Effects of visual information on decision making during way-finding in emergency and non-emergency situations

Abstract

Finding the way out of a building during evacuation is not an easy task. Ideally, instructions provide clear and unambiguous information to occupants about the best means to evacuate. However, many times, building occupants may find the best course of action is not always clear. Conflicting or ambiguous cues can make a process that requires a quick response, slow and possibly more dangerous. Emergency signage may be vague, conflicting with other cues, or easily overlooked. The egress route directed by signage may appear difficult to traverse or dangerous. It is crucial then to best understand how evacuees find, interpret, and act upon visual information provided by emergency signage and egress routes in emergency situations. We tested the way visual information of signage and routes is used when an occupant needs to evacuate a building. In a virtual reality experiment, conflicting visual cues were pitted against each other in order to best understand how participants use visual information.

1. Introduction

Emergency situations in which building occupants must exit for their safety can be difficult and complex. In a perfect world, instructions provide clear and unambiguous information about the best means to evacuate. However, in that sudden moment of evacuation, a building occupant may find the best course of action is not always clear. Conflicting or ambiguous cues can make a process that requires a quick response, slow and possibly more dangerous [1]. Emergency signage may be vague, conflicting with other cues, or easily overlooked [2]. The egress route directed by signage may appear difficult to traverse or dangerous. It is crucial then to best understand how evacuees find, interpret, and act upon visual information provided by emergency signage and egress routes in emergency situations.

Another important cue providing visual information during evacuation might be illumination. During fire emergencies, illumination of egress routes may be deteriorated. Previous studies showed that reduced illumination may play a role during evacuation [3]. One study gave first indication that corridor brightness and emergency exit signs affected spatial decision making while navigating with a joystick through a virtual environment. The authors found that when participants were facing competing environmental information, exit signs became less effective [4].

The visual information available should also influence the locomotor behaviors of pedestrians during evacuations. The Behavioral Dynamics Framework [5] gives a good foundation for this perception and action interaction, that can be expanded to include behaviors during building egress. Previous work has found the effect of visual information on a pedestrians’ walking speed [6], their direction of heading [7], and their interactions with groups of other people [8].

Along with the rise of high-resolution, mobile virtual reality (VR) technology, experimental research on evacuation behavior became ever more feasible. VR allows participants to experience simulated, complex, and dangerous situations with rigorous experimental control and high levels of immersion in the safe space of a laboratory [9–11], VR has been used to investigate emergency training [12, 13], pre-evacuation behavior [14–16], and effectiveness of way-finding installations during evacuation [4,17,18].

The first goal of the present study was to test whether illumination and directional information from emergency signage have a stronger influence on participants’ spatial decision making. The second goal was to compare the spatial decision making as well as other movement parameters such as position, speed, and heading in simulated emergency and non-emergency conditions.

In the present study, we tested the way visual information of signage and illumination is used when an occupant needs to evacuate a building. In a virtual reality experiment, we pitted conflicting visual cues against each other in order to best understand how participants use the given visual information.

2. Methods

2.1. Participants

Twenty participants, 13 female and 7 male (mean age23.7), were recruited for this experiment. All reported having no visual, auditory, or motor impairments. Informed consent was obtained from all volunteers, who were paid for their participation. The experiment was approved by the Brown University Institutional Review Board.

2.2. Apparatus

The research was conducted in the Virtual Environment Navigation Laboratory (VENLab) at Brown University. Participants were able to freely physically walk around in a 12 × 14 m room (as opposed to navigating with an input device, such as a joystick) while viewing the virtual environment through a wireless stereoscopic head-mounted display (Oculus Rift DK1) with a 111° diagonal field of view with a resolution of 640 × 800 pixels in each eye. Displays were generated on a Dell XPS workstation at a framerate of 60 fps, using the Vizard 4 software package (WorldViz). Head position and orientation were recorded at 60 Hz.

2.3. Display

The virtual environment of every trial was a hallway T-junction. The corridors were 1.79 m wide by 2.44 m tall, with the body of the T-junction measuring 12 m long, and each arm measuring 4.86 m long. On some trials, an exit sign was placed below the ceiling on the wall of the intersection. The sign (when visible) pointed left or right. On some trials, either or both arms of the T-junction could appear dimly illuminated or normally illuminated. Participants in the emergency evacuation exit scenario could hear a fire alarm sounding during each trial, as well as visibly see a flashing fire indicator on the ceiling above the intersection of corridors.

2.4. Design

The experiment was a 3×4×2 design. The following variables were manipulated:

• Exit Signage Direction: 3 levels, within subjects; left pointing, right pointing, or no sign.

• Corridor Illumination: 4 levels, within subjects; both corridors brightly lit, both corridors dimly lit, left bright and right dim, left dim and right bright.

• Exit Scenario: 2 levels, between subjects (10 participants per group); non-emergency exit, emergency evacuation.

The participant’sexit choice and their trajectories through the virtual corridors were recorded. There were 48 trials per participant, with 4 replications per trial condition.

2.5. Procedure

In the instructions, participants were told they are inside a building and must decide which direction to go at the intersection to leave the building. At the start of every trial, participants were situated in the virtual environment of a hallway T-junction, in the body of the T (see Fig. 1). At the end of each arm of the T was a door that opened automatically as the participants approached it. Once they passed through the doorway, the trial would end and the corridor would disappear. The participants would then orient to a new starting location, where the T-junction would appear again and the next trial would begin.

Fig. 1.

Schematic overview of the virtual T-intersection.

3. Results

3.1. Exit Choice

In 49.62% of the trials participants decided to exit through the left door, i. e. there was no bias in exit choice, p = 84. Fig. 2 illustrates participants’ exit choice as a function of the experimental conditions.

Fig. 2.

Exit choice as a function of exit signs, illumination, and alarm condition. Data from illumination conditions in which both arms are equally illuminated (either both dark or both bright) are combined (labeled light: no info).

The categorical exit choice data were analyzed using logistic regression (Table 1), based on a model in which the participant’s decision (left/right) was predicted by signage, illumination, and alarm conditions. We combined the illumination conditions in which both arms of the T corridor were either bright or dark, since in both cases the illumination did not provide information whether left or right was the better option. The reference category for the logistic regression model was the signage and illumination conditions which provided no information regarding direction in the no alarm group. The model predicted whether participants would go right.

Table1.

Coefficients and confidence intervals for logistic regression predicting exit choice at the T junction by signage, illumination and alarm condition

	b	exp(b) = Odds right	lower 195% CI	upper 95% CI	p
Intercept	−0.13	0.88	0.61	1.26	.49
Signage
Left	−2.65	0.07	0.04	0.12	<.001
Right	3.36	28.71	15.40	58.94	<.001
Illumination
Left	−2.60	0.07	0.04	0.14	<.001
Right	2.22	9.28	5.42	16.68	<.001
Fire alarm	0.13	1.15	0.77	1.70	.50

Note: b = regression coefficient (logit odds), CI = confidence interval, R² = .49 (Hosmer-Lemeshow), . 51 (Cox-Snell), . 67 (Nagelkerke).

Adding an interaction term to the model did not improve the fit, X²(12) = 13.22, p = .35. Signage and illumination were significantly predicting exit choice (Table 1). The main effects of signage and illumination showed that participants were more likely to follow the direction indicated by the information provided in the condition compared to the control condition, independent of the fire alarm condition. The effects of signage were stronger than the effects of illumination.

3.2. Speed & Heading Time Series

In order to analyze the differences in how the participant dynamically responded over the course of the trial, the grand means of time series for position, speed, and heading were plotted over time (Fig. 3). The plot for position (collapsed across left and right) shows that participants made a tighter turn in the emergency evacuation alarmed scenario compared with the non-emergency scenario (Fig. 3.a).

Fig. 3.

The grand means of time series for (a) position, (b) speed, and (c) heading.

In the time series for speed (Fig. 3. b), participants walked faster during the first section of the trial (before the turn down an arm of the intersection) in the emergency evacuation scenario. After the turn, as they approached the door at the end of the corridor, they slowed down to similar speeds between scenarios. The mean of participant’s speed (in m/s) across the time series was used to analyse the difference between scenarios. This difference in speed was significant, t (18) = −3.92, p < .001.

In the time series for heading (Fig. 3. c), the participants began trials facing 0° (straight ahead), walked forward, and then turned 90° down an arm of the intersection (collapsed across left and right). While there was a qualitative difference between the emergency evacuation and non-emergency exit scenarios, this difference was not significant, t (18) = −0.95, p = 3.96, when using the mean heading ( degrees ) across the participant’s time series. It is likely accounted for by the increase in speed during the first half of the trial causing the participants to turn earlier.

4. Conclusion

In the present study, we pitted conflicting visual cues from emergency exit signs and corridor illumination against each other in a simulated fire alarm scenario. The results show that both signage and illumination affect exit choice with stronger effects of the exit signs, with no interaction effects. Interestingly, exit choice was not modulated by the fire alarm compared to the control condition and are in line with previous research [4], However, participants walked significantly faster in the fire alarm group compared to the control group, indicating that locomotor behaviors are not only guided by visual information but also goal demands, such as evacuating during an emergency.

Future work should integrate these findings into evacuation models as well as empirical studies, testing more environmental variables, e. g., corridor width [4], reduced visibility due to smoke or social influence [19]. The goal of the present and future work is to contribute to a more generalized empirically driven framework describing pedestrian evacuation based on rigorously controlled experiments.