Article Text

Download PDFPDF

Technology-related distracted walking behaviours in Manhattan's most dangerous intersections
  1. Corey H Basch1,
  2. Danna Ethan2,
  3. Sonali Rajan3,
  4. Charles E Basch4
  1. 1William Paterson University, Wayne, New Jersey, USA
  2. 2Health Education and Promotion, Department of Health Sciences, Lehman College, The City University of New York, New York, Bronx, USA
  3. 3Department of Health and Behavior Studies, Teachers College, Columbia University, New York, New York, USA
  4. 4Department of Health and Behavior Studies, Teachers College, Columbia University, New York, New York, USA
  1. Correspondence to Dr Corey H Basch, Department of Public Health, William Paterson University, Wing 150, Wayne, NJ 07470, USA; baschc{at}


Use of mobile devices has been cited as a distraction while driving, and more recently, among pedestrians crossing urban streets. In 2010, over half of New York City traffic fatalities were pedestrians. The aim of this study was to estimate the prevalence of distracted walking due to pedestrians’ use of headphones, mobile phones, or both. Data were gathered by direct observations at the 10 intersections in Manhattan with the highest frequency of pedestrian–motor vehicle collisions. More than 1 in 4 of the >3500 pedestrians observed were distracted by mobile electronic devices while crossing during the ‘walk’ (28.8%) and ‘don't walk’ (26.3%) signals. Poisson regression analyses established there was a significant difference in individuals talking on a mobile device during the ‘walk’ signal versus the ‘don't walk’ signal; however, no other significant differences in other distracted walking behaviours were observed. This study contributes to the emerging literature on distracted walking behaviour by pedestrians in busy urban areas and can help to inform pedestrian-focused safety efforts.

Statistics from


Ninety-one per cent of the US population owns a mobile phone,1 and nearly half own a smartphone (a trend that is increasing).2 Smartphones allow users to engage in handheld computing tasks, watch or listen to audio and/or visual media. The use of mobile devices and smartphones results in increased rates of distraction among both automobile drivers and pedestrians. Mobile phone use while walking in busy urban streets can increase risk for injury or death. A recent nationwide study of hospital records revealed that the number of mobile-phone-related pedestrian injuries increased dramatically from 2007 (1.11%) to 2010 (3.67%), and that the majority of all pedestrian injuries related to mobile phone use (78.6%) involved the pedestrian either talking on the phone or texting.3 In 2010, more than 1500 pedestrians were treated for injuries related to mobile phone use while walking.4

In New York City (NYC), pedestrians comprised 52% of all NYC traffic fatalities between 2005 and 2009.5 Pedestrian crashes from motor vehicles in NYC have resulted in estimated expenses of $1.38 billion annually.5 The causes of pedestrian–motor vehicle crashes are not fully understood. However, distraction from technology-related devices appears to be a growing cause of injuries and fatalities.3 ,4 ,6

The few published studies on distracted walking behaviours suggest adverse effects on both youth and adults. In simulations, 10-to11-year-old participants answering a mobile phone were found to pay less attention to traffic, have less time to cross the street and to be at higher risk of getting ‘hit’ by a car.7 Compared with their undistracted peers, college students who were texting or listening to music were more likely to be ‘struck’ by a vehicle while crossing a virtual street.8 An Australian study revealed that male pedestrians talking on a mobile phone crossed more slowly than matched controls at ‘unsignalised’ crosswalks, and that females using phones not only crossed streets more slowly but were also less likely to look for traffic before or while crossing.9 A study of pedestrian behaviour at 12 intersections in San Francisco found that up to 18% of pedestrians were using a mobile device when crossing the street.10 In another recent study in Seattle that included observations of over 1100 pedestrians crossing at 20 high-risk intersections, almost 30% engaged in at least one distracting behaviour, including text messaging, listening to music or talking on a phone.11 Pedestrians who were texting took longer to cross the street and were almost four times more likely to engage in at least one additional hazardous behaviour, such as not looking both ways before crossing or disobeying the traffic signal.11 In a Nevada-based study of 866 pedestrians, only 13.5% of pedestrians looked both ways and crossed the street at the appropriate time; these researchers also identified several distractions when walking, including cell phone use and listening to music.12 Collectively, these studies contribute to understanding about the nature and scope of distracted behaviours of pedestrians.

We did not identify any published studies of technology-related distracted walking (hereafter stated as ‘distracted walking’) in NYC or any that observed distracted walking during the ‘don't walk’ signal, when there is greater risk of a pedestrian–motor vehicle injury occurring. We therefore observed the prevalence of distracted walking during designated ‘walk’ and ‘don't walk’ signals in the 10 most dangerous intersections in Manhattan, the borough in NYC with the highest prevalence of pedestrian fatalities.13


Study design

In this cross-sectional study, data were collected by two of the authors (CHB and DE) at the 10 most dangerous intersections in Manhattan, NYC. Based on data from 1995 to 2009, these 10 intersections had the highest frequency of pedestrian–motor vehicle collisions in Manhattan where pedestrians were most likely to be injured by a car.13 The coding instrument and protocol were developed, pilot tested and used to record distracted walking during (1) the ‘walk’ and blinking ‘don’t walk’ signals (when the light was changing from green to red) and (2) the non-blinking ‘don't walk’ signal.

Instrument development, coding and pilot testing

Development and pilot testing the coding protocol involved two phases. First, the observers assessed the feasibility of recording various kinds of information at two busy intersections, which resulted in the development of categories of distracted walking that were deemed feasible to accurately observe and record: (1) talking on a mobile phone or smartphone; (2) wearing headphones; (3) looking down and/or interacting with a mobile phone or smartphone (which may have included texting, web browsing or using other functions of the mobile device); and (4) engaging in more than one of these three behaviours. In addition to coding distracted walking, the total number of pedestrians crossing the street was recorded. Pedestrians not engaging in distracted walking were calculated by subtracting the number engaged in any distracted walking from the total number of pedestrians observed. Second, prior to collecting data at each of the 10 intersections under study, the observers collected data from two busy intersections and independently observed and recorded the total number of distracted walkers and the total number of pedestrians from one corner for five cycles of ‘walk’ (including blinking ‘don't walk’) and non-blinking ‘don't walk’ signals. There was complete agreement at each intersection.

Data collection

Individuals were identified as ‘pedestrians’ only if they were observed walking in the crosswalk. Those crossing the street outside of the crosswalk during the light changes were excluded. All individuals that walked independently through the crosswalk were counted, regardless of age. Individuals being pushed by someone else (eg, in a wheelchair or stroller) were excluded, as were those on bicycles, skateboards or other modes of transportation.

Data were collected at the 10 aforementioned intersections as follows. On each of the four corners of each intersection, distracted walking and total pedestrians crossing were observed and recorded for seven cycles of signal changes, only documenting results during the ‘walk’ signal; then observations and recordings were repeated for four cycles, only observing and recording data during the ‘don't walk’ signal. To help ensure accuracy, one coder tallied the total number of pedestrians, while the other coded pedestrians engaging in each of the four categories of distracted walking outlined above. On average, the two coders observed and recorded data for ∼ 11 min (SD=2.1) at each corner of each intersection. A signal change cycle was demarcated by the traffic light running through a full cycle of red (stop), changing to green (go; thus permitting pedestrians to enter the crosswalk), then to yellow (yield) and returning to red. Observations were conducted on weekdays only, between 7:00 and 16:00, and from May through June 2013. Five of the intersections were observed prior to 13:00 and five were observed after 13:00. This study was reviewed by the Institutional Review Boards at William Paterson University, Lehman College, The City University of New York, and Teachers College, Columbia University, and deemed exempt at each.

Data analysis

Data analyses involved calculating descriptive statistics and subsequently using Poisson regression analyses to evaluate differences in distracted walking behaviours between the ‘walk’ and ‘don't walk’ signals. The prevalence of pedestrians engaging in each category of distracted walking behaviours as well as engaging in any distracted walking behaviour was computed for each round of data collection at each corner of the intersections by dividing the number observed engaging in the respective behaviour by the total number observed, and then calculating the mean of these means. These analyses were calculated separately for the time frames corresponding to the ‘walk’ and ‘don't walk’ signals. In addition, 95% CIs were determined. The Poisson regression analyses allowed for the use of each observation site (n=10) as the unit of analysis to account for clustering and determine whether the observed frequency of distracted walking behaviours during the ‘walk’ signal differed significantly from the observed frequency of distracted walking behaviours during the ‘don't walk’ signal. All data were analysed in SPSS (V.20.0).


A total of 3784 pedestrians were observed, of whom more than 1 in 4 exhibited distracted walking (table 1). On average, 13 pedestrians were observed during each ‘walk’ signal and two were observed during each ‘don't walk’ signal. During both ‘walk’ and ‘don't walk’ signals, headphone use was the most frequently observed distracted walking behaviour (16.3% and 15.6%, respectively). Similar prevalence estimates of distracted walking behaviours were observed during the ‘walk’ and ‘don't walk’ signals.

Table 1

Prevalence (95% CI) of distracted walking behaviours at the 10 most dangerous intersections in New York City during ‘walk’ and ‘don't walk’ signals

We initially used a Poisson regression analysis, using each intersection as the model factor, to determine whether the overall variability in distracted walking behaviours varied significantly between each site. Results indicated that there was no statistically significant difference in type of distracted walking behaviours between each intersection (Wald χ2=1.59, p=0.208). We subsequently calculated Poisson regression analyses using each signal type as the model factor, while controlling for intersection, to determine whether the observed frequency of distracted walking behaviours during the ‘walk’ signal differed significantly from the observed frequency of distracted walking behaviours during the ‘don't walk’ signal. Results indicated that a statistically significant difference in talking on the mobile device was observed between the ‘walk’ and ‘don't walk’ signal groups. No other significant differences were observed (table 2). The relative risk of talking on one's mobile device during the ‘walk’ signal versus the ‘don't walk’ signal was estimated by taking the inverse natural log of the corresponding regression model's β coefficient (β=1.58 (95% CI 1.02 to 2.15)). The relative risk was estimated to be 4.85 (95% CI 2.77 to 8.58), indicating that the probability of talking on one's mobile device is approximately five times more likely to occur during the ‘walk’ versus the ‘don't walk’ signal.

Table 2

Poisson regression results comparing distracted walking behaviour frequency between ‘don't walk’ and ‘walk’ signals across the 10 most dangerous intersections in New York City


Distracted walking, as defined in this study, comprises behaviours exhibited by a substantial proportion of pedestrians crossing high-risk intersections in NYC. The observed rates of distracted walking are consistent with those reported by others.10 ,11 The high prevalence of these behaviours may pose emerging injury risks in urban areas as mobile technology increases in popularity1 ,2 and increasing proportions of the US14 ,15 and global16 population migrate to urban areas. The frequency with which distracted walking was observed during the ‘don't walk’ signal is particularly pertinent given that over half of pedestrian fatalities (56%) occurring from crashes in NYC involved crossing during the ‘don't walk’ signal.5 This could be a result of the transporting nature of reading on a screen or texting in that it impacts situational awareness or other cognitive distractions for the pedestrian.17 ,18

Wearing headphones was the most frequently observed distracted walking behaviour. Additionally, significantly more participants were observed talking on their mobile device during the ‘walk’ signal than the ‘don't walk’ signal; however, no other significant differences in distracted walking behaviours were observed between the two signal types. A recent study analysed 116 reports of pedestrian death and injury related to headphone use. The overwhelming majority of cases (89%) were in urban areas, and more than a quarter of all cases (29%) indicated the presence of an audible warning before the crash.19 Nevertheless, the nature and extent to which pedestrians’ use of headphones and other forms of technology contribute to pedestrian–motor vehicle injury is not fully known.20

There is a paucity of knowledge about the nature and scope of distracted walking, including its role as a behavioural risk factor for pedestrian–motor vehicle injuries and no intervention research. This small study was limited in substantial ways, but nevertheless begins to help fill gaps in current knowledge about the nature and extent of distracted walking in densely populated urban areas. The cross-sectional design precludes inferences about the stability of the estimates over time; additional studies are needed to replicate the observed estimates. The sample size was small, especially for observations pertaining to pedestrian behaviour during the ‘don't walk’ signal (and only included 4 vs 7 cycles at each corner of each intersection), which adversely affected the precision of estimates for the respective distracted walking behaviours. Studies with larger samples are needed. The intersections were purposively sampled because of the prior high frequency of pedestrian–motor vehicle injuries, and it is not possible to generalise from this sample of intersections to others. The consistency between our results and those reported by others is reassuring, but additional studies with larger and more varied samples are clearly needed. Since the unit of sampling was the intersection, we acknowledge that clustering effects are likely. Additional research is needed to identify locations where distracted walking is more and less likely to occur, and why. We did not measure the amount of distracted walking behaviour based on proportion of time or distance, so from that perspective, our rates may be overestimates. Such a measure would yield a more precise estimate of the true level of distraction, but would require videotaping. We also did not measure gender, estimated age or other demographic information of the pedestrians, which limits the descriptive findings that could be calculated. Such data collection would also require video recording. Despite these limitations, this study adds to the paucity of research on technology-related distracted walking in high-risk, high-frequency pedestrian–motor vehicle collision sites commonly found in urban centres like Manhattan.

What is already known on the subject

  • Mobile phone use while walking can result in injury or death.

  • A 2010 study reports that in New York City (NYC), there were 11 266 pedestrian injuries from motor vehicles, of which 149 resulted in fatality.

What this study adds

  • In this study, more than 1 in 4 pedestrians were distracted with their mobile device while crossing a dangerous intersection, even when crossing during the ‘don't walk’ signal.

  • The most common form of distracted walking observed was wearing headphones (∼16%), but more than 1 in 20 of the 3490 pedestrians observed crossing during the ‘walk’ signal were talking on or looking down at a mobile phone.

  • This study contributes to the emerging literature on distracted walking behaviour by pedestrians in busy urban areas and can help to inform pedestrian-focused safety efforts.

British woman receives award from RoSPA

The mother of an infant who strangled on a blind cord has received an award from the Royal Society for the Prevention of Accidents (RoSPA) in the UK for her tireless work to prevent these tragic deaths. Her efforts, with those of RoSPA have yielded a new European standard that changes the way all window furnishings across the European Union are made and sold. Editor’s comment: Long overdue but welcome nonetheless (noted by IBP).


The authors would like to thank Glen Johnson, Ph.D., Associate Professor, Department of Health Sciences at Lehman College for statistical consultation.



  • Contributors CHB, DE and CEB conceptualised the study. SR completed the data analysis. All authors contributed to writing the manuscript.

  • Competing interests None.

  • Ethics approval Teachers College, Columbia University; William Paterson University; Lehman College.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.