An Experiment in Wayfinding in a Subway Station Based on Eye Tracker Analytical Techniques for Universal and Age-Friendly Design

Abstract

The complexity of subway station space can impact the efficiency of passenger navigation. The subway spatial environment is a key factor affecting indoor wayfinding for pedestrians; however, the research framework that examines how various environment factors influence pedestrians during different stages of wayfinding remains ambiguous. This study examines how environmental elements may affect users to varying degrees at different stages of wayfinding, which in turn affects their wayfinding efficiency, recording and analyzing the wayfinding performance and eye-tracking data of 32 participants. The findings reveal that different environment factors exert varying degrees of influence on pedestrians at different stages of wayfinding. Significantly, signage (p = 0.000) proves to have the most substantial impact on wayfinding, followed by stairs/escalators (p < 0.05), but the participants walked to the wrong platform in the TD2 scenario because they were guided by the Line 2 signs in front of the stairs/escalators. Thus, the influence of signage is not entirely positive. This study contributes to an understanding of the differences in the influence of environmental elements on wayfinding during different wayfinding stages, and provides suggestions for the spatial environmental design of subway stations and the improvement of wayfinding efficiency.

1. Introduction

In recent years, the development of comprehensive rail transit systems has profoundly influenced urban growth and transportation choices. However, the internal spatial layout of some large-scale transportation interchanges has become increasingly complex. Numerous studies suggest that the complex architectural spaces confuse users and negatively affect their wayfinding behavior. In subway stations, being underground, passengers often face challenges in wayfinding, making research on wayfinding in subway stations a trending topic in recent years’ studies.

Previous studies have identified several factors that influence passengers’ wayfinding, including spatial knowledge [1], individual passenger characteristics [2], signage systems [3,4], and surrounding crowds [5,6]. Among these, spatial knowledge is considered a crucial factor affecting an individual’s wayfinding. Studies indicate that passengers with varying levels of spatial knowledge exhibit different decision-making behaviors when confronted with the same environment, and the characteristics of the spatial environment significantly impact passengers’ route selection [7,8,9,10]. Key architectural factors influencing spatial cognition include internal environmental information, layout characteristics, and accessibility, manifested in elements such as the location of facilities, landmarks, signage, spatial differentiation, maps, and lighting [11]. However, existing research on the wayfinding process primarily focuses on the perception of spatial environments from a behavioral psychology perspective, leaving the specific environment factors that influence wayfinding somewhat unclear. In unfamiliar or unknown environments, environmental information serves as a critical basis for individuals’ spatial cognition. Spatial behavior arises from individuals’ experiences and understanding of their environments, leading to variations in route selection. Regarding environmental information, its influence on wayfinding decisions manifests in two forms: first, as implicit information that subtly suggests direction to way finders; second, as explicit information that clearly indicates direction, such as signage [12]. The location of escalators, and continuous guiding light strips can be a clear indication of direction. By studying route selection in specific contexts, we can gain insights into how environmental conditions affect the interaction between wayfinders and indoor environments. However, few studies have explored the interaction between passengers’ significantly different perceptions of various environment factors and their wayfinding observations and judgments through fixation data in spatial environments, which constitutes the second issue addressed in this study.

Previous studies on wayfinding have primarily employed virtual reality environments to investigate factors such as the distance and speed of way finders [13,14,15]. However, the light, material, sound and other details in a virtual reality scene may have a certain separation from the real world, resulting in passengers in the wayfinding process finding it difficult to generate a sufficient sense of realism and trust, affecting the effect of wayfinding and failing to present the wayfinder’s psychological cognitive representations. The acquisition of spatial information largely occurs visually. Given the complexity and subjectivity of wayfinding behavior and spatial cognition, eye tracking, as a technology that accurately records human visual data, can capture fixation information during passengers’ wayfinding process. This allows for the quantitative analysis of human behavior and psychology [16]. In recent years, eye tracker analytical techniques have gradually been applied to research on subway stations, offering unique advantages in studying passengers’ visual attention and spatial behavior.

In light of these advantages, and to address the aforementioned research issues while gaining a better understanding of passengers’ wayfinding behavior, this study employs an on-site eye-tracking experiment. On the basis of the collected and analyzed eye-tracking visual data of subway passengers, we explore the impact of environmental elements in subway stations on passenger wayfinding efficiency, specifically including information on eight types of environmental elements: direction signs, glossy finish, spatial decoration, gates, stairs/escalators, elevators, pillars, and wall openings. We establish a research framework for influencing factors of subway passengers’ wayfinding behavior based on eye movement data. This study takes into account a variety of factors affecting wayfinding, such as spatial knowledge, signage, and personal attributes, comprehensively explores the role of each factor in different stages of wayfinding, and employs a variety of research methods to validate the results, demonstrating the rigor of the study. In addition, eye tracking records the gaze information of passengers during the wayfinding process, realizing the quantitative research on the wayfinding process. Starting from visual behavioral characteristics, the study of the influence of environmental factors on wayfinding in different scenarios, which is less involved in previous studies of subway wayfinding, provides a new research perspective for the field, fills the gaps in related research, and provides new ideas and methods for subsequent research and optimization of subway spatial design.

2. Related Work

2.1. Wayfinding Decision Making

“Wayfinding”, as a behavior, refers to the process by which individuals navigate to a destination in familiar or unfamiliar environments using sensory information from the external environment [17]. For example, through visual information, wayfinding passengers in a subway station need to visually observe the overall layout, elevator locations, passageway locations, etc., and through auditory information, passengers in a subway station need to listen to announcements to obtain important wayfinding information. Wayfinding in an unfamiliar environment can be quite challenging, as participants must be aware of their spatial position and identify the direction they need to follow [18]. In wayfinding research, social influence [19], environmental characteristics [20], and spatial cognition [1,21] are major factors that affect pedestrians’ wayfinding decisions. Some studies suggest that social influence is a key influencing factor in route selection during pedestrian evacuation [22]. Fu [5] discovered through virtual reality experiments that participants’ route selection during evacuation are influenced by surrounding crowds, indicating a tendency for individuals to follow others when making risk decisions, though not blindly. Yu [23] studied wayfinding behavior under stress conditions and found that the development of spatial knowledge under normal conditions is superior to that under stress conditions, and as spatial knowledge improves, individuals’ reliance on signage decreases. Tahir [24] et al. conducted a controlled experiment on the effects of navigation style and prior experience on wayfinding in a campus environment. Individuals were found to rely less on signage and navigate significantly more efficiently the more familiar they were with the environment. Researchers have noted that wayfinding in indoor environments presents more spatial cognitive challenges, with spatial knowledge being a crucial component of the wayfinding process. Wayfinders can determine the approximate location of their destination by processing and evaluating information in their spatial cognition [7]. Therefore, acquiring spatial knowledge is essential for environmental exploration. Kinateder [25] and Lin [1] found in their studies that participants tend to prefer familiar routes and exits, and those with a comprehensive understanding of the space are more likely to find the shortest route. Furthermore, Lin [15] proposed that repeated exposure significantly influences participants’ wayfinding by reducing the time required. However, time and psychological stress negatively affect wayfinding, although such stresses can be mitigated through repeated exposure to the space. While considerable research exists on the influence of spatial knowledge on wayfinding, the influence of specific spatial knowledge, namely spatial environment factors, on wayfinding warrants our research.

2.2. Spatial Environment Factor

Visual cues act as a mediating link between the external environment and user behavior, providing support for behavioral decision-making. Consequently, many researchers have defined and classified spatial visual environmental cues within the context of wayfinding. In architectural space, researchers such as Arthur and Hölscher [26,27] have suggested that escalators, entrances and exits, as well as symbols such as signs and text, are key environmental elements that influence wayfinding. Stairs are a major obstacle to wayfinding, as wayfinders assume that each floor of a multi-story building has the same spatial structure based on the location of the stairs. This assumption leads to the construction of cognitive maps of each floor and prevents the formation of a correct mental map of the entire environment. Li [28] proposed that both the types and combinations of subway signage can influence pedestrians’ short-term memory of these signs. Additionally, some researchers have highlighted the impact of sign visibility on passengers’ wayfinding [29,30]. Shi [20] considered various environment factors, such as illumination, color combinations, sign height, and observation angles, which affect the readability of subway station signs. Their findings indicated that brighter lighting conditions enhance readability for passengers, while achromatic combinations outperform colored ones. Observation angles have a marked impact on readability. Designers can give priority to brighter lighting conditions in the design of the spatial environment; in the color scheme, as far as possible, the use of a more concise combination of achromatic colors can enhance the legibility of logos, while logos can be set up at suitable viewing angles. Vilar [12] examined how corridor brightness and width, as well as the direction of signs, influence participants’ route selection in both everyday and emergency situations. The results showed that in everyday situations without signage, participants predominantly chose wider and brighter corridors, indicating their strong reliance on environmental information. Vilar et al. found through studies that focusing solely on signage in a complex building cannot provide good wayfinding performance. During indoor wayfinding, environmental variables such as corridor width and brightness can affect pedestrians’ route selection [10]. Environment factors are sometimes implicit information influencing pedestrians’ wayfinding, and researchers have long found that building layout significantly affects wayfinding performance despite the use of signage [31]. Arthur and Passini [32] pointed out that environmental cues such as signs sometimes potentially conflict with complex architectural spaces, people tend to get lost due to the existence of signage, hindering the formation of cognitive maps in the brain and leading to errors in wayfinding. Therefore, when studying wayfinding, we need to consider the negative impact of external environmental information such as some improperly designed or used environment factors on wayfinding. Although there are a number of studies on the influencing factors of wayfinding, the significant differences in specific environmental influencing factors in subway space are still a research gap and lack a systematic research framework. This study aims to fill the corresponding gap while investigating the above issues.

2.3. Eye-Tracking Equipment

In recent years, eye-tracking has been gradually applied to wayfinding experiments in architectural spaces. Previous studies have shown that using eye-tracking equipment to explore the impact of indoor spatial environment on human visual perception has become an important method [33]. As a visual cognitive assessment tool, eye-tracking makes studies on wayfinding cognition no longer limited to wayfinding results but extended to quantitative studies of the wayfinding process [34,35]. Helmut [36] proposed a method that integrates immersive virtual environments and eye-tracking technology to evaluate the role of navigation systems in indoor wayfinding in traffic buildings. At present, eye-tracking experiments include laboratory environment experiment and on-site experiment. Due to the complexity of human senses and scene space, the experimental approach in the laboratory environment, which generally involves participants using an eye-tracker to view photos or videos on a computer, does not allow participants to develop a true sense of the structural layout of the subway space or experience realistic wayfinding behavior, which can lead to the accuracy of the data collected through the eye-tracker being compromised. On-site experiments, compared to laboratory experiment, can, on one hand, provide more accurate data analysis, and on the other hand, help us explore wayfinding performance under different conditions. Some researchers have used eye-tracking technology to study the impact of different colors of signage on people’s cognitive behavior during evacuation [16,33]. In recent years, eye-tracking has gradually been applied to studies on subway space, but only a few scholars have studied subway evacuation, signage design, etc., with the use of eye-tracking technology, and an in-depth analysis of the influencing factors and fixation metrics of pedestrians’ route selection in subway stations is lacking. Therefore, this study will analyze the relevant fixation metrics from the perspective of data collection, including five gaze metrics: total fixation duration, fixation count, time to first fixation, first fixation duration, and visit count. This is more convincing and novel than previous studies and speculations.

3. Methods

In order to explore the factors that cause differences in route selection and spatial experience in subway stations, we conducted an on-site eye-tracking experiment on wayfinding in subway stations and collected objective eye-tracking data and behavioral data from participants. The following describes the details of this eye-tracking experiment, including the introduction of participants, equipment used, selection of experimental areas, experimental design and data collection.

3.1. Participants

A total of 32 participants (15 males and 17 females), aged between 12 and 65 years, were recruited for this study. All participants had normal or corrected vision, achieving a fractional visual acuity of more than 1.0 and a logarithmic visual acuity of more than 5.0, with corrections made by either wearing glasses or contact lenses. Participants had no strabismus, color blindness or color deficiency, and were in good physical and mental health. All participants volunteered to take part in this experiment, and then participants all put on eye-tracking glasses to start the wayfinding task from the same location. Prior to the experiment, participants were required to sign a consent form for the experiment.

We recruited a wide range of people from the society, including adults, the elderly and children, and excluded children under 12 years old and the elderly over 65 years old, because the elderly over 65 years old have fixed travelling habits and wayfinding methods developed over a long time, and when they encounter confusion in wayfinding, they are more inclined to ask for help directly from staff in the station, and have less initiative to think and make decisions on their own. Children under the age of 12 usually travel with adults, tending to be passive followers in the metro station; since they do not have the opportunity to make independent decisions, it is difficult for them to form a complete independent wayfinding strategy, and so even if they participated in the experiment, their data cannot truly reflect the general situation of independent individuals in metro station wayfinding, so there is no need for them to be included in the experiment separately. This choice is an important basis for creating an all-age-friendly metro station space, which helps us to gain a deeper understanding of the differences in the wayfinding needs and behaviors of people of different age groups in metro stations. Additionally, these participants included frequent underground riders, occasional underground riders, and people who had never travelled on the underground.

3.2. Apparatus

The apparatus used for the experiment mainly included eye-trackers and laptops. In order to assess participants’ visual attention to environment factors in subway station space, participants put on Tobii Pro Glasses 2 wearable eye-trackers with wireless real-time observation function to record the eye movement at a frequency of 50 Hz or 100 Hz, as shown in

Table 1. Basic information of equipment.

Equipment	Frequency	Manufacturer	Composition
Tobii Pro Glasses 2	50–100 Hz	Swedish Tobii AB (Stockholm, Sweden)	Head-mounted module, recording module, Tobii Glasses Controller 2.4.3 software, Tobii Pro Lab software (version 24.21)

. Frequency recordings can vary depending on the participant and the experimental environment in which they are made, but relevant eye-tracking studies have shown that data sampling rates of 80% or more are considered valid [37]. Figure 1 shows a participant using the Tobii Pro Glasses 2 eye-tracker during the experiment. Tobii Pro Glasses 2 captures accurate eye movement data using corneal reflex technology and dark pupil tracking technology. Eye-trackers capture high-definition images and learn participants’ fixation directions in real time with the use of cameras. A laptop is used for the on-site experiment to control Tobii Glasses Controller 2.4.3 software to calibrate and record eyeball metrics, while a single circular calibration plate is used to calibrate the fixation point and pupil position of each participant. The eye-tracking data of each participant were recorded for approximately 20 min, and then Tobii Pro Lab software (version 24.21) was used to analyze fixation heat map and eye-tracking metrics.

3.3. Experimental Area

This study aims at exploring the spatial environment factors that influence route selection by subway passengers. Based on the characteristics of subway station layout and surrounding environment and crowds, Taidong Station in Qingdao was selected as the site for on-site experiment, as shown in Figure 2. The first reason is that Taidong Station is a transfer station that requires intricate wayfinding in subway space. In addition, in terms of the layout, Taidong Station is a typical cross-shaped transfer station, with complex transfer routes. Moreover, it is the station with the highest passenger flow in Qingdao, receiving varying passengers from commercial centers, residential areas, offices, and leisure facilities, making it highly representative. In addition, Qingdao Metro’s peak hour traffic is around 1000–3000 passengers per hour, and a few hundred to 1000 during the weekdays. In order to avoid the randomness of experimental data and the impact of excessive passenger flow on eye-tracking experimenters on site, the experiment was conducted during off-peak hours on weekdays from 13:30 to 16:30 in December 2023.

3.4. Experiment Design

By conducting the experiment in a subway station, the impact of environment factors on wayfinding and the visual attention distribution of participants in scenes at decision-making nodes are analyzed. Before the formal experiment, a preliminary experiment was conducted, with the following aims: firstly, familiarization with the experimental process, timely identification of any problems in the experimental design, and making necessary modifications to the experiment; secondly, exploring participants’ attention to spatial environment factors as a basis for eye-tracking analysis in formal experimental scenes.

The preliminary eye-tracking experiment was conducted in the subway station on an afternoon in November 2023, with two participants involved and tasked to find their way after entering the station. To simulate the behavior of subway station passengers in real scenarios, the experimental design only specified the starting point and the ending point, and did not specify specific routes. The two participants started from an entrance of the subway station, found the designated target platform following their usual subway riding habits, and then found their way to leave the station, and returned to the starting point of the experiment to end the task. During the preliminary experiment process, troubles of different degrees were encountered: firstly, as the designated starting and ending points were both at the same exit, participants proceeding with the task of exiting the station immediately after having completed the task of entering the station could choose to exit along the route they entered the station because of their familiarity with the station environment, which had a negative impact on our study of spatial environment factors. Secondly, the participants reached different positions when completing the task of entering the station, resulting in a small sample volume for scenes at individual nodes in the later stage, and increasing the difficulty of eye-tracking analysis of such scenes. Therefore, the following modifications were made before the formal experiment.

The formal experiment was conducted in December 2023 during off-peak hours on weekdays. Three participants participated in the experiment every day. By modifying the preliminary experiment, in the formal experiment stage, the starting and ending points were set at different exits, and the route design was changed to entry–transfer–exit, with the intermediate target node set at the middle position of the platform. Figure 3 shows the route design of the experiment. The entry task was to find the middle position of the platform in the specified direction from the starting point (i.e., an entrance), the transfer task was to find the middle position of the platform of another line in the specified direction from the ending point of the entry task, and the exit task was to find another designated exit from the ending point of the transfer task. This route design covered the whole subway station environment. Based on the shortest route, the approximate distances for the three stages of wayfinding were 130 m into the station, 50 m for the transfer, and 115 m out of the station. To avoid interference from experimenters during the experiment, participants were required to independently complete the tasks at all stages, and experimenters were not allowed to follow them.

3.5. Procedure

In this study, on-site survey on the environmental conditions of Taidong Subway Station was conducted first, and then the rationality of the experimental design was tested through a preliminary experiment. After the shortcomings were corrected, the formal experiment was carried out. The procedure of the formal experiment is shown in Figure 4. Before the experiment began, participants signed a letter of consent, and the experiment could be stopped at any time if they felt uncomfortable during the experiment. After that, the participants followed the experimenters to the designated task starting point, where the experimental process and precautions were introduced. Upon fully understanding and agreeing with the experimental content and process, they began filling out the basic information and subjective assessment questionnaires. These questionnaires were designed to investigate, before the experiment, the impact of indoor environment factors on the participants’ wayfinding process using a 7-point Likert scale, based on the participants’ usual experience in riding the subway. Thereafter, the experimenters assisted the participants in wearing the eye tracker and calibrated it for them. When the calibration was successfully completed, the formal experiment began. The experimenters used Tobii Glasses Controller software 2.4.3 to start recording. At the end of the experiment, participants completed a post-experiment subjective assessment questionnaire, which was divided into two parts. The first part assessed whether there were any differences in the influence of indoor environment factors on participants’ wayfinding after the experiment compared to before. The second part investigated the difficulties participants faced in locating their targets, based on specific conditions encountered during the wayfinding task, such as positions where they were interfered and reasons for their confusion. After the questionnaire was completed, the next participant began the experiment. All these data were then analyzed and compared by researchers to obtain the most comprehensive information about the selected route.

3.6. Data Analysis

During the experiment, data affected by equipment issues or signal interference, such as low eye-tracking sampling rates and incomplete records, were excluded. Only data with a sampling rate above 80% were retained. After analysis, there were 32 participants in the experiment, resulting in 31 sets of valid eye-tracking data and 32 valid wayfinding trajectories. Eye-tracking data processing was divided into three parts. First, scenes that frequently required route selection and generated numerous behavioral factors were screened out based on wayfinding trajectories and behavior factors; next, eye-tracking data mapping was conducted for the scene node photos, with the data from all participants superimposed to create a fixation heat map. The color scheme of the heat map reflects the duration of participants’ fixation on specific stimuli, with fixation duration represented sequentially from green to yellow, orange, and red. Red indicates the longest and most frequent gaze in that space, while green represents areas with relatively fewer gazes, but there is no fixed temporal threshold defined between the colors. The final part involved defining areas of interest (AOI) by categorizing the spatial node photos into scenes based on environment factors. Subsequently, eye-tracking metric data for these scenes were exported and imported into SPSS 27.0 for data analysis.

First, a subjective questionnaire was used to analyze participants’ subjective perceptions of how indoor environment factors helped their wayfinding process. Next, the participants’ wayfinding trajectory superimposition map and wayfinding behavior distribution map were created from the eye-tracking video, which included hesitation and turn-back behaviors. This analysis helped identify the subjective reasons for differences in route distribution results, laying the foundation for a further understanding of the wayfinding cognitive process. Following this, the fixation heat map derived from the eye-tracking data was used to analyze the objective environment factors affecting participants’ wayfinding. Finally, a difference analysis was conducted on the eye-tracking metrics for various environment factors at each scene node. This analysis aimed to further understand the visual characteristics of participants observing environment factors across different routes, as well as their interest and cognition related to these factors. First, a normality test was performed on the eye-tracking metrics data, which did not meet the requirements for normality. Next, a difference analysis was conducted on the eye-tracking metrics in each scene. The Kruskal–Wallis test was used to assess the significant differences in eye-tracking metrics and to analyze the influence of various environment factors on the wayfinding process. The significance level was set at 0.05 for all analyses, which were conducted using SPSS 27 software.

Since the primary focus of this study is the visually perceptible spatial environment factors, the environmental information observed by participants in the eye-tracking video was categorized into eight types of physical environment factors: direction signs, glossy finish, spatial decoration, gates, stairs/escalators, elevators, pillars, and wall openings. Additionally, pedestrian factors were also part of the observation category; however, due to equipment limitations, the pedestrians observed by each participant and their locations varied. Moreover, the numerous uncontrollable factors related to pedestrians made data collection challenging, so these factors were not considered in this study. Tobii Pro Lab (version 24.21), the software used to process eye-tracking data, can derive two categories of metrics: “fixation” and “visit”. The fixation metrics focus on the processing and level of interest in a specific factor, while the visit metrics represent the observation and cognitive processes related to that factor. Five eye-tracking metrics were chosen for analysis in the following article: total fixation duration, fixation count, time to first fixation, first fixation duration, and visit count.

Table 2. Definitions of eye-tracking metrics.

	Variable	Abbreviation	Description
Eye-tracking metrics	Total fixation duration	TFD	TFD refers to the total duration of fixations on a specific point within the area of interest (AOI) from the moment it appears until the participant moves out of the AOI. A longer TFD indicates that participants paid more attention to the area and found it more challenging to process the information.
	Fixation count	FC	FC refers to participants’ ability to process the scene, the difficulty of the scene, and their interest in what they observe. Areas with a higher number of fixations are generally those that participants are more interested in.
	Time to first fixation	TFF	TFF refers to the duration it takes for a participant’s eye to perceive the content inside the area of interest (AOI). A shorter TFF indicates that an element is easier to notice. TFF is commonly used to measure visual saliency.
	First fixation duration	FFD	FFD refers to the time taken for a participant to process and form a preliminary cognition of the information in the area of interest (AOI). This metric reflects the participant’s attention to the stimulus target. Its value may be influenced by the attractiveness of the stimulus or the participant’s ability to process and comprehend the information.
	Visit count	VC	VC refers to the process of information processing and cognition of the participant, including memory retrieval, comparison, and judgment. It represents the results of comparing, inspecting, and confirming different areas of interest.

provides definitions for these metrics.

4. Results and Discussions

4.1. Wayfinding Assessment

To explore the influence of various environment factors in the subway spatial environment on participants’ wayfinding processes, eight environmental cues representing the spatial environment were assessed through a subjective questionnaire in the experiment.

Table 3. Summary of subjective questionnaire.

Investigation of Wayfinding Experiment in Subway Station
Date	19 December 2023 to 11 January 2024
Location	Taidong Subway Station
Sample	32
Participant	Personnel participating in the wayfinding experiment
Investigation items	Evaluate the contribution of environment factors to the wayfinding process. Before the experiment (according to the experience in riding subway at ordinary times) and after the experiment (according to this experiment process).

and

Table 4. Subjective evaluation of environment factors.

Environment Factor	Significant Interference	Major Interference	Minor Interference	No Effect	Slightly Helpful	Greatly Helpful	Significantly Helpful
Direction signs	−3	−2	−1	0	1	2	3
Glossy finish	−3	−2	−1	0	1	2	3
Spatial decoration	−3	−2	−1	0	1	2	3
Gates	−3	−2	−1	0	1	2	3
Stairs/Escalators	−3	−2	−1	0	1	2	3
Elevator	−3	−2	−1	0	1	2	3
Pillars	−3	−2	−1	0	1	2	3
Wall openings	−3	−2	−1	0	1	2	3

show the basic contents of the questionnaire. The assessment scale consists of seven levels, with participants filling out the questionnaire before and after the experiment. The average scores of the assessment results for environment factors before and after the experiment are as follows: direction signs (2.72, 2.88), glossy finish (−0.31, 0), spatial decoration (0.09, 0.19), gates (0.47, 0.53), stair/escalators (2.19, 2.31), elevators (1.41, 1.50), pillars (0.78, −0.38), and wall openings (0.81, 0.09). These results show that participants’ evaluations of the role of the influence of these environmental elements were almost always positive, both before and after the experiment, but there were negative evaluations of two environmental elements. In one case, participants rated the bright surface more negatively before the experiment because they thought it would interfere with their pathfinding, while this result improved significantly after the experiment. Participants’ evaluations of the columns were skewed positive before the experiment, but the evaluations became negative after the experiment, probably because participants found that the arrangement of the columns sometimes interfered with their sight of wayfinding as well as their judgment of the direction of wayfinding during the experiment.

As shown in Figure 5, the presence of more colors on the right side of the figure indicates stronger positive guidance from environment factors, while certain colors signify stronger negative interference. Comparing the results from the two stages, it is evident that in participants’ daily psychological expectations and actual wayfinding experiences, the guiding effects of direction signs (100%, 100%) and stairs/escalators (97% and 96.8%) are the most significant. These two elements establish the determination criteria for the alignment between horizontal and vertical streamlines in space, allowing participants to quickly familiarize themselves with the overall spatial environment and form cognitive maps. Participants expected that elevators (81.3%), pillars (59.4%), and wall openings (68.7%) would provide greater positive assistance. However, in actual wayfinding situations, the positive effects of pillars (15.6%) and wall openings (40.7%) were significantly lower than expected, resulting in some interference in the wayfinding process. Participants disagreed on the guiding roles of the two elements in wayfinding. This is likely because pillars can obstruct the line of sight, making navigation more difficult, while wall openings might lead participants off the correct route, resulting in further challenges. In both stages, glossy finish (21.9%, 34.4%) and spatial decoration (34.4%, 28.1%) had the least influence on wayfinding. Participants felt that, aside from glossy finish with signage, other elements would attract their attention to some extent but would not significantly impact the specific wayfinding task. Some participants noted that similar spatial decorations could sometimes complicate cognitive judgments related to wayfinding.

4.2. Movement Path

The logical basis for an individual’s wayfinding decision-making is usually based on their understanding of the external environment. Therefore, path selection may objectively reflect the individual’s decision-making result arising from the wayfinding behavior process and the cognitive process to a certain extent. Figure 6 shows the overall path trajectory of all participants when they complete the task according to the wayfinding process. Each color as shown represents the movement trajectory of different participants, so that you can directly see the path selection of the participants. Moreover, in combination with eye-tracking video record and post-interview, the subjective reasons for the differences in route distribution results may be analyzed in order to lay a sound foundation for further understanding of the cognitive process.

As shown in Figure 6a, participants made their first path choice after they entered the station from Exit F. Most participants chose the nearest gate, while one-third of the participants chose the farther path. We can see from the eye-tracking video record that since this stage involves the subject finding their way after entering the station and the first target point is the platform of Line 1, some participants were directed by the signage of Line 1 to the farther gate after entering the station, thus missing the nearest path. This result is also shown by the higher gaze data of subjects on signage (59.67) in the TD1 scenario of Table 5. It has been stated that if pedestrians see signs during wayfinding, they usually choose to move according to the information of the signs [38]. Upon the arrival at the platform of Line 1, the path selection results to the Taishan Road platform of Line 2 are different to a certain extent. Some participants did not walk to the middle position of the platform as specified, but chose to return to the station hall by the escalator from which they came and find the way to Taishan Road. A study shows that pedestrians choose familiar routes in an emergency and still do so under normal circumstances [1,23].

In addition, some participants walked to the middle position of Line 1 platform and directly chose to take the nearest escalator to the Licun Park platform. It is suspected that participants are misguided by the signage of Line 2 in front of the escalator, while the platform of Line 2 in Taidong Station is a side platform, and the platforms of the two lines are separated, resulting in movement to the wrong platform. Part of the reason may be that participants were misguided by cognitive maps in their brains. They first saw the opposite layout of Line 1 platform, intuitively believed that the platform layouts of Line 2 and Line 1 were the same, and thus made the wrong choice. The same is true in Figure 6c. Some participants’ route trajectories at the Licun Park platform were relatively chaotic due to incorrect wayfinding. However, five participants chose the escalator leading directly to the Taishan Road platform. According to video recordings and post-interviews, they said they saw the signage leading to Taishan Road or were familiar with the platform, and knew the shortest path themselves. This indicates that environmental familiarity will affect the participants’ route selection results. Participants who are familiar with the environment may move directly to the escalator or elevator which they are familiar with, but may not necessarily be the one nearest to them, to find their way.

As shown in Figure 6d, the route selection of Taishan Road platform is similar to that of Line 1 platform. Figure 6e shows that after participants walked to the station hall from Line 1 platform or Licun Park platform of Line 2, they found the transport facilities to Taishan Road again, and thus completed the transfer of the second stage. Whether they move to the wrong platform or return directly to the station hall from Line 1, these two lines increase the wayfinding time of participants. If passengers choose these two routes in reality, it may cause a travel delay or other problems. Figure 6f shows the route of the last stage of this wayfinding process, i.e., wayfinding to exit the station. As shown in the figure, participants have different paths after arriving at the station hall from traffic facilities at different positions. From the route map and wayfinding video, most participants successfully found Exit A to complete the wayfinding. Only five participants chose to move to the station hall by the northernmost escalator and failed to find the indicative signage of Exit A in a timely manner because their line of sight was blocked, resulting in detour and turn-back behaviors. This allows us to consider whether passengers from all directions should be taken into account when the signage is arranged.

In the on-site wayfinding experiment, the wayfinding behaviors of participants involve various scene changes and shifts of line of sight. In combination with the participants’ behavioral data, spatial nodes with high frequencies of behavioral elements are identified. This allows the further analysis of the relationship between the physical behavior and route selection of participants, thereby facilitating the analysis of factors influencing the wayfinding process. As shown in Figure 7, behaviors such as stopping, turning around and looking around during the wayfinding process are defined as hesitant behaviors, and altering the route or repeating the route is defined as turn-back behaviors. Based on the eye-tracking video record of participants, the hesitant and turn-back behaviors and interested areas are identified and marked frame by frame in order to draw the wayfinding behavior distribution map.

Figure 7 shows that there are many hesitant behaviors in the behavior statistical diagram of TD1, which is consistent with the above results. Participants saw the direction sign of Line 1, nearest gate and security checkpoint after entering the station, so they hesitated when making a decision. In the process in which participants are finding the platform of Line 1, they had less hesitant behaviors, but some participants had temporary hesitant behaviors in TD2 (i.e., in the scene of Line 1 platform). The latter may be caused by the path selection for each traffic facility here in combination with the route trajectory diagram in Figure 6. As shown in Figure 7, TD3 and TD4 are two scenes with the largest number of hesitant and turn-back behaviors, that is, the space scene of station hall at Licun Park platform and at the time of departure. In TD3, because the participants chose the wrong escalator, they did not find the target point Taishan Road platform after arriving at the platform, and the visual search habit and urgency of finding the target point led to more hesitant behaviors (e.g., turning around and looking around) and turn-back behaviors (e.g., going back and forth to find the target). In TD4, the participants looked for Exit A from different routes. Many spatial elements in the station hall attracted more visual attention of the participants, meaning that they had obvious hesitant behaviors. In addition, the eye-tracking video record shows that the spatial structure of the station hall results in cutting effect on the line of sight of participants, leading to alternative visual effect and behavioral hesitation. Figure 7 shows that, in order to ensure the accuracy of the direction in the wayfinding process, participants frequently had short-term hesitant behaviors throughout the wayfinding decision-making process. Through the scene with many wayfinding problems in this figure and the scene with changed route selection in the route trajectory map, the four decision-making point scene diagrams in Figure 7 are obtained. The visual information perception process will be analyzed according to the eye-tracking data in the four scenes to further explore the environment impact factors in the wayfinding process of participants.

4.3. Attention Distribution Under Different Behaviors

Table 5 shows the fixation heat map and histogram of four scene nodes in Taidong Subway Station. It intends to analyze the cumulative distribution of participants’ attention in the four scenes, and explore which environment factors the participants pay more attention to and which environment factors receive more dispersed attention in the scenes. In the heat map, red represents the areas of concentrated attention, i.e., the areas with the longest fixation time; green represents the areas with dispersed attention; and the areas that do not attract participants’ attention are transparent. In this case, the attention duration of the most focused region is the region with greater than 60 s of attention, and the attention duration of the more dispersed region is the region with greater than 5 s and less than 50 s of attention.

Table 5. Fixation characteristics of environment factors in each scene.

Scene Node	Fixation Heat Map	Total Fixation Duration of Environment Factors
TD1
TD2
TD3
TD4

Table 5. Fixation characteristics of environment factors in each scene.

Scene Node	Fixation Heat Map	Total Fixation Duration of Environment Factors
TD1
TD2
TD3
TD4

Table 5 shows that, in the TD1 scene, participants focused most on the signboard (59.67) and the gate security checkpoint (63.92), because they needed them to access the station and find the target point. Attention to wall openings (22.63), pillars (35.94), and glossy finish (11.11) is relatively dispersed, and little attention is paid to the elevator (5.02) and the escalator (13.03). The histogram shows that the longest total fixation duration (97.84) and fixation count (382) are given to the decorative elements, but due to their large coverage, participants, when they just enter the underground space, will need some time to cognize this space and therefore will look around, resulting in the longest fixation duration to the decorative elements. In the TD2 scene, signage (85.94) and stairs/escalators (37.04) attracted the most attention from participants. The attention for elevators (32.56), glossy finish (44.05), and stairs/escalators (37.04) was relatively even. At this stage, the signage and stairs/escalators helped participants locate the next target point, which is why they received more attention. In contrast, wall openings (9.20) were not very helpful for wayfinding during this stage, so they receive the least attention.

In the TD3 scene, the focus was primarily on the signage (136.44) and the transportation facilities like elevators (26.28) and stairs/escalators (28.60), and attention towards other elements was dispersed. This may be due to participants experiencing wayfinding errors during this stage, leading them to urgently seek the correct target point. As a result, they tended to repeatedly confirm the signage and transportation facilities to find the correct route, generating increased attention towards these elements. In the TD4 scene, besides the highest attention on the signage (155.69), participants also paid considerable attention to pillars (60.48) and glossy finish (42.49). This increased focus may be attributed to the fact that this stage involved wayfinding out of the station, where participants encountered visual obstructions caused by pillars while searching for the exit. Conversely, the attention towards elevators (3.87) and stairs/escalators (16.49) was minimal, as these two elements were no longer relevant during the exit process, failing to capture the participants’ attention. Through the analysis of the heat maps for these four scenes, it is evident that participants’ primary focus on environment factors varies across different stages; however, signage consistently holds an important position. Studies indicate that when pedestrians’ preferences contradict the content of the signage, they tend to follow the signage [12]. Additionally, as mentioned in the analysis above, the visual memory possessed by pedestrians can also influence their wayfinding behavior [39].

4.4. Visual Analysis by AOI

Tobii Pro Lab software (version 24.21)was used to divide and encode AOI environment factors of four scene nodes, including eight types of physical environment factors: direction signs (H1), elevators (H2), glossy finish (H3), stairs/escalators (H4), wall openings (H5), gates (H6), pillars (H7) and decorations (H8). Five eye-tracking metrics (TFD, FC, TFF, FFD, VC) for each area of interest were exported to perform a normality test on the five eye-tracking metrics with SPSS27.0. As shown in

Table 6. Normality test for experimental data of scenes.

Node	Degree of Freedom df	TFD		FC		TFF		FFD		VC
		KS Statistics	Sig.	KS Statistics	Sig.	KS Statistics	Sig.	KS Statistics	Sig.	KS Statistics	Sig.
TD1	169	0.771	0.000	0.793	0.000	0.817	0.000	0.663	0.000	0.856	0.000
TD2	182	0.815	0.000	0.819	0.000	0.813	0.000	0.631	0.000	0.866	0.000
TD3	145	0.752	0.000	0.886	0.000	0.894	0.000	0.740	0.000	0.850	0.000
TD4	92	0.699	0.000	0.749	0.000	0.929	0.000	0.605	0.000	0.839	0.000

Note: p value (sig.) > 0.05 indicates that the data fits a normal distribution.

, none of the eye-tracking metrics met the assumption of normal distribution; therefore, the Kruskal–Wallis H test was used to analyze the data as a whole.

Table 7. Kruskal–Wallis test for eye-tracking metrics of various environment factors in different scenes.

Node	N	TFD		FC		TFF		FFD		VC
		X2	Sig.	X2	Sig.	X2	Sig.	X2	Sig.	X2	Sig.
TD1	31	95.04	0.000 **	105.67	0.000 **	51.73	0.000 **	12.41	0.088	103.30	0.000 **
TD2	30	45.92	0.000 **	48.35	0.000 **	58.07	0.000 **	8.97	0.356	66.13	0.000 **
TD3	24	46.73	0.000 **	43.60	0.000 **	60.82	0.000 **	18.65	0.176	42.25	0.000 **
TD4	14	139.13	0.000 **	141.53	0.000 **	86.80	0.000 **	7.75	0.005 **	144.46	0.000 **

Note: ** represents significance level less than 0.01.

shows the results of Kruskal–Wallis H test across the four scenes. From Table 7, it can be observed that the differences in eye-tracking metrics for participants across scenes TD1-TD4 are similar. Notably, the eye-tracking metrics TFD, FC, TFF, and VC showed significant overall differences (p = 0.000), while the difference in FFD was not significant. TFD represents the participants’ level of interest in environmental information and the difficulty of information processing, and the differences in visual interest indicators among various environment factors and their influencing factors were analyzed by multiple comparisons of the TFD metrics across these four scenes, as shown in

Table 8. Multiple comparative differences of TED for various environment factors in different scenes.

Node	Pairwise Comparison	Sig.	Pairwise Comparison	Sig.	Pairwise Comparison	Sig.	Pairwise Comparison	Sig.	Difference Between Groups
TD1	H2–H7 H2–H1 H2–H6 H2–H8	0.000 ** 0.000 ** 0.000 ** 0.000 **	H3–H7 H3–H1 H3–H6 H3–H8	0.015 * 0.001 ** 0.000 ** 0.000 **	H4–H1 H4–H6 H4–H8	0.002 ** 0.000 ** 0.000 **	H5–H6 H5–H8	0.004 ** 0.000 **	H7 > H2, H3 H1 > H2, H3, H4 H6 > H2, H3, H4,H5 H8 > H2, H3, H4, H5
TD2	H5–H4 H5–H8	0.031 * 0.021 *	H5–H3 H5–H7	0.000 ** 0.000 **	H5–H1	0.000 **	H2H1	0.038 *	H1 > H2, H5 H3 > H5 H4 > H5 H7 > H5 H8 > H5
TD3	H7–H8 H7–H1	0.005 ** 0.000 **	H3–H8 H3–H1	0.043 * 0.000 **	H2–H1	0.000 **	H5–H1 H4–H1	0.000 ** 0.007 **	H1 > H2, H3, H4, H5, H7 H8 > H3, H7
TD4	H2H3 H2–H8 H2–H7 H2–H1	0.000 ** 0.000 ** 0.000 ** 0.000 **	H6–H3 H6–H8 H6–H7 H6–H1	0.013 * 0.000 ** 0.000 ** 0.000 **	H4–H8 H4–H7 H4–H1 H5–H8 H5–H7	0.002 ** 0.001 ** 0.000 ** 0.012 * 0.009 **	H5–H1 H3–H1 H8–H1 H7–H1	0.000 ** 0.000 ** 0.030 * 0.040 *	H1 > H2, H3, H4, H5, H6, H7, H8 H3 > H2, H6, H7 > H2, H4, H5, H6 H8 > H2, H4, H5, H6

Note: * represents significance level less than 0.05, ** represents significance level less than 0.01.

.

In the TD1 scene, visual interest differences were observed in H2 (p = 0.000), H3 (p = 0.015, p = 0.001, p = 0.000), H4 (p = 0.002, p = 0.000), and H5 (p = 0.004, p = 0.000). There were no significant differences among H1, H6, and H8. As can be seen from the results of the multiple comparisons of the environmental elements in Table 8, in the TD1 scene, the gaze times for the elevator (H2), the bright surface (H3), stairs/escalators (H4), and the wall opening (H5) were significantly lower than those for the other environmental elements. Participants, having just entered this scene, focused more on the signage (H1) that provided guiding cues, as their prominent colors and positions made them easier to notice, leading to higher attention levels. The high attention to pillars (H7) is attributed to their obstruction of participants’ lines of sight, which consequently resulted in lower attention to the elevator (H2) and stair/escalator (H4) in that area. In the TD2 scene, visual interest differences were found in H2 (p = 0.038) and H5 (p = 0.031, p = 0.021, p = 0.000). As can be seen from the results of the multiple comparisons of the environmental elements in Table 8, the visual attention of the elevator (H2) and the wall opening (H5) is significantly lower than that of the other environmental elements, while the signage continues to have the highest level of attention. Participants focused more on the signage (H1) that led to the next target point in this scene. Additionally, although the elevator (H2) was closer to participants compared to the stair/escalator (H4), the stair/escalator still garnered higher attention. This indicates that even though the stair/escalator was farther away from pedestrians, most people still considered it their first choice. This might be due to the longer waiting time for the elevator, which could increase travel time.

In the TD3 scene, the differences in visual interest are reflected in H1 (p = 0.000, p = 0.007) and H8 (p = 0.005, p = 0.043). From the multiple comparisons in Table 8, it can be seen that the gaze time for signage (H1) and decoration (H8) is significantly higher than the other environmental elements, and there is no significant difference between the other environmental elements. Most participants mistakenly entered TD3 due to misreading the signage at TD2. The unclear signage in this area led to brief feelings of anxiety and stress among the participants. Additionally, the route is relatively complex. Due to the presence of stress, pedestrians tend not to linger in one spot for too long; instead, they quickly gather scene information to satisfy their visual needs [40]. Furthermore, under stressful conditions, individuals process information more rapidly and are less likely to carefully connect various pieces of information to form a complete information chain [41]. Studies indicate that stress can improve wayfinding performance in less complex environments [42]. Participants tend to focus more on observing signage and spatial information to find the correct route, so the attention time of the signs (H1) and decorations (H8) will be higher, which in turn also leads to a decrease in the level of interest in H3, H5, and H6. The TD4 scene is similar to the TD1 scene, where participants primarily focus on the signage to locate the exit. However, the pillars in the station hall obstruct some participants’ views. Additionally, some pillars feature signage, which can attract participants’ attention. Since this stage is focused on finding the exit, the interest in elevators and stairs/escalators also decreases, which aligns with the results discussed in Section 4.3.

As shown in Figure 8, this study takes the TD1 scene as an example to analyze different eye-tracking indicators in that scene. TFD and FC indicate that participants need to repeatedly understand and process the labeling information to find the correct route. Additionally, upon initial entry into this scene, the spatial decoration also has a significant visual appeal to the participants. TFF shows that participants noticed the signage in the shortest time, suggesting that signage is the most easily detectable environmental elements in this scene. In contrast, participants took the longest to notice the elevators and stairs/escalators, indicating that these two environment factors are the least detectable in this scene, reflecting their low visual salience. For subway stations, stairs/escalators and their signage need to enhance readability through appropriate sizes and colors [43], thereby improving the wayfinding efficiency of pedestrians. VC represents the process of pedestrians’ information processing and cognition, and it is used to confirm results by comparing different AOIs. Participants validate their chosen routes by repeatedly checking the signage, while confirming decorations and other environment factors helps them quickly form cognitive maps to aid in wayfinding.

4.5. Nodal Scenario Optimization Recommendations

As shown in Figure 9, the following design recommendations are made based on the results of the study:

(1)

At the entrance of the subway station hall, it is necessary to avoid providing guidance with multi-directional signs for the same target platform. As shown in the TD1 hotspot map in Table 5 above, the presence of two directions directed towards the platform of Line 1 at the node of this scenario resulted in some participants choosing a longer path to the platform, delaying the efficiency of the passage. Therefore, for the guidance of the target point at the same spatial node, it should mainly indicate the fastest passage route. Even if both routes can lead to the destination, the signboard for the farther route should be placed later to prevent causing route interference to passengers. In addition, in order to cater to people of all age groups, signs in multiple forms should be adopted. For example, dynamic signs and signs with sound prompts should be added, which makes it easier for the elderly or children to obtain information.

(2)

It is necessary to increase the salience of the stairs/escalators and gate entrance of the station hall. In the TD1 scenario above, some participants increased their passage time precisely because of the inconspicuous location of the stairs/escalators and gate entrance, and the obscuring of the gate entrance by an advertising standing sign, which was not seen by the participants at first glance. This can be realized through the prominent design of the top surface shape or the surrounding interface in which stairs/escalators are located, so that the color of the top shape of each escalator or staircase is consistent, deepening the attention of passengers and their impression in the space, as shown in Figure 10. By contrasting the color with the surrounding environment, some interesting decorations can be added to the design to attract younger passengers for wayfinding. At the same time, it reduces the blocking of movable objects such as advertising billboards to the gate machine, and increases the “exposure” of the gate machine in the space.

In the platform space, the visibility of escalators and stairs should also be increased. In the design of subway stations, stairs/escalators are often wrapped with walls, so that passengers cannot find the stairs/escalators they want to go to in time. For example, in Table 5 above, the stairs/escalators in the TD2 platform scene are surrounded by a wall, resulting in passengers not noticing the target stairs/escalators in the first place, thus increasing gaze time. The walls should be removed or the stairs/escalators should be designed differently using space decoration such as lighting design and wall modeling to enhance the visibility of passengers, as shown in Figure 11.

(3)

Taidong Station, given its cross-shaped spatial layout, is more complicated for passengers who are unfamiliar with metro stations. In the design of the station, the spatial layout of the station should be optimized, making the layout of the upper and lower floors as consistent as possible, reducing the number of complex passages and corners, facilitating the rapid establishment of cognitive maps for passengers, reducing the possibility of passengers getting lost, and making it easier for elderly people and children with weak spatial cognitive ability to find their way around. In addition, one can set up a sufficient rest area in the space, which are also convenient for the elderly and children to rest in the process of finding their way; the location of the rest area should also be obvious and easy to reach.

(4)

Improved signage and enhanced spatial guidance. In Figure 7 above, it can be seen that participants showed more hesitation and folding behavior at the TD3 scenario, precisely because there was no continuous guidance to the signage of the Taishan Road platform at that scenario, which resulted in participants not being able to quickly find the target building escalator for wayfinding. The hotspot map in Table 5 also illustrates the problem. Therefore, in the design of metro stations, the legibility and coherence of signage should be enhanced to ensure that signs can be quickly recognized by passengers of all ages. For example, brightly colored signs with cartoon images should be designed for children to attract their attention, and larger, high-contrast signs should be provided for the elderly to make it easier for them to view, thus improving the efficiency of passengers’ wayfinding. In addition, the regional division of station space should be optimized, including the design of spatial guidance on the top surface or other interfaces, the station hall space can be combined with the flow of passengers into the station and out of the station for the guidance and division of interfaces, including by highlighting the main axis of the way, the main axis of the area using different shapes and colors and other spatial areas separate, forming a dynamic spatial guidance, to give passengers spatial psychological guidance hints.

5. Conclusions

This study utilized eye-tracking technology to explore the visual behavior characteristics of individuals regarding spatial environment factors at three different wayfinding phases of entering, transferring, and exiting a subway station, as well as the differences in how these factors influence pedestrians’ wayfinding behavior. From the perspective of visual behavioral characteristics, we collected and analyzed five eye-tracking metrics from all participants, along with path selection, fixation heat maps, and participants’ subjective evaluations. The results indicate that signage is a significant influencing factor in the wayfinding process; for example, the total gaze time of the logo was consistently very high across the four scene nodes. Among them, the total gaze time of the logo in the TD4 scene reached 155.69 s. In the subsequent analysis of significant differences between environmental elements, the difference between the logo and other environmental elements was also very significant (p = 0.000). However, it may mislead passengers unfamiliar with the spatial environment, causing them to take incorrect routes. This finding highlights the degree of influence of spatial layout on participants’ wayfinding. It was also found that different environmental elements had different degrees of influence at different stages of wayfinding: for example, at the station hall scene at the station entry stage, the total attention time for signs and decorations was higher; at the platform scene, vertical transportation facilities such as elevators (32.56) and stairs/escalators (37.04) had higher attention, e.g., in Table 5, it can be seen that, in the TD2 scenario, the participant was at the stage of going to the elevator or the stairs/escalators to transfer to another one, and compared with other scenarios, there is a significant increase in the total attention time for the elevator and stairs/escalators in this scene. The gaze hotspot map also shows that the color at the stairs/escalators gradually becomes red, and participants had a more concentrated gaze on the stairs/escalators at the TD2 scene. This suggests that they are also very important for participants’ wayfinding during the transfer phase. In addition, in the four spatial scenes, environmental elements with higher visual saliency are more easily noticed by the participants, and the visual saliency of some environmental elements should be enhanced accordingly to the visual attention of passengers to different elements in different stages.

In this study, environment factors such as structural pillars that obstruct the line of sight also influenced participants’ wayfinding. In addition, as mentioned in the above text, participants’ attention to signs (136.44) in the TD3 scenario was much higher than the other environmental elements, precisely because the participants did not find the target station here, and thus a certain amount of stress may have been generated. This may indicate that under a certain amount of stress, pedestrians are likely to block out some of the complex information by focusing their attention more. Additionally, participants showed a high level of attention to spatial decorations in various scenes. In the TD1 scene, the attention time to the space decoration reached 97.84 s, which may be due to the fact that the participants needed to look around to quickly recognize the space when they first entered the subway station. Therefore, when designing future subway stations, we can consider using spatial decorations to create clearer spatial guidance, for example, in Taitung Station, which has a large space and a complex layout, continuous decorative lines with directional indications or salient colors can be set on the top surface or the ground to guide the passengers to the target platforms or exits, which will help to increase the efficiency of pedestrians’ wayfinding in the subway station space. Based on the above suggestions for the design optimization of current or future subway stations, they can be directly applied to the design and transformation of subway station space, which can help to improve the legibility and orientation of subway station space, reduce the troubles of passengers in searching for directions, improve the efficiency of searching for directions, and then improve the travel experience of passengers.

Finally, although this study explored the effects of environmental elements on wayfinding at different wayfinding stages, the potential effects of factors such as participants’ prior knowledge and education on wayfinding behavior were not fully considered and analyzed. In addition, the study mainly used on-site eye-tracking experiments, but the method has some limitations and cannot analyze dynamic pedestrian flow. Although a representative subway station was selected for the field experiment, the experimental setting did not cover all types of subway station spatial features. In subsequent studies, the effects of multiple factors on wayfinding behavior can be further explored to clarify the complex interactions between these factors and environmental elements, including controlled experiments with people of different educational backgrounds and familiarity levels. A wider range of subway station types, such as subway stations within business districts and subway stations with relatively large differences in passenger flow, are chosen to improve the generalizability of the findings. Additionally, stimuli such as pedestrians’ facial features and gaze direction can affect eye-beating trajectories and have an impact on the processing of spatial cues [44]. Therefore, future research on wayfinding in subway stations could incorporate dynamic pedestrian flows, including the effects of the presence and movement of other pedestrians on wayfinding behavior during the process.