Effects of a frontal brake light on pedestrians' willingness to cross the street

Abstract

Effects of a frontal brake light (FBL, a potential external human–machine interface for automated vehicles) on participants’ self-reported willingness to cross a vehicle’s path were investigated. In a mixed design online study (vehicles in the experimental group were equipped with FBLs, there were no FBLs in the control group), participants observed videos of a vehicle approaching at different speeds from the perspective of a pedestrian standing at the curb. The vehicles exhibited either yielding behavior (braking onset 55 m or 32 m before standstill in front of the pedestrian ’s position) or non-yielding behavior (approach speed was maintained). Participants specified their willingness to cross the vehicle’s path at different distances. When the vehicle yielded (i.e., FBL was activated), willingness to cross was significantly higher in the experimental group than the control group. Notably, we further observed a significantly lower willingness to cross in the experimental group than the control group when the vehicle did not yield (i.e., FBL was deactivated). Novel external human–machine interfaces might therefore influence the interaction with vehicles not only when they are activated but also when they are deactivated.

Full text

Transportation Research Interdisciplinary Perspectives 23 (2024) 100990
Available online 16 December 2023
2590-1982/© 2023 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Effects of a frontal brake light on pedestrians’ willingness to cross the street
Daniel Eisele
1
,
*
, Tibor Petzoldt
2
Chair of Trafc and Transportation Psychology, Technische Universit¨
at Dresden, 01062 Dresden, Germany
ARTICLE INFO
Keywords:
Automated vehicle
AV
eHMI
Communication
Implicit
Interaction
ABSTRACT
Effects of a frontal brake light (FBL, a potential external human–machine interface for automated vehicles) on
participants’ self-reported willingness to cross a vehicle’s path were investigated. In a mixed design online study
(vehicles in the experimental group were equipped with FBLs, there were no FBLs in the control group), par-
ticipants observed videos of a vehicle approaching at different speeds from the perspective of a pedestrian
standing at the curb. The vehicles exhibited either yielding behavior (braking onset 55 m or 32 m before
standstill in front of the pedestrian’s position) or non-yielding behavior (approach speed was maintained).
Participants specied their willingness to cross the vehicle’s path at different distances. When the vehicle yielded
(i.e., FBL was activated), willingness to cross was signicantly higher in the experimental group than the control
group. Notably, we further observed a signicantly lower willingness to cross in the experimental group than the
control group when the vehicle did not yield (i.e., FBL was deactivated). Novel external human–machine in-
terfaces might therefore inuence the interaction with vehicles not only when they are activated but also when
they are deactivated.
Introduction
Along with the introduction of increasingly automated vehicles
(AVs) on public roads, human-robot interactions are introduced to
collaborative tasks which currently take place between humans. This
has sparked a growing body of research investigating how AVs should
interact with vulnerable road users. The goal is to ensure successful
human-AV interactions in terms of safety, performance, and human
satisfaction (Markkula & Dogar, 2022). A key step towards this goal is to
nd out how humans understand and react to the behavior and the
communicative cues of AVs. This is especially important, as mis-
understandings might lead to potentially fatal consequences. In order to
explicitly inform surrounding road users about an AV’s state and in-
tentions, efforts have been made to develop prototypes of so-called
external human–machine-interfaces (eHMIs). While the proposed de-
signs vary considerably (Wilbrink et al., 2023), most use light-emitting
displays to communicate their messages. The negotiation of the right-
of-way is considered a central topic (Dey et al., 2020a, p. 8). Most
prototypes communicate aspects of the AV’s intention (e.g., “I intend to
yield to you”). Notably, communicating aspects of the current behavior of
an AV (e.g., “I am decelerating”) has received little attention so far
(Clamann et al., 2017; Wenjun et al., 2023).
A frontal brake light (FBL) which communicates that the vehicle is
decelerating to road users ahead of (and potentially oblique to) the
vehicle, can be a simple approach to support the human-AV interaction
(Petzoldt et al., 2018). The FBL activates as soon as the vehicle begins
decelerating, remains activated during the deceleration and deactivates
as soon as the vehicle stops decelerating. The FBL itself does neither
provide a reason for the deceleration nor communicate the AV’s future
intentions (e.g., whether it will continue decelerating and come to a
standstill or start accelerating again). While such intentions could
possibly be inferred from the situational context (e.g., when the FBL
lights up, as an AV approaches a zebra crossing with a pedestrian
waiting at the curb), the FBL lacks the certainty that messages like “I am
yielding to you” can provide. However, in contrast to “I am yielding to
you”, the FBL does not require the vehicle to have a continuously reli-
able and comprehensive understanding of its surroundings in order to
work. From a technological perspective, the FBL is as simple as the brake
lights on the back of current motorized vehicles. Accordingly, it is not
surprising that patents for similar ideas date back nearly a century
(Douglass, 1924; Pirkey, 1925). Moreover, the FBL has the strength that
its message holds true for any observer (including other drivers), in
* Corresponding author.
E-mail addresses: daniel.eisele@tu-dresden.de (D. Eisele), tibor.petzoldt@tu-dresden.de (T. Petzoldt).
1
https://orcid.org/0000-0002-1751-3342
2
https://orcid.org/0000-0003-3162-9656
Contents lists available at ScienceDirect
Transportation Research Interdisciplinary Perspectives
journal homepage: www.sciencedirect.com/journal/transportation-
research-interdisciplinary-perspectives
https://doi.org/10.1016/j.trip.2023.100990
Received 22 August 2023; Accepted 5 December 2023

Transportation Research Interdisciplinary Perspectives 23 (2024) 100990
2
contrast to messages such as “I am yielding to you”, which have specic
and exclusive addressees. The latter might lead to safety relevant mis-
understandings if a road user incorrectly assumed to be the intended
addressee (Tabone et al., 2021). The FBL therefore remains functional in
complex everyday trafc scenarios with multiple road users present (see
Dey et al., 2020b, Dey et al., 2021).
Effects of FBLs have been studied in few empirical studies so far.
Recent evidence from a video based lab experiment suggests that an FBL
helps to identify decelerations considerably earlier (Petzoldt et al.,
2018). In another recent longitudinal eld study, 102 (non-automated)
vehicles on the apron of an airport were equipped with FBLs for 3.5
months. The surveyed staff members rated the FBL overall positively and
stated that it improved communication and safety (Monzel, 2018). In an
earlier study, participants used an FBL on their private vehicle for a
month. They thought that the FBL had merit and were more willing to
buy an FBL than participants from the control group who were not
familiar with the concept of an FBL before being surveyed. The latter
group also, in theory, saw merit in the concept of an FBL (Post & Mor-
timer, 1971). In conclusion, researchers have plausibly argued that at-
titudes towards the FBL are positive, both in theory and after real life
experiences, and that it facilitates the perception of a vehicle’s decel-
eration. It remains unclear, however, whether the FBL causes actual
behavioral changes in interactions with vehicles (like crossing a street
across their path) since the identication of deceleration can certainly
play an important role in the decision making process but “is neither
identical nor necessarily perfectly correlated with the actual initiation of
a crossing” (Petzoldt et al., 2018, p. 4). Therefore, potential effects of the
FBL on behavioral intent like a pedestrian’s willingness to cross should
be investigated in a next step.
Of course, the detection of a deceleration (vehicle decelerates: yes/
no), which is facilitated by an FBL, is just one factor in pedestrians’
decision-making on whether to cross. It has been shown that variations
in vehicle behavior like its deceleration rate and approach speed, the
distance from the pedestrian at braking onset, and the remaining
physical distance between vehicle and pedestrian inuence street-
crossing greatly (Dey et al., 2019; Ezzati Amini et al., 2019). The FBL
and other eHMIs that have been proposed so far are meant to only
communicate certain aspects of a vehicle’s state, behavior or intention
explicitly while the vast majority of cues remains implicit. Findings from
prior studies point to the ongoing importance of implicit cues that are
communicated by the vehicle’s kinematics when eHMIs are present (e.
g., Domeyer et al., 2020; Lee et al., 2022). It has been argued that they
will remain the basis for on-road communication and should be
augmented by eHMIs whenever they provide an advantage (de Winter &
Dodou, 2022). It seems plausible that effects on pedestrian behavior
caused by an eHMI can interact with the effects caused by a vehicle’s
(implicit) kinematic cues. Therefore, such interactions should be
considered in eHMI development in order to predict possible ramica-
tions caused by the introduction of an eHMI as precisely as possible.
It should be noted that the effects of eHMIs might go beyond situa-
tions in which they are actually activated. The fact that actively
communicating yielding intent via an eHMI has effects on crossing de-
cisions (e.g., Faas et al., 2020) raises the obvious question of how, in
similar scenarios, road users would respond to vehicles that are equip-
ped with an eHMI, but don’t communicate that message. It seems
plausible to assume that knowing a vehicle is equipped with a certain
eHMI from prior interactions can also inuence interactions in situations
in which the eHMI is not activated / an eHMI is completely absent.
Indeed, in a condition of mixed trafc (some vehicles braking with FBL
activated, some vehicles braking without), observers were slower to
identify deceleration for non-FBL vehicles compared to a condition in
which none of the vehicles was equipped with any eHMI (Petzoldt et al.,
2018). It seemed as if pedestrians were waiting for some explicit infor-
mation on deceleration, potentially resulting in more cautious decisions
when this explicit information is not available. As such possible sec-
ondary effects can be undesirable, they should also be considered in
eHMI development.
Based on the above considerations, we designed a study in which
both the presence of the FBL (i.e., the explicit information on the fact
that the vehicle is braking and when/where the braking onset begins)
and kinematic cues (approach speed and deceleration rate) were
manipulated in order to study their effects on pedestrian’s willingness to
cross. We considered situations in which the FBL is active (i.e., vehicle
brakes) as well as non-active (i.e., vehicle does not brake). Based on
prior research, we expected a dominant effect of the vehicle’s kinematics
on willingness to cross. We further expected effects of the FBL on will-
ingness to cross. During decelerations, we hypothesized to observe a
higher willingness to cross in front of a vehicle with an FBL than
without. We further assumed that knowing about the concept of an FBL
and interacting with it might lead to a lower willingness to cross when
the FBL is not activated (i.e., when the vehicle does not decelerate).
Methods
Design
To answer whether an FBL inuences pedestrians’ willingness to
cross, we conducted a randomized online study with a mixed design
featuring two groups. Participants were shown videos (from the
perspective of a pedestrian standing at the curb) of approaching vehicles
on a straight one-lane street. Following Dey et al. (2019), the videos
ended at different points in time, when the vehicle’s physical distance
from the observer was either 45, 30, 15, 5 or 1.5 m. Within groups, we
varied the vehicle’s initial approach speed (30/50 km/h, both common
speeds in German urban trafc) and the vehicle’s behavior (non-
yielding/yielding). In non-yielding conditions, the approach speed was
maintained until the vehicle passed the observer’s position. In yielding
conditions, the vehicle decelerated for either 55 or 32 m and came to a
standstill at the observer’s position. We refer to this variable as braking
onset. The combination of the factor levels resulted in 28 unique varia-
tions. Each of these was shown twice (to better estimate the “true”
willingness to cross in the respective variation) in randomized order,
amounting to 56 videos per participant. Whether vehicles were equip-
ped with an FBL, was varied between groups. In the experimental group
(EG) all vehicles were equipped with an FBL. In the control group (CG),
vehicles were not equipped with an FBL and there was no mention of the
existence of FBLs. Table 1 provides an overview over the factors and
factor levels.
The participants’ task was to specify their willingness to cross on a
Likert scale ranging from 1 (very low) to 10 (very high) at the moment
the video ended. In Germany, where the experiment was conducted, it is
the driver’s decision whether to yield to a pedestrian intending to cross
in places where this is not explicitly regulated. There is no obligation to
do so. However, of course, for a pedestrian, (safely) crossing an
approaching vehicle’s path is generally permitted and common in
Table 1
Overview over factors and factor levels.
Between
Subjects
Within Subjects
Presence of
FBL
on vehicle
Approach
speed
Yielding Behavior Distance from
Pedestrian
when video cut
off
FBL present 50 km/h no yielding, approach
speed maintained
1.5 m
FBL absent 30 km/h yielding, braking onset
55 m from pedestrian
5 m
yielding, braking onset
32 m from pedestrian
15 m
30 m
45 m
D. Eisele and T. Petzoldt

Transportation Research Interdisciplinary Perspectives 23 (2024) 100990
3
everyday trafc.
Participants
50 German residents took part in the study. Initially, a convenience
sample was recruited from the rst author’s personal network. To ach-
ieve the aspired sample size, additional participants were recruited
through Prolic, an online platform, which distributes online experi-
ments to paid research participants for a service fee (Prolic, 2022).
These participants received 4 GBP as a compensation. Screenings
revealed no signicant differences regarding street-crossing willingness
between the two subsamples. The experimental software (Labvanced,
2022) randomly assigned 30 participants to the EG and 20 to the CG.
Two participants (one in each group) were excluded from analysis based
on their unsystematic, inconsistent ratings regarding their willingness to
cross. The remaining 48, (27f, 21 m) were between 20 and 69 years of
age (M =31.6, SD =11.5).
Material
The videos of a vehicle approaching the camera perspective were
rendered using the VICOM Editor (2021). The perspective was set to an
average German eye-height of 1.6 m (Windel, 2019). The top part of
Fig. 1 illustrates the different behaviors of the vehicle. The different lines
depict the vehicle’s speed (y-axis) relative to the physical distance from
the pedestrian (x-axis). Every clip started when the vehicle was 100 m
away from the pedestrian. The vehicle either maintained its speed
continuously (horizontal lines) or started to decelerate in a linear
fashion at the two marked points (dashed lines) until it came to a
standstill at the participant’s position. This corresponded to deceleration
rates between 0.6 m/s
2
and 3.0 m/s
2
, which cover the spectrum of
service braking in manual driving (i.e., everyday braking, no emergency
maneuvers) according to Schnabel and Lohse (2011) as well as preferred
rates in automated vehicles (Scherer et al., 2016). Apart from being of
theoretical interest, the differences in the vehicle’s kinematics made the
vehicle’s behavior less predictable for the participants. At standstill, we
left enough space between the vehicle and the viewer’s perspective to
convey the impression that a pedestrian could cross in front of the ve-
hicle’s hood. The dotted lines in Fig. 1 mark the ve points where the
different video segments were cut.
Apart from the vehicle’s behavior (and the FBL in the experimental
group), there was no difference in the video clips. The FBL was added
using DaVinci Resolve (2022). Fig. 2 shows screenshots from a sample
clip (top: control group, bottom: experimental group). The FBL’s color
(magenta; RGB 225, 0, 225) was chosen for pragmatic reasons, as it is
unusual in German trafc and was well visible (Werner, 2019). The
stimuli were purely visual; there was no audio track. The clips were
presented in 50 fps and a maximum resolution of 1920 x 1080 px.
Procedure
Participants received a link to the experiment via e-mail or the
Prolic platform and used their own hardware to take part. Smartphones
and devices with a screen resolution lower than 1280 ×720 pixels were
excluded from participation. After they gave informed consent, they
were familiarized with the trafc situation by a screenshot from the
stimuli and the following text (in German): “You are a pedestrian on the
sidewalk of a one-way street. You are standing at the curb, as you intend
to cross the street. You have a clear destination. You are on the way to
your workplace/university. You are not in a hurry. This experiment
includes several video clips in which a vehicle will approach you. The
camera’s perspective depicts your visual eld as the pedestrian. These
videos will stop at predened points in time.” The participants in the EG
were further explicitly informed about the FBL’s functionality and
messages: “You have surely noticed the magenta-colored light on the
vehicle’s grill. This is a frontal brake light. It indicates that the vehicle is
braking. It activates as soon as the vehicle begins to brake, it continu-
ously remains activated as long as the vehicle is braking and deactivates
at the moment the vehicle stops braking. So, the frontal brake light
works just as ordinary brake lights at the rear of a motorized vehicle do.”
After watching a group-specic example video of a vehicle braking to
standstill, their task (specify their willingness to cross on a Likert scale
ranging from 1 to 10 at the moment the video ended) was explained.
They then had the opportunity to practice the task three times, before
Fig. 1. The different behaviors of the vehicle represented by its speed at different distances from the pedestrian (top) and the distances from the pedestrian’s position
at which the different segments ended (bottom).
D. Eisele and T. Petzoldt

Transportation Research Interdisciplinary Perspectives 23 (2024) 100990
4
the 56 experimental trials began. Afterwards, they rated the FBL on the
short version of the User Experience Scale (UEQ-S, Schrepp et al., 2017),
which measures user experience on a pragmatic and hedonic dimension
and provides an overall score. They then rated their level of (dis-)
agreement with several statements regarding the potential usefulness
and safety effects of the FBL on a 5 point Likert scale (Petzoldt et al.,
2018). At the end, they indicated their age and gender. After nishing,
they had the opportunity to download a debrieng that explained the
experimental mechanics of the two groups and the goal of the study. The
experiment took an average of 16 min to complete. This study complied
with the tenets of the Declaration of Helsinki (a set of ethical principles
for experiments involving human subjects).
Analysis
Two Mixed RMANOVAs were calculated in Jasp Team (2022) to
analyze the variances in the yielding and the non-yielding conditions
separately. Whether the vehicles were equipped with an FBL was the
between-subjects factor. Within-subject factors were the vehicle’s
approach speed, the distance when deceleration began (braking onset)
and the vehicle’s physical distance from the pedestrian at the time the
video ended. Willingness to cross was the dependent variable. The dis-
tance 45 m was excluded from the analysis of yielding conditions, as
yielding vehicles were not yet decelerating at that point when the
braking onset was 32 m. This does not apply to non-yielding conditions,
where all distances were analyzed.
An inspection of the data subsets revealed violations of the RMA-
NOVA’s assumptions. Where Mauchly’s test indicated a violation of
sphericity, Huynh-Feldt corrected F-Statistics are reported for yielding
data (
ε
was above 0.75) and Greenhouse-Geisser corrected values for
non-yielding data (
ε
was below 0.75), following Field (2017). Levene’s
test revealed heteroscedasticity in half of the yielding conditions (which
seems to be a desirable effect of the FBL, discussed below). Following
Pituch and Stevens (2016), we assume our ANOVAs remained robust
despite this violation, as our group size ratio falls within the commonly
referred to cutoff. Ratings of the FBL on the UEQ-S were analyzed using
the tool provided by its authors (Schrepp, 2018).
Results
Yielding conditions (FBL active in experimental group)
Fig. 3 depicts willingness to cross relative to the different distances
between vehicle and pedestrian during the yielding process. Values of
the two groups are depicted separately. As one would expect, willing-
ness to cross was highest at the furthest distance in both groups. Will-
ingness decreased at closer distances and rose again at the smallest
distance, shortly before the vehicle came to a standstill. This difference
between distances was signicant, F(3, 282) =11.8, p <.001,
η
p
2
=0.11.
Fig. 3 further shows that there was a between-groups difference in
willingness to cross. At every distance, willingness to cross was higher in
the experimental group (active FBL) than the control group (no FBL),
suggesting a main effect of the FBL on willingness to cross in yielding
conditions. This effect was indeed signicant, F(3, 282) =11.8, p <.001,
η
p
2
=0.11. Simple main effects indicated that the difference was
Fig. 2. Screenshots of a video clip in the control group (top) and experimental group (bottom).
Fig. 3. Yielding Conditions: Willingness to cross relative to physical distance
between vehicle and pedestrian, split by groups. Note. Values are averaged
across approach speeds (30/50 km/h) and braking onsets points (32/55 m); see
Table 2 for detailed descriptive values. As the vehicle was braking in these
conditions, the FBL was active in the EG; there was no FBL in the CG. Error bars
display the normalized 95 % condence interval of the mean (Morey, 2008).
D. Eisele and T. Petzoldt

Transportation Research Interdisciplinary Perspectives 23 (2024) 100990
5
signicant at every distance (every p <.008). There was no signicant
interaction between distance and group membership. An inspection of
the descriptive values showed that the variances in the experimental
group in 30 km/h conditions were consistently smaller than those in the
control group (see SDs in Table 2). These differences were indeed sig-
nicant in 7 of 8 conditions (every p <.03). The differences in conditions
with an approach speed of 50 km/h were only signicant at a distance of
1.5 m (p =.02).
As one would expect, we observed signicantly lower willingness to
cross in trials with a 50 km/h than 30 km/h approach speed (F(1,94) =
228.7, p <.001,
η
p
2
=0.71) and a lower willingness in trials with the 32 m
than the 55 m braking onset, F(1,94) =70.4, p <.001,
η
p
2
=0.43. There
was no interaction between group membership and these factors.
Descriptive values of willingness split by approach speed, distance and
group membership are supplied in Table 2.
Non-yielding conditions (FBL inactive in experimental group)
Fig. 4 depicts willingness to cross relative to the vehicle’s distance
from the pedestrian in non-yielding conditions (i.e., the approach speed
was maintained). Values of the two groups are depicted separately. As
one would expect, willingness to cross was highest when the vehicle was
the furthest from the pedestrian and decreased at closer distances,
suggesting a main effect of distance on willingness to cross. This effect
was signicant, F(2.5, 236.9) =169.9, p <.001,
η
p
2
=0.64. Bonferroni-
corrected post-hoc tests showed that the stepwise differences between
45 m and 5 m were signicant (every p <.001). The difference between
5 m and 1.5 m was not signicant.
We again observed a between-groups difference, pointing to an effect
of the FBL in non-yielding conditions. Fig. 4 shows that willingness to
cross was lower in the experimental group than the control group (in
contrast to yielding conditions, see Fig. 3), a difference that, although
comparatively small, was found to be signicant, F(1, 94) =4.1, p
=.045,
η
p
2
=0.04. There was no interaction between group membership
and distance.
As one would expect, we observed a signicantly lower willingness
to cross at a speed of 50 km/h than 30 km/h, F(1,94) =383.7, p <.001,
η
p
2
=0.43. There was no interaction between group membership and
speed. Descriptive values of willingness to cross split by approach speed,
distance and group membership are provided in Table 3.
Subjective ratings of the FBL
Overall, participants in the experimental group reported a positive
view of the FBL. With regards to user experience (as measured by the
UEQ-S, possible values range from −3 to +3) the pragmatic quality
received an average rating of 2.3 (SD =0.7), which is labeled “excellent”
on the UEQ-S benchmark. The hedonic quality was 1.4 (SD =0.9), which
corresponds to “good”. This amounts to an “excellent” overall rating of
1.8 (SD =0.7).
Table 4 shows the participants’ assessment of the frontal brake light
as measured by agreement to general statements. Most participants liked
the general idea of an FBL. They also felt that the FBL might contribute
to road safety, while they were less (but still somewhat) optimistic
regarding its potential to increases trafc efciency.
Discussion
In this experiment, effects of a frontal brake light and different
vehicle behaviors on pedestrians’ willingness to cross the street across
the vehicle’s path were studied. We focused on main effects of the FBL
and possible interactions with vehicle behavior.
Findings and conclusions
The results of our investigation show that the use of an FBL can
Table 2
Willingness to cross (M, SD) in yielding conditions at different distances split by vehicle behavior (approach speed, braking onset) and group.
Distance Vehicle Behavior
30 km/h 50 km/h
32 m 55 m 32 m 55 m
EG CG EG CG EG CG EG CG
1.5 m 7.7 (2.1) 5.7 (3.6) 8.6 (2.5) 6.6 (3.6) 5.7 (3.1) 4.5 (3.1) 4.9 (3.1) 3.3 (2.4)
5 m 6.8 (2.6) 5.2 (3.5) 8.1 (2.5) 6.6 (3.6) 4.2 (2.5) 3.3 (2.2) 5.2 (3.0) 3.7 (2.5)
15 m 6.5 (2.6) 5.7 (3.2) 8.2 (2.2) 8.6 (3.2) 4.7 (2.5) 3.3 (2.4) 5.7 (2.7) 4.2 (2.5)
30 m 7.1 (2.3) 5.8 (2.7) 8.4 (2.0) 7.1 (3.0) 5.4 (2.3) 4.3 (2.1) 6.0 (2.5) 5.4 (2.2)
Note. EG =experimental group; CG =control group. SDs are in brackets.
Fig. 4. Non-yielding conditions: Willingness to cross relative to physical dis-
tance between vehicle and pedestrian, split by groups. Note. Values are aver-
aged across approach speeds (30/50 km/h); see Table 3 for detailed descriptive
values. As the vehicle was not braking in these conditions, the FBL was inactive
in the EG; there was no FBL in the CG. Error bars display the normalized 95 %
condence interval of the mean (Morey, 2008).
Table 3
Willingness to cross (M, SD) in non-yielding conditions at different distances
split by vehicle behavior (approach speed) and group.
Distance Vehicle Behavior
30 km/h 50 km/h
EG CG EG CG
1.5 m 2.1 (2.0) 2.5 (2.7) 1.4 (1.0) 1.5 (1.2)
5 m 2.2 (2.6) 2.8 (2.5) 1.5 (1.5) 1.9 (1.7)
15 m 3.7 (2.4) 4.2 (2.7) 1.8 (1.2) 2.6 (2.0)
30 m 5.6 (2.6) 6.6 (3.0) 3.8 (2.3) 4.5 (2.4)
45 m 6.4 (2.4) 7.3 (3.0) 5.4 (2.7) 6.1 (2.3)
Note. EG =experimental group; CG =control group. SDs are in brackets.
D. Eisele and T. Petzoldt

Transportation Research Interdisciplinary Perspectives 23 (2024) 100990
6
inuence the willingness to cross. This extends on prior research that
pointed to its potential in speeding up the decision to cross, as it leads to
considerable improvements in the identication of a vehicle deceler-
ating (Petzoldt et al., 2018). In particular, we observed a higher will-
ingness to cross in front of a yielding vehicle in participants who saw
vehicles equipped with an FBL than those who were neither introduced
nor exposed to the FBL. Notably, the variance in willingness to cross
when the vehicle was yielding was lower in the subjects who saw the
FBL. The willingness to cross of those not exposed to the FBL was
signicantly more dispersed. Interestingly, this was only observed at the
lower approach speed of 30 km/h. These observations suggest that
eHMIs can not only help to raise the overall probability of a certain
behavioral reaction to an AV (in our case a higher willingness to cross in
reaction to the FBL) but also cause more uniformity in reactions (in our
case a lower interindividual variability in willingness to cross when the
FBL was present).
Perhaps our most important nding was that the FBL also inuenced
interactions with a vehicle in situations where it was not activated.
Those who knew about the concept and had interacted with the FBL in
prior trials, were more conservative in their willingness to cross when
the vehicle did not yield (i.e., when the FBL was inactive, as there was no
braking) than those who did know about the FBL. We argue this nding
is important in the broader sense, as it shows that introducing vehicles
equipped with novel eHMIs into trafc potentially inuences in-
teractions with these vehicles not exclusively in situations the eHMI is
activated in (which has been the main goal and focus of research so far)
but also in situations they are not activated in (in a possibly unintended
manner).
In line with prior research (e.g., Dey et al., 2019; Ezzati Amini et al.,
2019; Tian et al., 2023; Zach Noonan et al., 2023), vehicle kinematics
greatly inuenced willingness to cross in our study. When the vehicle
maintained its approach speed, the willingness to cross was highest at
the furthest distance between vehicle and pedestrian and decreased as
the distance decreased. At the higher approach speed, willingness to
cross was consistently smaller than at the lower speed. When the vehicle
was yielding on the other hand, willingness to cross also decreased as the
vehicle came closer, starting at an initially higher value. In contrast to
non-yielding behavior however, willingness to cross started to increase
again when the vehicle was very close. Dey et al. (2019) plausibly argue
that although the vehicle is decelerating, the situation is conceived as
ambiguous, which causes willingness to cross to go down from a certain
distance. As soon as the vehicle has nearly come to a standstill close to
the pedestrian however, it seems more certain that it will indeed yield to
the pedestrian - which increases willingness to cross. As one would
expect, we found that willingness was lower both when the initial
approach speed was higher and braking onset was later. This observa-
tion corroborates the notion that early braking can be used as a
communicative system, both on its own and in combination with an
eHMI (Dey et al., 2020a). This would simultaneously result in lower
braking intensities which are preferred by passengers in AVs (de Winkel
et al., 2023). These ndings add to the body of evidence that vehicle
kinematics (which are intuitively understandable), are an important
factor for the interaction with (automated) vehicles. The effects of
vehicle kinematics were independent from whether participants were
subjected to the FBL in the course of the experiment. It is possible
therefore that the FBL has a consistent effect across different vehicle
kinematics.
Regarding subjective ratings, the participants reacted positively to
the FBL. They mostly agreed that the FBL is a good idea and can enhance
trafc safety. With regards to user experience, it is not surprising that
our participants rated the pragmatic value of the FBL highly in a scenario
where the FBL’s message always indicated that the vehicle would yield
to the pedestrian. Interestingly, they also ascribed a good hedonic
quality to the FBL. These positive reactions are in line with prior studies
(Monzel, 2018; Petzoldt et al., 2018; Post & Mortimer, 1971). However,
as our study only investigated one very simple scenario, this positive
reaction cannot be generalized to all possible interactions with FBLs in
trafc.
Taken together, our ndings suggest that the FBL has the potential to
inuence a pedestrian’s likely response to an AV’s behavior in a pre-
dictable manner. If it can be ensured that the response can be inuenced
in a desirable way in every conceivable trafc constellation, the frontal
brake light (or similar eHMIs that communicate aspects of a vehicle’s
current behavior) might help in achieving intended interaction out-
comes. This can be regarded as a central goal in research on vehicle
automation (Markkula & Dogar, 2022).
Limitations
While our methods enabled us to answer whether the FBL can in-
uence pedestrians’ willingness to cross, it is important to point out the
study’s scope and limitations. We made use of a simple controlled setting
in which the pedestrian interacted with one vehicle at a time. This does
obviously not portray the complexity of real trafc. Potentially unde-
sirable effects of the FBL (e.g., when falsely understood as a yielding
signal when a vehicle is braking but has no yielding intent) are beyond
the scope of this study. Furthermore, while we did manipulate the ve-
hicle’s kinematics which are regarded as central in crossing decisions,
we made the conscious decision to forgo other factors in order to
maintain the desired level of experimental control. We did not consider
factors like vehicle characteristics (e. g., size, appearance) or heuristics
which inuence our decision-making like illumination of a trafc scene
(e.g., day vs. night), the social environment including other trafc
participants, and the trafc infrastructure and formal trafc rules, which
likely inuence the reactions to an eHMI.
We realize the video-based apparatus (as opposed to a real-world
investigation) limited both the possible physical dynamics of both ac-
tors as well as the physical danger that would arise from actual crossing
behavior. However, we argue that the required level of experimental
control (and safety) over conditions and participants would have been
difcult to achieve in the eld. One might argue that a simulation setup
with a higher ecological validity should have been used. However,
recent evidence suggests that participants perceive street-crossing situ-
ations in video-based online experiments similarly to those in a CAVE
setup that allows for a higher level of immersion and more natural in-
teractions with the environment (Tabone et al., 2023). As Recarte et al.
(2005) showed that estimated arrival time of a vehicle in video material
is highly equivalent to vehicles experienced in real life, the chosen
method seems to be suitable for our purpose.
Future research
In a next empirical step, possible undesirable effects of a frontal
brake light in more complex scenarios should be investigated (e.g., sit-
uations in which the reason for braking is not clear or confusing to an
observer) to allow weighing its pros and cons. Additionally, we propose
that future research should consider (possibly undesirable) effects of
Table 4
Participants’ general assessment of the FBL by agreement to a set of statements,
relative frequencies in %.
The frontal
brake light…
completely
disagree
disagree
somewhat
neither
nor
agree
somewhat
completely
agree
… is a good
idea
0 0 3.5 41.4 55.2
… can
facilitate
getting
ahead in
trafc
0 17.2 34.5 24.1 24.1
… can make
trafc safer
0 0 6.9 48.3 44.8
D. Eisele and T. Petzoldt

Transportation Research Interdisciplinary Perspectives 23 (2024) 100990
7
eHMIs in situations in which the eHMIs are not activated and investigate
the interaction with vehicles that are not equipped with these eHMIs (e.
g., in mixed trafc).
Funding
This work was supported by the German Federal Ministry of Edu-
cation and Research (BMBF) as part of the project CADJapanGermany:
HF [funding number 16ES1035].
CRediT authorship contribution statement
Daniel Eisele: Conceptualization, Methodology, Formal analysis,
Investigation, Writing – original draft. Tibor Petzoldt: Conceptualiza-
tion, Methodology, Writing – review & editing, Funding acquisition.
Declaration of competing interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Data availability
Data will be made available on request.
Acknowledgments
We thank our research assistant Carla Bernadette Bubeck for sup-
porting the realization of this study.
References
Clamann, M., Aubert, M., & Cummings, M. L. (2017). Evaluation of vehicle-to-pedestrian
communication displays for autonomous vehicles. Transportation Research Board
96th Annual Meeting, Washington DC.
de Winkel, K.N., Irmak, T., Happee, R., Shyrokau, B., 2023. Standards for passenger
comfort in automated vehicles: Acceleration and jerk. Appl. Ergon. 106, 103881
https://doi.org/10.1016/j.apergo.2022.103881.
de Winter, J., Dodou, D., 2022. External human–machine interfaces: Gimmick or
necessity? Transportation Research Interdisciplinary Perspectives 15, 100643.
https://doi.org/10.1016/j.trip.2022.100643.
Dey, D., Matviienko, A., Berger, M., Peging, B., Martens, M., & Terken, J. (2020).
Communicating the intention of an automated vehicle to pedestrians: The
contributions of eHMI and vehicle behavior. It - Information Technology, 0(0),
000010151520200025. https://doi.org/10.1515/itit-2020-0025.
Dey, D., Martens, M., Eggen, B., Terken, J., 2019. Pedestrian road-crossing willingness as
a function of vehicle automation, external appearance, and driving behaviour.
Transport. Res. F: Trafc Psychol. Behav. 65, 191–205. https://doi.org/10.1016/j.
trf.2019.07.027.
Dey, D., Habibovic, A., L¨
ocken, A., Wintersberger, P., Peging, B., Riener, A.,
Martens, M., Terken, J., 2020a. Taming the eHMI jungle: A classication taxonomy
to guide, compare, and assess the design principles of automated vehicles’ external
human-machine interfaces. Transportation Research Interdisciplinary Perspectives
7. https://doi.org/10.1016/j.trip.2020.100174.
Dey, D., van Vastenhoven, A., Cuijpers, R.H., Martens, M., Peging, B., 2021. Towards
Scalable eHMIs: Designing for AV-VRU Communication Beyond One Pedestrian. In:
13th International Conference on Automotive User Interfaces and Interactive
Vehicular Applications, pp. 274–286. https://doi.org/10.1145/3409118.3475129.
Domeyer, J.E., Lee, J.D., Toyoda, H., 2020. Vehicle automation–other road user
communication and coordination: theory and mechanisms. IEEE Access 8,
19860–19872. https://doi.org/10.1109/ACCESS.2020.2969233.
Douglass, S. F. (1924). Motor vehicle signal (United States Patent Ofce Patent
1,519,980).
VICOM Editor (1.11). (2021). [Computer software]. TÜV | DEKRA arge tp21 GbR. www.
vicomeditor.de.
Ezzati Amini, R., Katrakazas, C., Antoniou, C., 2019. Negotiation and decision-making
for a pedestrian roadway crossing: A literature review. Sustainability 11 (23), 6713.
https://doi.org/10.3390/su11236713.
Faas, S. M., Kao, A. C., & Baumann, M. (2020). A Longitudinal Video Study on
Communicating Status and Intent for Self-Driving Vehicle – Pedestrian Interaction.
14.
Field, A., 2017. Discovering statistics using IBM SPSS statistics, (5th edition). SAGE
Publications.
Jasp Team. (2022). JASP (0.16.4) [Computer software]. https://jasp-stats.org.
Labvanced., 2022. Computer software. Scicovery GmbH. https://www.labvanced.com.
Lee, Y.M., Madigan, R., Uzondu, C., Garcia, J., Romano, R., Markkula, G., Merat, N.,
2022. Learning to interpret novel eHMI: The effect of vehicle kinematics and eHMI
familiarity on pedestrian’crossing behavior. J. Saf. Res. 80, 270–280.
Markkula, G., Dogar, M., 2022. How accurate models of human behavior are needed for
human-robot interaction? For Automated Driving? arXiv:2202.06123, arXiv. htt
p://arxiv.org/abs/2202.06123.
Monzel, M., 2018. Feldstudie zur Erprobung einer Vorderen Bremsleuchte am Flughafen
Berlin-Tegel. Zeitschrift Für Verkehrssicherheit 64, 210–218.
Morey, R.D., 2008. Condence intervals from normalized data: a correction to cousineau
(2005). Tutorials in Quantitative Methods for Psychology 4 (2), 61–64. https://doi.
org/10.20982/tqmp.04.2.p061.
Petzoldt, T., Schleinitz, K., Banse, R., 2018. Potential safety effects of a frontal brake light
for motor vehicles. IET Intel. Transport Syst. 12 (6), 449–453.
Pirkey, O. S. (1925). Signal for automobiles (United States Patent Ofce Patent
1,553,959).
Pituch, K.A., Stevens, J., 2016. Applied multivariate statistics for the social sciences:
Analyses with SAS and IBM’s SPSS, (6th edition). Routledge/Taylor & Francis Group.
Post, D.V., Mortimer, R.G., 1971. Subjective evaluation of the front-mounted braking
signal. University of Michigan, Highway Safety Research Institute.
Prolic. (2022). [Computer software]. https://www.prolic.co.
Recarte, M.´
A., Conchillo, ´
A., Nunes, L.M., 2005. Estimation of arrival time in vehicle and
video. Psicomthema 17 (1), 112–117.
DaVinci Resolve (Version 18). (2022). [Computer software]. Blackmagicdesign. https://
www.blackmagicdesign.com/de/products/davinciresolve.
Scherer, S., Schubert, D., Dettmann, A., Hartwich, F., & Bullinger, A. C. (2016). Wie will
der „Fahrer“ automatisiert gefahren werden? Überprüfung verschiedener Fahrstile
hinsichtlich des Komforterlebens. In VDI Wissensforum GmbH (Ed.), Fahrerassistenz
und Integrierte Sicherheit 2016 (pp. 299–310). VDI Verlag. https://doi.org/
10.51202/9783181022887-299.
Schnabel, W., & Lohse, D. (2011). Grundlagen der Straßenverkehrstechnik und der
Verkehrsplanung. 1: Straßenverkehrstechnik (3., vollst. überarb. Au). Beuth.
Schrepp, M., Hinderks, A., Thomaschewski, J., 2017. Design and evaluation of a short
version of the User Experience Questionnaire (UEQ-S). International Journal of
Interactive Multimedia and Articial Intelligence 4 (6), 103. https://doi.org/
10.9781/ijimai.2017.09.001.
Schrepp, M. (2018). UEQ-S Data Analysis Tool [Computer software]. https://www.ueq-
online.org/Material/Short_UEQ_Data_Analysis_Tool.xlsx.
Tabone, W., de Winter, J.C.F., Ackermann, C., B¨
argman, J., Baumann, M., Deb, S.,
Emmenegger, C., Habibovic, A., Hagenzieker, M., Hancock, P.A., Happee, R.,
Krems, J., Lee, J.D., Martens, M., Merat, N., Norman, D., Sheridan, T.B., Stanton, N.
A., 2021. Vulnerable road users and the coming wave of automated vehicles: expert
perspectives. Transportation Research Interdisciplinary Perspectives 9, 100293.
https://doi.org/10.1016/j.trip.2020.100293.
Tabone, W., Happee, R., Yang, Y., Sadraei, E., García, J., Lee, Y. M., Merat, N., & de
Winter, J. (2023). Immersive Insights: Evaluating Augmented Reality Interfaces for
Pedestrians in a CAVE-Based Experiment. https://www.researchgate.net/
publication/370160064_Immersive_Insights_Evaluating_Augmented_Reality_
Interfaces_for_Pedestrians_in_a_CAVE-Based_Experiment.
Tian, K., Tzigieras, A., Wei, C., Lee, Y.M., Holmes, C., Leonetti, M., Merat, N.,
Romano, R., Markkula, G., 2023. Deceleration parameters as implicit
communication signals for pedestrians’ crossing decisions and estimations of
automated vehicle behaviour. Accid. Anal. Prev. 190, 107173 https://doi.org/
10.1016/j.aap.2023.107173.
Wenjun, P., Li, F., Xiang, J., 2023. Effect of vehicle external acceleration signal lights on
pedestrian-vehicle interaction [Preprint]. In Review. https://doi.org/10.21203/rs.3.
rs-2900757/v1.
Werner, A., 2019. New colours for autonomous driving: an evaluation of chromaticities
for the external lighting equipment of autonomous vehicles. Colour Turn 1–14.
https://doi.org/10.25538/TCT.V0I1.692.
Wilbrink, M., Cieler, S., Weiß, S.L., Beggiato, M., Joisten, P., Feierle, A., Oehl, M., 2023.
Principles for external human-machine interfaces. Information 14 (8), 463. https://
doi.org/10.3390/info14080463.
Windel, A., 2019. Kleine Ergonomische Datensammlung, (17th ed.). TÜV Rheinland.
Zach Noonan, T., Gershon, P., Domeyer, J., Mehler, B., Reimer, B., 2023. Kinematic cues
in driver-pedestrian communication to support safe road crossing. Accid. Anal. Prev.
192, 107236 https://doi.org/10.1016/j.aap.2023.107236.
D. Eisele and T. Petzoldt