The effectiveness of evacuation signs in buildings based on eye tracking experiment

Abstract

It is of crucial importance whether or not evacuees follow the evacuation signs during building evacuation. It is necessary for effective evacuation signs to be designed in a way that helps the occupants follow building safety guidelines. In this paper, 658 evacuation experiments with wearable eye tracking devices were carried out to test the effectiveness of both the setting positions and design of evacuation signs in buildings. There are four factors considered in this paper: (1) position, (2) colour, (3) graphics, and (4) (or flashing). The results show that the effect of the green “arrow” evacuation sign is the best. The effect of low (corridor) signs works better than those of high (room) signs, but the signs of low (corridor) could be blocked by the front evacuees. Obtaining the ratio of people obeying the evacuation signs under different conditions provides an effective basis for the improvement in design for building safety, and provides data support for computer simulation modelling of crowd evacuation.

Full text

1 23
Natural Hazards
ISSN 0921-030X
Nat Hazards
DOI 10.1007/s11069-020-04030-8
The effectiveness of evacuation signs in
buildings based on eye tracking experiment
Ning Ding

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer
Nature B.V.. This e-offprint is for personal
use only and shall not be self-archived
in electronic repositories. If you wish to
self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com”.

Vol.:(0123456789)
Natural Hazards
https://doi.org/10.1007/s11069-020-04030-8
1 3
ORIGINAL PAPER
The eectiveness ofevacuation signs inbuildings based
oneye tracking experiment
NingDing1,2
Received: 22 August 2019 / Accepted: 25 April 2020
© Springer Nature B.V. 2020
Abstract
It is of crucial importance whether or not evacuees follow the evacuation signs during
building evacuation. It is necessary for effective evacuation signs to be designed in a way
that helps the occupants follow building safety guidelines. In this paper, 658 evacuation
experiments with wearable eye tracking devices were carried out to test the effectiveness
of both the setting positions and design of evacuation signs in buildings. There are four
factors considered in this paper: (1) position, (2) colour, (3) graphics, and (4) (or flashing).
The results show that the effect of the green “arrow” evacuation sign is the best. The effect
of low (corridor) signs works better than those of high (room) signs, but the signs of low
(corridor) could be blocked by the front evacuees. Obtaining the ratio of people obeying
the evacuation signs under different conditions provides an effective basis for the improve-
ment in design for building safety, and provides data support for computer simulation mod-
elling of crowd evacuation.
Keywords Evacuation signs· Experiment· Eye movement· Design of signs
1 Introduction
With the continuous development of modern architecture, both the structure and the func-
tion of a building are gradually becoming more complicated. Evacuation under emergency
situations in such buildings has become a critical issue. Complex indoor places, such as
supermarkets, movie theatres, and gymnasiums, are usually complex enclosed environ-
ments with internal obstacles and multiple exits. They have features such as large scene
area, complex internal structures, and multiple exits. Unlike pedestrian evacuation model-
ling in a simple environment (access, barrier-free rooms) whose focus is on the research of
operational level, pedestrian evacuation modelling in complex buildings needs to take into
consideration behavioural characteristics at the tactical level (Hoogendoorn and Daamen
* Ning Ding
dingning_thu@126.com
1 School ofCriminal Investigation andForensic Science, People’s Public Security University
ofChina, Beijing, China
2 Public Security Behaviral Science Lab, People’s Public Security University ofChina, Beijing,
China
Author's personal copy

Natural Hazards
1 3
2005). How do evacuees choose the route in the evacuation process? How do they make
decision? Do they follow the evacuation signs? Only by solving these problems can a
strong basis for the design of building evacuation signs be provided and evacuation simula-
tion models be established.
In general, it is necessary to hold evacuation experiments to study the evacuation behav-
iour of the occupants. The reason is that the evacuation data in fire disasters cannot be
collected or the collection is incomplete, and researchers cannot burn a building to con-
duct evacuation experiments. To analyse the evacuation behaviours, the experiments can
be divided into five categories: (1) total evacuation drills (Peacock etal. 2012; Huo etal.
2016; Xudong etal. 2009; Kobes etal. 2010), (2) part evacuation experiments (Ma etal.
2012; Fang etal. 2010; Fang etal. 2012; Liao etal. 2014), (3) controlled pedestrian experi-
ments (Isobe etal. 2004; Kretz etal. 2006; Seyfried etal. 2009; Zhang et al. 2013), (4)
computer simulation evacuation experiments (Liu etal. 2015; Wei etal. 2015; Helbing
etal. 2005; Hu etal. 2015), and (5) virtual reality experiments (Kinateder etal. 2014; Ron-
chi etal. 2016; Marsh etal. 2012; Lu etal. 2014). Beyond these methods, it also includes
the method of utilizing animals and insects for evacuation behaviours (Lin etal. 2016), and
some research particularly focuses on evacuation in panic situations (Shiwakoti and Sarvi
2013; Shiwakoti etal. 2011). Olander etal. (2017) analysed exit signage design based on
the theory of affordances. It is found that, when compared to a red background, recom-
mended background colour for dissuasive signage is green (with red LED X-markings). Fu
etal. (2019) tested the influence of emergency signage on building evacuation behaviour
based on both experiments and questionnaire survey. In addition to the research focused
on buildings, several studies tested the effectiveness of the disaster signage in public areas.
However, these experimental studies mainly focused on the microscopic and macroscopic
characteristics of pedestrian flow, and they did not fully grasp the decision-making process
of the evacuees. In order to further analyse the psychology and behaviour of the evacuees,
eye tracking devices were introduced into the evacuation experiments.
Eyes are an important way for humans to obtain information, and they can reflect the
psychological activities of individuals. When quantifying reading behaviour, there are three
basic types of human eye movement: fixation, saccades, and pursuit movement (Yan 2004).
Saccades are sudden changes in the gaze point or gaze orientation, which are often uncon-
scious by individuals. The saccade is fast, up to 450 degrees per second. Pursuit movement
refers to the relative movement of the observed object and the eye. In order to ensure that
the eye always looks at the object, the eyeball follows the object (Liu and Yuan 2000).
Fixation is the condition when the gaze remains fixed during an interval ranging between
100 and 1000ms on a surface < 144 mm2; however, during the fixation of a motionless
stimulus, the eye is not entirely motionless itself; micro-saccades and micro-tremors can
be registered (Larmande and Larmande 1990). The eye tracking device is used to extract
eye movement data—such as the fixation position, the fixation duration, the fixation count,
the saccade frequency, and the pupil size—so as to study the intrinsic cognitive process
of individuals. SMI’s BeGaze 3.0 eye movement data analysis software can automatically
generate heatmap, gaze plot, duster, bee swarm, etc., providing intuitive results.
Eye tracking methods can be used for cognitive research in reading and information pro-
cessing (Rayner 1998). Researchers have used eye trackers to experimentally explore the
relationship between map design and map reading. Williams used eye movement experi-
ments to explore the impact of different visual characteristics (size, colour, shape, etc.) on
map users when searching (Steinke 1987a). Dobson studied the differences in visual pro-
cesses between map readings of the test subjects in the case of marked and unmarked area
boundaries. And he also evaluated the main information content of visual acquisition, and
Author's personal copy

Natural Hazards
1 3
the estimation and decision-making process in visual process (Dobson 1979). Antes etal.
studies the impact of map attribute design on map reading behaviour, and effects of experi-
ence on the performance of map-reading tasks. They emphasized the role of map-reading
tasks and user experience and found that experienced map users seemed to be better able to
explore map content, and the duration of fixations was shortest for sensitized users and long-
est for laypersons (Antes etal. 1985). Castner and Eastman studied the impact of map com-
plexity on map reading and concluded that perceived complexity of maps affects the way
maps are read and understood (Castner and Eastman 1985). Focusing on natural hazards and
hazard and risk mapping, eye movement tracking has been increasingly applied to the study
of map perception and design (Steinke 1987b). Wolfe and Horowitz listed colour (hue), size,
and direction as variables that attracted visual attention (Wolfe and Horowitz 2004). Cau-
vin etal. (2007) found that different users also present different recognition patterns for the
same map. Hornsby etal. (2009) conducted experimented on the four most commonly used
visual variables: size, hue, lightness, and direction, and found that size is the most effective
visual variable and direction is the most ineffective visual variable in the flicker paradigm.
Netzel etal. (2016) studied the effects of subway map colour, complexity, and task difficulty
on map-reading strategies and reading results. Brychtova and Coltekin used eye movement
experiments to analyse the effect of colour distance and font size on the readability of the
map. The research results showed that the larger the colour distance, the better the readabil-
ity of the map, and the medium font had a higher search efficiency (Brychtova and Coltekin
2016). Wang etal. (2016) used eye tracking technology to find that maps designed with
image symbols are more readable. In recent years, eye tracking devices have also received
further attention in evacuation experiments. Cosma combined virtual reality (VR) technol-
ogy with an eye tracker to build a smoke-filled railway tunnel evacuations scene (Cosma
et al. 2016). A green LED lighting system was installed on the road surface. The crowd
path selection behaviour under different lighting conditions was studied through VR evacu-
ation experiments. It was found that the lighting system has a significant positive impact on
the safety evacuation of the crowd. Andree etal. (2016) built a virtual high-rise building
through VR technology. Through experiments, Andree etal. studied the export selection
behaviour of crowds during the evacuation process and the waiting time when using eleva-
tors for evacuation. It was found that green evacuation indicators could effectively guide
evacuees to choose elevators. However, there are still few experimental studies using eye
trackers in evacuation. Tang etal. (2009) test the effectiveness of three scenarios based on
VR tools: one without emergency signs, another with an old-version emergency sign, and
the third with a new-version emergency sign, and the results showed that both old and new
signs can help the participants to find the best way. Xie etal. (2012) tested the effective-
ness of the emergency signage based on a questionnaire survey, and the results are used to
enhance the existing signage model within the building EXODUS software.
In this paper, individual evacuation experiments with wearable eye tracking devices
were carried out to test the effectiveness of the evacuation signs, as well as help solve the
following problems:
1. How can an evacuation sign be placed to be noticed by more evacuees?
2. How can an evacuation sign be designed to be noticed by more evacuees? The design
factors include colour, graphics, and twinkling.
3. How many evacuees can follow the evacuation signs?
Author's personal copy

Natural Hazards
1 3
In order to study these issues, 658 experiments were carried out including 130 pre-
experiments and 528 formal experiments. The building layout and the procedure of the
experiments are introduced in Sect.2. The results of these experiments, including the ratio
of seeing and following the evacuation signs, and fixation duration average, are shown in
Sect.3. In the Sect. 4, the comparison of these results and some interesting phenomena
discovered during the experiments are discussed.
2 Experiment design
2.1 Building layout andexperiment devices
The experiments were carried out in a complex building in People’s Public Security Uni-
versity of China. This building is used for police and student training, and there are several
rooms and staircases. Most of the participants were not familiar with the building since the
building was not used for normal class education. Room 8 on second floor was used for
the experiments, and the layout of the second floor is presented in Fig.1. When a partici-
pant arrived at this floor, he/she was blindfolded, and an experimenter guided him/her to a
room. The eye mask was then removed, and the experimenter helped him/her wear the eye
tracking glasses, and the device is mobile eye tracking glasses produced by SensoMotoric
Instruments (the type of the device is SMI ETG). The eye tracking glasses are shown in
Fig. 1 Layout of the second floor
in the experimental building
Author's personal copy

Natural Hazards
1 3
Fig.2a, and the glasses are connected with a mobile phone that can be held in the hand
of the participant (shown in Fig.2b), which is used to calibrate the eye movement of the
participant and store experimental data. The participant can easily wear the glasses on his/
her face with the mobile phone in his/her pocket. After calibrating the eye movement, the
participant was asked to stand with their back to the door by 3m. When he/she heard the
alarm sound, the experiment was started immediately, and when the evacuee found the
stairwell, the experiment was terminated. Before the experiment, the participant was not
told to notice the evacuation signs. The data were analysed in SMI Begaze software (shown
in Fig.2c).
Before the formal experiments, several pre-experiments were carried out. There were 5
sub-experiments in the pre-experiment. Most of the evacuation signs are set either high or
low on the wall of a corridor. Evacuation signs need to be visible at night, so they usually
have built-in lights or self-illumination. The pre-experiments were used to test the posi-
tions of the evacuation signs and how the illumination influences the effect of the signs.
After pre-experiments, formal experiments were carried out to analyse the following fac-
tors: (1) the colour of the evacuation sign, (2) the graphics of the sign, and (3) whether it
was twinkling or not.
The participants of this experiment were all undergraduates in a university, and there
were 30–50 individuals that took part in each sub-experiment. In the formal experiments,
there were 16 sub-experiments with 528 participants in total. Before each experiment
began, the participant was asked if he knew which room he was in. If he knew the room,
the experiment was considered invalid.
aEye tracking device bParticipant with the glasses cSMI Begaze software
Fig. 2 Eye tracking devices and SMI Begaze software
Table 1 Procedure of pre-experiment
Pre-experi-
ment 0
Position Illumination Colour Graphic Twinkling
0.1 Low (corridor) No Green Running man No
0.2 High (corridor) No Green Running man No
0.3 Low (corridor) Yes Green Running man No
0.4 High (corridor) Ye s Green Running man No
0.5 High (room) Yes Green Running man No
Author's personal copy

Natural Hazards
1 3
2.2 Pre‑experiment andresults
As is shown in Table1, Experiment 0.1–0.2 was used to test the position factor without
illumination. The low position is 30cm higher than the ground, and the sign is right oppo-
site to the door of the experimental room. The high position is 200 cm higher than the
ground. After calibrating the eye movement, the participant was asked to stand 3m away
from the door of the experimental room. During the experiment, the door was kept open.
Accordingly, the participant was facing to the sign at the beginning of the experiment. The
sign used in the pre-experiment is as shown in Fig.3a, and this green running man sign
was commonly utilized in China. The positions are shown in Fig.3b.
The results of Experiment 0.1–0.2 are shown in Table2. For the low and high posi-
tion, the percentages of the participants who could see the signs were 42.55% and
16.67%, respectively. In Experiment 0.3 and 0.4, a light was added into the signs of
both low and high location to be compared with Experiment 0.1. As it was difficult to
notice the signs with high position in the corridor, the position of an illuminate evacu-
ation sign was changed to the high position in the room in Experiment 0.5. There were
50, 50, 33, 30, and 33 participants in Experiment 0.1–0.5, respectively. The number of
valid participants (NP), number of the participants who saw the evacuation sign (NS),
a
Evacuation sign used in the pre -evacuation
b
Positions of the evacuation signs
Fig. 3 Positions of the evacuation signs
Table 2 Results of pre-experiments
Exp NP Position Illumination Sign NS NF Fixation
duration(s)
Mean S2
0.1 47 Low (corridor) No Running man 20
42.55%
12
60.00%
0.684 0.219
0.2 42 High (corridor) No Running man 7
16.67%
5
71.43%
0.526 0.005
0.3 33 Low (corridor) Ye s Running man 26
78.79% 19
73.08%
0.685 0.113
0.4 30 High (corridor) Yes Running man 4
13.33%
2
50%
0.349 0.215
0.5 33 High (room) Yes Running man 11
33.33%
11
100%
0.518 0.061
Author's personal copy

Natural Hazards
1 3
number of the participants who saw the evacuation sign and followed (NF), as well as
the average and variance of fixation duration are shown in Table2. It was found that the
sign of low position with illumination is the best in the pre-experiments, and 78.79% of
the participants can see the evacuation signs. Among them, 73.08% chose to follow the
guidance of signs. Although the ratio of participants who saw the signs in Experiment
0.5 is 33.33%, the ratio of following is 100%. As a result, in the formal experiments, two
positions are tested: low position in the corridor and high position in the room.
2.3 Experiment design
In the pre-experiment, it was found that the evacuation signs in the low (corridor) and
the high (room) set in the room door frame are more easily noticed by the evacuation
participants. Therefore, in the formal experiment, the position of the evacuation signs
was only considered for two situations: low (corridor) and high (room). In the formal
experiment, all of the built-in lights within the evacuation signs were turned on, it was
conducted during the daytime, and the light was sufficient. Three factors were taken into
consideration: (1) the colour of the evacuation sign, (2) the graphics of the sign, and
(3) whether or not it was twinkling. Among them, the colours were red and green, and
the graphics of the sign were “running man” and “arrow”. When twinkling, the built-in
lights of the signs were 0.5s off and 0.5s on. Therefore, for the low (corridor) and the
high (room) signs, 8 sets of experiments were, respectively, carried out. In those 16
experiments, each group was performed in one class, which generally consisted of about
Table 3 Experiment design
Experiment 1 Colour Sign Twinkling Position Illumination
1.1 (0.3) Green Running man No Low (corridor) Yes
1.2 Red Running man No Low (corridor) Yes
1.3 Green Arrow No Low (corridor) Yes
1.4 Red Arrow No Low (corridor) Yes
1.5 Green Running man Yes Low (corridor) Yes
1.6 Red Running man Yes Low (corridor) Ye s
1.7 Green Arrow Ye s Low (corridor) Yes
1.8 Red Arrow Yes Low (corridor) Ye s
Experiment 2 Colour Sign Twinkling Position Illumination
2.1 (0.4) Green Running man No High (room) Yes
2.2 Red Running man No High (room) Yes
2.3 Green Arrow No High (room) Yes
2.4 Red Arrow No High (room) Yes
2.5 Green Running man Yes High (room) Ye s
2.6 Red Running man Yes High (room) Ye s
2.7 Green Arrow Ye s High (room) Ye s
2.8 Red Arrow Yes High (room) Ye s
Author's personal copy

Natural Hazards
1 3
33 people. The experimental sequence is shown in Table3. The signs used are shown in
Fig.4.
3 Results
3.1 Analysis standard
Eye tracking devices are used during evacuation. Because the background is dynamic, the
analysis method is different from the static background. In general, in Experiments 1 and
2, eye movement characteristics of evacuees during evacuation can be divided into two
categories: (1) the evacuee did not see the evacuation signs; and (2) the evacuee saw the
evacuation signs. Then, the evacuees who saw the evacuation signs can be divided into
whether or not they followed the guidance direction of the signs. The data recorded by
the eye tracking device are when and how long did a participant gaze at the sign. In the
experiments, some of the eye movement data from the participants were invalid for two
main reasons: (1) the eye tracking glasses moved during evacuation, or (2) they knew the
location of the experimental room. When the eye movement falls on the indicator signs, it
is considered as the evacuee has seen the sign.
3.2 Results ofexperiment 1 and2
The experimental results include the following: the number of valid participants (NP), the
ratio of people who saw the signs (NS), the ratio of people who obeyed the direction of the
signs (NF), mean, the average time, and variance of the participants seeing the signs (S2).
The experimental results are shown in Table4. A total of 528 people participated in the
experiment, of which 483 were valid and 45 were invalid.
As can be seen from Table4, among the evacuation signs of low (corridor), the green
non-twinkling “running man” signs were seen at the highest ratio: 86.67%, and then, the
green twinkling “arrow” sign received the second highest ratio of 79.31%. In the follow-
ing proportions or NF, the highest is also the green non-twinkling “arrow” sign, whose
proportion is 96%, but its fixation duration average is not too long. The sign that had the
longest fixation duration average was the red twinkling “running man” one, but its NF is
only 69.57%.
In Experiment 2, the NS is lower than that of Experiment 1, so considering the pre-
experiment of the evacuation signs placed above the corridor, it could indicate that the
evacuation sign placed above is not easily seen by the participants. The sign having the
highest NS in Experiment 2 was the green twinkling “arrow” sign (76.67%), but its NF
aGreen running man b Red running man cGreed arrow dRed arrow
Fig. 4 Evacuation signs
Author's personal copy

Natural Hazards
1 3
ratio was not the highest (91.30%). For some signs, although the ratio of NS in Experiment
2 is not high, their NF ratio is high: the ratio of the green non-twinkling “running man”
sign, the green non-twinkling “arrow” sign, and the red non-twinkling “arrow” sign were
all 100%.
The fixation duration average in Experiment 2 was shorter than that in Experiment 1.
The longest fixation duration average was the red twinkling “running man” sign, which
is consistent in both two experiments. The fixation duration average in Experiment 2
was 0.46s, which was 35.62% shorter than that in Experiment 1. It is noticed that the S2
Table 4 Results of experiment 1 and 2
Exp NP Colour Twinkling Sign NS NF Fixation duration (s)
Mean S2
1.1 30 Green No Running man 26
86.67% 19
73.08%
0.67 0.14
1.2 31 Red No Running man 19
61.29%
15
78.95%
0.74 0.12
1.3 30 Green No Arrow 25
83.33%
24
96.00%
0.66 0.12
1.4 32 Red No Arrow 20
62.50%
17
85.00%
0.72 0.61
1.5 27 Green Yes Running man 20
74.07%
18
90.00%
0.63 0.14
1.6 31 Red Yes Running man 23
74.19%
16
69.57%
0.85 0.32
1.7 29 Green Yes Ar row 23
79.31%
21
91.30%
0.65 0.17
1.8 32 Red Yes Arrow 25
78.13%
20
80.00%
0.78 0.25
Exp NP Colour Twinkling Sign NS NF Fixation
duration (s)
Experiment
Mean S2
2.1 30 Green No Running man 11
36.67%
11
100%
0.52 0.06
2.2 30 Red No Running man 15
50.00%
14
93.33%
0.51 0.12
2.3 29 Green No Arrow 20
68.97%
20
100%
0.35 0.02
2.4 32 Red No Arrow 15
46.88%
15
100.00%
0.34 0.03
2.5 29 Green Yes Running man 14
48.28%
9
64.29%
0.40 0.07
2.6 29 Red Yes Running man 15
51.72%
12
80.00%
0.70 0.31
2.7 30 Green Yes Ar row 23
76.67% 21
91.30%
0.42 0.03
2.8 32 Red Yes Arrow 20
62.50%
16
80.00%
0.43 0.06
Author's personal copy

Natural Hazards
1 3
of 1.6 and 2.6 is larger than the rest of the experiments, respectively. The reason maybe
“red” means do not enter in people’s mind, on the contrary, the arrow shows a clear
direction to the participant. Some of the participants were not sure how to make deci-
sions. As a result, this type of signage is stable.
4 Comparison anddiscussion
4.1 Comparison
(1) For the sign below, the green non-twinkling “running man” sign works best.
The comparison of the ratio of seeing the evacuation signs in Experiment 1 can be
observed in Fig.5. For the “running man” and “arrow” sign, both of the two green signs
could be easily seen when there was no twinkling. The twinkling “arrow” sign was
and the non-twinkling green sign were easier to see. In the case of non-twinkling, the
ratio of the green sign being seen is significantly higher than that of the red sign, and
the difference of the gaze ratio between the “running man” sign and the “arrow” sign is
small. In the case of twinkling, the proportions in which the green and red evacuation
signs were gazed at were substantially the same, while the gaze ratio of the “arrow”
sign was higher than the “running man” sign.
For the low sign, the green non-twinkling “arrow” sign was seen significantly more
easily than the green twinkling “running man” sign, and the evacuator is more likely
to follow the green evacuation sign with a clear direction. In Experiment 1, the ratio
of obeying the evacuation signs (NF) is shown in Fig.6. Among the people who saw
the evacuation signs, the ratio of following the green “arrow” sign was higher than the
ratio of the red “arrow” sign. The green non-twinkling “arrow” sign had the highest
ratio of obeying the sign (NF). Among the non-twinkling signs, the green “arrow”
is significantly higher than the green “running man” in NF ratio. The reason may be
Fig. 5 Comparison on the ratio of seeing the evacuation signs in experiment 1
Author's personal copy

Natural Hazards
1 3
attributable to the fact that the “arrow” can give a clear direction indication. When the
decision time is short, clear direction signs worked better.
The NS and NF in Experiment 1 are shown in Fig.7. It is obvious that the green
non-twinkling “arrow” sign in Sub-experiment 1.3 works best, and both of the NS and
NF were the highest. Interestingly, although the NS of green non-twinkling sign was
higher, the NF is not good, which is the lowest of all NSs. Once again, this validates the
effect of clear direction signs on obeying evacuation instructions. In all experiments,
Fig. 6 Comparison on the ratio of following the guidance in experiment 1
Fig. 7 Comparison on the ratio of seeing sign and following the guidance in experiment 1
Author's personal copy

Natural Hazards
1 3
whether the signs were twinkling or not, the ratio of the red “running man” being gazed
at was low, and its evacuation indication effect was relatively poor.
Compared to the green sign,the red one did not work out so well, which is consist-
ent with the conclusion of Ref. (Olander etal. 2017). It was determined that, when
compared to a red background, the recommended background colour for dissuasive
signage is green (with red LED X-markings) (Olander etal. 2017).
(2) For the signs above, the green “arrow” sign was noticed at a higher ratio
In Experiment 2, the ratios of the NSs are shown in Fig.8. Among the evacuation
signs set above the door, the NS of green “arrow” sign was high, and the highest ratio
is the green twinkling “arrow” sign (76.67%). For the green evacuation sign, twinkling
could increase the NS. Among the “running man” signs, red was more likely to be seen
than green. In contrast, among the “arrow” signs, green was more likely to be seen than
red.
(3) For the signs above, the ratio of obeying the evacuation signs is higher in the non-
twinkling state.
In Experiment 2, the proportion of obeying the evacuation signs is shown in Fig.9.
In the non-twinkling state, the proportion of obeying the evacuation signs was higher.
Among them, the green “running man”, the green “arrow”, and the red “arrow” all
had a ratio of 100%. In Experiment 2, the twinkling green “running man” sign had the
lowest obeying rate of only 64.29%.
(4) For the signs above, the green “arrow” worked better than the green “running man”
The NS and NF in Experiment 2 are presented in Fig.10. Sub-experiment 2.3 and
Sub-experiment 2.7 had better results, which consisted of the green non-twinkling
“arrow” and the green twinkling “arrow”, whereas the green twinkling “running man”
had the worst effect. For the NF, the green non-twinkling “running man” had the worst
effect. As such, it can be concluded that the green “running man” sign is inferior. The
evacuation indications of the green “arrow” and the ‘running man” are in stark contrast.
The most commonly used evacuation signs in China (running man) are not effective.
Fig. 8 Comparison on ratio of seeing the evacuation signs in experiment 2
Author's personal copy

Natural Hazards
1 3
A χ2 test was performed between the number of participants gazing at the signs and
the influencing factors under different conditions of the experiment. The results are
shown in Table5. In Experiment 1, there are no factors where the p value was less than
0.05. From the value of χ2, the values of Sub-experiment 1 and 2 and Sub-experiment
3 and 4 are high. For the non-twinkling sign, the colour has a certain influence.
In Experiment 2, the difference between Sub-experiment 1 and 3 and Sub-experiment
5 and 7 is significant. Under the condition that the green sign is not twinkling, there is
a significant difference between “running man” and “arrow”; under the condition of
twinkling the green sign, “running man” and “arrow” also have significant differences.
Fig. 9 Comparison on the ratio of following the guidance in experiment 2
Fig. 10 Comparison on the ratio of seeing sign and following the guidance in experiment 2
Author's personal copy

Natural Hazards
1 3
This result further adds evidence to the previous conclusions. In addition, the χ2 of
sub-experiments 3 and 4 are high, which means colour has an impact on the “running
man” evacuation sign without twinkling.
(5) The fixation duration average is shorter, and the proportion of obeying signs is higher.
By comparing Experiment 1 and Experiment 2, it was found that the fixation duration
average in Experiment 2 was significantly shorter than the fixation duration average in
Experiment 1, while the proportion of obeying signs in Experiment 2 was higher than
that in Experiment 1 (Fig.11). The time of gaze reflects the decision-making time of
the evacuees to a certain extent. The shorter time for making decisions makes it easier
to obey the signs “without thinking”.
4.2 Logistic regression
In order to explore what kind of evacuation signs have the greatest impact on participants,
Ys is a 0–1 variable which is used to indicate whether the participants see the evacuation
signs and Yf is also a 0–1 variable which is used to indicate whether the participants fol-
low the evacuation signs. Also, the position, colour, sign, and twinkling of the evacuation
Table 5 χ2 test in experiment 1
and 2 Exp1 Exp2
Sub-Exp χ2p Sub-eXP χ2p
(1, 2) 5.074 0.240 (1, 2) 1.086 0.297
(3, 4) 3.377 0.066 (3, 4) 3.036 0.081
(5, 6) 0.000 0.992 (5, 6) 0.069 0.793
(7, 8) 0.013 0.910 (7, 8) 1.462 0.227
(1, 3) 0.131 0.718 (1, 3) 6.169 0.013
(2, 4) 0.010 0.921 (2, 4) 0.061 0.806
(5, 7) 0.215 0.643 (5, 7) 5.083 0.024
(6, 8) 0.134 0.714 (6, 8) 0.722 0.395
(1, 5) 1.447 0.229 (1, 5) 0.814 0.367
(2, 6) 1.181 0.277 (2, 6) 0.018 0.985
(3, 7) 0.157 0.692 (3, 7) 0.442 0.506
(4, 8) 1.871 0.171 (4, 8) 1.576 0.209
Fig. 11 Fixation duration aver-
age and ratio of following the
guidance
y = -0.8853x + 1.4599
y = -0.2244x + 0.989
50.00%
60.00%
70.00%
80.00%
90.00%
10
0.00%
110.00%
0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
Author's personal copy

Natural Hazards
1 3
signs are recorded as Xp, Xc, Xs, and Xt, respectively. For these two experiments, the soft-
ware SPSS 24 is used for binary logistic regression analysis. The parameter settings are: Ys
(yes–no: 1–0); Yf (yes–no: 1–0); position (high–low: 1–2); colour (green–red: 1–2); sign
(running man-arrow: 1–2); twinkling (yes–no:1–0). The values of the variables above were
extracted from the 483 valid experiments. The analysis results are shown in Tables6 and 7.
Two formulas can be obtained through analysis. For Table6:
For Table7:
From the values and plus-minus signs of the coefficient (B) in Tables6 and 7, the
conclusions were drawn: (1) In Experiment 1, colour and sign are the two main factors
(1)
Ys
=−
0.680Xc
+
0.337Xg
−
0.008Xc
+
1.525
(2)
Yf
=−
0.737Xc
=
1.015Xg
+
0.051Xc
+
1.242
(3)
Ys
=−
0.238Xc
+
0.819Xg
−
0.27Xc
−
0.863
(4)
Yf
=−
1.229Xc
+
1.729Xg
−
2.337Xc
+
3.091
Table 6 Variables in Experiment
1B S.E. Wald df Sig. Exp(B)
Ys
Xc− 0.680 0.310 4.818 1 0.028 0.507
Xg0.337 0.305 1.218 1 0.270 1.401
Xt− 0.008 0.304 0.001 1 0.979 0.992
Constant 1.525 0.652 5.470 1 0.019 4.594
Yf
Xc− 0.737 0.441 2.789 1 0.095 0.479
Xg1.015 0.453 5.022 1 0.025 2.758
Xt0.051 0.436 0.014 1 0.906 1.053
Constant 1.242 0.890 1.948 1 0.163 3.462
Table 7 Variables in Experiment
2B S.E. Wald df Sig. Exp(B)
Ys
Xc− 0.238 0.277 0.737 1 0.391 0.788
Xg0.819 0.277 8.730 1 0.003 2.269
Xt0.270 0.276 0.959 1 0.327 1.310
Constant − 0.863 0.578 2.234 1 0.135 0.422
Yf
Xc− 1.229 0.707 3.025 1 0.082 0.293
Xg1.729 0.704 6.034 1 0.014 5.635
Xt− 2.337 0.848 7.589 1 0.006 0.097
Constant 3.091 1.527 4.100 1 0.043 22.009
Author's personal copy

Natural Hazards
1 3
affecting Ys and Yf, and the effect of twinkling was very small; however, In Experiment
2, all three factors play an important role and the factor of twinkling becomes one of the
most important factors. (2) All positions in Experiment 1 are low and all positions in
Experiment 2 are high; therefore, the influence of position on Ys and Yf cannot be tested.
(3) When the evacuation signs are green and running man, the results of Ys and Yf are
most inclined to Yes, and the factor of twinkling is difficult to define. These conclusions
are consistent with the results obtained above.
4.3 Limitations
There is a gap between experiments and real emergency evacuation events. The partici-
pants in our experiments were all Chinese students and well educated. The type and gen-
der of participants and Chinese culture may have impact on the experimental results. For
instance, the green evacuation sign is utilized well in China. The frequency of twinkling of
the evacuation sign may be a variable. The experiment in this paper is an individual experi-
ment, and other individuals may influence the evacuees during emergency evacuation.
5 Conclusion
In this paper, 658 experiments with eye tracking devices were conducted. According to
the results of the experiments, the following was determined: (1) the effect of the green
“arrow” evacuation sign is the best; (2) the effect of the low (corridor) sign worked better
than that of the high (room) sign, but the sign of the low (corridor) could be blocked by the
front evacuee; (3) obtaining the ratio of people obeying the evacuation signs under differ-
ent conditions provides an effective basis for the improvement of the safety design of build-
ings (Rayner 1998), and provides data support for computer simulation modelling of crowd
evacuation (Zhang etal. 2017; Yenumula etal. 2015); and (4) evacuees with a short gaze
are more likely to obey the evacuation indicator. In our future work, additional features
should be tested within the experiments, such as social influence or social bond.
Acknowledgements This work is supported by the National Key R&D Program of China (No.
2017YFC0803300) andthe National Natural Science Foundation of China (71904194).
References
Andrée K, Nilsson D, Eriksson J (2016) Evacuation experiments in a virtual reality high-rise building: exit
choice and waiting time for evacuation elevators. Fire Mater 40:554–567
Antes J, Chang K-T, Mullis C (1985) The visual effect of map design: an eye movement analysis. Am Car-
togr 12(2):143–155
Brychtova A, Coltekin A (2016) An empirical user study for measuring the influence of colour distance and
font size in map reading using eye tracking. Cartogr J 53(3):202–212
Castner H, Eastman J (1985) Eye-movement parameters and perceived map complexity: iI. Am Cartogr
12(1):29–40
Cauvin C, Escobar F, Serradj A (2007) Cartographie thématique 1: une nouvelle démarche. Hermes Science
Publications, London
Author's personal copy

Natural Hazards
1 3
Cosma G, Ronchi E, Nilsson D (2016) Way-finding lighting systems for rail tunnel evacuation: a virtual
reality experiment with Oculus Rift®. J Transp Saf Secur 8:101–117
Dobson M (1979) The influence of map information on fixation localization. Am Cartogr 6(1):51–65
Fang Z, Song W, Zhang J, Wu H (2010) Experiment and modeling of exit-selecting behaviors during a
building evacuation. Phys Stat Mech Its Appl 389:815–824
Fang Z-M, Song W-G, Li Z-J, Tian W, Lv W, Ma J, Xiao X (2012) Experimental study on evacuation pro-
cess in a stairwell of a high-rise building. Build Environ 47:316–321
Fu L, Cao S, Song W, Fang J (2019) The influence of emergency signage on building evacuation behavior:
an experimental study. Fire Mater 43:22–33
Helbing D, Buzna L, Johansson A, Werner T (2005) Self-organized pedestrian crowd dynamics: experi-
ments, simulations, and design solutions. Transp. Sci. 39:1–24
Hoogendoorn S, Daamen W (2005) Self-organization in pedestrian flow. In: Traffic granul. Flow’03,
Springer, pp 373–382
Hornsby KS, Claramunt C, Denis M etal (2009) Evaluating the effectiveness and efficiency of visual vari-
ables for geographic information visualization. vol 12, pp. 195-211
Hu J, Li Z, Zhang H, Wei J, You L, Chen P (2015) Experiment and simulation of the bidirectional pedes-
trian flow model with overtaking and herding behavior. Int JMod Phys C 26:1550131
Huo F, Song W, Chen L, Liu C, Liew KM (2016) Experimental study on characteristics of pedestrian evacu-
ation on stairs in a high-rise building. Saf Sci 86:165–173
Isobe M, Adachi T, Nagatani T (2004) Experiment and simulation of pedestrian counter flow. Phys Stat
Mech Its Appl 336:638–650
Kinateder M, Ronchi E, Nilsson D, Kobes M, Müller M, Pauli P, Mühlberger A (2014) Virtual reality for
fire evacuation research. In: 2014 Fed. conference computer science information systems, pp. 313–321.
IEEE
Kobes M, Helsloot I, De Vries B, Post JG (2010) Building safety and human behaviour in fire: a literature
review. Fire Saf J 45:1–11
Kretz T, Grünebohm A, Schreckenberg M (2006) Experimental study of pedestrian flow through a bottle-
neck. J Stat Mech Theory Exp 2006:P10014
Larmande P, Larmande A (1990) Neuro-ophtalmologie. Masson, Paris
Liao YJ, Lo SM, Ma J, Liu SB, Liao GX (2014) A study on people’s attitude to the use of elevators for fire
escape. Fire Technol 50:363–378
Lin P, Ma J, Liu T, Ran T, Si Y, Li T (2016) An experimental study of the “faster-is-slower” effect using
mice under panic. Phys Stat Mech Its Appl 452:157–166
Liu W, Yuan XG (2000) Research on human vision-eye movement system. Chin Ergon 04:41–44 (in
Chinese)
Liu SB, Lo SM, Tsui KL, Wang WL (2015) Modeling movement direction choice and collision avoidance in
agent-based model for pedestrian flow. J Transp Eng 141:04015001
Lu X, Luh PB, Marsh KL, Gifford T, Tucker A (2014) Guidance optimization of building evacuation con-
sidering psychological features in route choice. In: Proceeding 11th World Congr. Intell. Control
Autom., IEEE, pp 2669–2674
Ma J, Song WG, Tian W, Lo SM, Liao GX (2012) Experimental study on an ultra high-rise building evacu-
ation in China. Saf Sci 50:1665–1674
Marsh KL, Wilkie CT, Luh PB, Zhang Z, Gifford T, Olderman N (2014) Crowd guidance in building emer-
gencies: using virtual reality experiments to confirm macroscopic mathematical modeling of psycho-
logical variables. In: Pedestr. Evacuation Dyn. 2012, Springer, pp. 197–212
Netzel R, Ohlhausen B, Kurzhals K etal (2016) User performance and reading strategies for metro maps: an
eye tracking study. Spatial Cognition Comput 17(1):1–28
Olander J, Ronchi E, Lovreglio R, Nilsson D (2017) Dissuasive exit signage for building fire evacuation.
Appl Ergon 59:84–93
Peacock RD, Hoskins BL, Kuligowski ED (2012) Overall and local movement speeds during fire drill evac-
uations in buildings up to 31 stories. Saf Sci 50:1655–1664
Rayner K (1998) Eye movements in reading and information processing: 20years of research. Psychol Bull
124(3):372–422
Ronchi E, Nilsson D, Kojić S, Eriksson J, Lovreglio R, Modig H, Walter AL (2016) A virtual reality experi-
ment on flashing lights at emergency exit portals for road tunnel evacuation. Fire Technol 52:623–647
Seyfried A, Passon O, Steffen B, Boltes M, Rupprecht T, Klingsch W (2009) New insights into pedestrian
flow through bottlenecks. Transp Sci 43:395–406
Shiwakoti N, Sarvi M (2013) Enhancing the panic escape of crowd through architectural design. Transp Res
Part C Emerg Technol 37:260–267
Author's personal copy

Natural Hazards
1 3
Shiwakoti N, Sarvi M, Rose G, Burd M (2011) Animal dynamics based approach for modeling pedestrian
crowd egress under panic conditions. Procedia Soc Behav Sci 17:438–461
Steinke TR (1987a) Eye movement studies in cartography and related fields. Cartographica Int J Geogr Inf
Geovis 24(2):40–73
Steinke T (1987b) Eye movement studies in cartography and related fields. Cartographica 24(2):40–73
Tang C, Wu W, Lin C (2009) Using virtual reality to determine how emergency signs facilitate way-finding.
Appl Ergon 40(4):722–730
Wang J, Lin L, Gao H, Zhang L (2016) Differences in college students’ spatial symbol cognition of tourism
map: based on experimental data from an eye-movement tracking system. Tourism Tribune 31(03):97–
105 (in Chinese)
Wei J, Zhang H, Guo Y, Gu M (2015) Experiment of bi-direction pedestrian flow with three-dimensional
cellular automata. Phys Lett A 379:1081–1086
Wolfe JM, Horowitz TS (2004) Opinion: what attributes guide the deployment of visual attention and how
do they do it. Nat Rev Neurosci 5(6):495–501
Xie H, Filippidis L, Galea ER, Blackshields D, Lawrence PJ (2012) Experimental analysis of the effective-
ness of emergency signage and its implementation in evacuation simulation. Fire Mater 36:367–382
Xudong C, Heping Z, Qiyuan X, Yong Z, Hongjiang Z, Chenjie Z (2009) Study of announced evacuation
drill from a retail store. Build Environ 44:864–870
Yan GL (2004) Eye Movements and Psychological Research. Tianjin Education Press, Tianjin (in Chinese)
Yenumula K, Kolmer C, Pan J, Su X (2015) BIM-controlled signage system for building evacuation ☆. Pro-
cedia Eng 118:284–289. https ://doi.org/10.1016/j.proen g.2015.08.428
Zhang J, Klingsch W, Schadschneider A, Seyfried A (2013) Experimental study of pedestrian flow through
a T-junction. In: Traffic granul. Flow11, Springer, pp 241–249
Zhang Z, Jia L, Qin Y (2017) Optimal number and location planning of evacuation signage in public space.
Saf Sci 91:132–147
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Author's personal copy