Impact of social and technological distraction on pedestrian crossing behaviour: an observational study
- 1Harborview Injury Prevention & Research Center, University of Washington, Seattle, Washington, USA
- 2Department of Pediatrics, University of Washington, Seattle, Washington, USA
- 3Seattle Children's Hospital and Seattle Children's Research Institute, Seattle, Washington, USA
- Correspondence to Dr Beth E Ebel, Harborview Injury Prevention & Research Center, University of Washington, 325 Ninth Ave, Box 359960, Seattle, Washington 98104-2499, USA;
- Accepted 2 November 2012
- Published Online First 13 December 2012
Objectives The objective of the present work was to study the impact of technological and social distraction on cautionary behaviours and crossing times in pedestrians.
Methods Pedestrians were observed at 20 high-risk intersections during 1 of 3 randomly assigned time windows in 2012. Observers recorded demographic and behavioural information, including use of a mobile device (talking on the phone, text messaging, or listening to music). We examined the association between distraction and crossing behaviours, adjusting for age and gender. All multivariate analyses were conducted with random effect logistic regression (binary outcomes) and random effect linear regression (continuous outcomes), accounting for clustering by site.
Results Observers recorded crossing behaviours for 1102 pedestrians. Nearly one-third (29.8%) of all pedestrians performed a distracting activity while crossing. Distractions included listening to music (11.2%), text messaging (7.3%) and using a handheld phone (6.2%). Text messaging, mobile phone use and talking with a companion increased crossing time. Texting pedestrians took 1.87 additional seconds (18.0%) to cross the average intersection (3.4 lanes), compared to undistracted pedestrians. Texting pedestrians were 3.9 times more likely than undistracted pedestrians to display at least 1 unsafe crossing behaviour (disobeying the lights, crossing mid-intersection, or failing to look both ways). Pedestrians listening to music walked more than half a second (0.54) faster across the average intersection than undistracted pedestrians.
Conclusions Distracting activity is common among pedestrians, even while crossing intersections. Technological and social distractions increase crossing times, with text messaging associated with the highest risk. Our findings suggest the need for intervention studies to reduce risk of pedestrian injury.
Motor vehicle–pedestrian crashes remain a significant source of serious injury, with an estimated 60 000 pedestrians injured and 4000 killed per year in the USA.1 Globally, pedestrians and other vulnerable road users account for almost half of road traffic deaths.2 In Washington, as in most US states, traffic laws require that vehicles yield to pedestrians in crosswalks and at intersections.3 ,4 Although the law assigns pedestrians the right of way, it does not relieve pedestrians of looking out for their own safety. Cautionary behaviours include using sidewalks and crosswalks when available, obeying traffic signals and looking both ways before entering the street.
While poor intersection design and dangerous driving account for some pedestrian fatalities, a recent study found that actions by pedestrians may account for as much as 15% of all deaths.5 In a study from Vancouver (British Columbia, Canada), 21% of pedestrians observed committed one or more crossing violations.6
Distracted walking, like distracted driving, is likely to increase in parallel with the penetration of electronic devices into the consumer market. As of 2011, there were more phones than people in the USA, and internationally, the number of mobile phone subscriptions is an estimated 5.9 billion.7–9 Distraction has been more frequently studied among drivers than pedestrians; up to 28% of driver crash risk is attributable to distraction from cell phone use or text messaging.10 Talking on the phone, texting, using an MP3 player or adjusting vehicle music controls diminish driver focus and increase the risk of a crash.11–18 Among drivers, cohort studies indicate that the use of voice/text devices is associated with crash risks ranging from 4–23 times above baseline levels.19
Less is known about the impact of distraction on pedestrian behaviour or risk. Pedestrians tend to act less cautiously when distracted, whether by cellphones, music players, food, or other people.5 ,20–22 Studies of distracted pedestrians have primarily focused on the behaviour of selected subjects in a simulated environment.22–25 Simulated environments offer more control of conditions and capture demographic and behavioural background on participants. However, removing pedestrians from their natural environment and making them aware of the focus on their walking behaviour may alter their actions. The few studies that have occurred in real environments have focused on individual locations, age groups, or distractions, or have tracked violations without examining their relationship to distracted walking.5 ,6 ,20 We sought to expand on prior research by studying distracting behaviours in a large group of pedestrians during the act of crossing the street.
We conducted a prospective observational study of pedestrian behaviour at intersections in Seattle, Washington in the summer of 2012. We chose the 20 intersections with the highest number of pedestrian injuries during the prior 3 years, based on data from the Seattle Department of Transportation. The study was reviewed and deemed exempt by the Seattle Children's Hospital institutional review board.
In order to maximise the flow of pedestrians at each intersection and observe a variety of pedestrians, three observation time windows were selected: 8:00–9:00 (morning commute), 12:00–13:00 (lunch break) and 16:00–17:00 (afternoon commute). We assigned each intersection to an observational time slot using random number generation. Data collection sheets were created for each intersection, mapping compass directions and landmarks.
At each intersection, two observers recorded pedestrian crossing behaviour for every pedestrian meeting the selection criteria. At each intersection of two streets, there were eight typical directions a pedestrian could walk. Each observer watched a given direction for 15 min and then rotated, to cover eight crossing directions in the span of the hour. Given the traffic activity at these high-risk intersections, crossing on the diagonal was almost never observed. To avoid possible selection bias, each observer carried a timer that vibrated once per minute. The observer recorded data for the first person who reached the curb after the timer went off. In cases where two individuals arrived completely simultaneously at the curb following the timer, the first person to step off the curb was selected as the ‘index pedestrian’.
Traffic volume data and collision data were obtained from the Seattle Department of Transportation. The width of the street was obtained by manually measuring the distances using a rolling measuring device. The number of lanes crossed for each street was recorded. Examples of data collection tools are available in appendices 1–3.
All observations were completed by two trained observers. Observers recorded demographic and behavioural information for each pedestrian. The demographic information recorded included the gender and estimated age of the pedestrian. The age categories were <18, 18–24, 25–44, 45–64 and 65+ years.
Observers additionally recorded whether the pedestrian was alone, in a group, or in a group talking to another person. Individuals were counted as being in a group if they appeared to be in a social interaction with another individual at the crosswalk, for example, holding hands. If an individual was talking to another person while crossing, they were counted as ‘group talking’.
For behaviours, the observer recorded the direction the pedestrian walked, whether the pedestrian crossed at the crosswalk, whether the pedestrian looked left and right, and whether he or she obeyed the intersection signal (if present). Crossing in the crosswalk meant that the pedestrian took no more than one step outside the painted crosswalk lines. Looking left and right meant that the observer had to see a noticeable turn of the chin left and right immediately prior to the pedestrian stepping into the roadway. Obeying the lights meant the pedestrian entered the street when the walk signal was lit. Pedestrians who entered the street after the ‘don't cross’ indicator had begun to flash were counted as disobeying the signal.
Observers also noted whether the pedestrian was using a mobile phone (phone to ear or earpiece), music player (earphones), or texting (manual use of mobile device) while crossing. Individuals with headphones in their ears connected to a device capable of playing music were counted as listening to music. The observer recorded the time it took each pedestrian to cross the intersection, measuring from the time both feet entered the street, to the time both feet stepped onto the sidewalk.
All data were entered into an Excel spreadsheet (Microsoft, Redmond, Washington, USA) and Stata 11 (Stata Statistical Software, College Station, Texas, USA) was used for data analysis.
Pedestrian distractions were categorised as follows: listening to music (headphones), handheld mobile phone, earpiece mobile phone, text messaging and talking with another person. Additional distractions (eg, carrying baby, pushing a stroller) were categorised as ‘other’. Age 18–24 was considered the reference category.
We examined the association between pedestrian crossing time (lane crossing time in seconds) and pedestrian distraction, adjusted for key confounders (age, gender). Next, we examined the association between pedestrian crossing behaviours (cross at crosswalk, obey lights, look both ways) and pedestrian distraction, adjusted for key confounders (age, gender). Finally, we examined the association between optimal pedestrian behaviour (crossed at crosswalk, obeyed lights, looked both ways) and pedestrian distraction. All multivariate analyses were conducted with random effect logistic regression (binary outcomes) and random effect linear regression (continuous outcomes), accounting for clustering by site.
Observers recorded crossing behaviours of 1102 pedestrians at intersections, with a mean of 55 observations at each site (range 19–88). Though observation times were randomly distributed among observation sites, nearly half of pedestrians (46.6%) were observed walking from 8:00–9:00 reflecting higher pedestrian density in the morning hours (table 1). The majority (54.3%) of pedestrians were in the 25–44 year age category. Slightly more than half the pedestrians observed were male. Nearly 80% of the pedestrians observed walked alone. Most pedestrians obeyed the lights (80.0%) and crossed at the crosswalk (94.4%), but only one-third (34.9%) of pedestrians looked left and right prior to entering the roadway. Approximately 30% of all pedestrians observed performed a distracting activity while crossing. Distractions while in the roadway included listening to music (11.2%), text messaging (7.3%) and using a handheld phone (6.2%) (table 1).
Distracted pedestrians took significantly longer to cross the intersection (table 2). The mean crossing time for an undistracted individual was 10.4 s. Across the average intersection length of 3.4 lanes, individuals using a handheld or hands-free phone took an additional 0.75 and 1.29 s to cross, respectively. Pedestrians who were text messaging took over half a second longer to cross each lane (0.55 s, 95% CI 0.36 to 0.75), adding an extra 18.0% to total crossing time. Pedestrians listening to music walked faster than undistracted pedestrians, by an average of 0.16 s per lane. Females walked slower than males, as did individuals talking in a group, compared to their counterparts who walked alone. Finally, as age increased, crossing time per lane also increased, with pedestrians 65 and older walking nearly a full second slower per lane than the reference group, 18–24 year olds (table 2)
Some forms of distraction were also associated with risky pedestrian behaviour (table 3). Pedestrians who were texting were significantly more likely to cross the street without looking both ways before crossing (OR=4.00, 95% CI 2.04 to 7.84). Listening to music (OR=1.69, 95% CI 1.06 to 2.66), talking with others (OR=1.69, 95% CI 1.08 to 2.59) and ‘other distractions’ (OR=2.86, 95% CI 1.19 to 6.98) were also associated with failure to look both ways. Nearly one-third of the ‘other distractions’ observed involved interactions with a child or a pet. Talking on the phone was not associated with failure to look both ways before crossing the street.
Female pedestrians, whether distracted or not, were somewhat less likely to look both ways before crossing the street.
Distracting behaviours, age, gender, or social grouping, were not associated with the likelihood that a pedestrian crossed at the crosswalk or obeyed the lights, as most pedestrians followed these safety behaviours. However, pedestrians who obeyed the traffic signals were 2.8 times more likely not to look both ways (table 3).
Finally, we examined the association between distracting behaviours and optimal crossing behaviour, defined as looking both ways, crossing at the crosswalk and obeying the traffic signals (table 4). Only text messaging and gender had a significant effect on optimal crossing behaviour. Only 26% of pedestrians exhibited all three optimal crossing behaviours. Walkers who were text messaging were 3.9 times more likely to exhibit at least 1 unsafe crossing behaviour. Controlling for distracting behaviours, females were twice as likely to exhibit at least one unsafe crossing behaviour, relative to male counterparts.
Our study found that many pedestrians send text messages or use mobile devices while crossing the street. Use of these devices is associated with slower crossing times. Text messaging appears particularly risky. Texting is associated with an 18% increase in crossing times and failure to perform routine pedestrian safety behaviours before stepping into the roadway. This is the first published observational study with sufficient power to examine the impact of texting on real-world pedestrian behaviour.20 As was found by earlier observational studies of pedestrian behaviour, individuals talking on a cell phone crossed more slowly than those who were undistracted.
In our study, nearly 30% of pedestrians continued distracting behaviours in the intersection, a proportion somewhat higher than the 20% figure from an observational study conducted in 2005.5 Pedestrians were chosen according to an algorithm to avoid risk of selection bias which may have been present in earlier studies.20 ,21 Fewer individuals were talking on mobile phones compared with a previous large observational trial which recorded the behaviour of all cell phone users to arrive at the crosswalk.20 This earlier study predated the widespread adoption of text messaging, and suggests changing patterns of mobile device use.
Laboratory studies have demonstrated that distractions impair pedestrian awareness of their surroundings. In simulation studies, children and college-age pedestrians behaved with less caution, experienced more ‘hits’ and close calls with virtual cars, and waited longer to cross than their undistracted counterparts, regardless of previous experience with mobile technology.22–25 Additionally, virtual studies of pedestrians corroborate our finding that text messaging increases high-risk pedestrian choices.22
Individuals walking with music crossed more speedily than those with no distractions; previous studies suggest that a musical beat may alter the natural gait speed of an individual.26 Though music listeners crossed more quickly, they were less likely to look both ways before crossing the street. Additionally, we were surprised to find that females were less likely to display optimal crossing behaviour than males, because males are disproportionately injured in crashes. However, previous observational studies suggest that before and during crossing, men were more focused on looking at vehicles than were women, who tended to focus on traffic lights and other individuals at the crosswalk.27 The higher rate of pedestrian injuries in males may be related to other risky behaviour, such as intoxication, crossing midblock and crossing arterials at night.
Recent work indicates that young people may display compulsive behaviours around mobile device use, which may make it more difficult to curb their use in contexts where it is unsafe, such as driving or crossing the street.28 This compulsive use of cell phones was also associated with higher reported crash rates.28
This study has strengths as well as limitations. We observed a large number of pedestrians under normal conditions, at a wide range of intersections. The unobtrusive nature of the study was designed to capture pedestrian behaviour. As a consequence, we estimated pedestrian age. The presence of an observer may have been noted by an alert pedestrian, who may have potentially altered his or her crossing behaviour. If this occurred, it is likely that the true prevalence of distraction may be higher than reported. The study was conducted in one Northwest city, and results may not be representative of other locales. Observation time windows may have favoured working-age individuals. This study analysed pedestrian risk behaviours such as disregarding traffic signals, but we are not able to link these behaviours to the risk of injury. We saw little evidence of highly risky pedestrian behaviour, such as running between moving vehicles, and conducted all observations during daylight hours. However, even moderately risky behaviours such as crossing against or close to the lights have been correlated with risk of injury eight times that of legal crossings.29
Pedestrian distraction in general, and text messaging in particular, is associated with slower crossing times and unsafe pedestrian behaviours. The steady rise in the prevalence of text messaging and the use of mobile devices for a wide range of functions such as playing games suggests that the risk of distraction will increase. Solutions are likely to include the three ‘Es’ of injury prevention: education of the public about risks, engineering and environmental modifications, and enforcement. Published surveys and the lay press suggest that drivers10 ,30–32 and pedestrians33 ,34 understand the risk of doing other activities while using mobile devices, believe that others should comply with the law, but continue to use devices. Individuals may feel they have ‘safer use’ than others, view commuting as ‘down time’, or have compulsive behaviours around mobile device use.10 ,28 ,31 ,32 Environmental modifications which separate pedestrians from traffic and promote safe crossing may be even more important in an era of growing distraction. While individuals do not feel ‘at risk’ for relatively rare events such as injury,35 they may feel ‘at risk’ for a distraction citation if there is visible evidence of effective enforcement, as is being considered in some US cities.36 ,37 Ultimately, a shift in normative attitudes about pedestrian behaviour, similar to efforts around drunk driving, will be important to limit the state-dependent risk of mobile device use.
What is already known on this topic
Pedestrian actions may account for 15% of pedestrian fatalities.
Talking on the phone while crossing the street impacts pedestrian behaviour. Cell phone use is associated with slower crossing times and fewer cautionary behaviours.
Simulation studies suggest text messaging also increases pedestrian risk.
What this study adds
A total of 29.8% of pedestrians were using a mobile device during street crossing, and 7.3% were actively texting.
Pedestrians who were text messaging displayed the highest risk of all distracted walkers, with slower crossing times and failure to display cautionary crossing behaviours.
Pedestrians who cross against the traffic signal are more likely to look left and right than those who follow the traffic signals.
We would like to thank Craig Moore and Vincent Prince from the Seattle Department of Transportation for their assistance in obtaining traffic flow and pedestrian crash data. Support provided by the Harborview Injury Prevention & Research Center, Seattle, Wash.
Contributors BEE conceived of the study. LLT, BEE and FPR designed data collection tools and planned the study. LLT pilot tested study instruments and LLT and RCA collected study data. BEE analysed the data, and the analytic plan and findings were reviewed with all coauthors. LLT drafted the manuscript. BEE, FPR and RCA revised the draft manuscript.
Funding This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None.
Ethics approval Seattle Children's Research Institute.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All collected study data have been summarised and published in this manuscript. Interested parties who wish to request access to disaggregated and deidentified data for meta-analysis or other purposes are welcome to request access from the corresponding author.