Relations between temperamental fear and risky pedestrian behavior

Abstract

Objective:

Child pedestrian injury poses a significant global public health challenge. This study examines relations between temperamental fear and children’s risky pedestrian behavior, utilizing mediation analytic strategies to study underlying mechanisms of the hypothesized relation.

Methods:

As part of a larger study, 240 seven- and eight-year-old children completed 30 crossings in a virtual reality (VR) pedestrian environment. Three pedestrian behaviors were considered: start gap (lag after a traffic gap appears before child initiates crossing into the gap), time to contact (TTC; gap between avatar and the lead oncoming vehicle), and hits (collisions with vehicles in simulated crossings). Temperamental fear was measured by parent report.

Results:

Fearful children were more likely to be struck by virtual vehicles, and the relation between fear and risky pedestrian behaviors was mediated by start gap and TTC. Specifically, children who were temperamentally more fearful were more likely to hesitate before initiating crossing, which led to shorter gaps between themselves and the oncoming vehicle, hence causing them to be more likely to be hit by virtual vehicles. Gender interacted with fear, such that fearful girls were most likely to be hit.

Conclusions:

Both temperamental fear and gender influenced the risk of child pedestrian injuries. Delayed entry into traffic and shorter gaps between children and oncoming vehicles may underlie this relation. Future research should explore how these factors might influence the effectiveness of prevention programs.

1. Introduction

Child pedestrian injury is a significant public health challenge both in the United States and globally. In the United States, 279 children younger than 14 years old died from pedestrian injuries in 2011, and another 33,518 children suffered non-fatal pedestrian injuries (NCIPC, 2014). Globally, it is estimated that over 30,000 child pedestrians are killed each year (Toroyan and Peden, 2007).

Multiple factors contribute to pedestrian injury risk among 7- and 8-year-olds, the target age group for this research. Environmental risk factors are prominent and include factors such as neighborhood street design and traffic engineering, the presence of sidewalks, and vehicle speed control (World Health Organization, 2013). Also relevant is the contextual environment in which many young children walk, which often involves busy and complex pedestrian environments that they negotiate without adult supervision (Macpherson et al., 1998; Martin et al., 2007; McDonald et al., 2011).

Another significant risk factor for child pedestrian safety, and the focus of the present investigation, is individual child characteristics. It has been long-recognized that children’s cognitive and perceptual development plays a significant role in pedestrian injury, with inferior cognitive-perceptual skills associated with increased risk (Schwebel et al., 2012). Similarly, children’s pedestrian safety is influenced by cognitively-influenced temperamental traits such as inhibitory control and risk-taking tendencies. Children who tend to be more inhibited and risk-averse have reduced injury risk (Barton and Schwebel, 2007; Hoffrage et al., 2003; Tabibi et al., 2012).

The role of other individual difference factors, such as emotion-based temperament, is less well understood (Barton and Schwebel, 2007; Schwebel et al., 2012). Fearful children have greater stress, react more anxiously to a range of situations, inaccurately perceive vulnerability in multiple settings, and are more often overwhelmed by new information at school (Rothbart and Jones, 1998; Talge et al., 2008). As road-users, fearful children might be affected in their ability to process information accurately and quickly compared to less fearful children. Information processing is critical to pedestrian safety, as a safe pedestrian must perceive a complex traffic environment, judge moving objects and the safety of entering the street, and time street entry so that he or she can cross the street within a traffic gap that permits sufficient time for the crossing (Schwebel et al., 2012). Reduced speed or accuracy of processing the perceived environment or in making decisions about traffic gap safety due to fearful reactions could lead to delayed entry into traffic gaps, entry into narrower gaps between oncoming traffic, and higher likelihood of injury when crossing streets.

Fear may also interact with other individual characteristics, such as gender, in influencing children’s risks of pedestrian injuries. In a study of bicycling safety, for example, Peterson et al., (1997) used a 5-point visual Likert-style scale to measure acute situational fear and found that girls reported more fear in entering simulated traffic gaps than boys, whether or not potential collisions were anticipated. Others have reported gender-related differences in street behavior. Barton and Schwebel (2007) found that girls tend to wait longer before entering a street to cross whereas boys were more likely to enter the near lane of traffic before the far lane was vacant. Stevens and colleagues (Stevens et al., 2013) found a similar pattern in a bicycling situation, where girls were more likely to wait for the lead car to completely pass the crosswalk before entering to cross, whereas boys were likely to start crossing the street while the lead car was still passing by.

Taken together, a picture emerges to suggest fear may be related to risky pedestrian behavior. Fearful children may delay entry into traffic gaps, creating a shorter gap between themselves and oncoming traffic and therefore increasing risk of pedestrian injury. This pattern may be true especially in girls, who are more likely to be fearful and also more likely to hesitate before entering traffic. Fig. 1 presents a conceptual model of our hypotheses, which we tested among 7–8 year old children who crossed streets within a virtual pedestrian environment. The squares in the figure represent variables of interest, and the arrows in the figure represent directional effects between two variables. Our hypotheses, represented conceptually in the figure, can be summarized as follows:

Fig. 1. Conceptual Model.

Note: TTC: time to contact.

1. Fearful children will be more likely to demonstrate risky pedestrian behaviors, as assessed by collisions with vehicles in the virtual pedestrian environment (hits), compared to less fearful children.

2. The relation between fear and risky pedestrian behavior will be explained partially by a delayed entry into traffic gaps (longer start gaps), and subsequently shorter gaps between the pedestrian and oncoming traffic (shorter TTC). Fearful children will have larger delays entering traffic, leading to shorter latencies between themselves and oncoming traffic while crossing the street and hence more risky pedestrian behaviors.

3. The relationship between fear and risky pedestrian behavior will differ by gender (interaction). Girls will wait longer before crossing streets than boys. Girls also will have higher levels of fear than boys.

4. Together, gender and fear will interact to amplify the relations between female gender and pedestrian safety and between fear and pedestrian safety. That is, the combination of fearfulness and female gender will create particularly elevated risk of pedestrian injury, as mediated by delayed entry into traffic and shorter times to contact. Low fear and male gender will be associated with reduced risk.

2. Methods

2.1. Participants

A total of 240 children ages 7–8 (M = 7.98 years, SD = .65) were recruited from community sources in the Birmingham, Alabama area, as part of a larger study (Schwebel et al., 2014). The sample was 43% male, 52% Caucasian, and 42% African–American.

2.2. Measures

2.2.1. Fear

Temperamental fear was measured with the parent-report Child Behavior Questionnaire (CBQ; Rothbart et al., 2001), a standard measure of child temperament. The fear subscale is comprised of 12 items (see Table 1) with adequate internal reliability (Cronbach’s alpha = .70; Rothbart et al., 2001) defined as “the amount of negative affectivity, including unease, worry or nervousness related to anticipated pain or distress and/or potentially threatening situations” (Rothbart et al., 2001), fear was assessed as the average of those 12 items, each answered by parents on a 7-point scale with higher scores indicating higher levels of fear.

Table 1.

List of fear items from CBQ (7-point scale, Rothbart et al., 2001).

Items

Is not afraid of large dogs and/or other animals. (R) Is afraid of burglars or the “boogie man." Is afraid of loud noises. Does not worry about injections by the doctor. (R) Is not afraid of the dark. (R) Is afraid of fire. Is very frightened by nightmares. Is afraid of the dark. Is rarely frightened by “monsters” seen on TV or at movies. (R) Is not afraid of heights. (R) Is rarely afraid of sleeping alone in a room. (R) Gets nervous about going to the dentist.

Note: (R) signifies the item was reversed before computing mean.

2.2.2. The virtual reality pedestrian environment

The virtual reality (VR) pedestrian environment used in this study, including hardware and software specifications, is detailed elsewhere (Schwebel et al., 2008). Briefly, the environment replicates an actual crosswalk near a local Birmingham, Alabama school. The crosswalk is mid-block and crosses a two-lane bi-directional road. Upon entering the VR, children stand atop a wooden curb and view the street environment on three monitors situated at eye-level in front of them. From a semi-immersive perspective, they view traffic moving bi-directionally and are instructed to step down off the curb when they deem it safe to cross. Children could wait as long as they wished to step down, and the average waiting time was 18.52 s (SD = 24.06 s). Traffic appeared at random intervals (using normal curve) following researcher-specified speeds and densities (see below). Vehicles that appeared were randomized and included over 10 types of vehicles (e.g., cars, minivans, trucks, school buses, fire engines). They appeared at a frequency representing typical vehicles on the actual road. Upon stepping down off the curb, children triggered the system to initiate a race- and gender-matched avatar to cross the simulated street, such that the environment switched from first person to third person and allowed children to learn whether their crossing was safe or not. The avatar walked at the child’s typical walking speed, as assessed previously in a different room as the average across five trials. Following a successful cross, children saw a cartoon character that offered positive feedback about the safe crossing. Following a close call (avatar was within 1 second of being struck by a simulated vehicle), children saw the cartoon character offer cautionary feedback. Following a collision with an oncoming vehicle, the simulator froze upon impact and the cartoon offered cautionary feedback.

The virtual environment includes ambient and traffic noise and was validated in a trial demonstrating behavior in the virtual world matched behavior in the actual street environment, both among children and adults (Schwebel et al., 2008). A short video demonstrating how children interacted with the virtual environment is available at: http://www.uab.edu/cas/psychology/psy-fac/33-primary-faculty/66-dr-david-c-schwebel.

For this study, children crossed the virtual street 30 times, 10 at each of three “difficulty” levels: 25 miles per hour (mph) traffic and light volume (8 vehicles/min); 30 mph traffic and moderate volume (12 vehicles/min); and 35 mph traffic and heavy volume (16 vehicles/min), in a randomized order. Prior to virtual reality assessment trials, children were oriented to the VR by completing 8 practice trials and received standardized instructions to cross when they perceived the virtual street environment to be safe.

2.2.3. Pedestrian safety measures

Three measures related to pedestrian safety were considered in this study: hits, start gap, and time to contact. All three variables are computed electronically by the virtual reality software.

The outcome for this study was risky pedestrian behavior, measured by hits. Very simply, hits reflect a count of crossings that were highly risky, in which the child would have been hit by a vehicle had it been an actual rather than a simulated crossing. Children crossed the virtual street for 3 trials with 10 times per trial, so results are expressed as the total count of hits across the 30 crossings.

Start gap and time to contact (TTC) were included as potential mediators. Start gap, defined as the temporal lag before initiation of crossing into a traffic gap, is considered an indicator of children's efficiency of cognitive processing in pedestrian situations (Barton, 2006; Plumert et al., 2004). It was measured only for the traffic gap that children entered, beginning when the last vehicle left the crosswalk and ending when the child's foot hit the platform on the lower curb. Inexperienced child pedestrians do not begin to process the safety of a traffic gap until that gap appears, causing a significant (e.g., 500–2000 ms) temporal delay before children enter a safe gap. Time to contact (TTC) was measured as the shortest time between the avatar and any approaching vehicle during the full crossing. Both start gap and TTC were computed as the average, in seconds, across children's 30 crossings.

2.2.4. Other measures

Basic demographic information was reported by parents. Ethnicity was coded as Caucasian or non-Caucasian. Body Mass Index (BMI) was measured as the child's body mass divided by the square of his/her height. Socioeconomic status (SES), as an important factor influencing child pedestrian injury risks (Graham et al., 2005), was computed as a standardized composite score of child’s parents’ education and family income. A verbal intelligence screening of children was conducted using the Peabody Picture Vocabulary Test-IV (PPVT-4; Dunn and Dunn, 2007). Normed on a large national sample, the PPVT-4 has excellent internal and test-retest reliability, and converges well with other measures of verbal intelligence (Dunn and Dunn, 2007; Pullen et al., 2010). These variables served as covariates in data analyses.

2.3. Data analysis

Data analysis was conducted in 4 steps. First, descriptive data were examined for all variables. Second, bivariate correlation analysis was performed to examine the intercorrelations between variables of interest. Third, for the full sample as well as stratified by gender, hierarchical regression analyses evaluated whether fear directly predicts risky pedestrian behavior (as measured by the number of hits in the virtual environment). Fourth, bias-corrected bootstrapping mediation analysis (MacKinnon et al., 2004; Preacher and Hayes, 2004, 2008) was performed to test the hypothesized mediational effect of start gap and TTC between fear and risky pedestrian behavior, with fear as the predictor and risky pedestrian behavior (hits) as the outcome variable. Age BMI, verbal intelligence, SES and ethnicity were controlled for; this analysis also was conducted for the full sample and stratified by gender.

3. Results

Descriptive data are presented in Table 2. Children on average were hit by virtual cars 4.50 (SD = 3.53) times during the 30 crossings, suggesting pedestrian injuries occurred about 15% of the time for our young sample. The average start gap was 1.31 (SD = 0.47) s, indicating children spent a little over a second on average processing street-crossing information before making the decision to cross within a particular gap. During crossings, the average shortest TTC with oncoming virtual vehicles was 3.14 s (SD = 0.95).

Table 2.

Descriptive statistics and Pearson correlations between variables of interest (N = 240).

	M (SD) or %	1	2	3	4	5	6	7	8	9	10
1. Fear	3.91 (1.01)	–	.17*	.14*	−.19**	.02	.20**	−.06	−.15*	−.02	−.05
2. Hits	4.50 (3.53)		–	.20**	−.73**	−.21**	.08	−.13*	−.15*	.11	−.16*
3. Start gap	1.31 (.47)			–	−.24**	−.04	.29**	−.08	−.12	−.06	−.12
4. TTC	3.14 (.95)				–	.16*	−.12	.05	.05	−.11	.06
5. Age (in years)	7.98 (.65)					–	.10	.01	.05	.07	.07
6. Gendera	43.3% boys						–	−.11	−.07	−.01	−.07
7. Ethnicityb	51.7% Caucasians							–	.65**	−.31**	.57**
8. SES	41.1% < $60,000								–	−.27**	.53**
9. BMI	18.23 (3.76)									–	−.17*
10. Intelligence	105.43 (17.32)										–

^aBoys were coded as 1 and girls were coded as 2.

^bCaucasians were coded as 1 and all others were coded as 0.

^*p < 0.05.

^**p < 0.01.

We next considered bivariate correlations between relevant variables. As shown in Table 2, fear correlated significantly with hits by virtual vehicles (r = .17, p < .05), start gap (r = .14, p < .05), and TTC (r = −.19, p < .01). We also found, as hypothesized, that gender correlated with fear (r = .20, p < .01); girls (M = 4.09, SD = 1.03) were rated by their parents to be more fearful than boys (M = 3.68, SD = .94). Gender also correlated significantly with start gap (r = .29, p < .01), with girls (M = 1.42, SD = .51) having longer delays entering traffic than boys (M = 1.15, SD = .36), but gender did not correlate significantly with hits or with TTC.

Next, we evaluated our primary hypothesis, that the association between fear and risky pedestrian behaviors is mediated sequentially by start gap and time to contact (TTC) both among the full sample and stratified by gender (see Figs. 2-4). The first step was to test the total effect between fear and risky pedestrian behaviors by conducting hierarchical linear regression analyses while controlling for potential covariates (age, BMI, IQ, SES and ethnicity; see Table 3). Next, the bootstrapping method with 95% bias-corrected confidence estimates and 10,000 bootstrap resamples (MacKinnon et al., 2004; Preacher and Hayes, 2004, 2008) was applied to test each mediation model, controlling for age, BMI, verbal intelligence, SES and ethnicity.

Fig. 2. Mediation Model 1: overall sample.

Note: Age Body Mass Index (BMI), verbal intelligence, Socioeconomic Status (SES) and ethnicity were controlled for in this analysis. TTC: time to contact.

Fig. 4. Mediation Model 3: boys only.

Note: Age Body Mass Index (BMI), verbal intelligence, Socioeconomic Status (SES) and ethnicity were controlled for in this analysis. TTC: time to contact.

Table 3.

Summary of hierarchical linear regression analyses of the effect of fear on the number of hits.^a

	Model 1. Full sample				Model 2. Girls only				Model 3. Boys only
	Variable	B	SE B	β b	Variable	B	SE B	β b	Variable	B	SE B	β b
Step 1	Age	−1.06**	.41	−.19	Age	−1.32*	.55	−.23	Age	−.91	.62	−.16
	Ethnicity	−.51	.76	−.07	Ethnicity	.26	1.05	.04	Ethnicity	−1.42	1.12	−.19
	SES	−.06	.41	−.01	SES	−.79	.55	−.19	SES	.89	.63	.19
	BMI	.06	.07	.06	BMI	.08	.11	.08	BMI	.06	.10	.07
	PPVT	−.03	.02	−.11	PPVT	.00	.03	.00	PPVT	−.05	.03	−.23
Step 2	Age	−1.10**	.40	−.19	Age	−.1.38*	.54	−.24	Age	−.91	.62	−.16
	Ethnicity	−.56	.75	−.08	Ethnicity	.05	1.03	.01	Ethnicity	−1.43	1.12	−.19
	SES	.07	.41	.02	SES	−.64	.54	−.16	SES	.97	.64	.21
	BMI	.07	.07	.07	BMI	.10	.11	.10	BMI	.06	.10	.06
	PPVT	−.02	.02	−.10	PPVT	.01	.03	.02	PPVT	−.05	.03	−.23
	Fear	.62*	.25	.18	Fear	.76*	.32	.22	Fear	.36	.42	.09

^aSES: Socioeconomic Status; BMI: Body Mass Index; PPVT: Peabody Picture Vocabulary Test, a verbal intelligence screen. Ethnicity coded as 1 for Caucasian and 0 for non-Caucasian.

^bβ is the standardized regression coefficient B.

^*p<.05.

^**p<.01.

Model 1 included the full sample of both boys and girls. Hierarchical linear regression analysis indicated that all predictors accounted for 11% of variance (R² = .11, F (6, 184) = 3.72, p < .01). As expected, the association between fear and risky pedestrian behaviors was statistically significant after all covariates were controlled for (B = .62, t = 2.50, p < .05). A two-mediator model was then constructed and tested (see Fig. 2). The mediational role of TTC was confirmed (B = .46, 95% CI [.11, .81]), with fearful children having shorter times to contact (B = −.18, t = −2.54, p < .05), which in turn led to more collisions with virtual vehicles (B = −2.59, t = −13.00, p < .001). Further, the mediational role of start gap and TTC together (in serial) was confirmed (B = .10, 95% CI [.02, .24]), suggesting the relation between fear and TTC is partially explained by start gap, a proxy for cognitive processing while deciding to cross the street. Fearful children had longer start gaps (B = .08, t = 2.29, p < .05). This delay in entering a traffic gap placed children closer to the oncoming traffic, creating a shorter TTC) (B = −.44, t = −3.10, p < .01). This smaller time window between children and oncoming traffic (lower TTC) subsequently led to more collisions with virtual vehicles (B = −2.59, t = −13.00, p < .001).

In Models 2 and 3, we stratified by gender and computed similar models for girls and for boys. For Model 2 with only girls, all predictors accounted for 14% of variance (R² = .14, F (6, 101) = 2.70, p < .05) and the association between fear and collisions with vehicles was statistically significant after controlling for covariates (B = .76, t = 2.36, p < .05). As shown in Figure 3, the mediational role of TTC was confirmed among the girls (B = .54, 95% CI [.09, 1.02]). Fearful girls had shorter TTC (B = −.21, t = −2.33, p < .05), which led to more hits in the simulated street environment (B = −2.58, t = −9.49, p < .001). However, the mediational role of start gap and TTC together (in serial) was not confirmed (B = .09, 95% CI [−.01, .30]), suggesting the relation between fear and TTC for girls may not be caused by differences in start gap.

Fig. 3. Mediation Model 2: girls only.

Note: Age Body Mass Index (BMI), verbal intelligence, Socioeconomic Status (SES) and ethnicity were controlled for in this analysis. TTC: time to contact.

Among boys, all predictors accounted for 14% of variance (R² = .14, F (6,76) = 2.04, p = .07). Unlike for girls, fear did not emerge as a significant predictor of hits for boys in the model (B = .36, t = .85 ns). Since the total effect from predictor (X) and outcome variable (Y) is influenced by a mixture of direct and indirect effects, a significant total effect is not required to demonstrate a mediation effect (e.g., Kenny and Judd, 2014; Zhao et al., 2010), thus we continued with the planned mediation analyses (see Fig. 4). The mediational role of TTC was not confirmed among the boys (B = .49, 95% CI [−.03, 1.13]). However, the mediational role was confirmed for start gap and TTC together (in serial) (B = −.21, 95% CI [−.59, −.04]), although in this case, after removing girls from the model, fearful boys surprisingly showed shorter start gaps (B = −.10, t = −2.11, p < .05). This shortened start gap put children further away from the oncoming traffic (longer TTC) (B = −.88, t = −2.87, p < .01). The longer time window between boys and oncoming traffic led to fewer collisions (B = −2.53, t = −8.02, p < .001).

4. Discussion

Parent-reported temperamental fear predicted 7–8 year-old children’s risky pedestrian behaviors in a computerized virtual reality environment. Fearful children were more likely to demonstrate risky pedestrian behaviors (getting hit by virtual cars) than less fearful children and this association was maintained after controlling for ethnicity, socioeconomic status, BMI and verbal intelligence. Increased negative affectivity and anticipated distress/pain might make children more likely to take risks when crossing a street. This result contradicts previous findings that more fearful children may act in more conservative rather than risky ways (Rudasill et al., 2010), perhaps because pedestrian settings represent a contextual environment with a large amount of stimuli that require rapid and accurate processing. In such situations, fearful children may be overwhelmed and thus incorrectly perceive high vulnerability in potentially risky situations (Rothbart and Jones, 1998; Talge et al., 2008).

The mediation analyses offer further evidence concerning possible mechanisms underlying the relation between fear and risky pedestrian behavior. Delayed entry into a traffic gap and shorter gaps between the child pedestrian and oncoming traffic partially explain why fearful children were struck more often by virtual vehicles. As hypothesized, children who were temperamentally more fearful had larger delays entering traffic in the VR environment. This may be because fearful children have an increased cognitive load emerging from their fearful cognitions and therefore need more time to process traffic information before making decision to cross a street. Our findings indicate that this delayed entry by fearful children led to shorter latencies between children and oncoming cars in the middle of the street, and hence they were more likely to be hit than fearless children.

The finding that fearful boys and fearful girls showed different behavioral patterns when crossing streets is noteworthy. Some previous research reports gender differences in street-crossing, with girls more likely to wait longer than boys before starting to cross the street (Barton and Schwebel, 2007; Stevens et al., 2013). The current study extends those findings and supports the hypothesis that the combination of female gender and temperamental fear could lead to particularly increased risk of pedestrian injury. Among girls, those who were more fearful tended to get hit more by virtual cars than less fearful girls, but no similar pattern emerged among boys.

Among boys, temperamental fear was negatively correlated with start gap, suggesting fearful boys may ‘think’ less and make decisions quickly before entering into traffic than less fearful boys. This result confirms previous findings that boys are generally greater risk-takers than girls (Barton and Schwebel, 2007) and may reflect gender-based cultural or societal expectations that pervade even to boys with fearful tendencies (Morrongiello and Hogg, 2004). The non-significant association between fear and risky pedestrian behaviors among boys also suggests there may be other confounding factors that were not considered in the current model. This non-significant association also reduced the power of the fearfulness to a modest level in explaining the overall variation among all children in safe crossing behaviors. Further investigation is warranted.

It is important to recognize that our results had modest effect sizes overall (.09–.23), and that fear is just one of many psychosocial, behavioral, environmental and other factors that influence a child's safety in pedestrian settings. In fact, a child’s temperamental fear is unlikely to play a highly significant role in most pedestrian injuries, but it may play a small and meaningful role in combination with other contextual factors.

Findings from this study have several implications for developing pedestrian safety intervention and prevention programs. First, replicating results from laboratory-based experiments (Morrongiello and Matheis, 2007), the findings suggest emotion-based temperament traits such as fear play a role in the likelihood of children taking risks as pedestrians. Pedestrian safety interventionists should consider the role of children's emotions in the design of prevention programs, and both the performance-degrading effects and the protective quality of fearfulness should be taken into consideration. Second, the role of gender must be considered. Emotion-based fear was more relevant to girls’ pedestrian safety than boys’. Intervention programs may need to be gender-specific to achieve the greatest effect. Finally, start gap and time to contact underlay the relation between fear and risky pedestrian behavior. Reducing delays in entering a gap seem particularly relevant to training programs, perhaps through strategies such as repeated practice at perceiving and choosing safe gaps quickly. This might be a particularly helpful strategy for fearful children, who could overcome fearful tendencies through repeated practice and success in real or simulated pedestrian environments.

This study adds to the existing literature in two main ways. First, the use of a virtual reality environment provided computer-level accuracy in measuring behavioral outcomes by our child pedestrians in a validated simulated environment. Second, the combination of bootstrapping mediation analysis technique and a relatively large sample offered sufficient power to study underlying mechanisms between children's temperamental fear and risky pedestrian behaviors.

Like all research, this study also had limitations. One limitation is the reliance on parent report to assess temperament. Temperament theorists recommend supplementing parent report with behavioral measures or reports from other informants (Seifer et al., 1994, 2004), but such measures were unavailable in this dataset. A second limitation lies in the nature of virtual reality environment. Although Schwebel et al. (2008) validated the virtual reality task with samples of both children and adults, behavior in simulated environments may never match behavior in real-world environments perfectly. Children may have recognized there were no actual risks in the VR environment and therefore behaved in riskier ways than they would have in an actual pedestrian environment. Further, the nature of this VR is such that children cannot adjust their walking speed once it is established, thus restricting the ability of children to “evade” oncoming traffic by quickening their pace mid-crosswalk. These two factors may explain the comparatively high rate of hits compared to what we might expect based on epidemiological evidence. A third limitation is the fact that the virtual environment offers only mid-block unsignalled crossings. Although mid-block crossing ranks among the highest risk of child pedestrian injuries (Malek et al., 1990), children's risky pedestrian behaviors at signaled intersections or in other traffic situations were not addressed. In conclusion, both temperamental fear and gender influenced the risk of child pedestrian injuries. Delayed entry into traffic and shorter gaps between children and oncoming vehicles may underlie this relation. Future research should explore how these factors might influence the effectiveness of prevention programs.