- Email: [email protected]
Driving strategies among younger and older drivers when encountering children
Acad. Anal. & Prev. Vol. 18. No. 4. pp. 315-324. Printed in Great Britain.
call-4595x% 0 1986 Pergamon
1986
s3.00+ .w Journals Ltd.
DRIVING STRATEGIES AMONG YOUNGER AND OLDER DRIVERS WHEN ENCOUNTERING CHILDREN* H. OUDE EGBERINK, P. F. LOURENS and H. H. VAN DER MOLEN Traffic Research Centre, University of Groningen, Rijksstraatweg 76, 9752 AH Haren, The Netherlands Abstract-Results are presented of a study into how drivers say they behave and how they actually behave in traffic situations in which children are involved. An analysis was made of the most important types of encounters in which drivers become involved in accidents with children. On the basis of accident surveys and psychological theories on information processing, it was assessed by means of a questionnaire what knowledge drivers have concerning their own behavior in these situations, as well as their expectations about typical child behavior. Actual behavior of drivers in these situations was investigated by assessing video recordings of their behavior in driving a one hour standard track through residential areas. From the questionnaire it appeared that younger drivers reported more frequently dangerous behavior than older drivers. The recordings showed that younger drivers also behaved more dangerously during child encounters. This iesult could not be explained by differences in speed, but could by the fact that younger drivers detected the children less frequently than older drivers. Implications for the contents of mass media campaigns and their evaluation are discussed. INTRODUCTION
In this paper the results of a cognitive and behavioral assessment study of driver-child encounters are presented. The study described here is the first part of a larger research project, aimed at developing the contents of Dutch mass media campaigns aimed at drivers. Another goal of the project is the development of evaluation instruments for such campaigns. Traffic accidents form a most serious threat to the lives and health of children. In 1978/79, one-third of Dutch child mortality between the ages of 5 and 14 was caused by traffic accidents (Weseman, 1983). The truly epidemic aspects of the child traffic accident problem have attracted attention on an international scale (Council of Europe, 1972; Havard, 1974; Jackson, 1978). From accident studies (Mattson and Lindensjo, 1971; Sandels, 1974) and from legal enquiries, it appears that drivers often state that they were “surprised” by the “sudden emergence” of children from behind visual obstacles, or that they had been aware of the children for some time but that the children had suddenly behaved quite differently from what the driver had expected. From conflict observation studies of drivers with children and adult pedestrians (Kockelke and Steinbrecher, 1983; Reichardt and Hehlen, 1980; Howarth and Lightburn, 1980), it appears that drivers show little avoidance behavior (swerving, braking, preparing to brake) when they encounter pedestrians. During encounters with very young children, there is no more avoidance behavior shown than with adult pedestrians (Kockelke and Steinbrecher, 1983; Howarth and Lightburn, 1980), despite the fact that encounters with children are the most difficult situations to keep under control because of children’s frequently unexpected behavior (Van der Molen, 1983). The authors mentioned above have strong doubts as to the existence of a safetyoriented attitude on the part of drivers towards children. However, little empirical evidence is available. Brown (1980) presented a theoretical analysis of the attention, perception and decision processes of drivers during child encounters. He concludes that it is specifically the inexperienced drivers who have very incomplete expectancy patterns with regard to the course of events during encounters with pedestrians. Guidance, training *This study was supported by the Directorate of Traffic Safety of the Dutch Ministry of Transport. An extended laboratory report is available in Dutcli (Lourens, P. F., Oude Egberink, H. and van der Molen, H. H., 1984). 315
H. OUDE EGBERINKet al.
316
and education are possible measures that may improve these expectancy patterns and, consequently, the drivers’ attention, perception and decision processes. In drivers, youth and inexperience generally go together. However, it appears that the relatively high accident involvement of young drivers is not only due to inexperience (Spoerer, 1977). In The Netherlands, about 80% of child victims are hit by male drivers. Similar percentages are found in other countries (Kirk, 1964; Sandels, 1974). Although lack of exposure data prevents the conclusion that mule drivers behave more dangerously, there is little doubt that young drivers are overrepresented in child traffic accidents (Skandia, 1971; Sandels, 1974). As a general hypothesis, it was expected to find a considerable amount of risky drivers’ behavior towards children, in actual driving as well as in questionnaire-responses of drivers. Moreover, younger drivers were expected to show more risky behavior than older drivers. METHOD
A selection of the most dangerous encounter situations was based on an analysis of accident literature, pertaining to children up to the age of 15 as pedestrians or cyclists. Nine traffic situations were selected: four with children as pedestrians and five with children as cyclists. To identify the kinds of behavior and psychological processes that may cause accidents in particular situations the function/event sequence from the accident causation model by Snyder and Knoblauch (1971) was used. A description of this model was also presented by van der Molen (1981). In this function/event sequence, actions and reactions on the part of the drivers, as well as the children are represented (see Fig. 1). The most important stages in the function/event sequence were operationalized into questionnaire-items to assess what drivers know about typical child behavior and their own behavior in particular situations. How drivers actually behave in such situations was assessed by having subjects drive for more than one hour through residential areas. The questionnaire The mode of questioning was based on a methodology for quantifying socalled “scripted
knowledge” (Galambos, 1983). The concept of scripted knowledge refers to the mental representation of the contents and structure of routinelike activities in daily situations. These mental representations enable us to function efficiently and properly when con-
Fig. 1. Generalized
function/event
sequence
once the driver and child are in a collision Knoblauch, 1971).
course
(Snyder
and
Driving strategies when encountering children
317
fronted with a task in a particular situation. Most questions in the questionnaire asked for what Galambos called “standardness” judgments. Such judgments given on a 7-point scale ranging from “(practically) never” to “(practically) always” indicate to what degree particular mental or physical actions are considered common by a person when he is carrying out a task in a particular situation. For each of the nine situations, a standard set of 13 questions was formulated, related to the following topics: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.
The usual speed (5-point scale). The frequency of behavior aimed at searching for children. The frequency of actual detection of children. The frequency of being distracted. Expectancy patterns with regard to children’s behavior. The frequency of sudden unexpected behavior on the part of children. The frequency of having to brake suddenly because of a child’s ‘dangerous behavior . The frequency of using the car’s horn in order to warn a child. The expectancy that a child will give way to the approaching vehicle. The perception of the risk of getting involved in an accident with a child. The feeling of having done everything necessary to avoid an accident involving a child. The estimation of the chance of a hypothetical accident involving a child in that specific type of situation (9-point scale). The age of the child in such a hypothetical accident (5-point scale).
When one is performing a certain activity, i.e. is confronted with a particular traffic situation, certain mental and physical actions will be more or less associated with the activity. By assessing to what extent subjects consider an action X to be standard when they are performing an activity Y, shortcomings in the behavior considered necessary from a safety point of view can be traced. In the questionnaire each situation was illustrated with a picture taken at locations on the experimental driving task route. The experimental driving task
The actual behavior of drivers when encountering children was studied experimentally by having the subjects drive a fixed route in a specific period of time and in a specially prepared car. The instrumentation in the car made it possible to register several aspects of the driving task: 1. a video system with a camera on the roof of the car to register the situation in front of the car; the camera picked up a 50” angle of view; 2. a microphone inside the car to register the spoken comments of the subject on the audio track of the video tape. 3. a speed radar to register the momentary speed of the car; samples made every second were collected and recorded with the video signal; 4. a micro switch connected to the accelerator which produced a signal in the time code of the video system whenever the accelerator was released, to register a potential willingness to reduce speed and preparedness to brake. No recordings were made of the first part of the route, i.e. about 20 minutes driving. This part served as a training phase to familiarize the subject with the car and the experimental situation. Subjects drove the total route of 38 kilometers in 75 to 90 minutes leading through several housing districts in the city of Groningen in the Netherlands. The districts differed in date and lay-out, ranging from more than 50 years old districts with narrow streets and many parked cars, to about 25 years old districts with high housing blocks and wide streets, to recently built districts with many one-family houses and curvy roads, frequently
318
H. OUDE EGBERINK et al.
surrounded by trees and bushes. The route consisted of 56 (parts of) streets which can be categorized into one of three functional street types within a housing district traffic system: 1. main access streets: wide streets with no nearby houses, connecting one district with another or with a highway; 2. access streets: wide streets within a district, with houses nearby; 3. residential streets: narrow streets with houses and often with many parked cars. A large number of primary and secondary schools are situated along the route. To ensure the presence of children in the streets, the driving task was performed between 11.30 and 13.00 hours, a period in which children are free from school for their lunch break. Subjects
Through an advertisement in a local newspaper inhabitants of the city of Groningen were invited to take part in an experiment on “personal driving styles”. From a total of 150 volunteers, 48 subjects were selected: 12 males and 12 females in the age group of 18-24 years and 12 males and 12 females in the age group of 30-56 years. Of each group of 12, half the number drove less than 10.000 km per year. Procedure
For each subject the sequence of activities was the same: 1. The experimental driving task 2. the completion of the questionnaire 3. a discussion of the encounters with children during the driving task, based on the video/audio recording of the trip. Before the driving task began, subjects were told that the research was being carried out to investigate the personal driving styles of car drivers. Attention was drawn to the fact that the trip should not be taken as some kind of examination, and that one should drive as one usually does in one’s own car. The experimenter was only present during the trip to indicate the route. Moreover, the subjects were aware of the video and audio recordings being made and were invited to talk aloud about what they considered to be of importance in the traffic situation at hand and their considerations in relation to the situation. No pressure was put upon the subject, however, to talk aloud. Subjects were, of course, not informed about the special interest taken in their behavior when encountering children in the street. Analysis of the questionnaire data
For each of the questions, a norm range of answers was determined which could be considered to represent safe behavior. Depending on their distance from this range (minimal range was 3), answers falling outside it were rated 1, 2, 3 or 4. The norm ranges for each question were determined on the basis of accident analysis studies, as well as on the judgments of a panel of experts from our Institute. Analysis of driving task data
After completion of the trip and the questionnaire, the specific goal of the experiment was explained. Then the video/audio recording of the trip was shown to the subject and discussed with the experimenter. Encounters with children were defined using two criteria: (1) the presence of a child on the monitor estimated to be younger than 15 years of age, and (2) the child being on a collision course with the experimental car or possibly getting onto a collision course due to an imaginary (sudden) change of behavior on the part of the child. Some aspects of the encounters with children were scored in the presence of the subject. For each encounter it was determined: 1. whether the subject could remember place, i.e. could recognize the situation;
that that particular encounter
had taken
Driving
strategies
319
when encountering children
2. what the speed of the car was when passing the child (displayed on the video monitor); 3. whether the subject made any change in his/her driving behavior in connection with the encounter, such as releasing the accelerator, swerving or paying extra attention to the situation at hand as reported by the subject. Evaluations of the driver’s behavior were determined by the experimenter at a later stage. To evaluate drivers’ behavior in relation to encountering children, the observed behavior was compared to a set of normative rules of behavior as drafted by the expert group at our Institute. The potential danger of an encounter and the type of stree’t in which the encounter took place were chosen as determining factors for normative driving behavior. No absolute validity is claimed with this set of rules, but it served well as a relative scale for measuring safety of driving. Potential danger was measured on a scale ranging from 0 to 5 by taking five dichotomous criteria into consideration: 1. the shortest distance between the car and the child was estimated to be less than five meters (score = 1) or not less than five meters (score = 0); 2. the child was on an actual collision course with the car (1) or it was not (0); 3. the child was estimated to be younger than eight years of age (1) or not younger than eight (0); 4. the driver’s view of the child was impeded, for instance by parked cars, (1) or it was not (0); 5. there was no possibility for the driver to make an evasive manoeuvre (1) or there was (0). Because no knowledge was available about the relative importance of each of the criteria it was decided to use a summation of these five criteria scores as an indication of the potential danger of an encounter. The set of normative rules for driving behavior when encountering children in different types of streets and with different degrees of potential danger is given in Table 1. Actual behavior during an encounter was scored on a nominal scale consisting of the categories “good” ( + + ), “satisfactory” ( + ), “unsatisfactory” ( - ), and “risky” ( - - ). Behavior within the normative rules was scored + + ; the scores + , - and - - were assigned, depending on the degree of deviation from the normative rules. Criteria for these distinctions were determined by the same group of experts that drafted the normative rules. The interraters reliability (percentage agreement) for the and - - scores was tested and came out to be 75%. All the traffic-relevant utterances of the subjects made during the trip were written down. Each uttered sentence was coded according to the four central concepts from the function/event sequence of Snyder and Knoblauch’s model (1971), i.e. searching, detection, evaluation and decision. The resulting numbers were transformed into proportiondata per subject to correct for the fact that some drivers made more utterances than others.
Table
1. Normative
driving
behavior
at encounters
with children
for different
types of streets
and degrees of
potential danger (A., maximal speed in km; B., other requirements) Potal-
A. Type of street
B. Other
tial
Main ecceee
Access
Residential
danger
streets
etreete
streets
requirements
for any type of street
15
15
15
verbal
25
25
20
remembered
detection,
released
accelerator
30
30
25
remembered
detection,
released
accelerator
released
accelerator
reaction,
released
40
35
30
remembered
detection,
45
40
33
remembered
detection
50
45
35
remembered
detection
accelerator
320
H. OUDE EGBERINKef al
Table 2. Evaluations of driving behavior at actual encounters with children for younger and older drivers, summed across subjects (in proportions) Dl-iVWS
Number
of
Evaluation
observations
++
+
younger
785
.29
.31
.26
.14
older
813
.34
.33
.22
.ll
1596
.32
.32
.24
.12
total
RESULTS
Questionnaire
In each of the nine situations, the group of younger subjects had a higher mean score than the older subjects, indicating behavior that was more dangerous in each situation. Mean scores over all situations were 35.5 and 24.2, respectively (t = 2.88, df = 46, p = .003). The differences between the situations were relatively small for both the younger and the older subjects. Meaningful differences between male and female subjects could not be found. A rank order of the situations was made with regard to the perceived probability of an hypothetical accident with a child in each situation. The frequently occurring situation with a child pedestrian on the pavement where there is no form of obscured vision (for instance due to parked cars) proved to be considered the least probable by the subjects. This contrasts sharply with the fact that this proves to be the most probable type of situation when accident data are examined. Subjects also considered the four situations with children as pedestrians to be potentially dangerous mainly for children up to nine years of age. The five situations with children as cyclists were mainly considered to be potentially dangerous for children older than nine years of age. Driving task
The 48 trips resulted in 1598 encounters with children. The behavior evaluations for younger and older drivers are given in Table 2. As, from a safety point of view, unsatisfactory and, especially, risky behavior deserves the most attention, it was decided to take as an individual’s driving score twice the proportion of risky behavior added with the proportion of unsatisfactory behavior. The mean score for younger drivers calculated in this way was .56, as compared to .45 for older drivers. It can, therefore, be concluded that younger drivers more frequently showed behavior below normative standards than older drivers (t = 1.70, df = 46, p < .05). The 1598 encounters cover both children as pedestrians and children riding on bicycles. Splitting the data into these two categories showed the same trend in each case. Especially during encounters with child pedestrians, the behavior of younger drivers is inferior to that of older drivers, as can be seen in Table 3. Table 3 shows that 78% of the encounters involved child pedestrians. Pedestrians are encountered by drivers when (possibly) crossing the road somewhere between inter-
Table 3. Mean individual driving scores for younger and older drivers at encounters with child pedestrians and child cyclists Younger
drivers
Encounter
number
of
with
observations
Older
drivers
mean
number
score
observations
of
mean
t-test
score
child
pedestrian
620
.51
624
.47
p =.06
child
cyclist
165
.48
189
.40
n.s.
Driving strategies when encountering children
321
crossing facilities. Cyclists are encountered in several circumstances; a car can come from behind, from the opposite direction, from the left or the right of the cyclists. Analysis of the data for these different kinds of encounters showed an important result. The most frequently occurring situationchild pedestrians (possibly) crossing somewhere in mid-block with or without parked cars-proved to result in the highest frequency of risky behavior. This particular kind of encounter made up 56% of all encounters. Actually, this is also the situation in which most of the accidents occur. The proportion of risky behavior in these cases was .17 for all subjects together, as compared to proportions between .06 and .lO for the other kinds of encounters. Especially when parked cars or other obstacles impeded the driver’s view of the child, the proportion of risky behavior was as high as .30! The trend in the difference between younger drivers and older drivers was the same for all kinds of encounters: younger drivers showed more risky behavior. It can consequently be assumed that there is no interaction between age and behavior during specific kinds of encounters. The two central normative aspects of driving behavior during encounters with children were detection and speed (see also Table 1): sections
or at intersections
with or without pedestrian
Did the driver report afterwards, when shown an encounter on the video-recording, that he or she noticed the presence of that child during the trip? What was the registered speed of the car when passing the child (as could be read from the video)? The difference found between younger and older drivers can be due to one of these factors or to both. Analysis showed that there was no difference in speed between the two age groups when passing a child: mean passing speed for younger drivers was 29.9 km/h, for older drivers 30.5 km/h. A clear difference, however, was found in detection. Older drivers detected the presence of children in 63% of the cases, while the younger drivers did so in only 51% of the cases (chi-square = 22.1, df = 1, p < .OOOl). The final step in the analysis of these data was made by determining whether detection and speed vary with the potential danger of an encounter in relation to the type of street in which the encounter took place. The results are presented in Figure 2. For younger drivers, detection improves with higher levels of potential danger in main access streets (chi-square = 21.5, df = 4,p < .OOl), and in access streets (chi-square = 11.7, df = 4, p < .02), but not in residential streets. For older drivers, there is no significant improvement in any of the street types, although the trends in the access streets and the residential streets clearly point towards improvement. These data indicate that, in general, encounters, with a higher potential danger attract more attention of the drivers than en-
70 60. 50 40
Potenti.
.d.npsr
PoLent‘~l
danger
Potent.*,
danger
Fig. 2. Effects of potential danger of an encounter on detection (-) and speed (----) for younger (n), and older (0) drivers in main access streets (A), access streets (B), and residential streets (C) (some of the levels of potential danger have been omitted because of insufficient data).
H. OUDE EGBERINK ef al.
322 Table 4. Classification
of verbal Snyder
search
of younger
utterances and
and older drivers (in proportions)
Knoblauch’s
components
detection
evaluation
.12
.53
.25
.11
.45
.35
decision
.ll 9
counters with a low potential danger. For both younger and older drivers, speed is significantly affected by the type of street (F = 24.5 and 52.7, respectively, df = 2, p < .OOl) and also by potential danger (F = 8.4 and 7.1, respectively, df = 4, p < .OOl). In general, the results show that drivers do respond to children in the streets. Yet, their responses are still frequently insufficient. The verbal utterances of the subjects during the trip, categorized according to the four components of Snyder and Knoblauch’s model, showed an interesting difference between younger and older drivers, as can be seen in Table 4. Older drivers make relatively more evaluation utterances than younger drivers (proportion-test. z = 6.90, p < .OOl). This could indicate that older drivers are more conscientiously occupied with the possible significance of what they detect, while for younger drivers the processing of information might more often be limited to a categorization of what they detect. A younger driver might say, for instance: “There is a child on a bicycle in front of me.” An older driver might more often add to this sentence the evaluation: “Children of this age sometimes make sudden swerving movements on their bicycles.”
Relation between the questionnaire and the driving task The predictive value of the questionnaire for actual behavior in the driving task turned out to be fairly high for encounters with child pedestrians. When the scores on the questionnaire for situations with child pedestrians were summed per subject and compared with the individual driving task scores for encounters with child pedestrians the correlations were .46 for younger drivers and .66 for older drivers. The questionnaire had, however, no predictive value for encounters with children as cyclists (- .07 for both age groups). This could be due to the fact that only relatively few encounters with children as cyclists could be observed during the driving task. The mean number of encounters with children as cyclists was 7, with a range of 1 to 17. Another method of predicting the driving task score would be speed measurements in circumstances where drivers can choose their own speed. The value of this method was analyzed by corrrelating the subjects’ actual speed at five different locations on the experimental route, where free speed choice was almost always possible, with the overall driving task scores for encounters with child pedestrians. For younger drivers the five correlations were - .06, .12, .23, .44 and .64. For older drivers the correlations were .37, .47, .48, .59 and .67. Compared with the earlier mentioned relationship between the questionnaire and the driving task (the correlation for all subjects was .57), these results lend support to the questionnaire as an evaluation instrument, rather than to speed measurements, for which the mean correlation was .40.
DISCUSSION
The major aim of the research project (of which the study reported here is the initial phase) was to develop the contents of Dutch mass media campaigns aimed at drivers. The results of our study supported the hypothesis that there is a considerable degree of risky behavior of drivers in child encounters. This was most obvious in two situations: the situation with a child on the pavement without vision on the child being obscured in any way, resulting in 16% risky behavior, and the situation with a child on the pavement where there is some form of obscured vision (mostly due to parked cars), which resulted
Driving strategies when encountering children
323
30% risky behavior. In the questionnaire, however, the scores for these two types of situations were not worse than those for the other situations, which suggests a lack of awareness about the possible impact of drivers’ own behavior in those two situations. Based on these results, the two situations mentioned were recommended for the contents of the mass media campaigns. Further study of the responses in the questionnaire for the two selected situations revealed a number of aspects which often deviated from the safety norms: in
1. 35% of the subjects showed a lack of awareness of the risk of becoming involved in an accident with a child. 2. When subjects did notice a child in such a situation, the expectations of 30% of them with respect to the child’s possible behavior were inadequate. 3. 23% of the subjects indicated that they were frequently lost in thought or distracted, which would result in too little attention being paid to children. 4. In the situation without obscured vision, 33% of the subjects indicated an usual speed in this type of situation which was normatively too high. That the habitual speed is also of importance in the situation with obscured vision may be derived from the fact that 50% of the subjects underestimated their actual speed in this type of situation. These four aspects: speed, attention, expectations and risk awareness will thus need special attention in the mass media campaigns. Another hypothesis was that younger drivers behave more risky than older drivers. The performance of younger drivers proved to be clearly inferior in all situations included in the questionnaire and was also significantly inferior in the driving task. Younger drivers have less risk awareness and detect fewer children. In contrast to several other studies as reviewed by Jonah (in this issue), no actual differences in speed were found between younger and older drivers. It can be concluded that search behavior and risk perception will need special attention in the campaigns. Although the performance of younger drivers was significantly inferior to that of older drivers, the quality of the performance of the older drivers does not permit us to aim the mass media campaigns at younger drivers only. It was therefore decided to aim the campaigns at all drivers. In the Introduction we mentioned that the existence of a positive attitude on the part of car drivers towards children is questioned by several researchers on the basis of their observational studies in this field (see, for instance, Howarth, 1984). Our study shows that, as Howarth put forward, many drivers behave inconsiderately towards children, or react insufficiently to their presence. On the other hand, our study also shows that-on average-drivers do take some avoiding action when they encounter children, and many of them do so to a considerable extent. From Fig. 2 it is evident that drivers also reduce their speed when encountering children, especially as the potential danger of these encounters increases. This indicates that the starting point of educational measures aimed at drivers is a better one than sometimes is suggested (e.g. Howarth, 1984). With some adaptations the methodology developed in this study is useful for the evaluation of mass media campaigns: The reliability of the driving task scores could be increased by refining the definitions of the criteria used. We are rather optimistic about the validity of the driving task: many of our subjects reported that their driving style in our experimental car did not differ from their driving in their own cars. This is in agreement with the findings of Kockelke and Steinbrecher (1983) who showed that drivers in their experiment drove at speeds similar to those of other drivers in the test-route streets. However, it might be interesting to compare our method of driving assessment with other methods, such as the unobtrusive observations made by Howarth and Lightburn (1980). The part of the questionnaire which deals with encounters with child pedestrians seems to have sufficient predictive power (for actual driving behavior) to be used as a valid evaluation instrument. The research instruments which have now been developed will be used as before and after measurements in a pre-post-control design to evaluate the impact of three regional mass-media campaigns. Three different lines of approach with regard to the contents will be used:
H. OUDE EGBERINK et al.
324
1. increase of risk perception by emphasizing the probability and the seriousness of accidents; 2. supply of adequate behavioral scenario information on three levels: -strategical. by stressing that driving in residential areas differs from driving on highways or bypass roads; -tactical, by giving information on adequate habitual speed and tendencies to underestimate one’s speed; -operational. by giving information on adequate search behavior; 3. a combination of the two approaches mentioned above. In view of our experimental data, all three of these lines of approach are interesting. From a theoretical point of view, we have no reason to assume that the risk perception line of approach will be more or less effective than the scenario line. The combination of these two lines seems the most promising one. During the evaluation we will, among other things, study any interactive effects between campaign approach and driver age. Hopefully, this will lead us to an answer to the question as to whether-as is sometimes argued-it is necessary to appeal in different ways to younger and to older drivers in mass media campaigns. REFERENCES and drivers really compatible’? In D. J. Obome and J. A. Lewis (Eds.), Human Facrors in
Brown J., Are pedestrians
Transport Research. Academic Press, London, 1980. Council of Europe, Accidents in Childhood as a Public Health Problem,
Strasbourg, 1972. studies of six characteristics of our knowledge of common activities. Behavior Research Methods and Instrumentation 15, 327-349, 1983. Havard J. D. J.. Child pedestrian casualties as a public health problem. Medicine, Science and the Law 14, 168-179, 1974. Howarth I., Interactions between drivers and pedestrians: some new approaches to pedestrian safety. In: Proceedings of the Symposium “Human Behaviour & T&tic Safety,” Detroit, 1984 (to be published). Howarth I. and Lightbum A., How drivers respond to pedestrians and vice versa. In: D. 1. Obome and J. A. Lewis (eds.), Human Factors in Transport Research. Academic Press, London, 1980. Jackson R. H., Hazards to children in traffic: a paediatrician looks at road accidents. Archives of Disease in Childhood
Galambos
J. A., Normative
53, 807-813,
1978.
Voice of the Pedestrian
60-65,
1979.
Jonah B. A., Accident risk and risk-taking behaviour among young drivers, Accid. Anal. & Prev., 18, 255-271, 1986. Kirk V., De jongsten in her verkeer: verkeersongevallen en verkeersgedragingen van O-6 jarigen. Dutch translation by the VBV, Society of Pedestrians, The Hague, 1964. Kockelke W. and Steinbrecher J., Sicherheit von Kindem und lltem Menschen im Verkehr-Umsetzung einer neuen Verhaltensvorschrift fur Kraftfahrer in die Verkehrspraxis, Zeitschrift fur Verkehrssicherheit 25, lo- 14, 1983. Lourens P. E, Oude Egberink H. and Van der Molen H. H., “Ik dacht dat ik wat zag: oh nee, het waren kinderen.” Een onderzoek naar automobilistengedrag ten opzichte van kinderen in het verkeer. Report VK 84-08, Traffic Research Center of the University of Groningen, Haren, The Netherlands, 1984. Mattson K. and Lindensjo S., A Report on Children in Traffic, Skandia Insurance Company, Stockholm, 1971. Reichart P. and Hehlen P., Verhalten von Automobilisten gegentiber Kindem im Strassenverkehr, Zeitschrtft fur Verkehrssicherheit 26, 81-82, 1980. Sandels S., Why are Children Injured in Tramc?, Skandia Report 2, Skandia Insurance Company, Stockholm, 1974. Skandia Insurance Company, The Skandia Report on Children- in Traffic, Stockholm, 197 I. Suoerer E.. Tvoische Fehlverhaltensweisen von Fahranfiingem. In: Typische Fehlverhaltensweisen von Fihranfaiigem ’ und Moghc’hkeiten gezielter Nachschulung. Kiiln, W. Germany, I977. Snyder M. B. and Knoblauch R. L., Pedestrian Safety: The Identification of Precipitating Factors and Possible Countermeasures, vols. I and 2. Operations Research Inc., Silversprings, MD, 1971. Van der Molen H. H., Child pedestrian’s exposure, accidents and behavior, Accid. Anal. & Prev. 13, 193-224, 1981. Van der Molen, H. H. ( Pedestrian ethology: unobtrusive observations of child and adult road crossing behaviour in theframework of the development of a child pedestrian training programma, dissertation. Published by the author, Haren, The Netherlands, 1983. Weseman P., Hoe onveilig is het voor kinderen op straat? In V. A. Gtittinger et al., Verkeersonveiligheid van Kinderen. Spruyt, van Mantgem & de Does, Leiden 1983.
call-4595x% 0 1986 Pergamon
1986
s3.00+ .w Journals Ltd.
DRIVING STRATEGIES AMONG YOUNGER AND OLDER DRIVERS WHEN ENCOUNTERING CHILDREN* H. OUDE EGBERINK, P. F. LOURENS and H. H. VAN DER MOLEN Traffic Research Centre, University of Groningen, Rijksstraatweg 76, 9752 AH Haren, The Netherlands Abstract-Results are presented of a study into how drivers say they behave and how they actually behave in traffic situations in which children are involved. An analysis was made of the most important types of encounters in which drivers become involved in accidents with children. On the basis of accident surveys and psychological theories on information processing, it was assessed by means of a questionnaire what knowledge drivers have concerning their own behavior in these situations, as well as their expectations about typical child behavior. Actual behavior of drivers in these situations was investigated by assessing video recordings of their behavior in driving a one hour standard track through residential areas. From the questionnaire it appeared that younger drivers reported more frequently dangerous behavior than older drivers. The recordings showed that younger drivers also behaved more dangerously during child encounters. This iesult could not be explained by differences in speed, but could by the fact that younger drivers detected the children less frequently than older drivers. Implications for the contents of mass media campaigns and their evaluation are discussed. INTRODUCTION
In this paper the results of a cognitive and behavioral assessment study of driver-child encounters are presented. The study described here is the first part of a larger research project, aimed at developing the contents of Dutch mass media campaigns aimed at drivers. Another goal of the project is the development of evaluation instruments for such campaigns. Traffic accidents form a most serious threat to the lives and health of children. In 1978/79, one-third of Dutch child mortality between the ages of 5 and 14 was caused by traffic accidents (Weseman, 1983). The truly epidemic aspects of the child traffic accident problem have attracted attention on an international scale (Council of Europe, 1972; Havard, 1974; Jackson, 1978). From accident studies (Mattson and Lindensjo, 1971; Sandels, 1974) and from legal enquiries, it appears that drivers often state that they were “surprised” by the “sudden emergence” of children from behind visual obstacles, or that they had been aware of the children for some time but that the children had suddenly behaved quite differently from what the driver had expected. From conflict observation studies of drivers with children and adult pedestrians (Kockelke and Steinbrecher, 1983; Reichardt and Hehlen, 1980; Howarth and Lightburn, 1980), it appears that drivers show little avoidance behavior (swerving, braking, preparing to brake) when they encounter pedestrians. During encounters with very young children, there is no more avoidance behavior shown than with adult pedestrians (Kockelke and Steinbrecher, 1983; Howarth and Lightburn, 1980), despite the fact that encounters with children are the most difficult situations to keep under control because of children’s frequently unexpected behavior (Van der Molen, 1983). The authors mentioned above have strong doubts as to the existence of a safetyoriented attitude on the part of drivers towards children. However, little empirical evidence is available. Brown (1980) presented a theoretical analysis of the attention, perception and decision processes of drivers during child encounters. He concludes that it is specifically the inexperienced drivers who have very incomplete expectancy patterns with regard to the course of events during encounters with pedestrians. Guidance, training *This study was supported by the Directorate of Traffic Safety of the Dutch Ministry of Transport. An extended laboratory report is available in Dutcli (Lourens, P. F., Oude Egberink, H. and van der Molen, H. H., 1984). 315
H. OUDE EGBERINKet al.
316
and education are possible measures that may improve these expectancy patterns and, consequently, the drivers’ attention, perception and decision processes. In drivers, youth and inexperience generally go together. However, it appears that the relatively high accident involvement of young drivers is not only due to inexperience (Spoerer, 1977). In The Netherlands, about 80% of child victims are hit by male drivers. Similar percentages are found in other countries (Kirk, 1964; Sandels, 1974). Although lack of exposure data prevents the conclusion that mule drivers behave more dangerously, there is little doubt that young drivers are overrepresented in child traffic accidents (Skandia, 1971; Sandels, 1974). As a general hypothesis, it was expected to find a considerable amount of risky drivers’ behavior towards children, in actual driving as well as in questionnaire-responses of drivers. Moreover, younger drivers were expected to show more risky behavior than older drivers. METHOD
A selection of the most dangerous encounter situations was based on an analysis of accident literature, pertaining to children up to the age of 15 as pedestrians or cyclists. Nine traffic situations were selected: four with children as pedestrians and five with children as cyclists. To identify the kinds of behavior and psychological processes that may cause accidents in particular situations the function/event sequence from the accident causation model by Snyder and Knoblauch (1971) was used. A description of this model was also presented by van der Molen (1981). In this function/event sequence, actions and reactions on the part of the drivers, as well as the children are represented (see Fig. 1). The most important stages in the function/event sequence were operationalized into questionnaire-items to assess what drivers know about typical child behavior and their own behavior in particular situations. How drivers actually behave in such situations was assessed by having subjects drive for more than one hour through residential areas. The questionnaire The mode of questioning was based on a methodology for quantifying socalled “scripted
knowledge” (Galambos, 1983). The concept of scripted knowledge refers to the mental representation of the contents and structure of routinelike activities in daily situations. These mental representations enable us to function efficiently and properly when con-
Fig. 1. Generalized
function/event
sequence
once the driver and child are in a collision Knoblauch, 1971).
course
(Snyder
and
Driving strategies when encountering children
317
fronted with a task in a particular situation. Most questions in the questionnaire asked for what Galambos called “standardness” judgments. Such judgments given on a 7-point scale ranging from “(practically) never” to “(practically) always” indicate to what degree particular mental or physical actions are considered common by a person when he is carrying out a task in a particular situation. For each of the nine situations, a standard set of 13 questions was formulated, related to the following topics: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13.
The usual speed (5-point scale). The frequency of behavior aimed at searching for children. The frequency of actual detection of children. The frequency of being distracted. Expectancy patterns with regard to children’s behavior. The frequency of sudden unexpected behavior on the part of children. The frequency of having to brake suddenly because of a child’s ‘dangerous behavior . The frequency of using the car’s horn in order to warn a child. The expectancy that a child will give way to the approaching vehicle. The perception of the risk of getting involved in an accident with a child. The feeling of having done everything necessary to avoid an accident involving a child. The estimation of the chance of a hypothetical accident involving a child in that specific type of situation (9-point scale). The age of the child in such a hypothetical accident (5-point scale).
When one is performing a certain activity, i.e. is confronted with a particular traffic situation, certain mental and physical actions will be more or less associated with the activity. By assessing to what extent subjects consider an action X to be standard when they are performing an activity Y, shortcomings in the behavior considered necessary from a safety point of view can be traced. In the questionnaire each situation was illustrated with a picture taken at locations on the experimental driving task route. The experimental driving task
The actual behavior of drivers when encountering children was studied experimentally by having the subjects drive a fixed route in a specific period of time and in a specially prepared car. The instrumentation in the car made it possible to register several aspects of the driving task: 1. a video system with a camera on the roof of the car to register the situation in front of the car; the camera picked up a 50” angle of view; 2. a microphone inside the car to register the spoken comments of the subject on the audio track of the video tape. 3. a speed radar to register the momentary speed of the car; samples made every second were collected and recorded with the video signal; 4. a micro switch connected to the accelerator which produced a signal in the time code of the video system whenever the accelerator was released, to register a potential willingness to reduce speed and preparedness to brake. No recordings were made of the first part of the route, i.e. about 20 minutes driving. This part served as a training phase to familiarize the subject with the car and the experimental situation. Subjects drove the total route of 38 kilometers in 75 to 90 minutes leading through several housing districts in the city of Groningen in the Netherlands. The districts differed in date and lay-out, ranging from more than 50 years old districts with narrow streets and many parked cars, to about 25 years old districts with high housing blocks and wide streets, to recently built districts with many one-family houses and curvy roads, frequently
318
H. OUDE EGBERINK et al.
surrounded by trees and bushes. The route consisted of 56 (parts of) streets which can be categorized into one of three functional street types within a housing district traffic system: 1. main access streets: wide streets with no nearby houses, connecting one district with another or with a highway; 2. access streets: wide streets within a district, with houses nearby; 3. residential streets: narrow streets with houses and often with many parked cars. A large number of primary and secondary schools are situated along the route. To ensure the presence of children in the streets, the driving task was performed between 11.30 and 13.00 hours, a period in which children are free from school for their lunch break. Subjects
Through an advertisement in a local newspaper inhabitants of the city of Groningen were invited to take part in an experiment on “personal driving styles”. From a total of 150 volunteers, 48 subjects were selected: 12 males and 12 females in the age group of 18-24 years and 12 males and 12 females in the age group of 30-56 years. Of each group of 12, half the number drove less than 10.000 km per year. Procedure
For each subject the sequence of activities was the same: 1. The experimental driving task 2. the completion of the questionnaire 3. a discussion of the encounters with children during the driving task, based on the video/audio recording of the trip. Before the driving task began, subjects were told that the research was being carried out to investigate the personal driving styles of car drivers. Attention was drawn to the fact that the trip should not be taken as some kind of examination, and that one should drive as one usually does in one’s own car. The experimenter was only present during the trip to indicate the route. Moreover, the subjects were aware of the video and audio recordings being made and were invited to talk aloud about what they considered to be of importance in the traffic situation at hand and their considerations in relation to the situation. No pressure was put upon the subject, however, to talk aloud. Subjects were, of course, not informed about the special interest taken in their behavior when encountering children in the street. Analysis of the questionnaire data
For each of the questions, a norm range of answers was determined which could be considered to represent safe behavior. Depending on their distance from this range (minimal range was 3), answers falling outside it were rated 1, 2, 3 or 4. The norm ranges for each question were determined on the basis of accident analysis studies, as well as on the judgments of a panel of experts from our Institute. Analysis of driving task data
After completion of the trip and the questionnaire, the specific goal of the experiment was explained. Then the video/audio recording of the trip was shown to the subject and discussed with the experimenter. Encounters with children were defined using two criteria: (1) the presence of a child on the monitor estimated to be younger than 15 years of age, and (2) the child being on a collision course with the experimental car or possibly getting onto a collision course due to an imaginary (sudden) change of behavior on the part of the child. Some aspects of the encounters with children were scored in the presence of the subject. For each encounter it was determined: 1. whether the subject could remember place, i.e. could recognize the situation;
that that particular encounter
had taken
Driving
strategies
319
when encountering children
2. what the speed of the car was when passing the child (displayed on the video monitor); 3. whether the subject made any change in his/her driving behavior in connection with the encounter, such as releasing the accelerator, swerving or paying extra attention to the situation at hand as reported by the subject. Evaluations of the driver’s behavior were determined by the experimenter at a later stage. To evaluate drivers’ behavior in relation to encountering children, the observed behavior was compared to a set of normative rules of behavior as drafted by the expert group at our Institute. The potential danger of an encounter and the type of stree’t in which the encounter took place were chosen as determining factors for normative driving behavior. No absolute validity is claimed with this set of rules, but it served well as a relative scale for measuring safety of driving. Potential danger was measured on a scale ranging from 0 to 5 by taking five dichotomous criteria into consideration: 1. the shortest distance between the car and the child was estimated to be less than five meters (score = 1) or not less than five meters (score = 0); 2. the child was on an actual collision course with the car (1) or it was not (0); 3. the child was estimated to be younger than eight years of age (1) or not younger than eight (0); 4. the driver’s view of the child was impeded, for instance by parked cars, (1) or it was not (0); 5. there was no possibility for the driver to make an evasive manoeuvre (1) or there was (0). Because no knowledge was available about the relative importance of each of the criteria it was decided to use a summation of these five criteria scores as an indication of the potential danger of an encounter. The set of normative rules for driving behavior when encountering children in different types of streets and with different degrees of potential danger is given in Table 1. Actual behavior during an encounter was scored on a nominal scale consisting of the categories “good” ( + + ), “satisfactory” ( + ), “unsatisfactory” ( - ), and “risky” ( - - ). Behavior within the normative rules was scored + + ; the scores + , - and - - were assigned, depending on the degree of deviation from the normative rules. Criteria for these distinctions were determined by the same group of experts that drafted the normative rules. The interraters reliability (percentage agreement) for the and - - scores was tested and came out to be 75%. All the traffic-relevant utterances of the subjects made during the trip were written down. Each uttered sentence was coded according to the four central concepts from the function/event sequence of Snyder and Knoblauch’s model (1971), i.e. searching, detection, evaluation and decision. The resulting numbers were transformed into proportiondata per subject to correct for the fact that some drivers made more utterances than others.
Table
1. Normative
driving
behavior
at encounters
with children
for different
types of streets
and degrees of
potential danger (A., maximal speed in km; B., other requirements) Potal-
A. Type of street
B. Other
tial
Main ecceee
Access
Residential
danger
streets
etreete
streets
requirements
for any type of street
15
15
15
verbal
25
25
20
remembered
detection,
released
accelerator
30
30
25
remembered
detection,
released
accelerator
released
accelerator
reaction,
released
40
35
30
remembered
detection,
45
40
33
remembered
detection
50
45
35
remembered
detection
accelerator
320
H. OUDE EGBERINKef al
Table 2. Evaluations of driving behavior at actual encounters with children for younger and older drivers, summed across subjects (in proportions) Dl-iVWS
Number
of
Evaluation
observations
++
+
younger
785
.29
.31
.26
.14
older
813
.34
.33
.22
.ll
1596
.32
.32
.24
.12
total
RESULTS
Questionnaire
In each of the nine situations, the group of younger subjects had a higher mean score than the older subjects, indicating behavior that was more dangerous in each situation. Mean scores over all situations were 35.5 and 24.2, respectively (t = 2.88, df = 46, p = .003). The differences between the situations were relatively small for both the younger and the older subjects. Meaningful differences between male and female subjects could not be found. A rank order of the situations was made with regard to the perceived probability of an hypothetical accident with a child in each situation. The frequently occurring situation with a child pedestrian on the pavement where there is no form of obscured vision (for instance due to parked cars) proved to be considered the least probable by the subjects. This contrasts sharply with the fact that this proves to be the most probable type of situation when accident data are examined. Subjects also considered the four situations with children as pedestrians to be potentially dangerous mainly for children up to nine years of age. The five situations with children as cyclists were mainly considered to be potentially dangerous for children older than nine years of age. Driving task
The 48 trips resulted in 1598 encounters with children. The behavior evaluations for younger and older drivers are given in Table 2. As, from a safety point of view, unsatisfactory and, especially, risky behavior deserves the most attention, it was decided to take as an individual’s driving score twice the proportion of risky behavior added with the proportion of unsatisfactory behavior. The mean score for younger drivers calculated in this way was .56, as compared to .45 for older drivers. It can, therefore, be concluded that younger drivers more frequently showed behavior below normative standards than older drivers (t = 1.70, df = 46, p < .05). The 1598 encounters cover both children as pedestrians and children riding on bicycles. Splitting the data into these two categories showed the same trend in each case. Especially during encounters with child pedestrians, the behavior of younger drivers is inferior to that of older drivers, as can be seen in Table 3. Table 3 shows that 78% of the encounters involved child pedestrians. Pedestrians are encountered by drivers when (possibly) crossing the road somewhere between inter-
Table 3. Mean individual driving scores for younger and older drivers at encounters with child pedestrians and child cyclists Younger
drivers
Encounter
number
of
with
observations
Older
drivers
mean
number
score
observations
of
mean
t-test
score
child
pedestrian
620
.51
624
.47
p =.06
child
cyclist
165
.48
189
.40
n.s.
Driving strategies when encountering children
321
crossing facilities. Cyclists are encountered in several circumstances; a car can come from behind, from the opposite direction, from the left or the right of the cyclists. Analysis of the data for these different kinds of encounters showed an important result. The most frequently occurring situationchild pedestrians (possibly) crossing somewhere in mid-block with or without parked cars-proved to result in the highest frequency of risky behavior. This particular kind of encounter made up 56% of all encounters. Actually, this is also the situation in which most of the accidents occur. The proportion of risky behavior in these cases was .17 for all subjects together, as compared to proportions between .06 and .lO for the other kinds of encounters. Especially when parked cars or other obstacles impeded the driver’s view of the child, the proportion of risky behavior was as high as .30! The trend in the difference between younger drivers and older drivers was the same for all kinds of encounters: younger drivers showed more risky behavior. It can consequently be assumed that there is no interaction between age and behavior during specific kinds of encounters. The two central normative aspects of driving behavior during encounters with children were detection and speed (see also Table 1): sections
or at intersections
with or without pedestrian
Did the driver report afterwards, when shown an encounter on the video-recording, that he or she noticed the presence of that child during the trip? What was the registered speed of the car when passing the child (as could be read from the video)? The difference found between younger and older drivers can be due to one of these factors or to both. Analysis showed that there was no difference in speed between the two age groups when passing a child: mean passing speed for younger drivers was 29.9 km/h, for older drivers 30.5 km/h. A clear difference, however, was found in detection. Older drivers detected the presence of children in 63% of the cases, while the younger drivers did so in only 51% of the cases (chi-square = 22.1, df = 1, p < .OOOl). The final step in the analysis of these data was made by determining whether detection and speed vary with the potential danger of an encounter in relation to the type of street in which the encounter took place. The results are presented in Figure 2. For younger drivers, detection improves with higher levels of potential danger in main access streets (chi-square = 21.5, df = 4,p < .OOl), and in access streets (chi-square = 11.7, df = 4, p < .02), but not in residential streets. For older drivers, there is no significant improvement in any of the street types, although the trends in the access streets and the residential streets clearly point towards improvement. These data indicate that, in general, encounters, with a higher potential danger attract more attention of the drivers than en-
70 60. 50 40
Potenti.
.d.npsr
PoLent‘~l
danger
Potent.*,
danger
Fig. 2. Effects of potential danger of an encounter on detection (-) and speed (----) for younger (n), and older (0) drivers in main access streets (A), access streets (B), and residential streets (C) (some of the levels of potential danger have been omitted because of insufficient data).
H. OUDE EGBERINK ef al.
322 Table 4. Classification
of verbal Snyder
search
of younger
utterances and
and older drivers (in proportions)
Knoblauch’s
components
detection
evaluation
.12
.53
.25
.11
.45
.35
decision
.ll 9
counters with a low potential danger. For both younger and older drivers, speed is significantly affected by the type of street (F = 24.5 and 52.7, respectively, df = 2, p < .OOl) and also by potential danger (F = 8.4 and 7.1, respectively, df = 4, p < .OOl). In general, the results show that drivers do respond to children in the streets. Yet, their responses are still frequently insufficient. The verbal utterances of the subjects during the trip, categorized according to the four components of Snyder and Knoblauch’s model, showed an interesting difference between younger and older drivers, as can be seen in Table 4. Older drivers make relatively more evaluation utterances than younger drivers (proportion-test. z = 6.90, p < .OOl). This could indicate that older drivers are more conscientiously occupied with the possible significance of what they detect, while for younger drivers the processing of information might more often be limited to a categorization of what they detect. A younger driver might say, for instance: “There is a child on a bicycle in front of me.” An older driver might more often add to this sentence the evaluation: “Children of this age sometimes make sudden swerving movements on their bicycles.”
Relation between the questionnaire and the driving task The predictive value of the questionnaire for actual behavior in the driving task turned out to be fairly high for encounters with child pedestrians. When the scores on the questionnaire for situations with child pedestrians were summed per subject and compared with the individual driving task scores for encounters with child pedestrians the correlations were .46 for younger drivers and .66 for older drivers. The questionnaire had, however, no predictive value for encounters with children as cyclists (- .07 for both age groups). This could be due to the fact that only relatively few encounters with children as cyclists could be observed during the driving task. The mean number of encounters with children as cyclists was 7, with a range of 1 to 17. Another method of predicting the driving task score would be speed measurements in circumstances where drivers can choose their own speed. The value of this method was analyzed by corrrelating the subjects’ actual speed at five different locations on the experimental route, where free speed choice was almost always possible, with the overall driving task scores for encounters with child pedestrians. For younger drivers the five correlations were - .06, .12, .23, .44 and .64. For older drivers the correlations were .37, .47, .48, .59 and .67. Compared with the earlier mentioned relationship between the questionnaire and the driving task (the correlation for all subjects was .57), these results lend support to the questionnaire as an evaluation instrument, rather than to speed measurements, for which the mean correlation was .40.
DISCUSSION
The major aim of the research project (of which the study reported here is the initial phase) was to develop the contents of Dutch mass media campaigns aimed at drivers. The results of our study supported the hypothesis that there is a considerable degree of risky behavior of drivers in child encounters. This was most obvious in two situations: the situation with a child on the pavement without vision on the child being obscured in any way, resulting in 16% risky behavior, and the situation with a child on the pavement where there is some form of obscured vision (mostly due to parked cars), which resulted
Driving strategies when encountering children
323
30% risky behavior. In the questionnaire, however, the scores for these two types of situations were not worse than those for the other situations, which suggests a lack of awareness about the possible impact of drivers’ own behavior in those two situations. Based on these results, the two situations mentioned were recommended for the contents of the mass media campaigns. Further study of the responses in the questionnaire for the two selected situations revealed a number of aspects which often deviated from the safety norms: in
1. 35% of the subjects showed a lack of awareness of the risk of becoming involved in an accident with a child. 2. When subjects did notice a child in such a situation, the expectations of 30% of them with respect to the child’s possible behavior were inadequate. 3. 23% of the subjects indicated that they were frequently lost in thought or distracted, which would result in too little attention being paid to children. 4. In the situation without obscured vision, 33% of the subjects indicated an usual speed in this type of situation which was normatively too high. That the habitual speed is also of importance in the situation with obscured vision may be derived from the fact that 50% of the subjects underestimated their actual speed in this type of situation. These four aspects: speed, attention, expectations and risk awareness will thus need special attention in the mass media campaigns. Another hypothesis was that younger drivers behave more risky than older drivers. The performance of younger drivers proved to be clearly inferior in all situations included in the questionnaire and was also significantly inferior in the driving task. Younger drivers have less risk awareness and detect fewer children. In contrast to several other studies as reviewed by Jonah (in this issue), no actual differences in speed were found between younger and older drivers. It can be concluded that search behavior and risk perception will need special attention in the campaigns. Although the performance of younger drivers was significantly inferior to that of older drivers, the quality of the performance of the older drivers does not permit us to aim the mass media campaigns at younger drivers only. It was therefore decided to aim the campaigns at all drivers. In the Introduction we mentioned that the existence of a positive attitude on the part of car drivers towards children is questioned by several researchers on the basis of their observational studies in this field (see, for instance, Howarth, 1984). Our study shows that, as Howarth put forward, many drivers behave inconsiderately towards children, or react insufficiently to their presence. On the other hand, our study also shows that-on average-drivers do take some avoiding action when they encounter children, and many of them do so to a considerable extent. From Fig. 2 it is evident that drivers also reduce their speed when encountering children, especially as the potential danger of these encounters increases. This indicates that the starting point of educational measures aimed at drivers is a better one than sometimes is suggested (e.g. Howarth, 1984). With some adaptations the methodology developed in this study is useful for the evaluation of mass media campaigns: The reliability of the driving task scores could be increased by refining the definitions of the criteria used. We are rather optimistic about the validity of the driving task: many of our subjects reported that their driving style in our experimental car did not differ from their driving in their own cars. This is in agreement with the findings of Kockelke and Steinbrecher (1983) who showed that drivers in their experiment drove at speeds similar to those of other drivers in the test-route streets. However, it might be interesting to compare our method of driving assessment with other methods, such as the unobtrusive observations made by Howarth and Lightburn (1980). The part of the questionnaire which deals with encounters with child pedestrians seems to have sufficient predictive power (for actual driving behavior) to be used as a valid evaluation instrument. The research instruments which have now been developed will be used as before and after measurements in a pre-post-control design to evaluate the impact of three regional mass-media campaigns. Three different lines of approach with regard to the contents will be used:
H. OUDE EGBERINK et al.
324
1. increase of risk perception by emphasizing the probability and the seriousness of accidents; 2. supply of adequate behavioral scenario information on three levels: -strategical. by stressing that driving in residential areas differs from driving on highways or bypass roads; -tactical, by giving information on adequate habitual speed and tendencies to underestimate one’s speed; -operational. by giving information on adequate search behavior; 3. a combination of the two approaches mentioned above. In view of our experimental data, all three of these lines of approach are interesting. From a theoretical point of view, we have no reason to assume that the risk perception line of approach will be more or less effective than the scenario line. The combination of these two lines seems the most promising one. During the evaluation we will, among other things, study any interactive effects between campaign approach and driver age. Hopefully, this will lead us to an answer to the question as to whether-as is sometimes argued-it is necessary to appeal in different ways to younger and to older drivers in mass media campaigns. REFERENCES and drivers really compatible’? In D. J. Obome and J. A. Lewis (Eds.), Human Facrors in
Brown J., Are pedestrians
Transport Research. Academic Press, London, 1980. Council of Europe, Accidents in Childhood as a Public Health Problem,
Strasbourg, 1972. studies of six characteristics of our knowledge of common activities. Behavior Research Methods and Instrumentation 15, 327-349, 1983. Havard J. D. J.. Child pedestrian casualties as a public health problem. Medicine, Science and the Law 14, 168-179, 1974. Howarth I., Interactions between drivers and pedestrians: some new approaches to pedestrian safety. In: Proceedings of the Symposium “Human Behaviour & T&tic Safety,” Detroit, 1984 (to be published). Howarth I. and Lightbum A., How drivers respond to pedestrians and vice versa. In: D. 1. Obome and J. A. Lewis (eds.), Human Factors in Transport Research. Academic Press, London, 1980. Jackson R. H., Hazards to children in traffic: a paediatrician looks at road accidents. Archives of Disease in Childhood
Galambos
J. A., Normative
53, 807-813,
1978.
Voice of the Pedestrian
60-65,
1979.
Jonah B. A., Accident risk and risk-taking behaviour among young drivers, Accid. Anal. & Prev., 18, 255-271, 1986. Kirk V., De jongsten in her verkeer: verkeersongevallen en verkeersgedragingen van O-6 jarigen. Dutch translation by the VBV, Society of Pedestrians, The Hague, 1964. Kockelke W. and Steinbrecher J., Sicherheit von Kindem und lltem Menschen im Verkehr-Umsetzung einer neuen Verhaltensvorschrift fur Kraftfahrer in die Verkehrspraxis, Zeitschrift fur Verkehrssicherheit 25, lo- 14, 1983. Lourens P. E, Oude Egberink H. and Van der Molen H. H., “Ik dacht dat ik wat zag: oh nee, het waren kinderen.” Een onderzoek naar automobilistengedrag ten opzichte van kinderen in het verkeer. Report VK 84-08, Traffic Research Center of the University of Groningen, Haren, The Netherlands, 1984. Mattson K. and Lindensjo S., A Report on Children in Traffic, Skandia Insurance Company, Stockholm, 1971. Reichart P. and Hehlen P., Verhalten von Automobilisten gegentiber Kindem im Strassenverkehr, Zeitschrtft fur Verkehrssicherheit 26, 81-82, 1980. Sandels S., Why are Children Injured in Tramc?, Skandia Report 2, Skandia Insurance Company, Stockholm, 1974. Skandia Insurance Company, The Skandia Report on Children- in Traffic, Stockholm, 197 I. Suoerer E.. Tvoische Fehlverhaltensweisen von Fahranfiingem. In: Typische Fehlverhaltensweisen von Fihranfaiigem ’ und Moghc’hkeiten gezielter Nachschulung. Kiiln, W. Germany, I977. Snyder M. B. and Knoblauch R. L., Pedestrian Safety: The Identification of Precipitating Factors and Possible Countermeasures, vols. I and 2. Operations Research Inc., Silversprings, MD, 1971. Van der Molen H. H., Child pedestrian’s exposure, accidents and behavior, Accid. Anal. & Prev. 13, 193-224, 1981. Van der Molen, H. H. ( Pedestrian ethology: unobtrusive observations of child and adult road crossing behaviour in theframework of the development of a child pedestrian training programma, dissertation. Published by the author, Haren, The Netherlands, 1983. Weseman P., Hoe onveilig is het voor kinderen op straat? In V. A. Gtittinger et al., Verkeersonveiligheid van Kinderen. Spruyt, van Mantgem & de Does, Leiden 1983.