Journal of Environmental Psychology 32 (2012) 342e352
Contents lists available at SciVerse ScienceDirect
Journal of Environmental Psychology
journal homepage: www.elsevier.com/locate/jep
Light distribution in dynamic street lighting: Two experimental studies on its
effects on perceived safety, prospect, concealment, and escape
Antal Haans*, Yvonne A.W. de Kort
Eindhoven University of Technology, The Netherlands
a r t i c l e i n f o
a b s t r a c t
Article history:
Available online 7 June 2012
The relationship between light and perceived safety at night is intuitively strong, yet theoretically and
empirically its workings are largely unknown. Intelligent dynamic road lighting, which continuously
adapts to the presence and behavior of users, can light the street only when and where it is needed. As
such, it offers a solution to the energy waste and luminous pollution associated with conventional road
lighting. With this innovation, however, new questions emerge about the effect of lighting on perceived
safety. We need to consider not only how much lighting pedestrians need to feel safe, but also which
parts of the street should be lit. In two experiments, we investigated the effect of different light distributions on perceived safety, and explored mediation by people’s appraisal of three safety-related cues
suggested in the literature: prospect (having an overview), escape (perceived escape possibilities), and
refuge/concealment (perceived hiding places for offenders). Both experiments, one with stationary and
one with walking participants, demonstrated that people prefer having light in their own immediate
surroundings rather than on the road that lies ahead. This could be explained, partially, by changes in
prospect, escape, and concealment. Against expectations, prospect was higher with lighting distributions
in which participants’ immediate surroundings, but not the more distant parts of the road, were most
strongly lit.
Ó 2012 Elsevier Ltd. All rights reserved.
Keywords:
Urban environments
Pedestrians
Fear of crime
Environmental attitudes
Road lighting
Gender
1. Introduction
Street lighting is ubiquitous in modern day urban life. It is
important for crime prevention, for orientation and obstacle
avoidance at night, and for providing a general sense of safety to
road users. As such, it supports nighttime commercial and leisure
activities, and is essential for the freedom to go out at night, in
particular to those vulnerable to or fearful of personal attacks (e.g.,
Keane, 1998). Lighting is important also for creating esthetically
pleasing urban environments, which in turn affect the prestige of
many modern cities (cf., Bouman, 1987). Despite these important
functions, an increasing number of people are concerned with the
possible drawbacks of excessive street lighting. This so-called photo
or luminous pollution affects not only the amateur astronomer who
is constrained by the city’s sky glow, but has a detrimental effect on
the health and well being of all humans (for an overview, see
Navara & Nelson, 2007) and animals (e.g., on bird migration;
Longcore & Rich, 2004).
* Corresponding author. Human-Technology Interaction group, Department of
Industrial Engineering & Innovation Sciences, Eindhoven University of Technology
(IPO 1.35), P.O. Box 513, NL-5600 MB Eindhoven, The Netherlands.
E-mail address: [email protected] (A. Haans).
0272-4944/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jenvp.2012.05.006
At the same time, there is an increasing awareness of climate
change and the impending shortage of energy sources. In the
Netherlands, about 823,000 MWh per year are used by municipalities for public lighting, accounting, on average, for 60% percent
of the local government’s energy consumption (Agentschap NL,
2010; Taskforce Verlichting, 2011). This includes the lighting of
streets at times when no street users are present, leading to energy
waste and unnecessary luminous pollution. Taken together, these
examples underline the clear and urgent need to reconsider the
way in which we illuminate our streets at night (also Boyce, Fotios,
& Richards, 2009).
1.1. Toward intelligent dynamic street lighting
Lighting technologies based on light emitting diodes (LEDs) are
a promising innovation for street lighting. Their energy efficiency is
rising steadily, and they offer better control over the illumination
output. As such, LED technology offers new possibilities for
dimming at periods with lower traffic densities, or adjustment of
the light output to weather conditions. Combined with appropriate
sensing technology to recognize the number, type, and location of
road users, LED technology also offers the possibility of intelligent
light-on-demand scenarios. Such intelligent dynamic lighting,
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
which adapts itself to the street user, can offer light only when and
where it is needed most. As such it has the potential of reducing
energy waste and luminous pollution without affecting
functionality.
However, important choices have to be made with respect to
how dynamic road lighting is implemented. When and where, for
example, do pedestrians benefit most from street lighting? Do
pedestrians prefer to have light in their immediate surroundings
(i.e., there where they are walking), or would they prefer more light
in the distance (i.e., on those parts of the street that lie ahead of
them). Issues like these are irrelevant for conventional lighting
where light is functioning all the time and everywhere, distributed
equally across the lampposts. However, they pose a major challenge
for dynamic road lighting scenarios. The distribution of light across
the lampposts is but one out of a wide range of variables of which
the effect on perceived safety is still unknown. As such, the possibility of dynamic road lighting requires, more than ever, a thorough
understanding of the effect of lighting on street users, in particular
with respect to their perceived personal safety.
1.2. Street lighting and perceived personal safety
We define perceived personal safety as a person’s immediate
sense of security, and an absence of anxiety of becoming victimized,
when traveling through a particular environment (cf., Blöbaum &
Hunecke, 2005). Street lighting is generally seen as the most
important physical feature of an environment to affect perceived
personal safety (e.g., Loewen, Steel, & Suedfeld, 1993; Nasar, Fisher,
& Grannis, 1993; Nasar & Jones, 1997). The amount and uniformity
of illuminance, and perhaps also light spectrum are found to affect
perceived personal safety (e.g., Boyce, Eklund, Hamilton, & Bruno,
2000; Knight, 2009). Improving street lighting is also an effective
means in combating crime. Although this has been subject to
considerable debate (see Pease, 1999), the general consensus
nowadays is that adequate street lighting can reduce crime rates in
a street (for a recent meta-analysis, see Welsh & Farrington, 2008).
1.3. Three safety cues: prospect, escape, and concealment
Researchers have focused in particular on how people’s sense of
safety is influenced by features of the physical environment. Fisher
and Nasar (1992) describe three such proximate cues: prospect,
refuge, and escape. These cues are based on Appleton’s (1975)
prospect-refuge theory in which he argues that evolution has
installed a preference for environments that offer shelter (refuge)
and a good outlook over what is happening in the environment
(prospect). People’s appraisal of prospect will be high in the
absence of physical features that hinder their field of view, such as
trees, buildings, or a bend in the road. Street lighting too is expected
to affect people’s appraisal of prospect, and thus safety feelings.
Adequate street lighting, for example, offers visibility, which is
a prerequisite for prospect (Loewen et al., 1993).
The term refuge is used somewhat ambiguously by Fisher and
Nasar (1992; also Loewen et al., 1993). For Appleton (1975), an
environment high in refuge offers plenty possibilities for shelter
(i.e., safe havens). For Fisher and Nasar, however, refuge is defined
as the ease with which potential offenders can conceal themselves
in a certain street. To avoid this ambiguity we use the term
concealment rather than refuge (cf., Blöbaum & Hunecke, 2005;
Nasar et al., 1993). Physical street characteristics that might
increase people’s appraisal of concealment include walls, bushes,
trees, and other objects that create blind spots in which offenders
might hide. Street lighting can reduce these blind spots, but might
also cast shadows in which potential offenders can hide (Nasar &
Jones, 1997).
343
The third and last proximate cue described by Fisher and Nasar
(1992) is escape, which refers to the extent to which the environment offers possibilities for evading a possible assault. People’s
perception of escape might be influenced by such physical features
as alleys and subway stations which might offer routes away from
the assaulter. In addition, accessibility to other people is important
for one’s perception of escape (also Loewen et al., 1993). In contrast,
physical features that might increase the possibility of being
entrapped, such as dead ends, have a negative effect on perceived
escape (e.g., Blöbaum & Hunecke, 2005). As such, adequate street
lighting might point street users to important possibilities for
escape, and light emitting from windows might point to social
activity, and thus support in case of an emergency.
To date the effects of lighting on prospect, concealment, and
escape have only been determined in quasi-experimental ways,
relying on the comparison of carefully selected outdoor environments or photos. This is problematic because these outdoor environments unavoidably differ not only in terms of the quality of the
lighting, but in many other physical features as well (e.g., the
specific placement of trees and bushes). Few, if any, researchers
have the opportunity to manipulate road lighting keeping all other
physical features constant (for an exception, see Vrij & Winkel,
1991). Novel lighting technologies based on LEDs, however, offer
more experimental rigor by allowing precise control of illumination
output, and thus the possibility to manipulate it independently of
other street, luminaire, or light characteristics.
1.4. Research goals
In the present paper, we explore where pedestrians benefit most
from street lighting, and thus how light can best be distributed over
the lampposts in dynamic road lighting scenarios. In particular, we
test whether people feel safer with more light in their immediate
surroundings, or with more light on those parts of the road that lie
ahead. At the same time, we aim to confirm experimentally that the
effect of lighting on perceived personal safety at night can be
explained by changes in people’s perceptions of prospect, escape,
and concealment. By using LED-based luminaires, we manipulate
light distribution independently from other physical street characteristics that might affect perceived personal safety. As far as we
know, this is the first time that theories around Appleton’s (1975)
prospect-refuge theory are tested in a truly experimental fashion.
We present data from two experiments: with stationary and
with walking participants. In Experiment 1, we focus on the safety
experience of young female pedestrians, since this is one specific
group of street users who might benefit most from adequate street
lighting. However, research suggests that one should focus not only
on biological sex, but on psychological gender as well. Blöbaum and
Hunecke (2005), for example, found that more feminine women
perceive a higher threat of crime when outdoors at night, as
compared to women with less feminine traits. We aim to confirm
this finding in Experiment 1.
2. Experiment 1
2.1. Method
2.1.1. Participants
Twenty-nine women participated in the experiment. The
participants’ mean age was 22.9 (SD ¼ 2.83; range 19e30). All
participants were relatively unfamiliar with the test site, with 17 of
the participants (i.e., 59%) visiting the street less than once a month,
and 25 (i.e., 86%) less than once a week. The test site was regarded
as relatively safe, with 21 (i.e., 72%) of the participants indicating
344
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
that they believed that the area was at least a bit safe or better. All
participants received V10.00 as compensation.
2.1.2. Setting and apparatus
The experiment was conducted at the road lighting test site on
De Zaale at the campus of Eindhoven University of Technology,
Eindhoven, The Netherlands (see Fig. 1). De Zaale is one of the
campus’ main connection roads for cars, cyclists and pedestrians.
The test site consists of several lampposts, each with a height of
10 m, positioned at a distance of about 30 m from each other. On
each lamppost, a Philips CitySoul BGP431 GRN88 LED luminaire
was installed next to the existing high-pressure sodium lamp (see
Fig. 2A). Each luminaire consists of 84 LEDs with a combined power
of 106 W, a luminous flux of 8820 lm at 100% brightness, and a color
temperature of 4000 K. The luminous intensity distribution of the
LED luminaire is provided in Fig. 2B. The luminance output of each
individual LED luminaire could be dimmed by means of power line
communication (Philips Starsense; for a more detailed description
of the system, see Atici, Özçelebi, & Lukkien, 2011).
For the present study, we divided the test site into two road
segments. The west segment is flanked by buildings and parking
lots on both sides. The east segment is flanked by buildings on one
side, and by a square and a lawn on the other side. Each segment
consists of about 150 m of road and included five LED luminaires.
Some of the other lampposts on and in the vicinity of the test site
were switched off during the experiments. Notable exceptions
included the luminaires situated on and near the parking lot south
of the test site (which acted as a temporary construction site) and
the lampposts near and on the T-junctions at both ends of De Zaale
(see Fig. 1). The study was conducted after sundown (around 22:30)
in the last two weeks of May 2010. During this period, weather
conditions varied from rainy to dry, from clear to cloudy skies, with
temperatures between 8 and 17 C. Being one of the campus’ main
connection roads, there was occasional traffic on De Zaale during
the experiments, and some cars were parked along both sides of De
Zaale.
2.1.3. Experimental design
We performed a 2 (road segment: west vs. east) by 3 (conventional, ascending, and descending light distributions) withinsubject experimental design. The conventional lighting distribution refers to a typical road lighting situation in which the amount
of illumination is distributed equally across the poles (see Fig. 3A).
For this condition, all six lampposts were set to 40% of their
maximum illuminance output E(%), resulting in an illuminance of
Eh ¼ 7.0 lux (horizontal illuminance measured at street level
straight underneath the lamppost with a Konica Minolta T10 illuminance meter). If classified as a pedestrian street, then the
conventional distribution would correspond to a class S4 in the
European EN 13201-2 standard (European Committee for
Standardization, 2003). In the two other light conditions, the
same amount of lighting was used, but this was distributed
differently over the five lampposts. By using a fixed total illumination budget, we assured that any differences in perceived safety
ratings were not due to differences in overall illumination levels. In
the ascending condition, we reduced the light in the participant’s
immediate surroundings in favor of illuminating more strongly the
road that lies ahead. For this purpose, the lamppost closest to the
participant was set to E(%) ¼ 1% (Eh ¼ .5 lux) and this gradually
increased for each next lamppost up to E(%) ¼ 80% for the luminaire
furthest away from the participant (Eh ¼ 12.5 lux; see Fig. 3B). The
opposite setting was used in the descending condition (see Fig. 3C).
Pictures of the east and west section of the test site are provided in
Fig. 4 for each light distribution. Since it was difficult to distinguish
the three lighting distributions, we feared that a serial evaluation of
each road lighting condition (i.e., one at a time) would be difficult
for participants. Therefore, we decided to use both serial and pairwise comparisons.
2.1.4. Procedure
In each experimental session, a maximum of four participants
could partake at the same time. During the experiment, the
participants were standing at the center of the test site (see Fig. 1).
The experiment consisted of two phases: pair-wise (Phase 1) and
serial comparison of the six different experimental conditions
(Phase 2). Participants always completed the pair-wise comparisons first. For this purpose, the experimenter presented the
participants with one of the three lighting distributions on the east
segment, and one of the three distributions on the west segment of
the test site. Participants were asked to indicate which light
distribution they preferred (i.e., west or east) with respect to
perceived personal safety. In total, participants made nine of these
pair-wise comparisons (i.e., all possible combinations). The order of
the pair-wise comparisons was counterbalanced across
participants.
After the pair-wise comparisons (i.e., Phase 1 of the experiment), the six road lighting conditions were once again presented
to the participant but this time one after the other. For each
condition, participants completed a questionnaire in which they
evaluated that particular street segment in terms of perceived
personal safety, and the proximate cues prospect, concealment, and
escape. Again, the order in which the conditions were presented to
the participants was counterbalanced.
After the second phase of the experiment (i.e., the serial
assessments), participants completed another questionnaire
probing the personality characteristics femininity and masculinity,
and several demographics. For each participant, the experiment
took about half an hour to complete.
2.1.5. Measures
In Phase 1 (i.e., the pair-wise comparisons), perceived personal
safety was assessed as the preference for one of each pair of light
Fig. 1. The street lighting test site at De Zaale on the campus of Eindhoven University of Technology, Eindhoven, The Netherlands. Circles represent trees. The black squares
represent the LED luminaires used in Experiment 1 and 2. In Experiment 2, the LED luminaire indicated with the black triangle was used in addition. The white squares outside the
test site represent lampposts that were on during the experiments. Lampposts on and in the vicinity of the test site that were turned off during the experiments are excluded from
the map. The black star indicates the position of the participants in Experiment 1, and the white stars indicate the start and end points of the route in Experiment 2.
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
Fig. 2. Panel A shows the Philips CitySoul BGP431 GRN88 LED luminaire (front) as
mounted beside the high-pressure sodium lamp (back), and Panel B shows the luminous intensity distribution of the CitySoul LED Luminaire (based on Royal Philips
Electronics, 2012).
345
with a 3), to “very well” (coded with a 5). The mean score across
these three items was used in the analysis. The average reliability
(Cronbach’s alpha) of this aggregated prospect measure was a ¼ .87
(across the six experimental conditions).
The proximate cue escape was measured with three self-report
items, such as “How small or large are your chances of escaping
from this street in case someone assaults you?” (for a complete
description of the items, see Table 1). Participants could respond on
a five-point scale ranging from, for example, “very small” (coded
with a 1), through “neutral” (coded with a 3), to “very large” (coded
with a 5). The mean score across these three items was used in the
analysis. The average reliability (Cronbach’s alpha) of this aggregated prospect measure was a ¼ .79 (across the six experimental
conditions).
The proximate cue concealment was measured with three selfreport items, such as “How easy or hard is it for a potential criminal to find a hiding place in this street?” (for a complete description
of the items, see Table 1). Participants could respond on a five-point
scale ranging from, for example, “very hard” (coded with a 1),
through “neutral” (coded with a 3), to “very easy” (coded with a 5).
The mean score across these three items was used in the analysis.
The average reliability (Cronbach’s alpha) of this aggregated prospect measure was a ¼ .80 (across the six experimental conditions).
The personality characteristics masculinity (or instrumentality)
and femininity (or expressiveness) were measured by means of
a shortened version of the Extended Personal Attributes Questionnaire (EPAQ; Runge, Frey, Gollwitzer, Helmreich, & Spence,
1981) as used by Blöbaum and Hunecke (2005). These selfreports consist of 16 five-point semantic differential items, eight
of which measure the positive masculine characteristic instrumentality (e.g., “Not at all self-confident” to “very self-confident”),
and eight the positive feminine characteristic expressiveness (e.g.,
“Not at all emotional” to “very emotional”).
2.2. Results and discussion
distributions presented on the east and west segments. For each
pair-wise comparison, participants were asked to indicate whether
they felt more comfortable having to walk into the east or the west
segment of the street when alone on this street during the night.
Responses were analyzed with the many-facet Rasch model (see
Bond & Fox, 2007), using the Facets software (Linacre, 2006). The
Facets software uses a joint maximum likelihood procedure for
calibrating a perceived safety estimate for each lighting condition.
The units of these estimates are called logits, or log odds units. This
method of analysis leads to similar results as conventional Thurstonian scaling (Andrich, 1978).
In phase 2 (i.e., the serial assessments), perceived personal safety
was measured with three self-report items, which were formulated
as statements: “I feel uncomfortable with the idea of having to walk
into this street”, “I would walk down this street in a higher pace
than I usually walk in”, and “I would rather avoid this street”.
Participants were asked to indicate whether they agreed or disagreed with these statements on a five-point response scale,
ranging from “disagree” (coded with a 1), through “neutral” (coded
with a 3), to agree (coded with a 5). The three statements were
based on an instrument used by Blöbaum and Hunecke (2005). The
mean score across these three items was used in the analysis. The
average reliability (Cronbach’s alpha) of this aggregated safety
measure was a ¼ .90 (across the six experimental conditions).
The proximate cue prospect was measured with three self-report
items, such as “How well or poorly can you see what is happening
in this street?” (for a complete description of the items, see Table 1).
Participants could respond on a five-point scale ranging from, for
example, “very poorly” (coded with a 1), through “neutral” (coded
2.2.1. Effect of light distribution on perceived safety
For the pair-wise comparisons in Phase 1, the estimated
perceived safety scores are summarized in Fig. 5A. We found
a significant main effect of light distribution on perceived personal
safety, with c2(2, N ¼ 29) ¼ 35.4, p < .01 (for computational details,
see Schumacker & Lunz, 1997). In contrast, there was neither
a significant main effect of road segment, nor a significant road
segment by lighting distribution interaction, with c2(1, N ¼ 29) < .1,
p ¼ .96, and c2(6, N ¼ 29) ¼ .4, p ¼ .99 respectively.
The perceived safety scores obtained with the serial evaluations
in Phase 2 are summarized in Fig. 5B. Despite our initial concerns
with such serial evaluations, the estimates were highly correlated
with those obtained with the pair-wise comparisons, with r ¼ .92
and p ¼ .01. Subsequently, we performed a 3 (light distribution) by
2 (road segment) repeated measures ANOVA with perceived
personal safety (measured during the serial evaluations in Phase 2)
as the dependent variable. Assumptions of normality and sphericity
were met. As before, we found a significant main effect of light
distribution, with F(2,56) ¼ 28.1, and p < .01. Further post-hoc
comparisons (Least Significant Difference) revealed that the
ascending distribution was rated significantly lower in terms of
perceived personal safety than the two other distributions with
p < .01. In contrast, there was only a marginally significant difference between the descending and the conventional distribution
with p ¼ .07. The main effect of road segment, and the road segment
by light distribution interaction were again not found to be statistically significant, with F(1,28) ¼ .4, p ¼ .51, and F(2,56) ¼ 1.6,
p ¼ .21, respectively.
346
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
Fig. 3. The conventional (A), ascending (B), and descending light distribution (C) in Experiment 1. E(%) is the percentage of the maximum output of a luminaire. Eh(lux) is the
horizontal illuminance at street level straight underneath the lamppost.
Taken together, these results demonstrate that, with respect to
their perceived personal safety, our participants preferred to have
more light in their own immediate surroundings, even if this meant
a reduction in the illumination of the more distant parts of the road.
We can thus conclude that illuminating pedestrians’ immediate
surroundings is more important than illuminating the road that lies
ahead. Importantly, the absence of a road segment by light distribution interaction indicates that the light emitting from the luminaires on the parking lot south of the test site did not confound
with the experimental effect of light distribution on perceived
safety (see Fig. 1). Since road segment did not affect perceived
personal safety in any way, we did not take this variable into
account in the remainder of the analyses.
2.2.2. Masculinity, femininity, and perceived safety
Our female sample had an average masculinity of M ¼ 3.66 with
SD ¼ .43, and an average femininity of M ¼ 3.80 with SD ¼ .66. The
two personality traits were moderately correlated with r ¼ .47
and p ¼ .01. Subsequently, we repeated the repeated measures
ANOVA twice: with femininity and with masculinity included as
a covariate. Assumptions of normality and sphericity were met.
Femininity was found to be negatively related to perceived
personal safety, with F(1,27) ¼ 7.1 and p ¼ .01. Masculinity, in
contrast, was not found to be related to perceived personal safety
with F(1,27) < .1 and p ¼ .81. These results corroborate the findings
of Blöbaum and Hunecke (2005). Neither femininity, nor masculinity, however, moderated the difference between the descending
and the ascending light distribution with F(2,54) .5 and p .63.
2.2.3. Light distribution, proximate cues, and perceived safety
For each light distribution, participants’ appraisals of the proximate cues prospect, concealment, and escape are summarized in
Table 2. As expected, prospect and escape were rated highest, and
concealment lowest with the lighting distribution that was deemed
most safe (i.e., the descending distribution). But can the effects of
light distribution on perceived personal safety indeed be explained
by changes in people’s appraisal of the three proximate cues? To
answer this question, we conducted a mediation analysis on the
two light conditions that differed most in perceived personal
safety: The descending versus the ascending light condition. For
this purpose, we ran a series of Linear Mixed Models (LMMs)
following the procedure for mediation analysis by Baron and Kenny
(1986), complemented with Sobel tests (e.g., Zhao, Lynch, & Chen,
2010). The results of this analysis are summarized in Table 3. All
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
347
Fig. 4. Photographs of the three light distribution conditions on the west and east segments of the test site for Experiment 1. Images are converted to gray scale.
three proximate cues were related to perceived personal safety in
the expected way: Whereas high prospect and high escape were
positively related to perceived personal safety, high concealment
was found to be negatively related to perceived safety (as indicated
by the second stages of the indirect effects). Our results also
confirm that the effect of light distribution on perceived safety is
mediated by changes in people’s appraisals of prospect, escape, and
concealment (as indicated by the significant Sobel tests in Table 3).
This mediation, however, was only partial as indicated by a significant direct effect of light distribution on safety estimates. In
general, offering more light in the participants’ immediate
surroundings, and thus less light on the more distant parts of the
road, increased their appraisal of escape, but reduced concealment
(as indicated by the first stage of the indirect effects). Against our
expectations, prospect increased when we offered more light in the
participant’s immediate surroundings. Strangely, participants
indicated to have a better overview over what was happening in the
street when they had less light in the distance, and thus more light
in their immediate surroundings.
3. Experiment 2
In Experiment 1, we found that participants, with respect to
their perceived personal safety, preferred having light in their
immediate surroundings (i.e., the descending condition) over
having light in the distance (i.e., the ascending condition). This
348
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
Table 1
Items of the prospect, escape, and concealment measures for Phase 2 of Experiment
1 and for Experiment 2.
Prospect
1. How well or poorly can [could] you see what is happening in this street?
2. How good or poor an overview do [did] you have over this street?
3. How good or bad can [could] you see the objects in this street?
Escape
1. How small or large are [would be] your chances of escaping from
this street in case someone would assault you?
2. How hard or easy would it be to bring yourself into safety in this street?
3. How hard or easy would it be for a criminal to entrap you in this street?
Concealment
1. How easy or hard is it for a potential criminal to find a hiding place
in this street?
2. Are [Were] there many or few places where a potential criminal can
hide in this street?
3. How big or small is the chance that a criminal might hide in this
street without being detected?
Note: Words in brackets refer to the wording of the items in Experiment 2.
effect of light distribution was explained mostly by an increase in
prospect. Strangely, prospect appears to increase not when more
light is offered on the more distant parts of the road, but when
one’s immediate surroundings are more strongly illuminated.
These findings, however, should be considered with two possible
limitations in mind. First, our participants evaluated the various
lighting conditions from a fixed position on the street. Since
pedestrians’ need for lighting might change when they start to
move actively through an environment, the findings are not
necessarily generalizable to more natural situations. In this second
experiment, we therefore aimed to confirm the findings of Experiment 1 with walking participants. Second, we could not switch off
the luminaires on and around both ends of De Zaale (i.e., near the Tjunctions in Fig. 1). As a result, participants always saw some
lighting in the far distance even in the descending condition. This
may have affected the results. To test whether pedestrians indeed
prefer having light in their immediate surroundings, we compared
safety perceptions when a trade-off was made between light in vs.
just outside their immediate surroundings. At the same time, we
further investigated individual differences in safety perceptions
and lighting needs. In contrast to Experiment I, we focused on sex,
rather than gender, and included two personality traits related to
safety feelings: Perceived attractiveness to criminals, and perceived
power. These are two traits from the sociopsychological fear of
crime model developed by van der Wurf, van Staalduinen, and
Stringer (1989). Attractiveness refers to the degree to which
people judge themselves and their possession to be an attractive
target to criminals. Power, in turn, refers to people’s selfassessments with respect to their capacity to avoid or at least
limit personal harm when assaulted by a possible offender.
3.1. Method
3.1.1. Participants
Fifty people participated in the experiment. The participants’
mean age was 21.8 (SD ¼ 2.5; range 18e27 years). Twenty-eight
(i.e., 56%) of the participants were men. Participants were mostly
undergraduate students from the Eindhoven University of Technology or surrounding institutes. Participants were rather familiar
with the test site, with 33 (i.e., 66%) of the participants visiting the
test-site more than once a week. The test site was regarded as
relatively safe, with 35 (i.e., 70%) of the participants indicating that
they believed that the area was at least a bit safe or better. All
participants received V10.00 as compensation.
3.1.2. Setting and apparatus
The experiment was conducted at the same test site as Experiment 1 (see Fig. 1). We used a so-called Wizard-of-Oz implementation of intelligent dynamic street lighting. A true dynamic
street lighting system would infer a pedestrian’s position in the
street on the basis of sensor data (e.g., cameras) and adjust automatically the illumination levels of individual lampposts. In our
setup, the experimenter in the control room kept track of the
participants’ position on the street by having them press a button
on a portable radio link each time they passed a lamppost. Upon
hearing this signal, he initiated a computer program that adjusted
the output of the luminaires in discrete steps, once every 3.75 s, for
15 s (which in pilot studies was determined to be the average time
it takes a person to walk from one lamppost to another). This
procedure was repeated when the participant indicated to have
passed the next lamppost, until reaching the end of the test site.
The experiment was conducted after sundown (around 19:30) in
the last week of September and the first week of October. During
this period, weather conditions varied from rainy to dry, from clear
to cloudy skies, with temperatures between 6 and 16 C. The test
site was temporarily closed to cars, but there were occasional
cyclists or pedestrians on the test site during the experiments.
3.1.3. Experimental design
We performed a three-condition (control, dark spot, and spotlight light distribution) within-subject experimental design. Again,
we used a fixed lighting budget for each of the three light distributions. The three lighting conditions are depicted in Fig. 6 and
Fig. 5. Effect of light distribution on perceived personal safety for the different road segments in Experiment 1 using pair wise (A) and serial comparisons (B), and for Experiment 2
(C).
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
Table 2
Mean prospect, escape, and concealment for each light distribution, and standard
error of estimates (SE), for Experiment 1 and 2.
Light distribution
Experiment 1
Conventional
Ascending
Descending
Experiment 2
Control
Dark spot
Spotlight
Prospect
Escape
Concealment
3.38 (.16)
2.73 (.18)
3.70 (.17)
2.96 (.15)
2.81 (.16)
3.25 (.13)
3.16 (.17)
3.50 (.16)
2.93 (.14)
3.63 (.13)
3.09 (.16)
3.85 (.13)
3.51 (.12)
3.35 (.13)
3.57 (.11)
3.08 (.10)
3.51 (.13)
3.03 (.13)
Note: Data for experiment 1 are aggregated across both road segments.
Figures between brackets are Standard Errors of estimates (SE).
show the illumination situation after the participant indicated to
have passed a lamppost. In the control condition, the available
illumination was distributed over three lampposts: one behind and
two lampposts in front of the participant (see Fig. 6A). We had
originally planned to include one additional lamppost behind the
participant, but due to a lack of bandwidth in the power line
communication we were unable to control reliably an extra luminaire. Each of the three luminaires, at that point in time, were set to
349
Table 3
Mediation of the effect of light distribution on perceived personal safety by prospect,
escape, and concealment for Experiment 1 and 2.
Independent
variable
Total
effect
Direct
effect
Mediator
Stage 1
Indirect
effect
Stage 2
Indirect
effect
Indirect
effecta
Experiment 1
descending vs.
ascending
1.05**
.31**
prospect
escape
concealment
.97**
.44**
.57**
.55**
.21*
.19*
4.64**
2.12*
1.89*
.70**
.21y
prospect
escape
concealment
.75**
.21y
.49**
.48**
.27**
.14y
3.30**
1.61y
1.54y
Experiment 2
spotlight vs.
dark spot
Note: All estimates are in unstandardized units.
yp < .10. *p < .05. **p < .01.
a
This column provides the Sobel statistics and one-sided significance tests.
E(%) ¼ 54% of their maximum power (Eh ¼ 9.5 lux). All other
lampposts were set to E(%) ¼ 1% (Eh ¼ .5 lux). In the dark spot
condition, participants were walking in a relatively dark area
between two lampposts both at E(%) ¼ 1%, but with the two
neighboring lampposts, one in front and one in the back, at a high
output level of E(%) ¼ 80% (Eh ¼ 12.5 lux; see Fig. 6B). This dark spot
Fig. 6. The control (A), dark spot (B), and spotlight conditions (C) in Experiment 2. E(%) is the percentage of the maximum output of a luminaire. Eh(lux) is the horizontal illuminance
at street level straight underneath the lamppost.
350
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
condition resembles the ascending condition in Experiment 1 in
that most of the available illumination is provided outside the
participant’s immediate surroundings. In contrast to Experiment 1,
however, we did not include the more distant lampposts which
were all set to E(%) ¼ 1% of their maximum power (Eh ¼ .5 lux). The
opposite scheme was used in the spotlight condition in which
participants were walking in a highly illuminated area between the
two nearest lampposts (see Fig. 6C). This condition resembles the
descending condition in Experiment 1.
3.1.4. Procedure
Participants were invited to the test site after sundown. In
contrast to Experiment 1, only a single participant could partake in
the experiment at one time. Each participant walked along the test
street three times under the different lighting conditions. The order
of the conditions was counterbalanced across participants. All
participants started on the east of the test site and were asked to
walk to the west side in their own pace (see Fig. 1). They were
instructed to press a button on a portable radio link each time they
passed a lamppost. At the end of the route they received a questionnaire from an experimenter which asked about their safetyrelated appraisals of the street. After completing the questionnaire, the experimenter changed the lighting scheme and the
participant walked the same route again but in opposite direction.
After the experiment, participants completed another questionnaire probing the personality characteristics perceived power and
attractiveness, and several demographics. For each participant, the
experiment took about half an hour to complete.
3.1.5. Measures
Perceived personal safety was measured with three self-report
items: “How safe or unsafe did you feel while walking down the
street?”, “How comfortable or uncomfortable were you with
walking down this street?”, and “To what extent would you normally avoid or pick this street during a nighttime walk?”. Participants could respond on a five-point scale ranging from, for
example, “very unsafe” (coded with a 1), through “neutral” (coded
with a 3), to “very safe” (coded with a 5). The mean score across
these three items was used in the analysis. The average reliability
(Cronbach’s alpha) of this aggregated safety measure was a ¼ .83
(across the three experimental conditions).
The proximate cues prospect, refuge, and escape were measured
with the same items as in Experiment 1, apart from slight changes
in wording (see Table 1). The average reliability (Cronbach’s alpha)
of the three measures were a ¼ .87, a ¼ .89, and a ¼ .76, respectively
(across the three experimental conditions).
The personality characteristic attractiveness was measured with
three self-report items based on the instrument by van der Wurf
et al. (1989): “To what extent do you regard yourself to be an
unattractive or attractive target for possible criminals?”, “To what
extent do regard your personal belongings to be an unattractive or
attractive target for possible criminals?”, and “How large or small is
the likelihood that a possible criminal will select you as a target?”.
Participants responded on a five-point response scale ranging, for
example, from “very unattractive” (coded with a 1), through
“neutral” (coded with a 3), to very attractive (coded with a 5). Since
responses to the second item did not correlate positively with the
other two items, the average response on the first and third item
was used in the analysis. The reliability (Cronbach’s alpha) of this
aggregated attractiveness measure was a ¼ .61.
The personality characteristic power was measured with three
self-report items based on the instrument by van der Wurf et al.
(1989): “To what extent do you regard yourself incapable or
capable of escaping from an attacker?”, “To what extent do you
regard yourself incapable or capable of chasing off an attacker?”,
“To what extent do you regard yourself incapable or capable of
defending yourself against a possible attacker?”. Participants
responded on a five-point response scale ranging, for example,
from “very incapable” (coded with a 1), through “neutral” (coded
with a 3), to very capable (coded with a 5). The mean score across
these three items was used in the analyses. The reliability (Cronbach’s alpha) of this aggregated power measure was a ¼ .88.
3.2. Results and discussion
3.2.1. Effect of light distribution on perceived safety
The perceived safety scores for each of the three lighting
distributions are summarized in Fig. 5C. We performed a 3 condition (control, dark spot, spotlight) repeated measures ANOVA with
perceived personal safety as the dependent variable. Assumptions
of normality and sphericity were met. As before, we found
a significant main effect of light distribution, with F(2,98) ¼ 14.0,
and p < .01. Further post-hoc comparisons (Least Significant
Difference) revealed that the dark spotlight distribution was rated
significantly lower in terms of perceived personal safety than the
two other distributions, with p < .01. However, we did not find
a statistically significant difference between the spotlight and
control distribution, with p ¼ .75. These findings corroborate, to
a large extent, those of Experiment 1: Participants prefer to have
light in their immediate surroundings.
3.2.2. Gender, attractiveness, power, and perceived safety
We repeated the ANOVA but this time with gender as an additional factor. Assumptions of normality, homogeneity of variance,
and sphericity were met. As expected, female participants had
lower perceived personal safety than male participants, with
F(1,48) ¼ 5.8 and p ¼ .02. In addition, there was a marginally
significant gender by light distribution interaction with
F(2,96) ¼ 2.9 and p ¼ .06. Can these gender effects be explained by
the personality characteristics attractiveness and power? As
expected, women, on average, rated themselves as more attractive
to criminals (M ¼ 3.3, SD ¼ .63) than men (M ¼ 2.0, SD ¼ .69) with
t(48) ¼ 6.6 and p < .01. Similarly, perceived power was lower for
women (M ¼ 2.4, SD ¼ .96) than for men (M ¼ 3.6, SD ¼ .88) with
t(48) ¼ 4.5 and p < .01. We found a moderate to strong correlation
between the two personality traits with r ¼ .63 and p < .01.
Subsequently we performed two additional repeated measures
ANOVAs, but this time with either attractiveness or power as
covariates. Assumptions of normality and sphericity were met. We
found attractiveness to be negatively related to perceived personal
safety, with F(1,48) ¼ 7.8, and p < .01. In addition, the attractiveness
by light distribution interaction was found to be statistically
significant with F(2,96) ¼ 3.3, and p ¼ .04, indicating that the light
distribution manipulation had a smaller effect on people that
perceive themselves to be less attractive to criminals. Neither the
main effect of power, nor the power by light distribution interaction
effect was found to be statistically significant, with p .11.
3.2.3. Light distribution, proximate cues, and perceived safety
For each light distribution, participants’ appraisals of the three
proximate cues are summarized in Table 2. Subsequently, we performed a series of LMMs to test for mediation of the effect of light
distribution on perceived personal safety by prospect, concealment
and escape. Again, we report only the mediation analysis for the
comparison of the two light conditions that differed most in
perceived personal safety: The spotlight versus the dark spot
condition. The results are summarized in Table 3, and largely
corroborate those of Experiment 1. Prospect was again found to be
the best mediator of the proximate cues. Moreover, prospect was
higher in the spotlight condition than the dark spot condition,
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
demonstrating, again, that prospect increases when participants
have more light in their immediate surroundings.
4. General discussion
Intelligent dynamic street lighting systems can reduce energy
consumption and light pollution by providing light to pedestrians
only when and where it is needed. In two experiments we investigated what areas of the road pedestrians prefer to have illuminated with respect to their perceived personal safety. The first
experiment showed that stationary pedestrians prefer having light
in their own immediate surroundings over light on the more
distant parts of the road. The importance of a well lit immediate
surrounding was confirmed with walking participants in Experiment 2: Again participants valued least the lighting distribution in
which their immediate surroundings were poorly lit (i.e., the dark
spot condition). Importantly, these findings cannot be ascribed to
differences in the overall illumination of the street as the average
illumination output per lamppost was kept equal for each condition. The use of such a lighting budget, however, prevented
comparing the lighting conditions in Experiment 2 with a conventional lighting setup. We can thus only speculate on how, for
walking participants, the spotlight type of light distribution would
perform when compared to conventional street lighting.
The effect of light distribution on perceived personal safety
could be explained, at least partially, by changes in people’s
appraisals of the safety-related street characteristics prospect,
escape, and concealment. This is the first time, at least to our
knowledge, that mediation of the effect of street lighting by these
proximate cues is demonstrated experimentally rather than quasiexperimentally. Whereas previous research relied on comparing
differently lit environments or pictures of environments, we
manipulated light distribution whilst keeping all other factors
constant. As such, our results cannot be confounded by other
factors than the light distribution, such as the exact type of luminaire, or the placement of trees, bushes, and buildings. Since
appraisals of the three proximate cues could only account for part
of the effect of light distribution on safety, we may conclude that
people take additional cues into account beside prospect,
concealment, and escape. Nasar and Jones (1997), for example,
argued for a differentiation between two types of concealment:
hiding places (i.e., the availability of objects behind which a criminal might hide) and dark spots (i.e., shadows in which criminals
might hide). Alternatively, it is not unlikely that the brightness of an
environment (or its darkness for that matter) is in itself a proximate
cue that affects one’s perceived personal safety (cf., Blöbaum &
Hunecke, 2005). Unfortunately, this alternative hypothesis is difficult to test empirically as brightness will be heavily confounded
with such cues as prospect, escape, and concealment.
As expected, a relatively high sense of safety was associated
with high prospect and escape, but low concealment. In both
experiments, prospect was found to be the most important
predictor for people’s perceived personal safety. Against our
expectations, however, prospect increased with more light being
offered in a participants’ immediate surroundings, rather than with
more light on the road that lies ahead. One possible explanation is
offered by Goffman (1971). He argued that people may focus
primarily on their immediate surroundings as potential threats
typically originate from events that occur near to the body. Good
prospect, for pedestrians, may thus mean having a good overview
over their immediate surrounding including the road sides.
Unfortunately, the items in our questionnaire that tapped into
prospect did not differentiate between having a good overview
over the road ahead and having a good overview over the road
sides. More research is needed to substantiate this. One suggestion
351
would be to use eye-tracking technology to determine the direction
of pedestrians’ gaze when walking along a certain street at night.
Relevant in this respect is the distinction between action and vista
space (e.g., Cutting & Vishton, 1995). Action space is the circular
area of about 30 m around the body in which a person’s public
actions take place. It is the area in which we can move around
quickly, speak easily, and pass around objects to other people.
Perhaps adequate illumination of a pedestrian’s immediate location
(i.e., action space) is important for perceived personal safety
because it allows people to respond effectively to dangerous events.
An escape possibility that is well illuminated may be highly visible
to a pedestrian, but of no immediate use when located 100 m
ahead. Vista space, in contrast, is defined as the area that extends
beyond action space. Adequate illumination of the road that lies
ahead (i.e., in vista space) might be more important for the planning of actions, and thus for orientation and way finding. Moreover,
the size of the area that must be illuminated in order for a road user
to feel safe may depend on his or her (potential) speed: The larger
one’s action space is, the larger may be the region of the road to
which one is attentive with respect to both concealment and escape
possibilities. As a result, we predict that cyclists may prefer a wider
spotlight, covering more of their immediate surroundings, than
pedestrians. More research is needed to test this hypothesis.
With respect to individual differences, we confirmed previous
research that femininity, but not masculinity, affects perceived
personal safety (e.g., Blöbaum & Hunecke, 2005). More specifically,
women with more feminine traits experienced a lower sense of
perceived safety than women with less feminine traits. In contrast
to Blöbaum and Hunecke, however, femininity was not found to
moderate the effect of light distribution on safety feelings.
However, this most likely reflects a lack of variance and thus
statistical power. More statistical power was achieved in experiment 2, where we included both male and female participants. As
expected, the more participants regarded themselves to be an
attractive target to criminals, the lower their perceived safety was
(e.g., van der Wurf et al., 1989). More importantly, perceived
attractiveness moderated the effect of light distribution on safety:
Lighting is more important for individuals who deem themselves
attractive targets. In contrast, individual differences in self-assessed
power, or the ability to defend oneself against an assailant, were not
found to affect perceived safety. This may be explained by the
absence of potential offenders on the test site, as people’s perceived
power is always in comparison to the strength of the possible
assailant (e.g., van der Wurf et al., 1989).
There were several limitations to the present experiment. First,
our participants perceived the test site as a relatively safe place.
Had participants felt more unsafe, or had the test site been located
in a more unsafe area, the results might have been different (cf.,
Nasar et al., 1993). Second, our participants in Experiment 2 had to
indicate their position on the street by pressing a button on
a portable radio link each time they passed a lamppost. By doing so,
we perhaps assigned a notion of control over the street lighting to
our participants, which might have affected their safety perceptions. Third, by using a Wizard-of-Oz implementation of a dynamic
lighting system in Experiment 2, the street lighting adapted only to
the participant, but not to an occasional other user of the test site.
This may have affected the results. More research is needed to
investigate how a dynamic lighting system should respond to the
presence of multiple road users, who depending on their transportation mode and traveling speed may have different lighting
needs. Whether and how a pedestrian’s need for lighting will
change in such situations is another question that stills need to be
addressed. Fourth, our research would have benefitted from
a better understanding of how the various light distributions affect
the visual system. How, for example, did a certain light distribution
352
A. Haans, Y.A.W. de Kort / Journal of Environmental Psychology 32 (2012) 342e352
affect what people could or could not see anymore? Describing the
light manipulations both functionally (e.g., in terms of the illuminance output) and perceptually might aid in the interpretation of
the results. Finally, we focused exclusively on the effect of light
distribution on perceived personal safety. Further research is
needed to determine how different light distributions affect other
street lighting functions, such as orientation and obstacle avoidance. For orientation and navigation, for example, adequate illumination of the road ahead (i.e., vista space; Cutting & Vishton,
1995) might well be of critical importance.
Despite these limitations, our study has provided insights into
the effect of light distribution on perceived personal safety. These
findings are important for future intelligent dynamic street lighting
systems that can adapt themselves to the road users. By illuminating only those parts of the street that are sufficient to ensure the
safety of a particular type of street user, these systems can avoid
energy waste and light pollution. In addition, we have demonstrated for the first time, at least to our knowledge, in a true
experiment that the effect of street lighting on perceived personal
safety is mediated by prospect, escape, and concealment. As such,
our research gives credibility to theoretical frameworks that are
based around Appleton’s (1975) prospect-refuge theory.
Acknowledgments
This research is affiliated to the Intelligent Lighting Institute
Eindhoven. The project was part of the ENSURE initiative, in which
Philips and Eindhoven University of Technology collaborated in
research on intelligent lighting solutions. We thank the Dutch
Ministry of Economic Affairs for their financial support. We thank
Thijs van Osch, Leonie Geerdinck, Meriete Horst, John Servaes, and
Doménique van Gennip for their assistance in conducting the
experiments. Also, we thank all members of the ENSURE road
lighting project, in particular Marco Haverlag, Gerrit Kroesen,
Cagdas Atici, Tanir Ozcelebi, Rob Kluijver, Herbert Fiedler, and Johan
Lukkien for their input and assistance, and Michael Linacre for
assisting with the analysis of the pair wise comparisons. Parts of the
present paper were presented at the LS12-WLED3 conference, July
11e16, 2011, Eindhoven, The Netherlands, and Environment2.0: The
9th Biennial Conference on Environmental Psychology, September
26e28, 2011, Eindhoven, The Netherlands.
References
Agentschap NL. (2010). Monitoring openbare verlichting 2010 [Monitoring public
lighting 2010]. Retrieved from Agentschap NL. http://www.agentschapnl.nl/
content/rapport-monitoring-openbare-verlichting-2010.
Andrich, D. (1978). Relationships between the Thurstone and Rasch approaches to
item scaling. Applied Psychological Measurement, 2, 449e460.
Appleton, J. (1975). The experience of place. London: Wiley.
Atici, C., Özçelebi, T., & Lukkien, J. J. (2011). System architecture for road lighting. In
Proceedings of the 2011 IEEE International Conference on Consumer Electronics
(pp. 681e682). http://dx.doi.org/10.1109/ICCE.2011.5722807.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in
social psychological research: conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology, 51, 1173e1182.
Blöbaum, A., & Hunecke, M. (2005). Perceived danger in urban public space: The
impact of physical features and personal factors. Environment and Behavior, 37,
465e486.
Bond, T. G., & Fox, C. M. (2007). Applying the Rasch model: Fundamental measurement
in the human sciences (2nd ed.). Mahwah, NJ: Erlbaum.
Bouman, M. J. (1987). Luxury and control: The urbanity of street lighting in nineteenth-century cities. Journal of Urban History, 14, 7e37.
Boyce, P. R., Eklund, N. H., Hamilton, B. J., & Bruno, L. D. (2000). Perceptions of safety
at night in different lighting conditions. Lighting Research and Technology, 32,
79e91.
Boyce, P. R., Fotios, S., & Richards, M. (2009). Road lighting and energy saving.
Lighting Research and Technology, 41, 245e260.
Cutting, J. E., & Vishton, P. M. (1995). Perceiving the layout: The integration, relative
potency, and contextual use of different information about depth. In W. Epstein,
& S. Rogers (Eds.), Handbook of perception and cognition. Perception of space and
motion, Vol. 5 (pp. 69e117). San Diego, CA: Academic Press.
European Committee for Standardization. (2003). European standard EN 13201-2,
road lighting, part 2: Performance requirements (Ref. No. EN 13201-2:2003).
Brussels, Belgium: European Committee for Standardization.
Fisher, B. S., & Nasar, J. L. (1992). Fear of crime in relation to three exterior site
features: Prospect, refuge and escape. Environment and Behavior, 24, 35e65.
Goffman, E. (1971). Relations in public: Microstudies of the public order. New York:
Basic Books.
Keane, C. (1998). Evaluating the influence of fear of crime as an environmental
mobility restrictor on women’s routine activities. Environment and Behavior, 30,
60e74.
Knight, C. (2009). Effect of lamp spectrum on perception of comfort and safety. In
Y. A. W. de Kort, W. A. IJsselsteijn, I. M. L. C. Vogels, M. P. J. Aarts, A. D. Tenner, &
K. C. H. J. Smolders (Eds.), Proceedings of experiencing light 2009: International
conference on the effects of light on wellbeing (pp. 22e30). Eindhoven, The
Netherlands: Eindhoven University of Technology.
Linacre, J. M. (2006). Facets (Version 3.60.0) [Computer software]. Chicago:
Winsteps.com.
Loewen, L. J., Steel, G. D., & Suedfeld, P. (1993). Perceived safety from crime in the
urban environment. Journal of Environmental Psychology, 13, 323e331.
Longcore, T., & Rich, C. (2004). Ecological light pollution. Frontiers in Ecology and the
Environment, 2, 191e198.
Nasar, J. L., Fisher, B., & Grannis, M. (1993). Proximate physical cues to fear of crime.
Landscape and Urban Planning, 26, 161e178.
Nasar, J. L., & Jones, K. M. (1997). Landscapes of fear and stress. Environment and
Behavior, 29, 291e323.
Navara, K. J., & Nelson, R. J. (2007). The dark side of light at night: Physiological,
epidemiological and ecological consequences. Journal of Pineal Research, 4,
215e224.
Pease, K. (1999). A review of street lighting evaluations: Crime reduction effects. In
K. A. Painter, & N. Tilley (Eds.), Crime prevention studies. Surveillance of public
space: CCTV, street lighting and crime prevention, Vol. 10 (pp. 47e76). Monsey, NY:
Criminal Justice Press.
Royal Philips Electronics. (2012). Product catalogue for CitySoul BGP431 GRN88 LED
luminaire.
Retrieved
from.
http://www.ecat.lighting.philips.com/l/led/
functional-lighting/citysoul-led/910503886418_eu/.
Runge, T. E., Frey, D., Gollwitzer, P. M., Helmreich, R. L., & Spence, J. (1981).
Masculine (instrumental) and feminine (expressive) traits: A comparison
between students in the United States and West Germany. Journal of CrossCultural Psychology, 12, 142e162.
Schumacker, R. E., & Lunz, M. E. (1997). Interpreting the chi-square statistics
reported in the many-faceted Rasch model. Journal of Outcome Measurement,
1(3), 239e257.
Taskforce Verlichting. (2011). Eindrapport Taskforce Verlichting: Zuinig licht, heldere
keuze [Final report Taskforce Lighting: Frugal light, bright choice] (Publication
No. 2TFVL1104). Utrecht, The Netherlands: Agentschap NL.
Vrij, A., & Winkel, F. W. (1991). Characteristics of the built environment and fear of
crime: A research note on interventions in unsafe locations. Deviant Behavior,
12, 49e67.
Welsh, B. P., & Farrington, D. C. (2008). Effects of improved street lighting on crime.
Campbell Systematic Reviews, 13.
van der Wurf, A., van Staalduinen, L., & Stringer, P. (1989). Fear of crime in residential environments: Testing a social psychological model. Journal of Social
Psychology, 129, 141e160.
Zhao, X., Lynch, J. G., Jr., & Chen, Q. (2010). Reconsidering Baron and Kenny: Myths
and truths about mediation analysis. Journal of Consumer Research, 37, 197e206.