Applied Ergonomics 59 (2017) 84e93

Contents lists available at ScienceDirect

Applied Ergonomics
journal homepage: www.elsevier.com/locate/apergo

Dissuasive exit signage for building fire evacuation

Joakim Olander a, Enrico Ronchi b, *, Ruggiero Lovreglio c, Daniel Nilsson b
a
Holmes Fire, New Zealand
b
Department of Fire Safety Engineering, Lund University, Sweden
c
Department of Civil and Environmental Engineering, Technical University of Bari, Italy

a r t i c l e i n f o a b s t r a c t

Article history: This work presents the result of a questionnaire study which investigates the design of dissuasive
Received 11 May 2016 emergency signage, i.e. signage conveying a message of not utilizing a specific exit door. The work an-
Received in revised form alyses and tests a set of key features of dissuasive emergency signage using the Theory of Affordances.
9 August 2016
The variables having the largest impact on observer preference, interpretation and noticeability of the
Accepted 26 August 2016
signage have been identified. Results show that features which clearly negate the exit-message of the
original positive exit signage are most effective, for instance a red X-marking placed across the entirety of
the exit signage conveys a clear dissuasive message. Other features of note are red flashing lights and
Keywords:
Dissuasive
alternation of colour. The sense of urgency conveyed by the sign is largely affected by sensory inputs such
Emergency signage as red flashing lights or other features which cause the signs to break the tendencies of normalcy.
Exit sign © 2016 Elsevier Ltd. All rights reserved.
Paired comparison
Evacuation
Dynamic signage
Route choice
Risk perception
Urgency

1. Introduction environmental features (Vilar et al., 2013).

The use of exit signage is an important component in people
way-finding in case of evacuation. In this context, evacuees can
have a tendency to move towards familiar routes (Sime, 1985). This
is associated with a possible sub-optimal use of emergency exits,
i.e. people tend to go towards the entrance of the building rather
than using emergency exits (Nilsson, 2009). However, evacuation
experiments in buildings have yielded results which show that
emergency signage can influence the behaviour of people during an
evacuation and that this depends on how visible the signs are.
Experimental work (Xie, 2011) showed that only 38% of people
actually detect conventional static emergency signage (e.g. based
on ISO standards (International Standards Organization, 2011)),
even if they are provided with unobstructed vision to signage
directly in front of them. The importance of emergency signage is
associated with the compliance of people with it (Duarte et al.,
2014; Vilar et al., 2014; Wogalter et al., 1993, 1989). Signage has
also been investigated in Virtual Reality in relationship with
* Corresponding author.
E-mail address: enricoronchi84@gmail.com (E. Ronchi).
A possible way to improve people awareness of the existence of
emergency signage is the use of flashing lights in the immediate
proximity of the emergency sign (Nilsson, 2009). This will alert
occupants of the specific emergency sign and in most cases influ-
ence the occupants to use that particular path for egress making it
possible to alter behaviours such as heading back down the known
path.
The use of for example flashing lights, or other dynamic features
incorporated into emergency signage is generally called dynamic
signage system. Experimental studies show that an increase from
38% to 77% in the visibility of the signage could be observed when
comparing dynamic signage systems to the more conventional
static emergency signage systems (Galea et al., 2014). This
conclusion was obtained by replicating an experiment performed
with static signs using a dynamic exit sign. The dynamic nature of
the sign was obtained using a set of LED lights which activated and
pointed towards the exit. In other words, a larger portion of the
egressing occupants is able to detect the sign and subsequently
follows the instructions provided by it. Processes that subsequently
impact observers' choice to follow signage are discussed frequently
within relevant literature. A common classification includes a four
step assessment based on the answer to these questions: Is the sign

http://dx.doi.org/10.1016/j.apergo.2016.08.029
0003-6870/© 2016 Elsevier Ltd. All rights reserved.
 J. Olander et al. / Applied Ergonomics 59 (2017) 84e93
visible to the observer? Is the sign noticed by the observer? Is the
sign correctly interpreted by the observer? Does action occur due to
the signage impacting the observer? (Ronchi et al., 2012; Xie, 2011).
The act of incorporating dynamic features into emergency
signage has been researched in different research projects. Exper-
imental studies investigated how egress route can be indicated
through the use of evacuation signage provided with lit green ar-
rows or red crosses (Bryant and Giachritsis, 2014). Another exper-
iment found that an egress route equipped with green flashing
lights was used more frequently than one without (Fridolf et al.,
2013). A virtual reality study showed that the design of flashing
lights impact the capability of exit signs to help people identifying
the existence of an egress route as well as route choice (Ronchi and
Nilsson, 2015). In addition, active dynamic emergency signage can
be used to both alert occupants of existing signage and also affect
their choice of egress path. These are systems capable of altering
which path is shown as the preferred egress route by altering the
meanings of the emergency signage (Galea et al., 2014). This can be
done by altering some characteristics of the sign to display a
negative or dissuasive message, with the goal of dissuading the use
of a particular path, and hence leading occupants down the egress
routes which are not marked as negative or dissuasive.
This paper aims to research and establish which type of emer-
gency signage is best suited to convey a negative or dissuasive
message to egressing occupants informing them that a certain
egress route should not be used. This is done by performing a
questionnaire study in which a set of features of selected dissuasive
signs have been evaluated. The questionnaire design has been
based on the Theory of Affordances (Gibson, 1977; Hartson, 2003).
The analysis of results has been done using both statistical tests and
a mixed ordered logit approach (Greene, 2010).
2. Methods
A study was carried out at Lund University (Olander, 2015) to
study the characteristics of dissuasive exit signage in which par-
ticipants had to evaluate dissuasive exit signage using a question-
naire. The questionnaire was designed using the Theory of
Affordances (Gibson, 1977; Hartson, 2003), specifically investi-
gating the affordances associated with different types of dissuasive
exit signage. Paired comparisons were used in the questionnaire to
assess the best layout for a dissuasive exit signage. This technique
allows having quantitatively comparison of different layouts
against each other and it provides the advantages of getting sta-
tistically robust results (Brown and Peterson, 2009; Noor, 2013).
Moreover, the questionnaire included a further task, in which
participants need to assess the functional affordance of several
dissuasive exit signage layouts using a 5-point Likert scale.
The evaluation of the impact of different features of the in-
stallations on the perceived affordance has been made using three
different methods, namely 1) binomial testing of the paired com-
parisons, 2) analysis of the Likert scale responses and 3) a mixed
ordered logit modelling approach.
2.1. The Theory of Affordances
The questionnaire that was used in this study is based on the
Theory of Affordances (Gibson, 1977). This theory assumes that an
object is perceived in relation to what it offers or affords an indi-
vidual. An affordance is, hence, what the object offers the individual
in relation to his or her goal. In other words, an affordance can be
used to explain how people perceive things that they sense. For
example, people would not only see an emergency exit as a door
with a sign on it, but that door is interpreted as a mean of reaching a
safe place. A modified version of this theory has been provided by
85
Hartson (Hartson, 2003). This later version assumes that the
assessment of what an object offers people can be made in relation
to different affordances, namely how an object can be (1) sensed
(i.e. Sensory Affordance), (2) understood (i.e. Cognitive Affordance),
(3) physically used (i.e. Physical Affordance), or (4) if it fulfil its
intended goal (i.e. Functional Affordance). Therefore, this extended
version of the theory allows researchers investigating the effec-
tiveness of different evacuation systems depending on the assis-
tance they offer to see, understand, use and fulfil the goal of the
systems.
By systematically exploring the sensory, cognitive, physical and
functional affordances provided by an evacuation system, it should
be possible to identify the advantages and disadvantages of an
evacuation system. Since the system under consideration is a visual
system (signage), physical affordance is not taken into consider-
ation in this instance.
The theory can be used to qualitatively analyse an array of
possible system designs to rule out the most appropriate system in
relation to a design goal. In addition, the theory has been used to
design an affordance-based questionnaire, which explores different
types of affordances in relation to the examined system.
2.2. Mixed ordered logit modelling
This work adopts a mixed ordered logit approach to investigate
the impact of different features of dissuasive exit signage on the
perceived functional affordances of this signage. Mixed ordered
logit models are a useful tool to achieve this goal since they provide
a quantitative statistical method to perform behavioural studies
aimed at obtaining a better understanding of the factors affecting
human behaviour and predicting behavioural responses (Lovreglio
et al., 2015). In particular, this modelling approach provides a
relevant methodology for capturing the sources of influence (in-
dependent variables) that explain the ordinal variable (dependent
variable) (Greene, 2010).
Let yi be an ordinal variable which can vary between 0 and m
(m 2 Nþ) for the i individual (i ¼ 1, …, k). It is possible to define an
unobserved (or latent) continuous variable (y*i ) defined in an utility
space characterized by threshold utility points (m0,…, mm1) which
is related to yi ordinal variable according to Equation (1). This latent
utility has an error component (ε) which is distributed as a logistic
random distribution with expected value (m) equal to zero and
variance (s2) equal to p2 =3 (Greene, 2010).
.pffiffiffi
y*i ¼ b$xi þ ε; ε  LðmjsÞ; m ¼ 0; s ¼ p 3
yi ¼ 0 if  ∞ < y*i  m0
(1)
yi ¼ 1 if m0 < y*i  m1
/
yi ¼ m if mm1 < y*i  þ∞
where xi ¼ {x1,i,x2,i,…,xn,i} is the n-components vector of indepen-
dent variables, b ¼ {b1,b2,…,bn} is the n-components vector of pa-
rameters (to be estimated), mi are threshold values for the m
ordered classes (to be estimated). Therefore, the parameters to be
estimated are n þ m  1 (one of the threshold values can be fixed as
reference point, i.e. m0 ¼ 0). Fig. 1 provides a graphical represen-
tation of the model described in Equation (1).
Since y*i is not a deterministic quantity, it is only possible to define
the probability that the i individual selects the y ordered value.
     
Pi ðyjbÞ ¼ P my1 < y*i < my ¼ L my  b$xi  L my1  b$xi ;
y ¼ 0; …; m
(2)
86 J. Olander et al. / Applied Ergonomics 59 (2017) 84e93
Fig. 1. Graphic representation of the ordered model described in Equation (1).
L() in Equation (2) is the logistic cumulative distribution func-
tion. The b parameters can have both fixed values and their own
probability distributions. The assumption of random parameters
allows estimating the mixed ordered logit model (or Random Pa-
rameters Ordered Logit model) (Greene, 2010). Considering the
probabilistic approach of this modelling analysis it is possible to
define a likelihood function to estimate the unknown parameters in
Equations (1) and (2) (Greene, 2010; Train, 2009). Therefore it is
possible an evaluation of whether and how a factor affects the
ordinal dependent variable depending on the p-values and the sign
of the b parameters.
3. Questionnaire study
An affordance-based questionnaire study was conducted to
investigate the features of dissuasive signage. This consisted in two
parts, namely 1) a set of paired comparisons of signage with
different features and 2) a Likert-scale type evaluation of different
installation setups.
3.1. Definition of signage features
The dissuasive emergency signage features were presented to
test participants in a virtually simulated environment. The envi-
ronment included dissuasive signs on the top of an egress door
placed within the corner of a hypothetical office space. The envi-
ronment was drawn using the 3D modelling software SketchUp™
(Trimble Navigation, Sunnyvale, USA), and imported into the Vir-
tual Reality game engine Unity3D™ (Unity Technologies, San
Francisco, USA) to incorporate lighting effects of the signage.
The selection of the specific installation setups was based on the
literature available on emergency signage designs (Arthur and
Passini, 1992; Duarte et al., 2014; Galea et al., 2014; Ronchi et al.,
2015, 2012; Xie, 2011). Further information on the selection of
the installation setups can be found in the report associated with
this paper (Olander, 2015). Additional features (e.g. auditory cues)
have not been taken into consideration in this work. The variables
under consideration during the experiment and the corresponding
installation setups are 1) Background colour (green plus red LED X-
markings), 2) Red flashing lights (yes or no) and 3) Sign (EXIT text
or running man pictogram).
The background colours under investigation were green
(including red LED X-markings) and red. Previous surveys have
shown that certain colours are associated with certain meanings
during egress scenarios (Nilsson et al., 2005; Troncoso et al., 2015)
and are highly dependent on the context upon which it is viewed
(Wickens, 2013). Green is generally associated with safety, red is
generally associated with danger while orange is generally associ-
ated with warning or nothing in particular (Nilsson et al., 2005).
Therefore, the colour red was further examined in this study within
various exit signage options to determine if it enhances the sign-
age's meaning. As the colour orange is most often associated with
warning but also with nothing in particular, it is presumed that red
offers better relation to the dissuasive message that is meant to be
conveyed. This is further enforced by virtual reality experiments
where only orange flashing lights were used to alert egressing
people of emergency signage, which showed no dissuasive results
or interpretation (Duarte et al., 2014). The impact of the green
background colour combined with red LED X-markings is also
investigated in this study. This combination has been studied in a
previous survey (Galea et al., 2014), showing that the marking
consisting of the X-marking covering the entirety of the sign pro-
duced the best results in terms of understanding that a certain
egress route was not to be used.
The survey also investigated the impact of red flashing lights.
This feature was included since incorporating dynamic elements
into conventional static emergency signage have shown potential
to increase the signage's visibility (Boer and van Wijngaarden,
2004; Galea et al., 2014; Nilsson et al., 2005; Ronchi et al., 2015).
Moreover, the flash rate selected was 1 Hz following the recom-
mendations presented in a previous research investigation (Ronchi
et al., 2015).
The selected pictograms for the signs included the running man
pictogram in line with the Swedish standard (AFS, 2008) and the
EXIT text layout in line with the text requirements of the New
Zealand Building Code, as found in NZBC Clause F8 (NZDBH, 2012).
The selected language for the text layout was English to create a
higher perceptible level of affordance due to the widespread use of
the English language within Sweden (where this experiment was
carried out) as well as allow people from various countries to
 J. Olander et al. / Applied Ergonomics 59 (2017) 84e93 87

participate in the survey.
3.2. Qualitative evaluation of installation setups
Considering the features previously presented, it was possible to
identify 8 (i.e. 23) different installation setups. In this study, only 5
installation setups were investigated (see Table 1). These installa-
tion setups were selected performing a qualitative evaluation of
their overall sensory, cognitive and functional affordance as indi-
cated in Table 1. This evaluation considered the way individual
features of the signage (e.g., type of message provided, colour cod-
ing, etc.) contributed to each individual affordance. A red sign with a
red X-mark was deemed to have very poor sensory affordance, so
this design has been discarded up-front given the lack of colour
contrast. When dynamic features (e.g. flashing lights) are used,
sensory affordance is expected to significantly increase due to
breaking the tendency of normalcy bias and drawing additional
attention to the signage. Sensory affordance is expected to heavily
rely on the signage's ability to stand out from the environment and
break the tendency of normality. The signs with no red flashing
lights have therefore been discarded since they were deemed to
have very low sensory affordance. The only exception was the sign in
which the running man was used in conjunction with the X-mark on
a green background since this was deemed to have a higher sensory
affordance than the corresponding case with a red background.
Detailed explanations on the choice of the five selected installations
are provided in the full report associated with this paper (Olander,
2015). The selected installation setups are illustrated in Fig. 2.
Test 2: IS3 vs. IS7. This test was performed for assessing the
impact of either providing a red background or a red LED X-mark
to the dissuasive emergency signage with running man
pictograms.
Test 3: IS3 vs. IS4. This test was performed for assessing the
impact of providing a running man pictogram or EXIT text to the
dissuasive emergency signage.
Test 4: IS4 vs. IS8. This test was performed for assessing the
impact of either providing a red background or a red LED X-mark
to the dissuasive emergency signage with EXIT text. This test
was also provided to prevent inconclusive outcomes if partici-
pants provide a specific set of answers to tests A, B and C.
The comparisons were chosen to investigate how a specific
feature has an impact on the perceived affordances. Comparisons
which would have investigated the same feature have not been
considered (for instance IS1 vs IS3 and IS1 vs IS4 would have
investigated the same feature, i.e. the impact of red flashing light,
thus only IS1 vs IS3 has been performed).
For each test, the participants were asked to answer five ques-
tions, assessing the efficiency of the signage, with grounds in the
Theory of Affordances.
The first question related to sensory affordance. The aim of this
question was to establish whether certain features provide a more
noticeable sign, as to heed the signage it must first be noticed. The
question was as follows:
(Question 1) Which of the two signs is the easiest to notice?

3.3. Questionnaire
Paired comparisons were performed to identify the impact of
different installation setups. This was achieved by administering
four pairs of installation setups to a set of volunteer test partici-
pants facing two screens in a lecture room. Virtually modelled signs
were shown in pairs on the screens. The paired comparison tests to
test the range of variables were as follow:
Test 1: IS1 vs IS3. This test was performed for assessing the
impact of either providing red flashing lights adjacent to the
dissuasive emergency signage or not.
The second question related to sensory affordance. The aim of
this question was to establish if any specific features enhance or
diminish the ability to clearly distinguish the markings on the
dissuasive emergency signage. The question was as follows:
(Question 2) In which of the two exit signs is it easiest to distin-
guish the details?
The third question related to cognitive affordance. The aim of
this question was to establish if any specific features increase the
dissuasive emergency signage's ability to convey a dissuasive
message causing observers to understand that evacuation should

Table 1
Selection of the setups (ISs) based on qualitative analysis of affordances (aff.).

IS Background Colour Red Flashing Lights Sign Sensory Aff. Cognitive Aff. Functional Aff. Screening

1 Green þ X-mark No Running Man Medium High Positive Selected

2 Green þ X-mark No EXIT Medium High Positive Discard
3 Green þ X-mark Yes Running Man Very High Very High Positive Selected
4 Green þ X-mark Yes EXIT Very High Very High Positive Selected
5 Red No Running Man Low Medium Positive Discard
6 Red No EXIT Low Medium Negative Discard
7 Red Yes Running Man High Medium Negative Selected
8 Red Yes EXIT High Medium Positive Selected

The selected installation setups are in Italic.

Fig. 2. The selected installation setups.

88 J. Olander et al. / Applied Ergonomics 59 (2017) 84e93
not occur via this specific route. The question was as follows:
(Question 3) Which of the two exit signs best conveys the message
of not using this exit route?
The fourth question related to functional affordance and pro-
vided information about which dissuasive emergency signage
design is most preferred for use by the observer. The question is as
follows:
(Question 4) Which of the two exit signs would you prefer be used
to advise you not to use an exit route?
The fifth question related to risk perception. The aim of the
question was to establish if any specific features cause a greater
sense of urgency or risk for the observer which would in turn cause
observers to be more aware of the risk of the emergency situation.
The word urgency was used instead of risk as it was expected that
the meaning of the term risk may vary greatly between participants
(Kinateder et al., 2014). The question is as follows:
(Question 5) Which of the two exit signs would result in the
greatest sense of urgency of the situation?
For all the above questions, participants were given the option
to choose sign A, sign B or that sign A and sign B are equivalent,
with the exception of the fourth question where only the option for
sign A or sign B is given. Participants were also given the option to
describe which features cause them to answer as they do.
In addition to these five questions, the participants were also
given the option to rate the tested exit signage's functional
affordance individually by answering the following Liker-scale type
question:
(Question 6) Please rate the signage options from 1 -5 based its
ability to tell you to not use an exit door (note that multiple signs
may receive same score) where 1 meant worse performance and 5
meant better performance.
The questionnaire also contained background questions to
establish participants' age, gender, nationality, colour blindness or
any eye-sight deficiencies and whether or not they have had pre-
vious experiences with fire drills or actual evacuations.
Pilot testing was done to validate the questionnaire, i.e. to
determine if participants would have any issues with understand-
ing the tests or questions asked. Comments from the pilot tests
were taken and incorporated into the final testing setup.
3.4. Participants
Participants were recruited by spreading the information of the
questionnaire study through social media and social groups. In
total, 46 participants' responses were collected. Participants con-
sisted majorly of students with only one identifying as a worker.
The age of the participants was on average 23 years old, with a
minimum of 19 and a maximum of 40 years. A total of 29 partici-
pants identified as Swedish while the remaining 17 identified as
people of other nationalities, with the second largest group being
Germans with 6 participants. On a scale of 1e5, where 1 is low and
5 is high, the participants rated their own English skill as an average
of 4.6, with a low of 3 and a high of 5. Previous fire emergency
experience of the participants was largely consisting of fire drills
(76%) with a few participants also having experience real scenario
evacuations (30%). Several participants stated having eye-sight
impairments such as near and far sightedness, however this was
in all cases corrected by glasses or contact lenses. A single partici-
pants noted colour blindness. Participants' demographics and
characteristics may have an impact on the perception and under-
standing of the dissuasive emergency signage. On one hand, this
aspect was not analysed in this work since the main objective was
the investigation of the impact of different features of dissuasive
exit signage on the perceived functional affordances of this signage.
On the other hand, the sample size was not sufficient to investigate
such an impact. Future studies are, therefore, necessary to inves-
tigate this issue.
3.5. Procedure
Participants were seated in a lecture hall facing two screens. On
the screens were projected images of the hypothetical emergency
scenario in which dissuasive emergency signage was present.
During the questionnaire study, the lighting in the room was turned
off for the shown images to be clearer. Virtually modelled signs
were shown in pairs on the screens and the participants were asked
to answer the questions described in Section 3.3 using a ques-
tionnaire sheets. Upon completion of the questionnaire study the
questionnaire sheets were collected from the participants. Partici-
pants were thanked and cinema tickets were distributed as
compensation for their time.
Three trials were carried out with different participants (Trial 1:
16 participants, Trial 2: 14 participants, Trial 3: 16 participants). The
order of tests was randomized for the different trials. In addition,
the position of the signage display on the screen (left or right) was
also randomized during the trials. This was done to minimize the
effects of questionnaire fatigue and possible sources of bias due to
the position of the signage on the screens.
For ethical purposes, participants of the questionnaire study
were notified of their right to withdraw from the questionnaire
study at the start of the questionnaire procedure. Upon with-
drawing, their questionnaire sheet would be discarded and not
used. No participants chose to withdraw from the questionnaire
study. Then, contact information gathered upon recruiting partic-
ipants was used only to remind participants of the time and place
of the questionnaire study in the days leading up to the appointed
time. This contact information was not shared and subsequently
destroyed when the questionnaire study procedure was
completed.
4. Results
Three independent methods were used to perform a thorough
investigation into the questionnaire responses. The first method
consisted of binomial testing of the paired comparisons. The second
method used Wilcoxon-signed rank testing techniques to evaluate
the Likert scale results. The third method adopted a mixed ordered
logit modelling approach. The use of three independent methods to
evaluate questionnaire results is deemed to provide multiple in-
sights into the responses provided by the test participants.
4.1. Binomial testing
The results of the paired comparisons were statistically analysed
using binomial testing. The null hypothesis in this case was that
there is no preference for any one sign over another, in other words,
that the two signs of each test are expected to be preferred equally
often. Table 2 shows a summary of the binomial testing results of
the questionnaire. The number of responses tested varies for
questions 1, 2, 3 and 5 due to responses where no choice was made
or because the cases in which signs were considered equal had
been removed.
 J. Olander et al. / Applied Ergonomics 59 (2017) 84e93 89

Table 2
Summary of binomial testing results, where c ¼ frequency of answer, % ¼ percentage of answer and p ¼ probability of null hypotheses. Statistical significant values are
presented in bold. The number of observations varies between tests.

Question 1 Question 2 Question 3 Question 4 Question 5

c % p c % p c % p c % p c % p

Test 1
IS1 0 0 .00 14 64 .29 16 38 .16 18 39 .18 0 0 .00
IS3 45 100 8 36 26 62 28 61 40 100
Test 2
IS3 9 26 .01 7 17 .00 34 79 .16 34 74 .18 6 25 .00
IS7 25 74 34 83 9 21 12 26 18 75
Test 3
IS3 4 36 .55 10 33 .10 12 46 .85 22 48 .88 5 45 1.00
IS4 7 64 20 67 14 54 24 52 6 55
Test 4
IS4 11 30 .02 5 12 .00 37 86 .00 37 80 .00 8 23 .00
IS8 26 70 37 88 6 14 9 20 27 77

As statistical significance is checked within a 95% confidence urgency than a sign with green background and red X-markings.
interval and a Bonferroni correction is applied, a p-value lower than Comments from the questionnaire include that red is associated
.0125 (ai ¼ 0.0125 ¼ 0.05/n, where n is the number of tests) results with danger and alarm.
in a statistically significant difference from the null hypotheses. A
summary of the statistically significant differences and a descrip-
tion of what this implies are presented here.
4.2. Likert-scale results and analysis
Test 1/Question 1: A sign with pair of flashing red lights placed
adjacent to it is considered easier to notice than a similar sign The questionnaire also included the option for participants to
not provided with flashing red lights. rate the installation setups individually on a Likert scale from 1 to 5,
Test 1/Question 5: A sign with pair of flashing red lights placed where 1 was related to worse performance and 5 was related to
adjacent to it is considered to create a greater sense of urgency better performance (i.e. functional affordance). Descriptive statis-
than a similar sign not provided with flashing red light. tics are presented in Table 3. Results were further analysed using
Test 2/Question 1: A sign with red background and no addi- Wilcoxon signed-rank tests (see Table 4) by using IS3 as benchmark
tional markings is considered easier to notice than a sign with design. The choice of IS3 as benchmark design is driven by the
green background and red X-markings. suggestion of its use as effective dissuasive signage in previous
Test 2/Question 2: It is considered easier to distinguish the research (Galea et al., 2014).
details of a sign with red background and no additional mark- Applying Bonferroni corrections (ai ¼ 0.0125), statistically sig-
ings, as opposed to a sign with green background and a red X- nificant differences can be observed in two cases. IS3 is rated to
marking. Comments from the questionnaire include that a red perform better than IS7 and IS8. This result indicates that the
X-marking causes the pictogram of IS3 to become difficult to signage with a green background combined with red LED X-
distinguish. markings provides higher functional affordance than a signage with
Test 2/Question 3: A sign with green background and a red X- red background. In addition, a marginal trend may be seen showing
marking is considered more appropriate at conveying the that IS3 also performs better than IS1 (this could not be proven
message of not utilizing a specific exit door as opposed to a sign statistically significant). In other words, the presence of red flashing
with red background and no additional markings. Comments lights improves the functional affordance of a signage. Finally, no
from the questionnaire include that a red sign with no addi- statistically significant difference was found between IS3 and IS4.
tional markings is easily mistaken for a regular but oddly This result indicates that the type of sign has negligible impact on
colored exit sign. the functional affordance.
Test 4/Question 2: It is considered easier to distinguish the The presence of red flashing lights and the impact of the type of
details of a sign with red background and no additional mark- sign and background colour on functional affordance were inves-
ings, as opposed to a sign with green background and a red X- tigated estimating a mixed ordered logit model. The independent
marking. Comments from the questionnaire include that a red variables included in this model are the dummy variables while the
X-marking causes the text of sign IS4 to become difficult to dependent variable is the rank stated for the functional affordance
distinguish. for each installation setups (see Question 6 in Section 3.3), which
Test 4/Question 3: A sign with green background and red X- identify five functional affordance classes (class 1: very low func-
marking is considered more apt at conveying the message of tional affordance, …, class 5: very high functional affordance). The
not utilizing a specific exit door as opposed to a sign with red dummy variable B is equal to 1 when the background colour is red,
background and no additional markings. Comments from the
questionnaire include that a red sign with no additional Table 3
markings is easily mistaken for a regular but oddly colored exit Descriptive statistics of Likert scale testing.
sign.
Installation Setups N Mean Standard Dev. Min Max
Test 4/Question 4: A sign with green background and red X-
marking is preferred over a sign with red background and no IS1 46 3.11 1.35 1 5
IS3 46 3.59 1.10 1 5
additional markings to defer from exit use. IS4 46 3.65 0.99 1 5
Test 4/Question 5: A sign with red background and no addi- IS7 46 2.20 1.22 1 5
tional markings is considered to create a greater sense of IS8 46 2.02 1.29 1 5
90 J. Olander et al. / Applied Ergonomics 59 (2017) 84e93

Table 4
Wilcoxon signed rank test results of Likert scale data and effect size (r).

Comparison N Mean Rank Sum of Ranks Z P r

IS3 - IS1 Negative Ranks 14 18.86 264 2.061a .04 .21

Positive Ranks 26 21.38 556
Ties 6
IS3 - IS4 Negative Ranks 16 14.22 227.5 0.224b .82 .02
Positive Ranks 13 15.96 207.5
Ties 17
IS3 - IS7 Negative Ranks 7 13.29 93 4.425a .00 .46
Positive Ranks 34 22.59 768
Ties 5
IS3 - IS8 Negative Ranks 7 13.14 92 4.665a .00 .49
Positive Ranks 36 23.72 854
Ties 3

Where:
Negative Ranks: IS3-ISx < 0; Positive Ranks: IS3-ISx > 0; Ties: IS3-ISx ¼ 0 with x ¼ 1,4,7,8.
a
Based on negative ranks.
b
Based on positive ranks.

the dummy variable R is equal to 1 when there are red flashing

lights and the dummy variable S is equal to 1 when the sign is EXIT
Text.
This means that 230 observations (i.e. 5 types of setups  46
participants) can be used to estimate the mixed ordered logit
model.
Equation (3) illustrates the specification of the latent utility
function which represents the functional affordance in relation to
the background colour (B), the presence of red flashing lights (R)
and the type of sign (S). The values for b, r and s re normally
distributed, i.e. b  Nðmb ; s2b Þ, r  Nðmr ; s2r Þ and s  Nðms ; s2s Þ.

y* ¼ bB þ rR þ sS þ const þ ε
The model specified in Equation (3) has been calibrated using a
number of 300 Halton draws to simulate random distributed pa-
rameters. Table 5 illustrates the modelling results as well as the
likelihood for a model including only a constant (L0), the likelihood
for the proposed model (LM). Using a likelihood ratio test to
compare the proposed model with the constant model, it is
possible to observe that the proposed model introduces a signifi-
cant improvement in terms of fitting (p ≪ 0.05).
Given the definition of the functional affordance classes, Table 5
indicates that the variables having positive parameters (i.e. means
for random parameters) increase the probability to have a higher
Table 5
Estimated mixed ordered logit model.
Number of Respondents: 46
Number of Observations: 230
L0 ¼ 367.801
LM ¼ 333.530
AdjR2 ¼ 0.07a
Variable Description Coeff.
(3)
Fig. 3. Probability Density Function (PDF) of the b parameter in the mixed ordered
logit model.
degree of the perceived functional affordance whereas a negative
one reduces it. In other words, the presence of red flashing lights
increases functional affordance while the use of a red background
instead of a green background combined with red LED X-markings
decreases functional affordance. Moreover, the use of EXIT text as
sign instead of a running man pictogram has negligible impact on
the functional affordance since the p-value of ms is greater than .05.
Therefore, this model confirms the results inferred using the Wil-
coxon signed-rank tests.
Finally, the model indicates that the influence of the background
colour on the functional affordance is not equal among all the re-
spondents (sb has a p lower than .05) and it has the normal dis-
tribution illustrated in Fig. 3. The distribution illustrates that there
is a percentage (i.e. grey area) of respondents for whom the use of a
red background increases the functional affordance.
p

const Constant 2.438 .00

mb Mean associated with b 2.648 .00
mr Mean associated with r 0.793 .00
5. Discussion
ms Mean associated with s 0.117 .49
sb Standard deviation associated with b 1.678 .00 Based on the affordance-based questionnaire, this paper rep-
sr Standard deviation associated with r 0.019 .84 resents a systematic attempt to evaluate the features of dissuasive
ss Standard deviation associated with s 0.011 .94
emergency signage and provide recommendations on their design.
m0 Threshold between Class 1 and 2 0.000 Fixed param.
m1 Threshold between Class 2 and 3 1.783 .00 The ability of an exit sign to dissuade from using the exit should
m2 Threshold between Class 3 and 4 2.825 .00 be considered in the assessment of a dissuasive emergency sign
m3 Threshold between Class 4 and 5 4.263 .00 design. In addition to this, to be considered a feasible design, the
a
The adjusted McFadden's R2 is not a measurement of fitting as the R2 of the dissuasive emergency signage must also be able to display a regular
multivariate analysis. exit sign when required, to allow its implementation into new
 J. Olander et al. / Applied Ergonomics 59 (2017) 84e93
sophisticated evacuation systems and procedural solutions.
The results of the affordance-based questionnaire showed that
the cognitive affordance of the dissuasive signage was of most
importance as it directly influenced the functional affordance (i.e.
whether you use the sign as intended or not, e.g., the results of Q3
and Q4 match) and the final preference of signage amongst par-
ticipants. Therefore, features which increase the cognitive afford-
ance of the signage should be considered a priority when designing
a dissuasive emergency sign. This is in agreement with the findings
of Galea et al. (2014) who found that signs with dissuasive markings
which were not able to convey a sufficient enough cognitive mes-
sage were misunderstood by observers (i.e. only crossing out
certain areas of a sign only negated certain parts of it, such as di-
rection or movement speed).
Since the whole act of creating a dissuasive emergency signage
starts with the use of a regular exit sign, the most important factor
becomes making observers understand that the sign no longer
conveys an EXIT HERE message, but instead is supposed to convey a
DO NOT EXIT HERE message. Sufficient alteration to the sign should
be provided to convey this message. Therefore, once again, the
cognitive affordance of the sign is proven to be of great importance
when designing a dissuasive emergency sign.
However, the need to also provide a sufficient amount of sen-
sory affordance to the dissuasive signs must not be disregarded. For
dissuasive emergency signage to also be effective the signage must
in some way break the tendency of normality to become visible to
the passing would-be observers (Nilsson, 2009). The importance of
sensory affordance plays a key role towards this goal, although it is
of secondary importance if compared to cognitive affordance in the
overall functional affordance scores. This confirms the findings of
the experiments conducted by Frantzich (2001), relating to how
people fail to notice exit signs and instead follow known paths.
The sensory affordance of the signs was attempted to be
increased by the use of red flashing lights placed adjacent to the
signage. Results show that an installation setup with a red back-
ground instead of a green background provided an increase in
sensory affordance due to being both more noticeable and easier to
see details. This does not conform to the findings of Wong and Lo
(2007) who found that a green and white sign was more visible
than a red and white sign. It may be the result of the red LED X
markings causing observers to have a more difficult time viewing
the green sign, than if these LED markings would not be present.
However, due to providing an unambiguous message this signage
option was quickly discarded as a preferred option for use, once
again enforcing the notion that cognitive affordance is of the most
importance for discerning functional affordance and preference of
use.
Within the present work, an attempt was made to increase the
cognitive affordance of the signage options in different ways: The
incorporation of a red LED X-marking over the signage, the use of a
red background instead of a green background and the use of red
flashing lights adjacent to the signage. Comments to the ques-
tionnaire revealed that quite a few participants would have wanted
the red LED X-markings to be more full, or even fully drawn lines
over the signs for increased visibility and understanding. However,
in an attempt to come up with a feasible solution for real dynamic
signage the ‘dotted’ LED X-marking was chosen, due to the possi-
bility to actually incorporate this into real current exit signage
without reducing their primary efficiency of showing current exit
routes. Providing fully drawn lines may cause the original message
of the sign to be obscured even when the dynamic portion of it is
not activated.
Only providing a red background instead of a green background
or only providing red flashing lights adjacent to the signage seem to
cause observers to rethink the message of the sign but the final
91
results vary heavily. This is thought to be because signs only pro-
vided with these features have not properly negated the original
EXIT HERE message and are instead providing an ambiguous
message, stating either danger or warning while at the same time
still stating EXIT HERE. A few participants also noted that this
configuration caused them to believe that the exit sign was urging
them to exit through the door faster, which is the opposite of what
was intended.
Another evident trend is that the use of green background
(including red LED X-markings) seems to perform better than a red
background in terms of functional affordance. In fact, both the
binomial analysis (see tests 2 and 3) and Likert-scale result analysis
(see comparisons IS3 - IS7 and IS3 e IS8 as well as the mb parameter
of the mixed ordered logit) illustrate that signage with green
background and red X-marking is considered more apt at
conveying the message of not utilizing a specific exit door. Despite
this trend, the results of the binomial analysis show that signage
with red background provide a higher sensory affordance, since it is
considered easier to notice and allows an easier identification of the
details. Moreover, the use of a red background creates a greater
sense of urgency than a sign with green background and red X-
markings. This last result can be explained by the fact that red
commonly is associated with danger and alarm (Nilsson et al.,
2005).
The Likert-scale result analysis suggests the use of red flashing
lights to increase functional affordance. Moreover, signage with red
flashing lights is considered easier to notice (i.e. higher sensory
affordance) and provide a greater sense of urgency.
The analysis of the signs under consideration demonstrated that
there is no significant difference between the use of EXIT text and a
running man pictogram as sign in terms of both functional and
sensory affordance.
It is important to note that the current paper has investigated
only the design of visual cues. Future research should consider
multi-modal warnings in line with the investigations performed in
the area of human factors (Baldwin et al., 2012; Haas and van Erp,
2014). In fact, in some instances auditory cues have been proven
useful in communicating urgency (Edworthy, 1994) or guiding
people to safety (Fridolf et al., 2013).
A discussion should be made on the impact of the sample under
consideration in the present study, i.e., it was mostly made by
Swedish students with a consistent experience in evacuation drills.
The scope of the study was the evaluation of visual evacuation
systems, thus the sample characteristics are deemed to allow a
generalizability of results for able-bodied adults. Nevertheless, re-
sults are deemed to have no external validity if the population
under consideration includes people with sight impairments (given
the low number of people declaring this type of impairment).
Similarly, since a European sample has been investigated in this
sample, the use of these findings should consider the fact that
colours may have different meanings in different cultures (Nilsson
et al., 2005; Wickens and Hollands, 2000), i.e., the applicability of
findings should be carefully reviewed in light of the cultural
background/context of the population (e.g., emergency exit signs
are written in red rather than green in the USA, with a potential
impact of this on the answers of a population including participants
from there). In addition, since the present study compared only
different designs for dissuasive signage, future research should also
investigate the use of those designs in contrast with signage which
encourage the use of a certain evacuation route.
The use of a questionnaire study as a research method permitted
the obtainment of a significant amount of data (46  4 binomial
comparison and 46  5 ranks) with good cost-effectiveness and
ecological validity. Moreover, the selected research method allowed
high experimental control and repeatability of the same scenarios
92 J. Olander et al. / Applied Ergonomics 59 (2017) 84e93
for different participants. Nevertheless, future studies should focus
on the collection of objective behavioural data in addition to the
subjective preference in order to further confirm the findings of the
paper. For instance, future efforts may focus on the investigation of
a change in meaning of a sign (from safe to not safe) adopting an
objective measures (e.g. reaction time and proportion of correct
answers). Such future research study is deemed to provide useful
insights into the features which facilitate the transition between
usable exits to unsafe ones.
Questionnaire results based on stated-preference presents a set
of limitations. In fact, the ecological validity of data collected in
virtual settings is lower than data from a real emergency. For
instance, respondents would probably be under stress during a real
evacuation scenario and this might affect their interpretation of the
environment. Participants' awareness of taking part in an experi-
ment could bias the answers. Indeed, the respondents could
conform their answers to what they think should the correct
answer for the researcher (Lovreglio, 2016). Nevertheless, it should
be noted that the type of experiments performed present higher
experimental control if compared with online surveys, where the
researchers have limited possibility to make sure that the re-
spondents follow the planned experimental procedure.
From a methodological perspective, the use of the Theory of
Affordances for questionnaire design has shown to be suitable for a
detailed evaluation of signage design. This is in line with the usage
of this framework for other types of evacuation systems (Ronchi
et al., 2016, 2015). The use of a mixed ordered logit model
allowed a deep investigation of the relationships between the
features of an installation setup and its functional affordance.
Future studies should focus on investigating affordances of
different types of dissuasive emergency signage to provide a deeper
understanding on the effectiveness of other installation setups as
well as deepening the cultural impact and other participants'
characteristics (e.g. previous fire emergency experiences) on the
understanding of such installation setups. This work provides re-
sults which can be used for the design of future surveys and
behavioural studies investigating the impact of dissuasive emer-
gency signage (e.g. virtual reality experiments, online surveys). In
fact, the results of this study can be used as a starting information
point to develop new stated-preference survey based on Efficient
Design (Lovreglio et al., 2016).
The final aim of the area of research discussed in this paper is
therefore design optimization through information technology
(allowing comparison and prioritization among different evacua-
tion systems). The current study represents a significant method-
ological contribute for the assessment of optimal signage design.
6. Conclusion
Based on the results of a questionnaire study with 46 partici-
pants, it has been possible to obtaining a set of recommendations
that signage designers can use during the design stage of dissuasive
signage. The following main recommendations can be given:
1) Installation setups affecting cognitive affordance should be
prioritized since this affordance plays a major role in the
effectiveness of dissuasive signage.
2) The recommended background colour for dissuasive signage is
green (with red LED X-markings), if compared to a red
background.
3) The use of red flashing lights is recommended since it increases
sensory affordance.
4) No significant differences have been found over the two types of
message coding (the written text EXIT or a running man
pictogram) under consideration, so these two features may both
be used to design dissuasive signage.
References
AFS, 2008. AFS 2008:13 Skyltar Och Signaler. Arbetsmiljo € verket.
Arthur, P., Passini, R., 1992. Wayfinding: People, Signs, and Architecture. McGraw-
Hill Book Co, New York.
Baldwin, C.L., Spence, C., Bliss, J.P., Brill, J.C., Wogalter, M.S., Mayhorn, C.B.,
Ferris, T.K., 2012. Multimodal cueing: the relative benefits of the auditory, vi-
sual, and tactile channels in complex environments. Proc. Hum. Factors Ergon.
Soc. Annu. Meet. 56, 1431e1435. http://dx.doi.org/10.1177/1071181312561404.
Boer, L.C., van Wijngaarden, S.J., 2004. Directional Sound Evacuation from Smoke-
filled Tunnels. Presented at the Safe & Reliable Tunnels. Innovative European
Achievements, Prague, Czech Republic.
Brown, T.C., Peterson, G.L., 2009. An Enquiry into the Method of Paired Comparison:
Reliability, Scaling, and Thurstone's Law of Comparative Judgment.
Bryant, P., Giachritsis, G.L., 2014. The GETAWAY Project - Improving Passenger
Evacuation Techniques in Railway Stations. Transport Reasearch Arena, Paris.
Duarte, E., Rebelo, F., Teles, J., Wogalter, M.S., 2014. Behavioral compliance for dy-
namic versus static signs in an immersive virtual environment. Appl. Ergon. 45,
1367e1375. http://dx.doi.org/10.1016/j.apergo.2013.10.004.
Edworthy, J., 1994. The design and implementation of non-verbal auditory warn-
ings. Appl. Ergon. 25, 202e210. http://dx.doi.org/10.1016/0003-6870(94)90001-
9.
Frantzich, H., 2001. Experiments, occupant behaviour and response time e results
from evacuation. In: Proceedings of the Second International Symposium
HBIF2001. Presented at the Human Behaviour in Fire. Interscience Communi-
cations, pp. 159e165.
Fridolf, K., Ronchi, E., Nilsson, D., Frantzich, H., 2013. Movement speed and exit
choice in smoke-filled rail tunnels. Fire Saf. J. 59, 8e21. http://dx.doi.org/
10.1016/j.firesaf.2013.03.007.
Galea, E., Xie, H., Lawrence, P., 2014. Experimental and survey studies on the
effectiveness of dynamic signage systems. Fire Saf. Sci. 11, 1129e1143. http://
dx.doi.org/10.3801/IAFSS.FSS.11-1129.
Gibson, J.J., 1977. Perceiving, Acting, and Knowing: toward an Ecological Psychology.
Lawrence Erlbaum Associates; distributed by the Halsted Press Division. Wiley,
Hillsdale, N.J. New York.
Greene, W.H., 2010. Modeling Ordered Choices: a Primer. Cambridge University
Press, Cambridge, New York.
Haas, E.C., van Erp, J.B.F., 2014. Multimodal warnings to enhance risk communica-
tion and safety. Saf. Sci. 61, 29e35. http://dx.doi.org/10.1016/j.ssci.2013.07.011.
Hartson, H.R., 2003. Cognitive, physical, sensory, and functional affordances in
interaction design. Behav. Inf. Technol. 22, 315e338.
International Standards Organization, 2011. ISO 3864-1:2011 Graphical Symbols -
Safety Colours and Safety Signs.
Kinateder, M.T., Kuligowski, E.D., Reneke, P.K., Peacock, R.D., 2014. A Review of Risk
Perception in Building Fire Evacuation (No. NIST TN 1840). National Institute of
Standards and Technology.
Lovreglio, R., 2016. Modelling Decision-making in Fire Evacuation Using the
Random Utility Theory. Politecnico di Bari, Milano e Torino, Italy.
Lovreglio, R., Fonzone, A., dell’Olio, L., 2016. A mixed logit model for predicting exit
choice during building evacuations. Transp. Res. Part Policy Pract. 92, 59e75.
http://dx.doi.org/10.1016/j.tra.2016.06.018.
Lovreglio, R., Ronchi, E., Nilsson, D., 2015. A Mixed-ordered Logit Approach to
Investigate Correlations among Different Affordances in Fire Evacuation. In:
Boyce, K. (Ed.), Presented at the 6th Human Behaviour in Fire Symposium 2015.
Interscience, Downing College, Cambridge, UK, pp. 409e420. ISBN 978-0-
9933933-0-3.
Nilsson, D., 2009. Exit Choice in Fire Emergencies: Influencing Choice of Exit with
Flashing Lights. Dept. of Fire Safety Engineering and Systems Safety, Lund
University, Lund, Sweden.
Nilsson, D., Frantzich, H., Saunders, W., 2005. Coloured flashing lights to mark
emergency exits - experiences from evacuation experiments. Fire Saf. Sci. 8,
569e579. http://dx.doi.org/10.3801/IAFSS.FSS.8-569.
Noor, F., 2013. A bayesian look at the rao-kupper model for paired comparisons. Pak.
J. Stat. Oper. Res. 9, 241. http://dx.doi.org/10.18187/pjsor.v9i3.569.
NZDBH, 2012. Compliance Document for New Zealand Building Code e Clause F8 e
Signs. New Zealand Department of Building and Housing, Wellington, New
Zealand.
Olander, J., 2015. Comparative Study of Dissuasive Emergency Signage (No. 5491).
LUTVDG/TVBB. Departement of Fire Safety Engineering, Lund University, Lund,
Sweden.
Ronchi, E., Nilsson, D., 2015. A Virtual Reality Experiment on the Design of Flashing
Lights at Emergency Exit Portals for Road Tunnel Evacuations (No. 3180).
Departement of Fire Safety Engineering, Lund University, Lund [Sweden].
Ronchi, E., Nilsson, D., Gwynne, S.M.V., 2012. Modelling the impact of emergency
exit signs in tunnels. Fire Technol. 48, 961e988. http://dx.doi.org/10.1007/
s10694-012-0256-y.
Ronchi, E., Nilsson, D., Koji c, S., Eriksson, J., Lovreglio, R., Modig, H., Walter, A.L.,
2015. A virtual reality experiment on flashing lights at emergency exit portals
for road tunnel evacuation. Fire Technol. http://dx.doi.org/10.1007/s10694-015-
0462-5.
 J. Olander et al. / Applied Ergonomics 59 (2017) 84e93
Ronchi, E., Nilsson, D., Modig, H., Walter, A.L., 2016. Variable Message Signs for road
tunnel emergency evacuations. Appl. Ergon. 52, 253e264. http://dx.doi.org/
10.1016/j.apergo.2015.07.025.
Sime, J.D., 1985. Movement toward the familiar: person and place affiliation in a fire
entrapment setting. Environ. Behav. 17, 697e724. http://dx.doi.org/10.1177/
0013916585176003.
Train, K., 2009. Discrete Choice Methods with Simulation, second ed. Cambridge
University Press, Cambridge, New York.
Troncoso, J., Nilsson, D., Ronchi, E., 2015. Response to Emergency Way-finding
Systems by People from Different Cultures. Presented at the Human Behav-
iour in Fire Symposium. Interscience Communications, Cambridge, UK.
Vilar, E., Rebelo, F., Noriega, P., Duarte, E., Mayhorn, C.B., 2014. Effects of competing
environmental variables and signage on route-choices in simulated everyday
and emergency wayfinding situations. Ergonomics 57, 511e524. http://
dx.doi.org/10.1080/00140139.2014.895054.
Vilar, E., Rebelo, F., Noriega, P., Teles, J., Mayhorn, C., 2013. The influence of
93
environmental features on route selection in an emergency situation. Appl.
Ergon. 44, 618e627. http://dx.doi.org/10.1016/j.apergo.2012.12.002.
Wickens, C.D., 2013. Engineering Psychology and Human Performance, fourth ed.
Pearson, Boston.
Wickens, C.D., Hollands, J.G., 2000. Engineering Psychology and Human Perfor-
mance, third ed. Prentice Hall, Upper Saddle River, NJ.
Wogalter, M.S., Allison, S., Mckenna, N., 1989. Effects of cost and social influence on
warning compliance. Hum. Factors J. Hum. Factors Ergon. Soc. 31, 133e140.
Wogalter, M.S., Kalsher, M.J., Racicot, B.M., 1993. Behavioral compliance with
warnings: effects of voice, context, and location. Saf. Sci. 16, 637e654. http://
dx.doi.org/10.1016/0925-7535(93)90028-C.
Wong, L.T., Lo, K.C., 2007. Experimental study on visibility of exit signs in buildings.
Build. Environ. 42, 1836e1842. http://dx.doi.org/10.1016/j.buildenv.2006.02.011.
Xie, H., 2011. Investigation into the Interaction of People with Signage Systems and
its Implementation within Evacuation Models (Ph.D. thesis).