Objective The aim of this study was to compare parents’ expectations for their children crossing streets with children's actual crossing behaviours and determine how accurately parents judge their own children's pedestrian behaviours to be.
Method Using a fully immersive virtual reality system interfaced with a 3D movement measurement system, younger (7–9 years) and older (10–12 years) children's crossing behaviours were assessed. The parent viewed the same traffic conditions and indicated if their child would cross and how successful she/he expected the child would be when doing so.
Results Comparing children's performance with what their parents expected they would do revealed that parents significantly overestimated the inter-vehicle gap threshold of their children, erroneously assuming that children would show safer pedestrian behaviours and select larger inter-vehicle gaps to cross into than they actually did; there were no effects of child age or sex. Child and parent scores were not correlated and a logistic regression indicated these were independent of one another.
Conclusions Parents were not accurate in estimating the traffic conditions under which their children would try and cross the street. If parents are not adequately supervising when children cross streets, they may be placing their children at risk of pedestrian injury because they are assuming their children will select larger (safer) inter-vehicle gaps when crossing than children actually do.
Statistics from Altmetric.com
Unintentional injuries are the leading cause of preventable child deaths in most developed countries.1 Although many causes of childhood injuries have been declining in recent years, pedestrian injuries remain an issue worldwide, killing over 30 000 children annually.2 Children in the age range of 5–9 years account for a disproportionate number of pedestrian injuries and are at an increased risk of injury because of their small stature and underdeveloped cognitive-perceptual capabilities.3–5 Although environmental factors, such as vehicle speed and traffic volume, influence children's risk of pedestrian injury,6 ,7 how children cross streets can also increase this risk.8 ,9 Generally, studies have shown that young children choose poor locations for crossing, misjudge the time needed to cross and select inter-vehicle gaps that result in cars coming dangerously close to them.10–12 It is essential, therefore, that caregivers accurately appraise their child's crossing behaviours, so that they can judge their readiness for safely crossing streets when unsupervised. The current study considered this issue and addressed two questions: how accurately parents judge the traffic conditions in which their child would select to cross and how safely they believe their child would do so.
There is very little research on this topic. One study presented photos of different traffic conditions and asked questions from which examiners could infer crossing behaviours (eg, would the child cross between parked cars). Children consistently endorsed crossing in riskier locations than their parents expected of them.13 A limitation of the study, however, is that children and parents were reacting to photos rather than dynamic traffic conditions that present a realistic experience (eg, volume, speed) and require in-the-moment decisions about crossing.
In the current study, children were tested within a fully immersive virtual reality (VR) pedestrian environment and they had to decide what gap sizes (inter-vehicle times) to cross in. The parents were presented videos, showing exactly the traffic conditions their child saw and asked to judge whether or not the child would cross and how successful they expected their child to be. Comparing the expectations of the parent with the actual crossing behaviours of their child provided information as to what extent parents accurately judged their children's crossing behaviours.
Study design and power calculation
A cross-sectional design was used, with respondent (child, parent) as a between-participant factor. For every child who participated, one parent did also so that parent judgements about children's pedestrian behaviours could be directly compared with their children's actual crossing behaviours in the virtual environment.
Based on past research, we anticipated obtaining at least medium effect sizes (eg, of 0.20–0.25) when conducting analysis of variance (ANOVA) tests comparing means across groups.14 With α level set at 0.05 and power set at 0.80, this resulted in a minimal overall sample size of 65.
The sample comprised 139 children who were in one of two age groups: young or 7–9 years (N=74, 49% male, M=8.46, SD=0.88 years) and old or 10–12 years (N=65, 55% male, M=11.68, SD=0.90 years). Children were recruited throughout the community; Guelph is a suburban setting, 45 min from Toronto, with a population of approximately 122 000, based on 2011 census data. All children were developing normally (as reported by parent) and no participant had any immediate family member who had ever been injured by a car as a pedestrian. The child sample comprised predominantly middle-upper income families (78% earned above $80 000), with 63% of parents having some/completed college/university. Nearly all of the participants were Caucasian (97%). All measures and procedures were reviewed and approved by the University Research Ethics Board, and all participants gave independent written consent before testing began.
The simulator sickness questionnaire was completed prior to booking an appointment to ensure that the participant was not at an increased risk for sickness while wearing the VR headset. The questionnaire assessed the person's history of migraine headaches, claustrophobia, motion sickness and dizziness/nausea,15 and anyone reporting these symptoms was excluded from the sample (N=2).
Virtual reality and movement tracking system
The system was constructed in an 8 m×5 m room using an 8-camera optical-motion tracking system (PPTH by Worldviz) to feed position data to specialised software (Vizard), using a high-level scripting language (Python) to accomplish many low-level graphics and hardware interfacing actions. Participants viewed the virtual environment through a Virtual Research Systems 1280×1024 resolution stereoscopic head-mounted display (HMD). Mounted on the HMD is an Inertia Cube 3, which is a 3 degrees of freedom (ie, X,Y and Z coordinates) orientation-tracking system that uses accelerometers, gyroscopic and magnetic sensors to track the orientation data of a participant's head such that changes in head orientation change the participant's view of the virtual environment virtually instantaneously. All movement and orientation data are captured at a rate of 60 times per second. The virtual environment is a two-lane street with sidewalks. Traffic approached in the closest lane and from the child's left.
The virtual environment's realism is enhanced visually by trees, shadows and textures, and aurally by realistic sounds of traffic movement (eg, engine sound becoming louder as cars get closer). Participants control the direction they walk, their speed of movement, and if they make a poor crossing decision they have the ability to step back to the curb or speed up to evade the approaching vehicle. A screen shot showing a scene presented to the child is shown in figure 1; the actual display was in colour.
Participants were tested at a laboratory on campus. Two trained research assistants were involved: one overseeing the operation of the computer that controlled the VR equipment and instructing the child participant on the VR street crossing trials, and the other remaining in the test room and available to assist if needed during completion of the trials. The parent watched the traffic presented to the child on a monitor in a nearby room and, after a given trial, made their judgement about how their child would behave; the parent saw the actual scene that was presented to the child in their 3D goggles, but in a non-3D format.
Each child participant completed two phases. In phase 1 (VR familiarisation), she/he was introduced to the virtual environment. First, a researcher demonstrated how to cross the street while wearing the headset as the child watched a computer monitor; this included demonstrating what would happen if one was to be hit by a car (ie, all vehicles disappear and a siren plays). Then the child was fitted with the VR headset. To ensure that the headset was fitted properly and the child could see clearly, a test screen appeared with letters and she/he had to correctly identify the letters before she/he was shown the street environment; knob adjustments allowed the viewer to adjust the headset to improve fit and make the letters clearly visible. Subsequently, the child was positioned to stand on the curb facing a two-way street (ie, one-lane each way); to prevent possible tripping hazards all curbs were visually presented but the child did not have to step down/up on these. The child was instructed to walk across with no traffic, turn around once they crossed near lanes, and then walk back to where they started from; the street was the actual size of a typical two-lane road. They proceeded to repeat this process 10 times with no cars appearing. This gave the child time to adjust to the VR environment, experience walking with the headset on and ask any questions and make any adjustments before traffic was presented. Note that previous research has shown that by the end of this phase individuals are fully accustomed to the VR equipment,16 and our own pilot data confirmed that walking speed stabilised (ie, the data reached asymptote) by 10 trials (ie, length of this phase).
In phase 2 (test trials), the child was introduced to traffic and had to monitor the traffic flow and walk across when she/he deemed it safe to do so. For all trials, all vehicles travelled at 50 km/h, which is a moderate speed that is associated with increased risk of pedestrian injury for children.6 ,7 Children were presented 15 trials in a random order, including three trials for each of five inter-vehicle gaps (2, 3, 4, 5, 6 s) and their gap threshold (ie, averaged gap sizes they crossed into) was determined. A similar gap threshold was computed for the parent based on their judgement of whether their child would try to cross in each gap size (yes, no). In addition, to compare parent judgements of whether or not the child would be safe with children's performance, we designated a ‘safe’ crossing as the child not trying to cross or crossing but having >1 s time left to spare,11 and a ‘risky’ crossing as one on which the child was hit or the car passed within 1 s gap from the child, and we then compared the proportion of children's and parents’ trials that met these criteria; on each trial that the parent judged the child would try to cross, they had to indicate what the outcome would be (make it without any concerns with the car beyond a 1 s gap from the child; have a close call as indicated by the car being within 1-s gap from the child; would likely be hit).
Descriptive and parametric statistics (ANOVA, logistic regression) were applied to characterise the data and compare the results from the parent with their child. Several preliminary data checking procedures were applied and addressed before analyses were completed,17 including examining variables for violations in normality, assessing for multivariate outliers based on Cook's distance and assessing for violations of sphericity for within-participant effects to determine if adjustment to the degrees of freedom (Greenhouse–Geisser correction) was warranted.
Analyses proceeded in two stages: the first focused on the yes/no responses of the parents (ie, would the child try to cross in a given gap) and the second on the extent of success they expected for the gap conditions their child would try to cross into (‘safe’ or ‘risky’=child is hit or has a close call of the car passing within 1 s gap from the child).
Were parents accurate in predicting the smallest gap size their child would try to cross in?
We considered parent and child reactions to each gap size presented and identified the smallest gap size that parents expected their children would cross in and compared this with children's gap threshold based on their actual crossing behaviour. An ANOVA was conducted on these scores with respondent (2: child, parent), child age group (2: young, old) and child gender (2: boy, girl) as between-participant factors. As shown in table 1, there were no main or interaction effects involving child age or sex; however, parents’ estimates of their children's gap threshold were significantly larger than children's actual gap threshold, F(1, 107)=76.92, p<0.001, =0.42. Parents expected that children would not try to cross into inter-vehicle gaps less than about 4 s, whereas children crossed into much tighter gaps, averaging about 3 s. Moreover, the parent and child scores were not significantly correlated; r(111)=0.08, p=0.20. Thus, parents were not accurate in judging the most risky traffic condition for which their children would try to cross, assuming that children would select a larger (safer) minimum gap size to cross into than they actually did.
Were parents accurate in judging the outcome of their child trying to cross?
Overall, children were hit on about 6% of trials when they tried to cross, and an ANOVA revealed no significant variation due to, child age or sex. To determine if we could use children's performance to predict parent judgements, we limited our focus to the 3 s gap condition because this fell between the child and parent gap threshold averages and it provided sufficient variance in the data to be analysed (ie, larger gap conditions most children did and smaller gap conditions most parents judged their children would not ever try). A logistic regression was then conducted to test if we could predict the probability that the parent's judgement would fall into the ‘safe’ (ie, child did not cross or did so safely with >1 s gap between the child and the car) versus ‘risky’ (ie, car passes within a gap of 1 s from the child or the child is hit) category based on their child's actual performance, with child age and sex controlled for in Block 1 of the regression. The child's performance was calculated as the proportion of trials on which they met safety criteria, with these scores standardised for ease of interpreting the regressions results (ie, the OR for a 1 unit change in the independent variable represented a 1 SD change). Block 2 of the regression was not statistically significant, χ2(1)=0.252, p=0.62, indicating that a 1 SD increase in the proportion of trials on which children crossed safely did not increase the odds that their parent indicated they would cross safely; Block 1 was also not statistically significant, confirming that child age and sex did not differentially impact parent judgements. Thus, the level of safety in children's crossing behaviour did not differentially predict parents’ judgements of whether or not their children would cross safely. Only 13% of parents were correct in their predictions.
Young children are at elevated risk of experiencing pedestrian injuries and past research demonstrates that this risk arises, at least in part, because of their crossing behaviours.8 ,9 Selecting too small a gap to try and cross into is one such risk behaviour that young children often display.11 ,18 ,19 The current findings indicate that parents are surprisingly inaccurate in judging the inter-vehicle gaps that their children would chose to cross into. Specifically, children selected gap sizes that posed greater risk of pedestrian injury than the gap sizes into which parents expected they would cross.
One way that parents can manage children's risk of pedestrian injury is by closely supervising their children's crossing of streets until they are certain that the child can manage this risk effectively without direct supervision. If parents are overestimating their child's cautiousness in crossing streets, then this may result in their granting the child greater autonomy and, therefore, producing the undesirable effect of compromising the child's safety as a pedestrian. Thus, efforts to reduce child pedestrian injuries may need to be broad and focus on children's skill level and understanding of traffic risks (eg, judging speed, inter-vehicle distance) as well as promoting caregiver supervision and improving the accuracy of parental expectations about their children's crossing behaviours.
Providing parents with guidelines for how to accurately appraise their child's readiness for crossing independently may foster more accurate judgements about their child's capabilities. Therefore, rather than advising parents of an age at which children can cross independently (ie, 10 years seems a popular one), educating parents about the skill set needed to cross safely would allow them to adjust for individual differences in children's readiness to cross unsupervised (eg, Does your child always look left, right, left before crossing? Does your child carefully monitor traffic as she/he crosses? Can your child ignore distractions when crossing?). In addition, raising parents’ awareness of their judgement biases and the inappropriateness of their expectations may evoke greater supervision of children, particularly if they agree that crossing streets constitutes a high injury-risk context for their children.20 Given the positive impact that supervision can have to reduce child injury risk,21–23 finding ways to convince parents of the need to supervise in pedestrian contexts would seem essential, particularly because this would also create opportunities for parents to observe and appraise their children's abilities to safely cross in traffic. Finally, past research has found that when parents are crossing streets with their children they do not typically seize opportunities to teach children about pedestrian safety.24 Determining ways to evoke more teaching when the opportunity arises also would be an important goal to assimilate in future interventions aimed at reducing child pedestrian injury risk. To the extent that experience crossing in traffic can serve to tutor the visual system about important safety parameters (eg, vehicle distance and speed relations) and create important opportunities for practicing the coordination of self-movements relative to vehicle movements, such experiences may be essential for evoking developmental advancements in the capacity to cross streets safety.25 Nonetheless, providing children these perceptual-motor experiences needs to be done in a way that ensures their safety as they develop the skills necessary for safely crossing streets in varying traffic conditions.
In addition to behaviour-directed interventions, incorporating built-environment changes that consider pedestrian safety in designing roads also is essential to create a comprehensive approach to pedestrian injury prevention. For example, traffic-calming solutions (eg, speed bumps) are effective to reduce vehicle speed, and speed is a significant risk factor for pedestrian injury.26 There are also a number of design features of roads that impact pedestrian safety (eg, intersections vs roundabouts, four-lane rather than three-lane roads, whether or not paved shoulders are provided)27 and that could be considered when designing or modifying roads.
Limitations and future research directions
Although the present research provides important insights into the accuracy of parents’ judgements of their children's crossing behaviours, there are some limitations that must be acknowledged and that may inform future studies on this topic. First, the sample tested was fairly homogeneous. Extending the research by recruiting families with more diverse income and ethnic backgrounds would enhance generalisability. Related to this, it could be that parent and child judgements would vary depending on urban and suburban traffic experiences, so considering this variable in future research might prove important. Second, although using a virtual street environment yields many benefits for studying pedestrian crossing behaviours, it is possible that children exhibited riskier behaviours with virtual traffic than they might in real traffic environments. Importantly, Schwebel and colleagues have shown that VR technology has construct, convergent and face validity,28 and this method has been rated highly for scientific merit.29 Nonetheless, in future research it may prove informative to reward children for successful crossings and penalise them for failing to do so, in order to explore if such differential consequences affect children's street crossing behaviours. Finally, expanding to present a broader range of traffic situations, both easier and harder, would provide insight into whether the extent of inaccuracy in parent judgement reflects a consistently applied bias or one that depends on the demands of the traffic conditions.
What is already known on the subject?
Young children are at high risk for pedestrian injury.
How children cross streets contributes to this risk.
Few studies have evaluated how accurately parents judge their children's crossing behaviours.
What this study adds?
Parents were highly inaccurate in judging the traffic conditions their children would cross in.
Children selected inter-vehicle gaps that were significantly more risky than what parents expected.
Parents may be placing their children at risk by assuming greater skill and cautiousness than children show while crossing streets.
Rediscovery is not news!
While preparing this issue's News and Notes, I found recent coverage of several studies that appear to restate what is already well known. Perhaps we do need to be reminded that seat belts are effective in preventing injuries or that child car seats are often not properly installed. But neither finding could be regarded as novel. I include the report from Qatar that reiterates the importance of seat belts; RoSPA's reminder that cleaning products should be locked away from children; child accident prevention trust's (CAPT) message regarding the dangers of liquitabs and button cell batteries. I also found reports highlighting dangers of railways, ingested hand sanitisers and carbon monoxide. (Noted by IBP)
The authors thank Dr Carolyn McGregor for advice and the team of extraordinary programmers, whose talents and tenaciousness made our dream of this system a reality, including Robin Vierich, Jonathon Beer and Tom Hall. Thanks also to the children for their enthusiastic participation, and to Jessica Switzer and Sarah Pyne for assistance with pilot data collection. Reprint requests can be sent to the first author at firstname.lastname@example.org
Contributors BAM conceptualised the study, developed the methodology and analysis plan, and prepared the manuscript for publication. MC collaborated on study design and methodology, supervised implementation of the study and conducted the analyses.
Funding This work was supported by grants from the Canadian Institutes for Health Research. The first author was supported by a Canada Research Chair award and this research was supported by grants from the Canadian Institutes for Health Research and the Canadian Foundation for Innovation Leaders Opportunity Fund.
Competing interests None.
Patient consent Obtained.
Ethics approval Research Ethics Board at the University of Guelph approved of the project.
Provenance and peer review Not commissioned; externally peer reviewed.
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.