1 Introduction

1.1 Background

Conditions of emergency situations, such as fire, in large built structures, often necessitate the adaption of escape routes to ensure a safe evacuation. Digital escape route signage allows adaptation to the specific emergency conditions. However, people are not familiar with digital escape route signage and might feel irritated if the displayed escape route direction changes right in front of them. As a result, they might follow other factors of influence on direction choices. There are three main categories of such influences, environmental influences, social influences and influences of familiarity with places and routes. Regarding environmental influences, corridor width and lightning, for example, exert an influence on route choice in the way that people tend to take routes that are wider and brighter, first shown by Taylor and Socov [1] and confirmed, for instance, by Vilar et al. [2]. Furthermore, they try to maintain the initial direction of travel, referred to as least-angle strategy by Conroy-Dalton [3] and Hochmair and Frank [4]. Social influences are mostly based on the affiliation theory by Mawson [5, 6], studies of Sime [7, 8] and normative social behaviors, differentiated from informative social behaviors by Deutsch and Gerard [9]. Various investigations have found that people tend to move towards other people, such as Mawson [5, 6], Nilsson et al. [10] and Kinateder et al. [11]. Influences of familiarity with places and routes have been found by Butcher and Parnell [12] and in evacuation field studies by Sime [7, 8] and Shields and Boyce [13]. Moreover, perceived risk, such as in emergency situations, leads to pronounced affective behaviors [14,15,16]. Analytic thoughts are hindered by time pressure, further leading to affective decisions [17]. Keinan et al. [18] found evidence for rash decisions under stress, supported by the eye-tracking study by Stankovic [19].

1.2 Cognitive Tunneling

Easterbrook [20] was the first to describe what became known as the effect of cognitive tunneling, also referred to by attentional tunneling and attentional bias. Attention was observed to be focused on specific cues, referred to as “central cues”, and less on other cues, referred to by “peripheral cues”, the more people experienced stress resulting in emotional arousal, referred to by Easterbrook [20] as “drive”. Easterbrook’s groundwork [20] was confirmed and refined by various studies. In the field of aviation, Yeh and Wickens [21] reported on attentional tunneling in pilots, missing out on unexpected targets, with the attention overly drawn by cued objects. Briggs et al. [22] reported on the risks of cognitive tunneling due to emotionally involving telephone conversations by drivers, unlikely to be solved by hands-free cell phones [23]. Dixon et al. [24] point to the problem of attentional tunneling in the context of surgery endoscopy with the aid of AR, though not drawing a clear line to inattentional blindness [25] with the famous example of the “invisible gorilla” [26].

1.3 Aim and Scope

Despite the variety of the aforementioned studies on cognitive tunneling, to the best of the authors’ knowledge, there has been no study investigating effects of cognitive tunneling in high-stress evacuation situations with regard to digital escape route signage. Only the study by Vilar et al. [2] referred to this phenomenon though with regard to the established static escape route signage and with the participants not moving in reality. Their findings indicate that the participants decided in compliance with the established static escape route signage, especially in emergency conditions, but payed more attention to environmental influences in their condition without signs. Vilar et al. [27] pointed to the high arousal experienced in emergency conditions. According the Yerkes-Dodson law [28, referred to by 29], the function of performance in dependence has an inverted U-shape. The influence of higher stress levels including physical demands were not investigated in these studies. Moreover, in a prior study, we found that the consistency of decisions in favor of specific digital escape route sign types decreased with mental, emotional and physical stress [30].

Therefore, the scope of the present paper is the investigation on the effect of cognitive tunneling in high-stress emergency situations with mental, emotional and physical stress and its meaning for the use of digital escape route signage. A study in a virtual environment of a corridor system with adaptive digital signage and competing environmental influences, i.e. corridor widths, lightning and angle between the current and the follow-up path, was conducted. In consideration of the ageing society in most western countries, an age-differentiated design was chosen. Furthermore, age effects were expected as we had found differences in preferences of digital escape route sign types in a prior age-differentiated study using paired comparisons [30]. Hence, an age-differentiated study design with 30 young and 30 elderly participants was used. There were three conditions varying in applied stressors, i.e. background noise and simulated fire in the virtual environment (VE), and physical demands in terms of walking speed, which was not only realized in the VE but also matched with the real world requirements using a tread mill. The decisions in favor of the digital escape route signs were analyzed with the focus on differences between low- and high-stress situations.

2 Method

2.1 Participants

60 participants were recruited externally. There were two age groups. The young group consisted of 30 participants (15 f, 15 m) aged between 20 and 30 years (\( M = 25.27 \), \( SD = 2.84 \)), the old group consisted of 30 participants (15 f, 15 m) aged between 60 and 79 years (\( M = 68.20 \), \( SD = 5.14 \)) to cover potential age effects. Ethical approval was obtained by the ethics committee of the University Hospital of the RWTH Aachen University (EK 190/16). No participant had to be dropped out of the recruited ones, as exclusion criteria were communicated beforehand. These were defined by severe illnesses, especially coronary heart diseases, cardiac pacemakers, bypasses, artificial arteries and joints, pregnancy as well as mental, mobility, visual and hearing impairments. No participant violated the minimum boundary for visual acuity of 0.8 or suffered from a red-green colour-deficiency. No participant had less than 80% correct responses in a stimulus-response pre-test (see Sect. 2.5).

2.2 Apparatus

The photos in Fig. 1 show the view over the shoulder of a participant onto the virtual environment (VE) and the experimental setup in the laboratory with a participant walking on the treadmill looking at the VE. The h/p cosmos mercury treadmill had a walking surface of 1.5 by 0.5 m. The preinstalled customizable safety harness system prevented participants from injures in case of stumbling, which was of particular importance as persons with an age up to 79 years were included in the study. The VE was displayed on a 65″ monitor in walking direction on the treadmill in a distance of 1.5 m. It was mounted on a height-adjustable table to ensure the best immersion possible. The eyes of the participant were on 2/3 of the monitor’s height. For direction choices, i.e. left of right, two Wii remotes, one for each hand, were used. Two external loudspeakers at either side of the monitor were played looped acoustic background noise during the evacuation conditions. There was no natural light to the laboratory to keep light conditions constant. Therefore, all windows were opaquely closed.

Fig. 1.
figure 1

Left: view over the shoulder of a participant onto a corridor intersection in the virtual environment. Right: view from behind on the participant secured in the harness on the treadmill in front of the 65″ monitor. The opaquely closed windows were slightly opened to improve light conditions for the photos.

Full size image

2.3 Virtual Environment

The virtual environment (VE) was built in Unity 3D. All potential paths of the participants were predefined by waypoints. The current position was linearly interpolated according to the individual speed of the participant walking on the treadmill. There were 40 models of rooms according to the 40 decision in 40 intersections. The room order was randomized, except for the first two rooms. These were always rooms without signs though with environmental cues potentially influencing direction choices, i.e. differences in lighting, corridor width and angle between the right and the left hand side. Each room model had several inner and outer opaque walls to create corridors and was based on a template room, which contained all possible waypoints, triggers and settings (lighting, sign type, sign direction, sign change, etc.). One corridor led from the starting point of the room to the intersection. The length of this corridor was adaptive to the speed of the participant to offer each participant the same potential decision time of 4000 ms and 2000 ms after sign display change, regardless of his/her individual speed. For the way from the intersection to the starting point of the next room, there was a left and a right path, consisting of several corridor parts. Taking the left and the right path after the T-intersection together, the path through every room model was a closed loop by design. However, the participant was prevented from going backwards by a wall, which was automatically inserted at the end of the path that was not chosen, not noticeable by the participant. On one side, the first part of the corridor after the T-intersection was narrower, i.e. 1.30 m, darker with only one light source at the far end. The path was sharply angled, inhibiting the participant to see how the path would go on. The corridor on the other side was wider, i.e. 2.60 m, and much brighter, as it was equipped with nine light sources. The path was less angled, allowing the participant to see around the corner and that the path almost led in the direction of his current direction of travel. The escape route sign displays, sized 0.37 m by 0.22 m, were placed in the top-middle position in the T-intersection at a height of 1.80 m above the floor in the VE to ensure the highest possible effectiveness of the signage [31]. The background of the sign displays was modelled as emissive light source.

Invisible barriers were used to start and end the decision input option. Each barrier was modeled as a single shot event, which was triggered when the virtual camera, i.e. the virtual position of the participant in the VE, passed the barrier. When crossing the barrier at the beginning of the corridor directly leading towards the next intersection, the participant’s decision was triggered by the rumble function, i.e. the Wii remotes in both hands started to vibrate. As soon as the participant decided for a direction, only the Wii remote in the corresponding hand continued to vibrate, while the other one turned still as feedback for the participant. For example, the participant pressed the button on the Wii remote in the left hand, deciding to turn left at the intersection, this Wii remote continued to vibrate, while the Wii remote in the right hand turned still. Technically, after the direction decision, the waypoints of the left or right path were added to the queue according to the choice of the participant, followed by adding the waypoints of the first path of the next room, leading towards the next intersection, to the queue. The conditions were highly controlled and the same for each participant. Hence, participants could not stand still or turn, not in the VE and, obviously, not on the treadmill. Hence, if the virtual camera reached an intersection but no input, i.e. direction decision, had been entered, the movement was continued “against the wall”. The whole monitor turned black for 3000 ms. Then, the camera was placed at the starting point of the next room. This procedure was shown in the VE training to avoid potential surprises and unintended behavioral responses by participants, such as stopping on the treadmill and stumbling during the actual experimental conditions.

2.4 Experimental Design

A mixed design was used in the present study. All participants completed 40 randomized decision-making tasks for directions at 40 T-intersections in each of three fully permuted experimental conditions varying in applied stressors and walking speed, individually matched between the VE and the treadmill (see Table 1).

Table 1. Experimental conditions varying in walking speed and applied stressors.
Full size table

Each intersection led onto a lighter, wider and less angled route on one side offering the view into the following corridors in the direction of initial travel, and onto a darker, narrower and sharper angled path on the other side. 32 of the 40 intersections were equipped with a digital escape route sign based on ISO 7010 [32] pointing to the right or the left. Four sign types were used, each one at eight intersections: Standard without further elements, Updated with temporal update information, Dynamic with a green flashing frame and Crossed with a cross for the opposite direction indicating that it is blocked. The direction of the displayed escape route sign was changed half of the times while the participant was approaching. The distance to the next T-intersection was always the same and independent of the direction decision of the participant, facilitated by the automatically adapting VE, not noticeable for the participant (see Sect. 2.3).

Each participant completed the walk through the automatically adapting corridor system three times under three experimental conditions varying in the situation, i.e. everyday vs. emergency, applied stressors and physical demands (see Table 1). The least demanding condition was the 40noS condition, defined as everyday condition. The walking speed was 40% of the maximum walking speed assessed in a 10-meter Fast Walk Test (10 mFWT [33], see Sect. 2.5) corresponding to a comfortable speed, calculated on the data basis provided by Bohannon [34]. The walking speed was realized on the treadmill and as well within the virtual environment. The speed was the same in the evacuation condition 40 + S. However, stressors in form of fire, coming from underneath the doors in the corridors, and continuous background noise, i.e. a looped sequence of echoed incomprehensible crowd mumbling, played at about 68 dB(A), were applied. The highest demanding condition was the evacuation condition 65 + S with the same stressors as in 40 + S but at 65% of the maximum walking speed in the 10mFWT, calculated as the ratio of the maximum evacuation speeds according to the MSC Circ. 1238 [35] and maximum walking speeds [34]. The dependent variable is the percentage of direction decisions in compliance with the escape route direction recommended by the digital signage.

2.5 Procedure

Upon arrival, participants were orally informed about the procedure, before they signed an informed consent form. They answered several questionnaires, for instance on demographics. The participants were tested for visual acuity, red-green colour-deficiency and reactions in a computer-based arrow stimulus-response (S-R) test. A 10-meter Fast Walk Test (10 mFWT [33]) was performed to assess the maximum walking speed of participants to adapt the treadmill speed to the participants’ gait abilities in the trials and to receive comparable results from persons with different characteristics, such as age and gender. The participants were instructed to walk as fast as they safely could without running from one marking on the floor to another, as this approach has established since Bohannon et al. [36]. The markings were three meters before and after the measuring distance to inhibit effects of acceleration and deceleration within the measuring distance of the ten meters. Furthermore, Watson [37] recommended the destination marking not to be the ending point of the measuring distance to avoid premature deceleration and also distraction from the walking test, while time is still measured. Participants were asked to perform the test twice. If the slower speed of the two tests deviated by more than 10% from the faster, the test was repeated and the results compared again. The average of the two tests served as reference value for the following parts.

As a treadmill was used in the experimental conditions, each participant completed a five-minute training on the treadmill. Thereby, every participant became used to walking on a treadmill and between-subjects effects of potential prior experiences were minimized. Safety was ensured by a harness system (see Fig. 1 right). It prevented participants from injuries in case of stumbling. Before the training, the harness was adjusted to the participant’s girth and body height. This adaptation not only offered optimal safety for the participant but also ensured that no body weight was loaded on the harness during the trials. The maximum speed during the experimental conditions was 65% of the walking speed achieved in the 10 mFWT. Thus, the walking speed was gradually increased to this maximum experimental speed reaching 25% of this speed after one minute, 50% after two minutes, 75% after three minutes and 100% of the maximum experimental speed after four minutes, i.e. 65% of the speed in the 10 mFWT, which was kept constant during the last minute. Afterwards, decision-making with the input devices, i.e. the Wii remotes, was practiced. The training in the VE through four intersections without environmental influences nor escape route signage ensured understanding and allowed a stronger focus on the task during the experimental conditions rather than on the operation.

Prior to the experimental conditions and in between them participants had ten-minute breaks in sitting position to minimize influences from the previous part on the following condition. The situation, i.e. everyday vs. evacuation, was described to the participants upon the start of each condition with a standardized text with the differentiation between an everyday and an evacuation situation. Participants were told to imagine that they were in corridor of passenger cabins inside a cruise ship. They were instructed to try get out. In the everyday scenario, this instruction was described as a general ship announcement, clearly indicating that there is no emergency, but as result of a fire alarm in the evacuation scenarios. Further instructions were that the participants could not stand still nor turn around in the VE and that they were supposed to make their decisions as they would in reality. Each experimental condition was concluded by a short semi-structured interview on strain and decision-making.

2.6 Statistical Analysis

All statistical analyses were conducted using IBM SPSS Statistics 23. The level of significance was set to α = .05. All tests were two-tailed. The data was not normally distributed. Therefore, the non-parametric Friedman’s ANOVA for repeated measures was used. Pairwise comparisons between the conditions were performed by the non-parametric Wilcoxon signed-rank test for dependent samples, reported according to Field [38] by the test statistic \( T \) as the lowest value of the two types of ranks, i.e. positive and negative ranks, and the effect size \( r = Z/\sqrt N \), with N for the number of observations. Bonferroni correction was applied to pairwise comparisons for evaluating significance, \( \alpha_{c} = .05/c \), with \( c \) for the number of comparisons. Marginally significant was defined as \( \alpha > \alpha_{marg } > \alpha_{c} \). Exact statistics instead of asymptotic statistics were used for all tests.

3 Results

At intersections with congruent environmental cues and displayed escape route directions, the compliance rate with the signage was very high, i.e. 89% \( (SE = 1.92) \). This is little surprising. However, the decisions at the intersections with conflicting cues are by far more meaningful. These were the intersections with escape route signage pointing in one direction but the environmental cues such as corridor width and illumination suggesting a decision for the opposite direction. A change in decision-making behavior, depicted in Fig. 2, was observed between the everyday condition and the emergency conditions with applied stressors in form of fire from underneath the doors in the VE and continuous background noise. The decisions at intersections with matched environmental cues and escape route sign directions as well as the comparison with these are not further discussed in this article.

Fig. 2.
figure 2

Decisions for directions in compliance with the digital escape route signs at intersections with the displayed direction competing with the direction suggested by the environmental influences, over the three experimental conditions varying in the walking speed between 40% and 65% of the maximum individual speed and the application of stressors in the two evacuation conditions 40 + S and 65 + S.

Full size image

There was a main effect of condition, indicated by Friedman’s ANOVA, \( \chi^{2} \left( 2 \right) = 22.34 \), \( p < .001 \). The depicted data in Fig. 2 suggests a stronger effect of condition on young than on elderly people, confirmed by separated Friedman’s ANOVAs, young group: \( \chi^{2} \left( 2 \right) = 20.18 \), \( p < .001 \); old group: \( \chi^{2} \left( 2 \right) = 4.89 \), \( p = .089 \). Hence, the influence of condition on decisions in favor of escape route sign recommended directions was highly significant in young participants, but not quite significant in elderly participants. Bonferroni corrected pairwise comparisons between the everyday condition and both emergency conditions revealed significant increases in decisions in compliance with escape route signage in the young group for both emergency conditions, condition 40 + S: \( T = 3 \), \( p < .001 \), \( r = - .51 \), condition 65 + S: \( T = 4 \), \( p < .001 \), \( r = - .48 \), indicating a large and a medium to large effect size. Regarding the old group, the increase in decisions in favor of the escape route sign direction from the 40noS condition, without any stressors at a comfortable walking speed, to the 65 + S condition, with fire in the VE and background noise at evacuation speed, was significant, \( T = 8, \) \( p = .021 \), \( r = - .29 \), while the increase to the 40 + S, with applied stressors but without increased speed, was only marginally significant, \( T = 8, \) \( p = .034 \), \( r = - .27 \), both indicating small effect sizes.

4 Discussion

In summary, we have conducted an age-differentiated study in a virtual environment with digital escape route signage and competing environmental influences, such as corridor width and lightning. The two age groups consisted each of 30 participants with the young group aged from 20–30 years and the old group aged from 60–79 years. Their decisions varied between the three conditions, i.e. a low-demanding everyday condition without stressors at a comfortable walking speed (40noS) and two evacuation conditions with simulated fire and background noise with the condition 40 + S of these two evacuation conditions at a comfortable walking speed and the other at a more demanding typical evacuation speed.

Significant differences in direction decisions were found between the everyday and the two evacuation conditions when sign directions conflict with environmental behavior, suggesting a pronounced tunneling effect in the high-stress situations. A high compliance of direction choices with the displayed directions on digital signage was observed in the evacuation conditions, especially in the young group. It is noteworthy that the effect was stable despite the randomized adaptation of the displayed escape route direction to – in reality – potentially changing emergency conditions, such as fires. Participants reported in the semi-structured interviews to have felt more stressed in the evacuation conditions and to have, hence, focused more strongly on the visual search for the digital escape route signs followed by the complying decisions. In contrast, in the everyday situation, participants payed more attention to the other environmental cues and made their decisions often in favor of the corridor brightness and width. This is particularly interesting, when taking into account that there would have been good reason to hypothesize the opposite as the influence of affect and limitation of logical reasoning had been found to increase, for example, with risk perception [14, 16] and perceived stress [39]. However, the attentional tunneling in favor of escape route signage is in line with the escape study by Vilar et al. [2], using the established static signs and finding high compliance rates in emergency conditions.

The observed, strong focus on digital escape route sings, especially in evacuation conditions emphasize the importance of up-to-date escape route signage. People are likely to miss out on information provided from other sources in evacuation situations, as they narrowly focus on escape route signs during escape, like the pilots in the study by Yeh and Wickens [21] who missed out on unexpected targets when focused on cued targets.

Nevertheless, the findings should be interpreted in light of the limitations of the conducted research. There are limitations regarding the sample as well as the virtual environment (VE). Strict exclusion criteria were defined. On the one hand, they ensured participants’ health and safety. On the other hand, these exclusion criteria might have biased the sample resulting in a healthier and fitter sample than the average population, especially regarding the elderly participants. Regarding the VE, real world conditions would, of course, have been more desirable, but were discarded for ethical and safety reasons.

5 Conclusion

Evacuation conditions were found to lead to cognitive tunneling and a strong attentional focus on digital escape route signs, especially in young persons. These findings emphasize the potential and importance of up-to-date digital escape route signage to support people’s way finding during evacuations. Obsolete direction information provided by traditional static signage is not only useless but dangerous because of the strong focus on signage, especially under evacuation conditions, in which correct direction choices are most critical.