Investigation of Cyclist’s Distraction Due to External Auditory and Visual Stimuli

Abstract

The impact of cycling-related traffic crashes on public health has increased significantly in recent decades, with cyclists being among the most vulnerable road users. The risk of severe injury in traffic crashes is notably high for cyclists, especially when distracted. Research indicates that distraction while cycling significantly increases the crash risk. This study investigates cycling distraction through a field operational test involving 100 participants. Riders followed a predetermined course while being exposed to external visual and auditory stimuli, including alarms, advertising signs, and car horns. Distraction levels were measured using eye-tracking technology. Data were analyzed by means of descriptive statistics, cluster, and correlation analyses. Our findings showed that auditory stimuli distract a higher percentage of cyclists, while audiovisual stimuli from road-related factors cause longer-lasting distractions. Additionally, five distraction clusters were identified based on stimulus duration. Lastly, it was found that males were more likely to belong to high-distraction clusters, whereas females and daily cyclists were more likely to fall into the lowest-distraction group.

1. Introduction

Road crashes are one of the leading causes of death worldwide, with approximately 1.35 million fatalities reported annually, according to the World Health Organization (WHO). Vulnerable road users, such as pedestrians, motorcyclists, and cyclists, account for more than half of these deaths [1]. WHO defines a road traffic crash as a collision involving at least one moving vehicle on a public or private road, resulting in injury or death [2]. These crashes can lead to property damage, injuries, or fatalities, and are typically classified as “fatal or non-fatal injuries resulting from a collision involving at least one moving vehicle” [3].

Crashes often result from a combination of circumstances, where the likelihood of a crash increases from low to high risk. For instance, two vehicles arriving at the same location at the same time could potentially lead to a collision. However, crashes do not always occur even in such situations, as factors like drivers’ response times, braking effectiveness, attention, and speed play a crucial role. Other situations, such as pedestrians crossing the road, objects on the road, or unexpected shifts in traffic flow, also increase the chances of a crash, albeit to a lesser extent. In most cases, drivers can take preventive measures such as slowing down, changing lanes, or honking the horn to reduce accident risk.

Hence, various factors can interact and result in a road crash [4]. According to this model (the Swiss cheese model), crashes are rarely the result of a single cause. Instead, they often arise from a complex interplay of multiple factors [5]. Understanding these interactions is crucial for effective accident prevention and safety management. Moreover, crash prevention can be achieved through the enforcement of traffic rules, increased driver awareness, and continuous improvements in road and vehicle safety.

However, the likelihood of a collision rises significantly if the driver is distracted, which may result in a road crash if necessary actions, such as applying the brakes, are not taken in time. Key factors affecting the probability of a crash include the driver’s attention and behavior, the condition of the vehicle, the road network, and other contributing elements. Crash prevention can be achieved through the enforcement of traffic rules, increased driver awareness, and continuous improvements in road and vehicle safety.

Crashes generally result from a mix of factors, categorized into three main groups [6]:

• Road and environmental factors: These include the geometric features of the road network, construction quality, maintenance, weather conditions, and visibility.
• Vehicle factors: These involve the technical aspects of the vehicle, including safety equipment, design, and maintenance.
• Human/road user factors: These encompass the physical, mental, and social characteristics of the driver, including behavior patterns, age, experience, fatigue, and sobriety.

It is important to note that the human factor is often the most significant cause of traffic crashes [6]. Therefore, preventive measures should focus on reducing human-related errors. One of the most dangerous threats to road safety is driver distraction, which shifts attention from critical driving tasks to secondary activities [7]. While studies on distractions have mostly focused on motor vehicles, the analysis of the impact of distractions on all road users, including cyclists, requires further exploration. Urban environments, with their multiple stimuli, exacerbate distractions, making them a hidden risk factor that can significantly increase the likelihood of crashes. Given cyclists’ heightened vulnerability, distractions are even more likely to result in serious injuries or fatalities [8].

While studies on distractions have mostly focused on motor vehicles, the impact of distractions on all road users, including cyclists, requires further exploration, as this topic is relatively less explored compared to studies on motor vehicle driver distraction. To address this gap, the present study aims to analyze cyclists’ behavior in response to various distractions using precise data collection methods such as video recording and frame-by-frame analysis. Hence, the goal of the present study is to provide insights into cycling distraction. In order to achieve the aim of the study, descriptive statistical analysis, as well as k-means cluster and correlation analysis, are applied. The practical implications of the paper include the need to enhance road infrastructure, gain a deeper understanding of distracted cycling behavior, and develop safer road environments to reduce crashes and fatalities.

2. Literature Review

2.1. Definition of the Term Distraction in Road Safety

Studies on distracted driving have used various criteria to explore its impacts. One key issue in such studies is the ambiguity of the term “driver distraction,” as it lacks a precise, universally accepted definition. The word “distraction” itself is defined broadly, often as “the distraction of the mind or attention from a specific object,” which fails to capture its full scope in scientific contexts. As [9] highlight, many studies fail to establish a clear structure for the term, which leads to a lack of consistency in its application. This is problematic, as the definition of distraction is critical for accurately assessing its effects on road safety.

Various definitions of driver distraction have emerged, emphasizing different aspects of the phenomenon ([9,10,11,12]). Some definitions focus on the shift of attention from critical driving stimuli to other, non-driving stimuli [13]. Others highlight the involvement of objects or events, both inside and outside the vehicle, that divert the driver’s attention [14]. Distraction is also seen as a misallocation of attention [15] or an adverse effect on a driver’s ability to process relevant information for safe driving [16]. Regan et al. [17] conducted a review of current definitions and classifications of driver distraction and inattention to better interpret and compare research outcomes related to specific types of driver inattention and defined distraction as a temporary diversion of attention from driving to an object, person, task, or event unrelated to driving, which reduces the driver’s awareness and performance, increasing the risk of corrective actions, near-crashes, or crashes.

In summary, the definitions of driver distraction vary in their focus, with some emphasizing its impact on driving performance, others addressing the causes of distraction, and most describing it as a disruptive force interfering with safe driving. In the context of this study, distraction was defined as the time during which the cyclist’s gaze was directed away from the road ahead, specifically toward the presented visual or auditory stimulus. Distraction was measured using gaze redirection data captured through eye-tracking technology. A participant was considered distracted when their gaze shifted away from the forward roadway and toward the location of the stimulus (visual or auditory). More specifically, in our study, two areas of interest (AOIs) were defined based on their functional relevance to the riding task and the position of experimental stimuli.

• Primary AOI: The road ahead (including dynamic road features such as upcoming turns, other road users, and surface conditions) was defined as the task-relevant AOI, reflecting the central visual field required for safe navigation.
• Peripheral AOIs: Elements such as sidewalks, bushes, or off-road areas were considered non-task-relevant AOIs, included for the purpose of identifying gaze deviations from the road.

Finally, it is noted that the primary gaze-related metric used was the duration of this deviation (in seconds), recorded for each stimulus event.

2.2. Motor Vehicle Driver’s Distraction: Types and Effects

Relevant literature on driver distraction as well as mobile phones and advertising signs is extensive (see, for example, [7,18,19,20,21,22,23]). The present literature review provides an overview of distraction-related factors in road safety. For a more exhaustive list of relevant studies and findings, the readers are encouraged to refer to [24,25].

Driver distraction can be categorized into various types, such as visual, cognitive, and auditory. Regan et al. [17] provided a comprehensive taxonomy and analysis of distraction types. On the other hand, Stutts et al. [26] identified and categorized distraction sources from naturalistic driving data. A major distraction factor is mobile phone use (both handheld and hands-free), which can significantly impair reaction times and situational awareness of drivers, as several studies highlight (e.g., [27]). Similarly, texting was found to significantly increase crash risk [28].

In-vehicle devices can cause cognitive distraction [29]. For example, Dingus et al. [28] conducted a large-scale study that observed over 3500 drivers, aged 16 to 98, over a period of 3 years, using on-board cameras and sensors to capture data on real-world driving conditions. The study aimed to evaluate the factors contributing to crashes, and it found that human-related factors such as errors, fatigue, and distraction accounted for almost 90% of crashes. Distraction, in particular, was responsible for more than half of all trips, with drivers engaging in distracting activities in 51.93% of journeys. Moreover, prolonged gazing at external stimuli significantly increased the risk of crashes, with the odds ratio of 7.1. The study revealed that distraction contributed to 68.3% of injury and property damage crashes.

The presence of passengers and passenger interaction is a common and significant source of in-vehicle distraction. It increases drivers’ mental workload and cognitive demands, which in turn diminishes their reflexes and slows both the recognition of events and the physical response to them, as noted in existing research ([22,26,30]).

Visual stimuli in the urban environment (e.g., advertising signs) were studied in literature ([21,31]). Another study by Lemonakis et al. [32] analyzed the visual behavior of motorcyclists using eye-tracking glasses to observe their gaze during real-world conditions in Western Greece. The study found that visual stimuli in the urban environment, such as billboards and road signs, caused significant distraction to motorcyclists, with distractions lasting over 2 s on average, which is dangerous for road safety. The study showed that distraction occurred across all three road environments (urban, suburban, and semi-urban), negatively impacting the time spent focused on driving and endangering both the motorcyclist and other road users. While much of the research has focused on driver distraction in motor vehicles, there is relatively limited research on cycling-related distractions. Wolfe et al. [33] conducted an observational study in Boston, collecting data from 1974 cyclists at four high-traffic intersections. They found that 31.2% of cyclists were distracted, with auditory distractions (17.7%) being more common than visual distractions (13.4%). Distractions were most frequent during evening commutes, emphasizing the need for further studies on cycling distractions. Terzano [34] analyzed the behavior of 1360 cyclists in The Hague and found that those using phones, listening to music, or talking exhibited more unsafe behaviors. Distracted cyclists also cause more evasive actions from others, like distracted driving; multitasking while cycling poses risks to both the rider and surrounding road users.

A cross-sectional study by Useche et al. [35] analyzed data from 1064 cyclists from 20 countries. They found that distractions were prevalent among cyclists, with incidents ranging from 34.7% to 83.6%. The study confirmed that distractions contributed to risky behaviors and traffic crashes, further linking them to crash rates. These findings indicate that distractions play a significant role in cycling crashes, affecting both the riders’ attention and overall safety. Møller et al. [36] conducted a study focusing on tactile distractions, which argued that secondary tasks, such as using a smartphone, do not necessarily worsen cyclists’ behavior. The study used a cycling simulator and found that cycling performance remained unaffected when secondary tasks requiring cognitive attention but not visual attention were performed.

In summary, research on cyclist distraction due to visual and auditory stimuli remains limited compared to studies on motor vehicle driver distraction. The behavior of non-motorized urban road users, such as cyclists, during their exposure to external stimuli remains underexplored. To address this gap, the present study aims to analyze cyclists’ behavior in response to various distractions using precise data collection methods such as video recording and frame-by-frame analysis. The goal is to improve road infrastructure, better understand distracted cycling behavior, and create safer road environments to reduce crashes and fatalities.

3. Methods and Data

To investigate cyclist distraction caused by exogenous auditory and visual stimuli, an experiment was conducted. Volunteer cyclists were invited to participate in the experiment, which took place in the city of Volos. More details about the recruitment process are presented in the following sections. Volos is a coastal city located approximately 330 km north of Athens, the capital of Greece, with a population of approximately 140,000 inhabitants. The city is characterized by a mild Mediterranean climate and a predominantly gentle topography with limited longitudinal gradients, which create favorable conditions for cycling. Furthermore, Volos has a well-developed cycling infrastructure and a long-standing tradition of cycling, enhancing the city’s suitability for bicycle use and promoting cycling as a viable mode of transport.

3.1. Route and Stimuli

3.1.1. Route Planning

A prerequisite for the execution of the experiment was the selection and definition of a specific route. The location was chosen based on the safety and comfort of the riders due to reduced traffic, as well as the ease of placing and activating the stimuli. The route is located in the parking area of the Panthessaliko Stadium, specifically on Sophia Spanoudi Street. There were no other users or obstacles that could affect cyclists’ behavior. The total length of the route is 500 m, and three exogenous stimuli—visual and auditory—were placed along it (Figure 1). The route features slight longitudinal slopes, a total width of 5.50 m, and an absence of vertical and horizontal signage.

3.1.2. Participants

The total number of cyclists participating in our experiment was 100. A prerequisite for participation was the ability to ride a bicycle and also to give their consent to participate after they were informed about the experiment. Participants were selected through their social networks, including friends, family, and work environment. To avoid influencing their cycling behavior, the cyclists were not given detailed explanations about the exact subject of the survey or the nature of the stimuli. Each participant who completed the experimental route was asked to fill out a questionnaire. The questionnaire included demographic information such as gender and age group, as well as questions about cycling experience, involvement in cycling crashes, the conditions under which the experiment was conducted, and the distraction phenomenon.

3.1.3. Visual and Auditory Stimuli

Other studies have used a variety of visual and auditory stimuli, such as traffic lights, printed traffic signs, the use of mobile phones, listening to music/podcasts, the use of a navigator, etc., to investigate the change in the driving behavior of cyclists ([37,38,39,40]). These stimuli were either internal or external to the cyclist/bicycle system. In our own research, we used exclusively external visual and auditory stimuli intended to represent frequent situations that a cyclist encounters in real traffic conditions.

The first stimulus, which was visual, was located 105 m from the start of the route and involved a parked vehicle that activated its emergency lights (alarm) and reversing lights as the cyclist approached (Figure 2). Emergency and reversing lights were activated when the participant was approximately 1 m from the stimulus location. This ensured that all riders encountered the stimuli under consistent conditions and allowed for accurate measurement of their gaze response and attention shift. For safety reasons and to avoid extreme reactions, the vehicle was positioned at a lateral distance of 3 m from the edge of the roadway.

The second visual stimulus was a 33 × 49 cm advertising sign (poster). It was an announcement for a student event and was considered suitable due to its vibrant colors and the strong contrast between the font and the background color. The poster was affixed to a streetlight pole near the riders’ crossing, at a height of 1.90 m and a distance of 220 m from the starting point (Figure 3).

As for the third stimulus of the experiment, it was auditory, namely, a wireless audio player (speaker) was used via a Bluetooth connection. The model was a Xiaomi Mi Bluetooth speaker with a total power output of 6 Watts (Figure 4). The sound produced was 78 decibels (at the crossing point) and simulated the sound of a car horn. As in the case of the “alarm”, the sound of the car horn was activated when the participant was approximately 1 m from the stimulus location. To simulate the horn, an application was used on a smartphone connected to the wireless speaker, which generated the sound. It is noted that auditory distraction in the experiment was assessed by monitoring the riders’ visual responses to auditory stimuli. While auditory stimuli do not generate visual data directly, the corresponding gaze redirection provided a measurable indicator of attentional shift.

3.2. Conducting the Experiment

3.2.1. Members-Organizers

In addition to the participating cyclists, three individuals, each with a distinct position and role, were required to ensure the smooth execution of the experiment. The first person, acting as the coordinator, gave the start signal to the participant. The second person was responsible for activating the emergency lights (alarm) and reversing the car as the cyclist approached. The third person was tasked with activating the horn sounding at the moment the cyclist passed the predetermined point.

3.2.2. Conditions

The experiment was conducted during April and May 2023, with temperatures ranging from 15 to 20 degrees Celsius. All experimental sessions were conducted during daylight hours, concluding before sunset, under clear and sunny weather conditions to ensure consistent visibility. These conditions ensured that the visual stimulus (poster) was clearly perceptible to all participants, without interference from low lighting or glare. No other vehicles interfered with the execution of the measurements.

3.3. Equipment

3.3.1. The Vehicle

For the purposes of the experiment, a bicycle was provided for use by all participants. However, participants were also allowed to use their own bicycles if they preferred. The provided bicycle, which was white in color, was manufactured by KTM and belongs to the trekking category. The model’s name is Manhattan, featuring 28-inch wheels and 24 gears.

3.3.2. Eye Tracking Device

The participants in the experiment used SensoMotoric Instruments eye-tracking glasses (Figure 5). The system consists of a pair of lenses with a front-facing camera, two smaller cameras to track the movement of each eye, and a microphone. The recorded data are stored on a laptop computer via a wired connection using a USB cable. The system also includes a lens replacement capability and unique fittings to adjust the system to the user’s nose, allowing the eye cameras to be raised or lowered to center the eyes for optimal system integration into the cyclist’s field of view.

3.3.3. Computer and iView Software

The iViewETG software was installed on the laptop carried by each rider in a backpack during the experiment and was used to record data. Using the device’s cameras (SMI eye-tracking glasses), the software recorded eye movements. Before the start of recording for each participant, the device had to be calibrated. To ensure measurement accuracy, this step involved calibrating both the eye cameras and the front camera of the device. Calibration was performed by directing the cyclist’s attention to three different fixed points, which the user acknowledged by focusing on them, while the program operator confirmed each point by clicking on it. After the calibration process was completed, the system was ready for recording. The data collected by the device were analyzed using the BeGaze program (version 3.7) (Figure 6) and subjected to frame-by-frame analysis.

4. Results

4.1. Sample Description

The study included 100 cyclists, comprising 66% males (66 participants) and 34% females (34 participants). The riders were grouped into seven age categories, with the 46–55 and 56–65 age groups having the highest representation. The 26–35 age group followed with 18 riders, while the under 18, 36–45, and over 65 age groups each had 12 participants. Most cyclists reported average to high levels of cycling experience.

Regarding frequency of use, 36% of participants cycled regularly, while 64% did not use their bicycles daily. Among regular cyclists, 11% rode 2 days a week, 22% cycled daily, and most used their bicycles 3 to 5 days a week. The primary reasons for cycling were recreation (66.7%), fitness (55.6%), and work/education (33.3%).

When asked about cycling locations, 20% of participants exclusively cycled in urban areas, 44% in suburban areas, and 34% in both urban and suburban routes. Only 2% cycled in all environments. Notably, just two participants had been involved in a cycling-related traffic crash.

Regarding their behavior during the experiment, half of the cyclists reported riding more cautiously, while the other half maintained their usual behavior. Comfort levels varied: 42% felt highly comfortable, and 46% felt very comfortable during the experiment. On distractions, 60% of cyclists believed they were not distracted while riding, while 40% felt they were. In terms of distraction’s importance to road safety, 50% considered it very important, 26% thought it was important, and 12% saw it as moderately important. These responses are summarized in

Table 1. Summary of participants’ responses.

Variable	Percentage
Participants’ gender	Male = 66% Female = 34%
Participants’ age group (years)	Up to 18 = 12% 18–25 = 2% 26–35 = 18% 36–45 = 12% 46–55 = 22% 55–65 = 22% Older than 65 = 12%
Cycling experience	Limited = 18% Average = 28% High = 54%
Frequent bicycle users	Yes = 64% No = 36%
Bicycle use: number of days per week	0–2 days = 12% 3–5 days = 66% 6–7 days = 22%
Reason for bicycle use: (more than one answer was acceptable)	Commuting = 34% Recreational = 66% Fitness = 56%
Type of route used	All = 2% City center = 20% City center and urban network = 34% Suburban network = 44%
Involvement in a bicycle crash in the past	Yes = 2% No = 98%
Level of caution during the measurements	More than usual = 50% Usual = 50%
Comfort during the experiment	Very low = 0% Low = 2% Average = 10% High = 42% Very high = 46 riders
Perceived distraction during the experiment	Yes = 40% No = 60%
How significant is distraction for road safety?	Very low = 8% Low = 4% Average = 12% High = 26% Very high = 50%

.

4.2. Descriptive Analysis

The descriptive statistics (percentages) of rider distraction stimuli in which the riders were distracted, per stimulus, are presented in Figure 7 that follows. It is noted that a rider was considered distracted if they redirected their gaze toward the source of the stimulus at any point during its appearance, even for a brief time interval. It is observed that the 3rd stimulus was the one that caused the highest distraction to the riders.

Understanding distraction patterns and duration is essential for assessing cyclists’ responses to different stimuli and identifying natural variation. To explore this further, the data were analyzed using the Jenks Natural Breaks method, which identifies natural cut-off points to minimize within-group variance and maximize between-group variance. The cut-off points were defined using the classInt package [41] in the R programming language [42]. Observations with zero distraction time were excluded; therefore, the levels were set as follows: <0.53 s, 0.53–1.14 s, 1.15–1.95 s, and greater than 1.95 s. Figure 8 presents the number of participants per distraction class for each of the three stimuli. This classification enables comparison of the strength and duration of attention shifts elicited by different stimulus types. For example, a higher count in the longer-duration class indicates that a specific stimulus led to more prolonged distraction episodes.

For Stimulus No. 1, the highest frequency of distraction was observed in the greater than 1.95 s range, indicating that this stimulus tends to cause more prolonged distractions. For Stimulus No. 2 (poster), the highest frequency was recorded in the 0.53–1.14 s range, suggesting that this stimulus causes more moderate distractions of short-to-medium duration. In the case of Stimulus No. 3 (horn), the highest frequency of distraction occurred in the 0.53–1.14 s range, but the 1.15–1.95 s range also has a high frequency, implying that this stimulus leads to medium-length distractions more frequently.

Overall, 27.6% of the observations ranged from 0.53 to 1.14 s, with an average duration of 0.74 s, while 30.6% were recorded within the 1.15–1.95 s range, with an average duration of 1.52 s. Additionally, 29% of the observations lasted more than 1.95 s, with an average duration of 2.83 s. The overall average distraction duration across all observations was 1.62 s.

Distraction was further classified as either continuous or intermittent based on gaze duration and behavior. Continuous distraction referred to a single, uninterrupted gaze toward the stimulus; in contrast, intermittent distraction involved multiple shorter glances toward the stimulus, interspersed with returns to forward gaze. The percentages of distracted riders, by category, are presented in Figure 9. According to the findings, the vast majority of riders (93.3%) showed continuous distraction in the 3rd stimulus (horn). This percentage is 59.0% and 74.2% for Stimuli 1 and 2, respectively, and in these, the dominant form of distraction observed is also continuous. As expected, drivers who experienced intermittent distraction had a higher average duration of distraction than those who experienced continuous distraction across all three stimuli. Specifically, the average duration of intermittent distraction for the 1st stimulus (emergency lights) was 2.99 s compared to 1.93 s for continuous distraction. For the 2nd stimulus (poster), it was 2.13 s versus 0.99 s for continuous distraction, while for the 3rd stimulus, it was calculated to be 2.05 s, which is longer than the corresponding 1.27 s for continuous distraction.

In terms of bicycle usage, riders who regularly use bicycles exhibited fewer instances of distraction across all three stimuli (Figure 10). Specifically, 22.2% of bicycle users were distracted by the 1st stimulus, compared to 48.4% of non-users. For the 2nd and 3rd stimuli, the percentages of distracted bicycle users were 27.8% and 55.6%, respectively.

4.3. Cluster Analysis

Cluster analysis is a widely used technique in data science and statistics for grouping data points with similar characteristics. In this study, cluster analysis was conducted to identify meaningful patterns in the dataset based on the distraction duration for each stimulus. The process involved several key steps, including data preprocessing, feature selection, clustering algorithm application, and evaluation of cluster validity.

Initially, the dataset was cleaned to handle missing values. A set of relevant features, i.e., duration of distraction from Stimulus No. 1 (activated cars’ emergency and reversing lights), No. 2 (poster), and No. 3 (car horn), was then selected. Finally, the k-means clustering algorithm was employed, which is effective for partitioning data into distinct groups based on similarity. The analysis was conducted in the R programming language [42], using the “cluster” package [43].

To determine the optimal number of bicycle riders’ clusters, a combination of the elbow method and the silhouette method was used. The elbow method involves plotting the within-cluster sum of squares (WCSS) as a function of the number of clusters. The point at which the curve exhibits an “elbow” or a sharp decline in WCSS indicates the most appropriate cluster count. As it is shown in Figure 11(left), the elbow was observed at k = 5, suggesting that five clusters provided a good balance between complexity and explanatory power.

To further validate this selection, the silhouette method was applied, which measures how well each data point fits within its assigned cluster compared to other clusters. The silhouette score ranges from −1 to 1, with higher values indicating well-separated and cohesive clusters. Silhouette scores were calculated for different values of k, and as shown in Figure 11(right), k = 5 results in a relatively high score compared with other alternatives. From the results of the two methods and for maintaining a reasonable and interpretable number of clusters, the extraction of five clusters was considered ideal.

Based on these findings, the clustering of bicycle riders was finalized with five distinct groups, which exhibited clear separation as shown in Figure 12. The combination of the elbow approach and the silhouette method ensured a robust and data-driven selection of the optimal number of clusters, leading to reliable insights from the dataset.

For further analysis and characterization of the five clusters of bicycle riders, for each of them, the average duration of distraction from Stimulus No. 1 (alarm), No. 2 (poster), and No. 3 (car horn) was calculated, and it is presented in

Table 2. Average distraction duration by stimulus type and cluster.

Cluster	Distraction Duration from Emergency and Reversing Lights (in s)		Distraction Duration from Poster (in s)		Distraction Duration from Car Horn (in s)
	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.
1	0.210	0.664	1.370	0.419	0.260	0.419
2	3.400	0.786	2.313	0.689	1.350	0.431
3	0.317	0.789	0.029	0.129	0.143	0.227
4	3.000	0.726	0.283	0.356	2.067	0.665
5	0.493	0.708	0.107	0.232	1.436	0.486
Total	0.924	1.347	0.398	0.749	0.844	0.841

.

Based on the results of Table 2, it is identified that the five clusters have the following characteristics:

• Cluster 1: Those belonging in this cluster are mostly distracted by visual stimuli that are exogenous road factors.
• Cluster 2: Those belonging in this cluster are heavily distracted by all stimulus types.
• Cluster 3: Those belonging in this cluster are distracted to a very low extent by all stimulus types.
• Cluster 4: Those belonging in this cluster are mostly distracted by endogenous road factors.
• Cluster 5: Those belonging in this cluster are mostly distracted by audio factors.

As the next step of the analysis, the correlation between the cluster type and the users’ characteristics was investigated. More specifically, it was explored whether gender, age, cycling experience, daily use of bicycle, and perceived importance of distraction in terms of road safety, affect the cluster to which each user belongs or not. The analysis showed that only the gender variable (at a 95% confidence level) and the daily use of a bicycle (at a 90% confidence level) seem to significantly affect cluster membership.

Table 3. Gender distribution of cluster participants.

Cluster	Male	Female
1	6	4
2	8	0
3	23	19
4	12	0
5	17	11
Total	66	34
Chi-square test result (p-value)	0.010

and

Table 4. Daily bicycle use distribution among cluster participants.

Cluster	Daily Bicycle Users	Non-Daily Bicycle Users
1	2	8
2	2	6
3	22	20
4	4	8
5	6	22
Total	36	64
Chi-square test result (p-value)	0.060

present the relationship between gender and cluster, as well as between daily bicycle use and cluster membership. It is identified that males are much more likely to belong in Cluster 2 or 4, which are two clusters that indicate high levels of distraction due to the various stimuli. On the other hand, females are more likely to belong to Cluster 3, which represents the cluster with the lowest levels of distraction. In this cluster, it is also more likely to identify people who are cycling on a daily basis. On the other hand, those who cycle less frequently are more likely to be found in Cluster 5, which indicates people who are distracted by audio factors.

Finally, the relationship between cluster type and perceived distraction was examined, revealing a strong correlation between the two variables (

Table 5. Perceived distraction of cluster participants.

Cluster	Did Not Feel Distracted During the Experiment	Felt Distracted During the Experiment
1	6	4
2	2	6
3	21	21
4	6	6
5	25	3
Total	60	40
Chi-square test result (p-value)	0.003

). Interestingly, participants in Cluster 3, characterized by the lowest levels of objective distraction, were evenly split between those who reported feeling distracted and those who did not. Conversely, individuals in Cluster 2, associated with the highest distraction levels, were more likely to subjectively perceive distraction during riding. The most notable finding, however, concerns Cluster 5, whose members were primarily affected by auditory stimuli. Despite their measurable distraction, the vast majority of these participants did not report feeling distracted during their ride, suggesting a potential disconnect between objective and perceived distraction in the presence of audio-based stimuli.

5. Conclusions

Investigating cyclist distraction caused by various visual and auditory stimuli is crucial for improving road safety and reducing crash risks in urban environments. Cyclists are vulnerable road users, and distractions from external factors, such as car horns, advertising signs, and alarms, can impair their focus, reaction time, and decision-making. Understanding how different stimuli may affect attention may assist in developing targeted safety measures, such as improved road design, better cycling infrastructure, and awareness campaigns. Identifying high-risk distraction patterns also enables policymakers to implement effective regulations and interventions. By studying cyclist distraction, safer cycling environments can be promoted for reducing traffic-related injuries and fatalities.

The analysis of our experimental data results concerning cyclists’ distraction, specifically focusing on the duration of distraction and the different participant categories, led to several significant conclusions. Among the visual stimuli, dynamic stimuli—particularly the alarm stimulus—caused a higher percentage of riders to experience distraction. This group of participants also demonstrated a notably longer average distraction time compared to those exposed to static stimuli (the second stimulus). The dynamic nature of the visual stimulus likely draws more attention and thus leads to greater levels of distraction, as drivers must continuously adjust their focus in response to the changing stimuli.

In contrast, auditory stimuli emerged as the most distracting to drivers, with a substantial proportion of participants showing consistent or continuous distraction when exposed to sound-related stimuli. This suggests that auditory stimuli, which demand attention and may trigger automatic responses, have a more pervasive effect on driver attention, drawing focus away from the road for extended periods. The results showed that continuous distraction was the most dominant form of distraction across all categories, further highlighting the significant impact of stimuli in maintaining attention or diverting it over time.

Interestingly, the study also found that riders who experienced intermittent distraction had a higher average duration of distraction compared to those who experienced continuous distraction. This finding may seem counterintuitive, as one would expect continuous distractions to lead to longer attention lapses. However, intermittent distractions, which occur in cycles, may cause more frequent shifts in attention, leading to more sustained periods of reduced focus during the intervals of distraction.

The study also explored the potential effect of cycling experience on distraction. Non-cyclists exhibited a higher rate of distraction in response to each stimulus. This group showed a greater tendency to be distracted, although the total duration of the distraction was not significantly impacted by their cycling habits. This suggests that prior cycling experience might not necessarily affect susceptibility to distraction in a car setting, but other variables, such as driving experience or familiarity with the road environment, could have a stronger influence.

The findings reveal distinct patterns of cyclist distraction influenced by external stimuli and individual characteristics. The identification of five clusters highlights varying susceptibility to visual, auditory, and road-related distractions. Notably, gender and daily cycling frequency were found to play a significant role in distraction levels. Males are more likely to be assigned to high-distraction clusters, whereas females and frequent cyclists demonstrate greater focus and lower susceptibility to distractions. These insights underscore the importance of targeted safety measures and awareness initiatives to mitigate risks and enhance road safety for all cyclists.

Overall, the findings highlight the need for targeted policy interventions to mitigate cyclist distraction and enhance road safety. Given that dynamic visual stimuli and auditory cues significantly increase distraction, policymakers should consider regulating roadside advertising signs, particularly animated or flashing signs, near cycling routes. Additionally, stricter noise control measures, such as limiting excessive honking in high-cyclist areas, can help reduce auditory distractions. Public awareness campaigns should emphasize distraction risks, particularly for male cyclists who are more prone to high-distraction clusters. Furthermore, integrating distraction-awareness training into cycling education programs can help riders develop strategies to maintain focus. Infrastructure improvements, such as dedicated cycling lanes with minimal external distractions, can further enhance cyclist safety and reduce crash risks.

The study is subject to some limitations that need to be mentioned. Although measures were taken to reduce potential biases—such as varying participant cycling experience—it is acknowledged that some residual bias may exist due to factors like prior exposure to similar environments or individual risk perception. This limitation should be considered when interpreting the results. Also, while this study focuses on visual distraction, defined as the time riders diverted their gaze from the forward roadway toward a stimulus, we acknowledge that not all gaze shifts necessarily imply reduced safety. In some contexts, such as reacting to an auditory warning or a flashing light, these glances may serve a protective function. However, in urban cycling environments, frequent or prolonged deviations from the forward view can compromise situational awareness, especially in areas with mixed traffic, pedestrian crossings, or obstacles. Thus, gaze diversion is widely recognized as a potential proxy for distraction and risk. Nevertheless, a key limitation of our study is the lack of performance-based safety metrics. Future research should incorporate additional data streams such as speed variation, braking behavior, and path deviation to directly assess the functional impact of distraction. This would allow a better distinction between adaptive attention shifts and those that compromise safety, enabling more refined conclusions about the implications of rider distraction. Finally, a limitation of the current study is that auditory distraction was measured indirectly through visual gaze shifts toward the auditory source. While this approach offers an observable proxy for attention redirection, it does not capture the full spectrum of cognitive or auditory processing. Future research should consider combining eye-tracking with additional data sources to more comprehensively assess how auditory cues impact situational awareness and cycling safety.