. 2001. Coping with poverty: Impacts of environ-
ment and attention in the inner city. Environment and
Behavior 33:5–34. doi:10.1177/00139160121972846.
Laumann, K., T. Gärling, and K. M. Stormark. 2003. Selective
attention and heart rate responses to natural and urban
environments. Journal of Environmental Psychology
23:125–34. doi:10.1016/S0272-4944(02)00110-X.
38 H. OHLY ET AL.

Mayer, F., S. C. McPherson Frantz, E. Bruehlman-Senecal,
and K. Dolliver. 2009. Why is nature beneficial?: The role
of connectedness to nature. Environment and Behavior
41:607–43. doi:10.1177/0013916508319745.
Ottosson, J., and P. Grahn. 2005. A comparison of leisure time
spent in a garden with leisure time spent indoors: On mea-
sures of restoration in residents in geriatric care. Landscape
Research 30:23–55. doi:10.1080/0142639042000324758.
Perkins, S., H. Searight, and S. Ratwik. 2011. Walking in a
natural winter setting to relieve attention fatigue: A pilot
study. Psychology 2:777–80. doi:10.4236/psych.2011.28119.
Pfeffer, R. I., T. T. Kurosaki, C. H. Harrah, J. M. Chance, D.
Bates, R. Detels, S. Filos, and C. Butzke. 1981. A survey
diagnostic-tool for senile dementia. American Journal of
Epidemiology 114:515–27.
Rich, D. 2008. Effects of exposure to plants and nature on
cognition and mood: A cognitive psychology perspective.
Dissertation Abstracts International: Section B: Sciences
Engineering 68:4911.
Robertson, I. H., T. Manly, J. Andrade, B. T. Baddeley, and J.
Yiend. 1997. `Oops!’: Performance correlates of everyday
attentional failures in traumatic brain injured and normal
subjects. Neuropsychologia 35:747–58. doi:10.1016/S0028-
3932(97)00015-8.
Schulz, K. F., D. G. Altman, and D. Moher. 2010. CONSORT
2010 Statement: Updated guidelines for reporting parallel
group randomised trials. British Medical Journal 340:c332.
doi:10.1136/bmj.c332.
Schulz, K. F., I. Chalmers, R. J. Hayes, and D. G. Altman. 1995.
Empirical evidence of bias: Dimensions of methodological
quality associated with estimates of treatment effects in con-
trolled trials. Journal of the American Medical Association
273:408–12. doi:10.1001/jama.1995.03520290060030.
Shin, W. S., C. S. Shin, P. S. Yeoun, and J. J. Kim. 2011. The
influence of interaction with forest on cognitive function.
Scandinavian Journal of Forest Research 26:595–98.
doi:10.1080/02827581.2011.585996.
Staats, H. 2012. Restorative environments. In The Oxford hand-
book of environmental and conservation psychology,ed.S.
Clayton, 445–58. New York, NY: Oxford University Press.
Stark, M. A. 2003. Restoring attention in pregnancy: The natural
environment. Clinical Nursing Research 12:246–65.
doi:10.1177/1054773803252995.
Sutton,A.J.,K.R.Abrams,D.R.Jones,T.A.Sheldon,andF.
Song. 2000. Methods for meta-analysis in medical research.
Wiley-Blackwell.
Taylor, A. F., and F. E. Kuo. 2009. Children with attention deficits
concentrate better after walk in the park. Journal of Attention
Disorders 12:402–09. doi:10.1177/1087054708323000.
Taylor, A. F., F. E. Kuo, and W. C. Sullivan. 2002. Views of
nature and self-discipline: Evidence from inner city chil-
dren. Journal of Environmental Psychology 22:49–63.
doi:10.1006/jevp.2001.0241.
Tennessen, C. M., and B. Cimprich. 1995. Views to nature: Effects
on attention. Journal of Environmental Psychology 15:77–85.
doi:10.1016/0272-4944(95)90016-0.
vandenBerg,A.E.,andC.G.vandenBerg.2011.A
comparison of children with ADHD in a natural and
built setting. Child: Care, Health Developments 37:430–
39.
van den Berg, A. E., S. L. Koole, and N. Y. Van Der Wulp.
2003. Environment preference and restoration: (How) are
they related? Journal of Environmental Psychology 23:135–
46. doi:10.1016/S0272-4944(02)00111-1.
Vickers, A. J., and D. G. Altman. 2001. Statistics notes:
Analysing controlled trials with baseline and follow up
measurements. British Medical Journal 323:1123–24.
doi:10.1136/bmj.323.7321.1123.
Vohs,K.D.,R.F.Baumeister,B.J.Schmeichel,J.M.
Twenge,N.M.Nelson,andD.M.Tice.2008.Making
choices impairs subsequent self-control: A limited-
resource account of decision making, self-regulation,
and active initiative. Journal of Personality and Social
Psychology 94:883–98. doi:10.1037/0022-3514.94.5.883.
Wu, S. H., C. L. Chang, J. H. Hsu, Y. J. Lin, and S. J. Tsao.
2008. The beneficial effects of horticultural activities on
patients’community skill and motivation in a public psy-
chiatric center. Proceedings of the International Symposium
on Horticultur Practices and Therapy for Human Well-
Being 775:55–70.
JOURNAL OF TOXICOLOGY AND ENVIRONMENTAL HEALTH, PART B 39