Br. J. educ. Psychol.

, 53, 152-168, 1983

PEDESTRIAN BEHAVIOUR OF CHILDREN AND

ACCOMPANYING PARENTS DURING SCHOOL
JOURNEYS: AN EVALUATION OF A TRAINING
PROGRAMME
BY H. H. VAN DER MOLEN, J. VAN DEN HERIK AND
C. VAN DER KLAAUW
(Traffic Research Centre, University of Groningen, The Netherlands)
SUMMARY. From earlier studies it seems that parents who accompany their children
when crossing roads often do so badly from the point of view of good protection and
traffic education. In this report the results are presented of video observations of road
crossing behaviour of 63 pre-school children and their parents, before and after the
application of a pedestrian training programme for children in which parents were
actively involved. Parents behaved far from ideally in various respects, but they
displayed better traffic behaviour than children, both before and after the programme.
After the programme, parents not only gave better examples when crossing, but they
also gave more verbal instruction, something which rarely occurred before. The
children displayed the trained behaviours more frequently after the programme.
Publicity campaigns of road safety organisations in many countries have been based for
many years on the belief that good examples of road crossing by parents will be followed
by their children. Empirical evidence for this belief is presented here. The results
demonstrate that it is not only desirable that parents set better examples for the children
they accompany, but that it is also possible to achieve this by means of specific training
programmes.

INTRODUCTION
THEfirst objective of this study was to analyse road crossing behaviour of adults
accompanying their children to or from kindergarten. In the light of this objective
the present study can be considered to be a more elaborate follow-up of an earlier one
(Van der Molen, 1981b).
The second objective was to evaluate the effects of an experimental traffic
education programme on pedestrian behaviour of parents and the children they
accompanied to or from kindergarten. This was done by comparing observations
carried out before and after application of the programme. The sections of the
present report dealing with the evaluation of programme effects are part of an
extensive study carried out in Hoogkerk at the end of 1979, evaluating the effectiveness
of a child pedestrian training programme. Other evaluation studies concerning this
programme were presented by Rothengatter (1981), who dealt-amongst other
things-with children’s road crossing behaviour under test conditions, and by Van
der Molen (1982b) who dealt with road crossing behaviour of children observed
unobtrusively when walking unsupervised. Those two studies can be considered as
“ product evaluations ” of the programme, i.e., they measured the effects of the
programme in terms of attainment of its objectives.
The present study is more concerned with “ process evaluation ” of the pro-
gramme, i.e., it provides information on the way in which mediators (i.e., parents)
applied the programme and how their children reacted to this (Rothengatter, 1981).
Although parents were instructed to train their children in the street and were advised
to repeat this regularly, they had never been instructed explicitly to d o this on the way
to or from kindergarten. However, if the ideas behind the programme did come
across to the parents, we would expect them to become better educational models and
instructors after the programme and to show this on the way to or from kindergarten.
Similarly one would expect the children to behave more according to what they had
152
 H. H. VAN DER MOLENET AL. 153
learned during the educational programme. This we may consider to be product
evaluation under specific social circumstances.
Earlier research
Many children become involved in pedestrian accidents despite adult supervision
(Mattson and Lindensjo, 1971; Sandels, 1974; Grayson, 1975b). Sandels (1974)
gave a detailed description of the circumstances of several accidents where children
escaped from the adult’s supervision suddenly and became involved in an accident.
Although these examples rarely involved children crossing at the same time as the
adults, it appeared from an analysis of 166 child accidents at pedestrian crossings, that
21 per cent of the children crossed together with adults (Mattson and Lindensjo,
1971). They reported that in almost half the number of cases the adults behaved just
like the children. In most cases the adult was near the child at the moment the
accident happened, but it also happened that the child had run ahead or had pulled
itself loose.
From an observational study, Grayson (1975a) concluded that “ supervision
could have been substantially improved . . . and . . . accompaniment was often less
than complete ”. Indications of incomplete accompaniment and less than optimal
protection were that children ran more than adults while crossing and often ran ahead
or lagged behind.
Sandels (1975) described several observations indicating that parents often
provide bad examples for their children : not stopping, only partial use of the pedes-
trian crossing, crossing at an angle, crossing when approaching cars are close. The
more quantitative data from Grayson’s (1975a) study indicate that the usual bad
example given by adults is that of a failure to stop at the kerb before crossing the road.
On the other hand they make many more head movements than their children and
they run much less.
From observation studies in England, Sweden and Germany it appears that
accompanied children rarely seem aware of the road crossing task at hand. They not
only appear less observant than their leaders (Grayson, 1975a; Michalik, 1975), but
accompanied children are also less observant than when crossing alone (Sandels,
1975; Michalik, 1975; Downing and Spendlove, 1981), and they stop less (Michalik,
1975). Sandels’ (1975) suggestion that children tend to transfer responsibility to the
adults seems a plausible explanation for these findings.
The empirical results summarised above were corroborated in an unobtrusive
observation study which was carried out at one zebra crossing in Groningen, where
adults would cross with their children travelling to or from kindergarten (Van der
Molen, 1981b). The road our subjects crossed was a two-way, undivided road with a
traffic lane in each direction and a total traffic volume of about 450 vehicles per hour
(veh./h.).
Only about half the number of accompanied children were led by the hand all the
way across the road. When not leading children by the hand, adults did not effec-
tively prevent children from running ahead, or from lagging behind and then catching
up with a run. Of the children who walked unrestrained, 58 per cent ran across part
of the road.
However, it appeared that holding hands led to stereotyped behaviour of the
children. The majority of children who held hands all the way made no correct head
movements and, as might be expected, stopped only when the adult did. In summary,
they showed no awareness of actively participating in the road crossing task. Children
who walked unrestrained all the way were twice as observant as those who were led by
hand across the road. Even so, less than one third of the unrestrained children made
correct head movements.
154 Parents and Children as Pedestrians
Entirely in line with Grayson's (1975a) and Michalik's (1975) data, we found that
adults made many more head movements than children and that they ran much less.
They stopped somewhat less frequently at the pavement and at the centre island.
Concerning correct head movements, adults set very good examples (89 per cent of
them looked both ways at the kerb), but they did not make their children look too
(19 per cent of them looked both ways at the kerb). The low percentage (47 per cent)
of adults who stopped at the kerb indicated that-in this respect-they ought to give
a much better example to the children they accompanied.
From the results of this study we could draw the following conclusions concerning
the topics which need more emphasis in educational programmes aimed at parents.
When they accompany their children to or from school :
(1) parents should stop before crossing;
(2) parents should make their children look as well as they do;
(3) in general parents should make children aware of the road crossing task at
hand to obtain their active participation, even when the child is led by the
hand;
(4) parents should make certain that the child does not lag behind or run ahead,
either by holding hands or by verbal instruction (depending on the situation
and the child's tendency to conform with verbal instruction).
These recommendations are in line with those of the Dutch Road Safety Organisation
(Veilig Verkeer Nederland) which are distributed among parents by means of their
leaflet " Lend them a hand '' (Help ze 'n handje).
The recommendations are based on the assumption that parents play an impor-
tant role in the acquisition and stabilisation of children's traffic behaviour. This
assumption seems plausible because of their close relationship and the frequency with
which they cross together. But, although it seems to be taken for granted in the
traffic educational literature that the example of the parent will be followed by the
child, there are thus far no empirical data in the traffic area to support that notion
(Gerber, 1978). From psychological theory validated in other areas one may conclude
that observing a model generally leads to learning. However, it is not clear under
what conditions a child imitates the model behaviour (Gerber, 1978; Zumkley-
Munkel, 1976).
Social learning theory distinguishes between three types of observational learning
(Rothengatter, 1981, p. 67):
(1) Observational learning proper or modelling: the observer acquires new
behaviour through observing the behaviour of a model.
(2) Inhibitory and dis-inhibitory effects : the rate of initiative behaviour increases
or decreases, because the observer observes the consequences (positive or
negative) of the modelled action.
(3) Facilitating effects: already learned behaviour is called upon by the behaviour
of the model.
The first type of learning seems the most relevant with respect to pre-school children
with little traffic experience. The other two types of learning indicate however that
setting a good example always remains important for the (traffic) education of
children. Further, a distinction can be made between the adult as a model and the
adult as an instructor. As an instructor the adult (and especially the parent) is
supposed to inform, protect and instruct the child in the real situation; as a model he
is supposed to perform the instructed (verbalised) behaviour him- or herself. Gerber
(1978) expects (with the research on altruistic behaviour of Bryan and Walbeck, 1970,
in mind) that children orient themselves generally more to the behaviour of the mother
than to her verbal instruction.
 H. H. VAN DER MOLENET AL. 155
In a review of British, German and Swedish studies on parental attitudes towards
children’s road safety problems and road safety education, Rothengatter (198 1, ch. 9)
reported that showing the child how to behave is mentioned as a frequently used
method to teach children traffic behaviour, but that parents reacted remarkably
adversely to practising together correct behaviour for various reasons. Most parents,
however, relied on verbal explanations of correct behaviour.
Limbourg and Gerber (1978) found a discrepancy between the verbalised instruc-
tion given by the parents and their own traffic behaviour: they taught rules which they
broke themselves. That parents do not show ideal pedestrian behaviour appeared also
from the observation studies summarised above. Despite this, Gerber (1978) found
in a study, at a pedestrian crossing with traffic lights, that parents, as a group, can be
distinguished from unmarried adults or married adults without children, with respect
to overt traffic behaviour, behaviour intentions and expectations of the social norms
concerning their own behaviour. Mothers especially showed a more normative
behaviour when they were in the company of their children. Compatible with these
results are those from Mooy (1980) who found that adults, crossing in the company
of children at a zebra with traffic lights, crossed more at the zebra and while the
pedestrian light was green, than adults or children crossing independently. From his
research Gerber (1978) concluded that there is a great readiness to set a good example
but that the problem is to convert those intentions into overt behaviour.
An outline of the trafic training programme
The traffic training programme was described in detail by Rothengatter (1981,
chapter 10). We will therefore limit our description to what is relevant for under-
standing the evaluation study dealt with here.
Three tasks were trained: (1) crossing in quiet streets, (2) crossing in the vicinity
of parked cars and (3) crossing at junctions. The first grade of kindergarten school
(4- and 5-year-old children) was trained to perform only tasks 1 and 2 and the second
grade (5- and 6-year-old children) only tasks 2 and 3. These three tasks were selected
from a set of 26 pedestrian tasks identified and described in detail in a theoretical
analysis of ideal (i.e., safe) pedestrian behaviour (Van der Molen, Rothengatter and
Vinjk. 1981). Selection of the tasks was made on the basis of importance and feasi-
bility. ‘ Importance ’ assessments of the three pedestrian tasks were based on an
extensive review of child pedestrian’s exposure, accidents and behaviour (Van der
Molen, 1981). ‘ Feasibility ’ assessments on training of the tasks were based on an
extensive review of children’s cognitive and sensory abilities and limitations with
regard to the functions necessary for pedestrian task performance (VinjC, 1981). On
the basis of the latter analysis it was decided to train children only to cross in quiet
streets, and only to cross when no traffic was approaching as young children were
considered unable to assess speeds and distances of oncoming cars properly.
The training procedure consisted of four stages:
(1) a “ modelling ” stage in which the trainer demonstrated and verbalised the
correct behaviour ;
(2) a “ practise together ” stage where the child is prompted and guided to
perform the behaviour elements displayed in the “ modelling ” stage; the
child is encouraged to verbalise its behaviour, and praised for each correctly
performed behaviour element;
(3) a “ practise alone ” stage during which the child performs the behaviour
alone and receives specific forms of verbal feedback;
(4) an “ observation and reward ” stage in which the trainer asks the child to
cross the street at a different place and follows the child at some distance;
after the child has completed the crossing procedure the trainer notes down
156 Parents and Children as Pedestrians
and verbalises the child’s performance and rewards the child for every cor-
rectly performed behaviour element. During the fourth stage the trainer
uses observation scoring forms and specially designed reward material.
Each training session lasted between 10 and 15 minutes. The programme allowed for
four training sessions per task, but advocated repeated monitoring of the child’s
performance after completion of these four sessions.
Before the training started the children had received classroom instruction with
video films in which a child model showed the correct behavioural elements of each
task about four times.
For the introduction of the programme to the parents, parent school evenings
were organised which consisted of three parts: (1) a lecture in which the problems of
pre-school children in traffic, their psychological limitations and the rationale of the
programme were outlined, (2) a video film which demonstrated the crossing pro-
cedures to be trained and the training principles to be followed and (3) a group
discussion and the opportunity to practise the behaviour training techniques.
All parents received a specially designed training booklet which provided
basically the same information as given in the school evening lecture and video film
presentation. The booklet explicitly described the three crossing tasks and how to
give instruction to the children.
Before the programme was introduced mothers were interviewed about their
opinions on children’s difficulties in traffic and their road safety education attitudes
and practices (Rothengatter, 1981, ch. 9). Almost all parents stated that they spent
time on traffic education. The topics they emphasised were more related to exposure
(staying on the pavement, where to play) than to specific crossing tasks. About half
of the parents indicated that they had demonstrated the desired behaviour and the
same number that they had practised with the child, which contrasts with the aversion
to practising reported in the literature.
For reasons of experimental design described elsewhere by Rothengatter (198 1)
the (roughly) 300 children and their parents were treated in three different ways,
described in our detailed research report (Van den Herik et al., 1981). Owing to this
design about 70 per cent of all pre-school children in Hoogkerk and at least 50 per
cent of their parents have been practising the training programme. In the present
evaluation study we could not single out parents and children who had been practising
the programme, so the effects reported here are bound to be conservative estimates
for those who did actually practise it.
METHOD
Hypothesis
Two types of hypothesis were formulated. Firstly, come those concerning pro-
gramme effects, i.e., comparisons between behaviour in the pre-observations and post-
observations. Secondly, are hypotheses that could be tested in the pre-observations
and a second time in the post-observations in order to check whether the effects found
in the pre-observations were stable ones. The arguments underlying the hypothesis
formulated were outlined in the introduction section.
Pre-observations and post-observations analysed separately
Hypotheses were formulated concerning the behaviour examples adults give to
the children and how their behaviour correlates with that of the children. Further-
more we sought replications of behavioural patterns in adult behaviour and in
child behaviour found in an earlier study (Van der Molen, 1981b) and described
below.
 H. H. VAN DER MOLENET AL. 157
We hypothesised that about an equal percentage of adults and children would
stop at the kerb, but that a greater percentage of adults would make head movements
left, right and backwards, both at the kerb and while crossing the road. Further, we
expected more adults to cross in a normal tempo. We hypothesised positive cor-
relations between adults and their charges with respect to all these behaviours. We
expected such correlations even when controlling for effects of the presence of traffic
on behaviour.
For adults we hypothesised positive correlations between stopping at the kerb
and making head movements in any direction. For children we predicted not such
correlations on the basis of our findings in our earlier study (Van der Molen, 1981b).
Based on this study we expected negative correlations between holding hands and
making head movements, both at the kerb and while crossing. Further, we expected
children who were holding hands to run less often ahead or to stay behind and to cross
more often in a normal tempo.

Programme effects: comparisons of pre- and post-observations

The product evaluation studies mentioned in the introduction (Rothengatter,
1981; Van der Molen, 1982b) focused on a limited set of behavioural items, which
were directly derived from the behavioural objectives for the children and were easily
observable: approaching at a normal speed, stopping at the kerb, head movements
left, right and backward (at junctions) at the kerb and crossing at right angles at a
normal speed. Since we were in this study interested not only in a strict product
evaluation in terms of the programme objectives but also in other aspects of behaviour
of both children and adults, we operationalised a broader range of behavioural items
that were instructed or recommended explicitly or implicitly. Hypotheses were
formulated concerning improvements in protective, instructive and model behaviour
of the parents and improvement of the children’s behaviour. As a result of better
protection we expected fewer children to walk ahead or to stay behind while crossing
the road. As a sign of better instruction we expected in the post-observation a larger
percentage of adults who communicated with the child with relation to the crossing
strategy, while the child reacted to this by making head movements. Our expectations
concerning an increase or decrease of holding hands were somewhat ambiguous. We
hypothesised that-for reasons of protection-parents would increase holding hands
during approach to the kerb, but that-for reasons of instruction-they would de-
crease the frequency of holding hands at the kerb and on the road and that they would
take less often the initiative to hold hands.
For both children and adults we expected improvements in road crossing be-
haviour, i.e., stopping at the kerb, head movements left, right, and backwards at the
kerb, no false starts (i.e., no stepping from the kerb and returning immediately),
crossing at right angles, making complete use of the zebra, crossing in a normal
tempo and making head movements left, right and backwards while on the road.

Observations
The study was carried out in Hoogkerk (population about 5,500), which is a
small suburban town in the vicinity of Groningen (population about 160,000).
Hoogkerk consists mainly of family residences in a varied, generally modern style.
Most houses have a garden in front of the house and at the back.
Traffic was counted four times at each location, just before and after school time,
during one quarter of an hour. Multiplication by four gave indices for the number
of vehicles per hour, which ranged from 12 to 1,OOO. Twenty per cent of the obser-
vations were made at locations with less than 75 veh./h., 56 per cent between 75 and
200 veh./h. and 24 per cent with more than 200 veh./h.
158 Parents and Children as Pedestrians
The observations finally used for analysis were made at 18 different locations,
mostly at junctions with pedestrian crossings (zebras), but without traffic lights. For
more details concerning the number of couples observed at specific locations or the
precise facilities and traffic intensities per location, we refer to our detailed research
report (Van den Herik et al., 1981).
We used a video set with two cameras and a mixer which made it possible to get
an overall view of the situation and at the same time a close-up of the zebra (separated
views). The equipment was installed in a delivery van (Renault Estafette) with
especially mounted small upper windows, big enough to observe with a camera from
the inside but too small to detect the camera easily from outside. Pre-observations
were carried out in the first two weeks of October, 1979, the post-observations in the
last week of November and the second week of December. We observed just before
and after school time, during about a quarter of an hour, four times a day.
We only analysed behaviour of " couples ", i.e., one pre-school child and its
supervisor, although it happened occasionally that they crossed together with others.
All adult subjects were females and in most cases we assume that they were mother and
child. From the fact that 37 per cent of the houses i n Hoogkerk is private property
(Groningen 20 per cent) we assume that the average income is relatively high as
compared with Groningen but comparable to the national average. We have no further
indications to estimate the socio-economic level of our subjects.
The video material was analysed and scored in the laboratory with a slow
motion recorder. In order to establish a reasonable scoring reliability, the first and
third authors checked the first batch of incoming protocols together until agreement
was very high, and afterwards occasional protocols at regular intervals. The same
procedure was followed with the post-observations. No formal reliability indices were
assessed. For a description of the observation protocols and definitions of the
variables we refer to Appendices A and B of our complete research report.
Design
As we knew from earlier studies that parked cars influence the crossing behaviour
considerably (Van der Molen, 1981a), we-eliminated the few cases where parked cars
were present. Other interfering variables could be the characteristics of the locations
themselves, so we matched for locations in such a way that in pre- and post-
observations the same number of couples were observed at each location. Finally we
had 63 couples in both observations. We intended to observe the same couples
(adult and child) in both the pre- and post-observations, but we managed to do this
for 15 couples only. For the statistical analysis of programme effects we therefore
divided our total number of observations into two independent samples (N = 2 x 48 =
96) in which the 48 couples before and after the education programme are different
ones, and a dependent sample (N = 15) in which the 15 couples are the same ones
in both observations. For some analyses (not concerning programme effects)
we used 48+15=63 couples of the pre-observations or 63 couples of the
post-observations.
The samples are not random with respect to crossing locations, because we
searched for places where people were expected to cross. Also they are non-random
with respect to accompanying adults and children. Nevertheless, we have no reason
to assume that the behaviour of these parents and children would differ from that of
other parents and children in Hoogkerk, when they cross at similar traffic intensities
on the way to or from school. Owing to the unobtrusive character of the observations
we could not assess which of the observed children and/or parents had actively taken
part in training programme (see paragraph on outline of the programme) but we have
no reason to assume that there was any bias for or against those who took part.
 H. H. VAN DER MOLENET AL. 159
Statistical analysis
As all our variables were dichotomies (yes/no), phi-coefficients of correlation were used
to measure the strength of the association between two variables (Guilford and Fruchter,
1978). The statistical significance of these correlations was assessed by means of the Fisher
exact test (Siegel, 1956). The Fisher exact test was also used to test differences in propor-
tions between adults and children and between independent groups before and after the
programme (Grizzle, 1967; Cohen, 1977). McNemar’s test (Siegel, 1956) was used to detect
significant improvements in behaviour in the dependent sample. With the small numbers in
our sample (the number of subjects that changed behaviour was always at most 10 out of 15)
this test becomes a symmetric (i.e., P = 0.5) binomial test (Siegel, 1956).
As we had generally clear directorial expectations and needed all the power we could
get (see below) we decided to test one way in most cases (i.e., when indicated by our
hypotheses). See Molenaar (1977) and Cohen (1977) for arguments in favour and against
this practice respectively.
Recently, several authors have shown both for Dutch and American journals in the
area of psychological and educational research that the power of the tests used is often very
low indeed, especially for small and medium effect sizes in the population (Elstrodt and
Mellenbergh, 1978; Hoogstraten et a/., 1980; Brewer, 1972; Brewer and Owen, 1973;
Chase and Chase, 1976). However, when N is small it is more advisable to seek a well
argued balance, which may change with the conditions of the studies, between a and /3 risks,
than to focus blindly on the convention to set a at 0.05 or less (Elstrodt and Mellenbergh,
1978).
Therefore we decided to set the one-sided a-risk at 0-05 for the Fisher exact tests and
at 0.20 for the McNemar tests. Power for the Fisher exact tests is generally good (i.e.,
3 0.80) for medium size population effects (comparable to tables with a phi-coefficient of
0.30) and often for smaller effects. The McNemar tests had even with this high a-risk good
power only for very large population effects, so in the dependent sample any small programme
effects could hardly be shown (see our extended report for details).
When dealing with the discrete distribution of the Fisher exact test it is sometimes
advised to reject not only when P exceeds the chosen a, but also for those table outcomes for
which the probability of exceeding is the closest possible to the predetermined a-value
(Snijders, 1980; see also Cohen, 1977, p. 150), be it sometimes a larger one. In order to
inform the reader one may also specify the summed P-value of the tables more extreme than
the one sampled (Snijders, 1980). We will follow these recommendations for both the
Fisher exact tests and the symmetrical binomial tests. When the P-value is insignificant but
still rather low (0*05<P< 0.10) we will specify this insignificant P value in order to help
when the reader cannot distinguish between outcomes with very high P-values (e.g., P = 0.50)
and rather low ones (Cohen, 1977).

RESULTS
Adults as models
The data of the pre-observation summarised in Table 1 show that o u r hypotheses
were corroborated. Adults stopped as often at the kerb as children, a greater per-
centage of adults made head movements in all directions, both at the kerb and while
crossing the road and more adults crossed in a normal tempo. The differences are
somewhat smaller in the post-observations. This is due to the fact that child behaviour
improved more after the training programme. From these data we conclude that
adults give good examples relative to the behaviour of their charges but their perfor-
mance is far from ideal. The improvements of adult and child behaviour after the
programme, apparent from Table 1, will be dealt with in the section on programme
effects.
I n Table 2 &coefficients are presented between adult and child behaviour. O u r
hypothesis that adult and child behaviour correlates positively is corroborated on the
whole. Stopping behaviour especially correlated very strongly although we regarded
B
160 Parents and Children as Pedestrians
TABLE 1
PERCENTAGES OF ADULTSAND CHILDREN DISPLAYING
SPECIFIC BEHAVIOURS
IN THE PRE- AND
POST-OBSERVATIONS
Pre-observation Post-observation
N=63 N=63 N=63 N=63
Behaviour Adults Children P-values Adults Children P-values
Kerb % % % %
Stopping 43 46 60 64
HM left 67 35 s 0.01 84 46 G 0.01
HM right 71 41 2 0.01 70 60
H M backward 10 3 18 18
Road
Normal tempo 98 78 d 0.01 98 79 d 0.01
H M left 52 33 0.02 44 33
HM right 67 51 0.05 65 46 0.02
HM backward 6 0 ** 8 10
*= the P-value = 0.14; the P-value of the next table is d 0.05.
**= this is the most extreme table possible; P = 0.06.
HM = head movement.
blanks = not significant when tested one-sided with a = 0.05 (in the notation of
Cohen, 1977: a1 = 0.05).
P-values refer to Fisher exact tests for the differences between children and adults.

8 TABLE 2
CORRELATIONS BETWEEN CHILD A N D ADULTBEHAVIOUR
DURING PRE- AND POST-OBSERVATIONS

Pre-observation Post-observation
N=63 N=43 N=20 N=63 N=34 N=29
traffic and
no traffic traffic no traffic traffic or traffic no traffic
Behaviour combined only only no traffic only only
Kerb 4 d d d 4 4
Stopping' 0.71** 071 0.66 0.89** I .oo 0.82
HM left 0.16 0.19 0.09 0.14 0.16 0.03
HM right 010 0.16 -0.04 0.31** 0.3 1 0.25
HM backwards -0.06 - 0.06 -0.05 045** 0.35 0.63
Road
Normal tempo1 - - - - - 1*00
HM left 0.20-f 0.23 0.15 0.1st 0'20 0.21
HM right 0.31** 0.40 0.09 0.34** 0.33 0.35
HM backwards 0.00 - - 0.30. 0.21 0.56
1 N = 23 in both pre- and post-observations, instead of 63, as the correlations were calculated
only for those pairs that were not holding hands.
* = significant with a1 d 0.05 (or the closest possible P-value) with the Fisher exact test.
** = significant with a1 < 0.01.
t = the P-value of the next extreme table is d 0.05
Indications for significance are given only in the first columns (N = 63).
- = Correlations are undefined owing to the fact that either children or adults showed no
variation in their behaviour.

only those couples who were not holding hands. Tempo on the road did not cor-
relate because of a lack of variation in behaviour. The correlations between head
movements of children and adults correlated more weakly, but on the whole they were
positive.
 H. H. VAN DER MOLENET AL. 161
Table 2 shows that there was no systematic drop in the correlations when
momentary traffic was held constant. Therefore we conclude that the correlations
were probably not caused by the environment in which they crossed together.
The correlations between child and adult behaviour during the pre-observations
are probably not due to instruction on the part of the adults either. Only two out of
63 couples displayed verbal communication to which the child reacted by making
head movements.
In summary we conclude that adults set relatively good, but not ideal, examples
and that this influenced the head movements of the children to some extent and
stopping of the children to a great extent.
Protection of the children
In the pre-observations 44 per cent, 59 per cent and 54 per cent of the children
were held by the hand during approach, at the kerb and while crossing the road
respectively. The initiative to hold hands came mostly (75 per cent) from the adults.
Holding hands correlated negatively (- 0.43) with walking ahead or staying behind
on the part of the children (significant at P < 0.01). Only two out of 34 children
who held hands while crossing (i.e., at least a major part of the road) did run ahead
or stay behind as opposed to roughly one third of those walking unrestrained. The
post-observations showed a similar picture. Holding hands correlated somewhat
positively with a normal tempo in the pre-observations but rather strongly in the post-
observations (Table 3). Of those restrained the greater part of the road, 18 per
cent and 6 per cent still ran part of the road during the pre- and post-observations
respectively. These percentages are indicative of incomplete protection which is
reflected by the fact that in the pre-observations 22 per cent of all children walked
ahead or stayed behind while crossing. These data are in agreement with the results
of our earlier study (Van der Molen, 1981b).
In contrast to our earlier findings holding hands had no serious detrimental
effect on the frequency of the children’s head movements in the present study (Table
3). Most correlations were negative (as expected) but very small.
Stopping and head movements
As shown in Table 4 both adults and children displayed positive correlations
between stopping and making head movements at the kerb. Thus for adults our
TABLE 3
+COEFFICIENTS
OF HOLDING
HANDSWITH CHILDREN’S
HEAD
MOVEMENTSAND OTHER
BEHAVIOURS
Pre-observation Post-observation
Behaviour N = 63 N = 63

Kerb 4 4
Holding hands-HM left - 06 - 16
Holding hands-HM right - 02 12
Holding hands-HM backwards -03 - 09
Road
Holding hands-HM left - 09 - 05
Holding hands-HM right - 08 12
Holding hands-HM backwards 00 -04
Holding hands-ahead/behind -43** -49**
Holding hands-tempo 12 41**
Zeros and decimal points omitted.
* = significant with 011 < 0.05; Fisher exact text
** = significant with 011 < 0.01.
 162 Parents and Children as Pedestrians
TABLE 4
+-COEFFICIENTS BETWEEN STOPPING AND MAKINGHEADMOVEMENTS
AT
THE KERBDURING PRE- AND POST-OBSERVATIONS

Adults (N = 63) Children (N = 63)

Related behaviours pre post pre post
Stopping-HM left 41;' 27; 46** 30.
Stopping-HM right 33** 46** 26; 33.
Stopping-HM backwards 16 29; 01 26;
Zeros and decimal points omitted.
* = significant with a1 g 0.05; Fisher exact test.
** = significant
' with a1 < 0.01.

hypotheses were corroborated, but for children we could not replicate the earlier (and
rather unique) finding that these correlations were near zero for accompanied children.
Programme effects
Protection and communication (Table 5 )
As shown in Table 5, holding hands during approach increased, but not signifi-
cantly. Holding hands at the kerb and while crossing the road did not decrease, nor
did initiatives for holding hands come less from the adults. On the contrary, the
changes were in the direction of holding hands more frequently (however, when tested
post hoc one-sidedly in the unpredicted direction these trends were not significant).
We conclude that parents regarded holding hands as a protective aspect, both during

TABLE 5
OF THE CHILDREN'S PROTECTION AND COMMUNICATION IN THE TWO SAMPLES TAKEN TOGETHER
IMPROVEMENTS
(N = 2 x 63), THE INDEPENDENT(N = 2 x 48) SAMPLE
AND THE DEPENDENT (N = 15) SAMPLE
Percentages or numbers of pedestrians displaying behaviour in pre- and
post-observations
Number and
% % % %
Significance directions Significance
pre- post- pre- post- at of changes at
N = 63 N = 63 N = 48 N = 48 ul = 0.05 N = 15 al = 0.20

Joint Independent (Same

Behaviour observations observations couples)
ND + -
Child walks ahead or stays
behind 22 16 23 13 3 2 1
Holding hands approach 44 59 42 58 * 7 4 3
Holding hands kerb 59 63 56 63 4 2 2
Holding hands road 54 56 54 54 7 4 3
Adult initiates holding
hands1 75 90 73 89 0 0 0
Communication with the
child to which it reacts
with making head
movements 3 17 2 21 < 0.01 3 2 1
1 N = 1 1 + 9 = 20 in the independent samples, 1 in the dependent sample and 12+ 10 = 22 in the
combined samples.
* P-value 0.8; for the next extreme table 0.04.
blanks = not significant at the specified one-sided a-level.
 H. H. VAN DER MOLENET AL. 163
approach, kerb and the road and that they were not inclined to give up the protective
value of holding hands for the sake of making the children more independent in their
road crossing activities. In the independent sample, communication between adults
and children concerning proper road crossing procedures improved significantly.

Improvement of adult behaviour (Table 6 )

In the independent sample three out of four types of kerb behaviour improved
significantly. In this sample only one behaviour type on the road improved signifi-
cantly (making complete use of the zebra). Several on-the-road behaviours could
hardly improve because of the high initial percentages (angle of crossing, no false
starts, normal tempo). Head movements while on the road did not significantly
improve. Thus, effects on behaviour which were not explicitly instructed or recom-
mended did not show up in the road behaviour, with the exception of adequate use of
the zebra. However, the behaviours the children were trained in did improve, except
those which already had high initial values (i.e., crossing at right angles and normal
tempo).
The dependent sample shows a consistent non-change or a slight change in the
opposite direction, which partly contradicts the results of the independent samples.
Therefore we have also presented the percentages for the combined samples (N =
48f 15 = 63) in Table 6 . This column shows that significant findings in the N = 48
samples are always clearly reflected in the N = 63 sample, despite opposite trends in
the N = 15 sample. We consider these deteriorations in the dependent sample as
error fluctuations in a small sample (even when tested post hoc with a one-sided c1 of
0.20, only two of the nine deteriorations were significant).

TABLE 6
IMPROVEMENTS I N ROADCROSSING
BEHAVIOUR
OF ADULTS FOR THE Two SAMPLES
TOGETHER
(N = 2 x 63), THE INDEPENDENTSAMPLES
(N = 2 x 48) AND THE DEPENDENTSAMPLE
( N = 15)
Percentages or numbers of adults displaying behaviour in pre- and
post-observations
Number and
”/, 76 % % Significance directions of
pre post pre post at changes**
Behaviour N = 63 N = 63 N = 48 N = 45 U I = 0.05 N = 15
ND + -
-~ _. -

Kerb
Stopping 43 60 40 63 0.02 4 2 2
HM left 67 84 63 88 < 0.01 5 2 3
HM right 71 70 69 75 8 2 6
HM backwards 10 18 8 19 * 0 0 0
Road
Crossing at right angles 95 89 94 92 3 0 3
Making no false starts 100 95 100 96 1 0 1
Complete use of zebra1 84 93 81 97 0.03 2 0 2
Normal tempo 98 98 98 100 1 0 1
HM left 52 44 50 48 4 0 4
HM right 67 65 67 69 4 1 3
HM backwards 6 8 6 10 1 0 1
** No changes occurred in the predicted and tested direction.
* P-value 0 1 2 ; for the next extreme table <0.05.
1 N = 36f35 = 71 in the independent sample, 9 in the dependent sample and 45+44 = 89 in
the combined samples. blanks = not significant at the specified U-value.
 1 64 Parents and Children as Pedestrians
TABLE 7
~PROVEMENTS IN ROADCROSSING BEHAVIOUR OF CHILDREN FOR THE TWO SAMPLES TOGETHER (N = 2 X 63),
THE INDEPENDENT SAMPLES (N = 2 X 48) AND THE DEPENDENTSAMPLE (N = 15)

Percentages or numbers of children displaying behaviour in pre- and

post-observations
Number and
% % % % Significance directions of Significance
pre- post- pre- post- at changes at
N = 63 N = 63 N = 48 N = 48 al = 0.05 N = 15 al = 0.20
Behaviour ND + -
Kerb
Stopping 46 64 44 67 0.02 4 2 2
HM left 35 46 35 42 10 7 3 0.17
HM right 41 60 42 60 0.05 5 4 1 0.19
HM backwards 3 18 2 13 0.06* 6 5 1 0 11
Road
Crossing at right angles 95 89 94 92 3 0 3
Making no false starts 98 95 98 96 1 0 1
Complete use of zebra' 87 91 83 94 ** 2 0 2
Normal tempo 78 79 81 83 6 3 3
HM left 33 30 35 33 7 4 3
HM right 51 46 56 44 7 5 2 0.23*
HM backwards 0 10 0 6 + 3 3 0 0.13
+ The most extreme table possible; P = 0.12.
The value of P as close as possible to u (see the section on statistical analysis).
** P-value 0.14; for the next extreme table <0.05.
1 N = 36+35 = 71 in the independent samples, N = 9 in the dependent sample and 45+44 = 89 in
the combined samples. blanks = not significant at the specified U-value.

Improvement of child behaviour (Table 7 )

The results of the independent sample and the dependent sample are more alike
than was the case with adults. Head movements to the left at the kerb did not signifi-
cantly change in the independent sample and stopping did not improve in the depen-
dent sample. Despite this, we may conclude with confidence that the four types of
kerb behaviour improved significantly. That the changes are considerable as well can
easily be seen from the percentages of the combined samples (N = 2 x 63). As two out
of six tests show that head movements on the road improved significantly, we conclude
that there was some carry-over effect from training head movements at the kerb to
head movements on the road. For the combined samples, however, this effect shows
up clearly only for head movements backwards. Just like the adults, children im-
proved in their use of the zebra (only in the independent sample and just not signifi-
cantly). As with the adults, the explicitly trained on-the-road behaviours (crossing at
right angles and normal tempo) did not improve.
Trafic influences
Momentary traffic was defined as present when bicyclists or motorised vehicles
passed when the pedestrian was at the kerb or when they approached while the
pedestrian was crossing the road. The presence of traffic correlated 0.38 and 0.36
with stopping of adults in the pre- and post-observations respectively (P < 0.01 with
one-sided Fisher exact test). Although the correlations are similar, inspection of the
two-fold tables showed that the programme was effective in changing, to some extent,
the adult habit of rarely stopping in the absence of traffic. During the pre-observa-
tions only 15 per cent of the adults stopped in the absence of traffic, which is only
one-third of what would be expected on the basis of independence of behaviour and
traffic. During the post-observations 41 per cent of them stopped in the absence of
 H. H. VAN DER MOLENET AL. 165
traffic, which is two-thirds of what would be expected on the basis of independence
(see our extended research report for details). Since there was less momentary traffic
during the post-observations (54 per cent) than during the pre-observations (68 per
cent), it can be argued that the positive effects of the training programme which we
found were in any case not induced by more momentary traffic in the post-observa-
tions. In fact we would have expected worse traffic behaviour in the post-observation
if no education programme had been administered.
The data presented have been based on observations at locations with a range of
traffic intensities between 12 and 1,OOO cars per hour. From Table 8 it becomes clear
that programme effects were not equally spread over locations with different in-
tensities. As far as stopping is concerned improvements occurred especially a t
locations with the lowest intensities, i.e., where stopping was not already elicited by
traffic during the pre-observations. This effect was stronger for children than for
adults. When considering head movements in all directions together (HM totals in
Table 8), a similar trend was found. However, when considering separate head
movement directions the results were less straightforward.
TABLE 8
PERCENTAGES IMPROVEMENT IN STOPPING AND HEADMOVEMENT^ FOR LOW, MEDIUM AND
HIGHTRAFFIC INTENSITIES (PRE-OBSERVATION PERCENTAGES IN BRACKETS)
Low intensity Medium High
Behaviour < 75 carslhour 75-200 carslhour z 200 carslhour
Adults (N = 13) (N = 35) (N = 15)
Stopping at kerb (8)+38% (49)+ 10% (60)+ 13%
HM-left at kerb (54)+ 15% +
(69) I3 % +
(73) 27 %
HM-right at kerb (46)+ 16% (71)- 6 % (93) 0 %
HM-backwards at kerb (8) 0% ( I ] ) + 18% (7)- 7 %
HM totals +31% +25% +20%
Children
Stopping at kerb
HM-left at kerb +
(8) 23
HM-right at kerb
HM-backwards at kerb
HM totals +61%

DISCUSSION
From this study we conclude that important aspects of the traffic behaviour of
both adults and pre-school children have improved after the child pedestrian training
programme, especially those aspects that were explicitly incorporated in the actual
training programme and which were not already performed quite well initially.
These effects showed up despite a decrease in momentary traffic and despite the fact
that only about 50 per cent of the parents and 70 per cent of the pre-school children
were actually involved in the training programme. However, the improvements in
behaviour did not reflect the entire repertoire of traffic behaviour and even the im-
proved behaviour was often far from ideal standards. Nevertheless, it is clear that
improvements could be induced and that pre-school children were not only able to
perform the correct traffic behaviour after application of the programme (Rothen-
gatter, 1981) but also that they showed more of the instructed behaviour in real life
situations afterwards, even when accompanied by adults. After the programme
166 Parents and Children as Pedestrians
adults not only set better examples but also gave significantly more verbal instructions
to which the children reacted by making head movements. Furthermore, trends
suggesting better protection of the children were found.
Compared with earlier empirical findings (Van der Molen, 1981b) we found in
the present study little evidence that accompanied children were dragged along by
their mothers as parcels, while completely unaware of the pedestrian task. We did
not corroborate our earlier findings that children who were led by the hand made
dramatically fewer head movements and that children’s stopping and making head
movements were unrelated. Explanations may be considered in terms of differences
in socio-economic level of parents in the two studies or in the development of public
awareness during the three years between the two studies.
Although adults appeared to behave far from ideally, they displayed better
traffic behaviour than children in many respects, both before and-to a somewhat
smaller extent-after the programme. We should bear in mind, however, that the
adults in this study were in most cases the mothers of the children they accompanied
and that recent research (Gerber, 1978; Mooy, 1980) indicates that mothers are
better pedestrians than other adults, especially in the company of children.
The question whether the adult is a model who is followed by the child she
accompanies can be answered affirmatively. We found positive correlations (12 out
of 14) between the same behaviour characteristics of adults and children they accom-
panied. These correlations were not artefacts of momentary traffic conditions.
During the pre-observations we found little evidence of communication about road
crossing procedures between adults and children (instruction). Therefore we conclude
that the correlations are indications of modelling.
It remains to be shown, however, that the children who are getting the better
examples will also display better behaviour when crossing alone or with friends. In
order to confirm this general belief (which is based on generalisations from psycho-
logical theory) observation studies are needed that relate modelling behaviour of
parents and behaviour of their children when walking unaccompanied.
In our detailed research report we have shown that several correlations between
child and adult behaviour in the post-observation (Table 2) were significantly greater
(tested post hoc, however) than in the pre-observations. These greater correlations in
the post-observation paralleled an increase in verbal communication (instruction
while crossing, and an increase of the level of performance for both children and
adults).
Rothengatter (1981) has argued that generalisation of his evaluation of the traffic
training programme to busier traffic environments is difficult. Our observations,
however, cover a broader range of traffic situations than the areas where his training
and evaluation procedures were carried out. Although generalisation from our
locations to other environments is still a rather complex issue, because traffic environ-
ments differ from each other in many respects, there is nevertheless some evidence
that behaviour patterns are rather consistent in different situations. This is apparent
from a comparison of earlier data (Van der Molen, 1981b) and the data found in this
study. On the other hand we found indications that learning effects are stronger in
quiet areas. Whether Hoogkerk is representative of other residential quarters in the
Netherlands or Great Britain is hard to prove, but we have no reason to believe that
the effects found would be different in other modern residential quarters with similar
area-design, traffic flow and socio-economic level of the population.
ACKNOWLEDGMENTs.-ThiS work was supported by a grant from the Directorate of
Traffic Safety-DVV and the institute for Road Safety Research-SWOV, as part of a
 H. H. VAN DER MOLENET AL. 167
research project on child traffic education. We wish t o thank the many colleagues, members
of the steering committee, assistants and students who contributed t o this study. They are
personally acknowledged in our detailed research report.

REFERENCES
BANDURA, A., and WALTERS, R. H. (1969). Social Learning and Personality Development. London1
New York: Holt, Rineliart and Winston.
BREWER, J. $. (1972). On the power of statistical tests in the ‘American Educational Research
Journal . Am. educ. Res. J., 9, 391-401.
BREWER, J. K., and OWEN,P. W. (1973). A note on the power of statistical tests in the ‘Journal of
Educational Measurement . J. edric. Measurement, 10, 71-74.
BRYAN, J., and WALBECK, N. H. (1970). The impact of words and deeds concerning altruism upon
children. Child Dev., 41, 747-757.
CHASE,L. J., and CHASE,R. B. (1976). A statistical power analysis of applied psychological research.
J. appl. Psychol., 61,234-237.
COHEN,J . (1977). Statistical Power Analysis for the Behavioural Sciences. New York: Academic
Press.
DOWNING, C. S., and SPENDLOVE, J. (1981). Effectiveness of a Campaign to Reduce Accidents Involving
Children Crossing Roads near Parked Cars. Transport and Road Research Laboratory Report
986, Crowthorne, U.K.
ELSTRODT,M., and MELLENBERGH, G. J. (1978). Een minus de vergeten fout. Nederlands Tijdschrift
voor de Psychologie, 33, 33-49.
GERBER, D. (1978). Der Erwachsene als Vorbild im Straszenverkehr: Einstellungeit und Verhalten.
Frankfurt am Main: Peter Lang.
GRAYSON, G. B. (1975a). Observations of Pedestrian Behaviour at Four Sites. TRRL Laboratory
Report 670, Transport and Road Research Laboratory, Crowthorne, U.K.
GRAYSON. J. B. (1975b). Hamoshire Child Pedestrian Accident Studv. TRRL Laboratow-~ Report
668, Transport and Road Research Laboratory, Crowthorne, U:K.
GRIZZLE,J. E. (1967). Continuity correction in the ,$-test for 2 x 2 tables. Am. Statistician,
October, 22-32.
GUILFORD, J. P., and FRUCHTER, B. (1978). Fundamental Statistics in Psychology and Education.
Tokyo: McGraw-Hill Kogakusha.
HERIK,J. VAN DEN, KLAAUW, C. VAN DER,and MOLEN,H. H. VAN DER (1981). Road Crossing
Behaviour of Children and AccomDanyinn
_ . _ Adults. Report VK 81-02, Traffic Research Centre,
University of Groningen.
HOOGSTRATEN, J., KOELE,P., and PLOEG.D. A. VAN DER (1980). Publikatie-gewoonten in Nederland.
Tddschrift voor Onderwijsresearch, 5 , 1, 3-8.
KROONENBERG, P. M., and VERBEEK,A. (1981). User’s Guide to Fischer. Department of Mathematics,
Preprint nr. 188. University of Utrecht.
LIMBOURG, M., and GERBER, D. (1978). Das Tubinger Elterntrainungsprogramm fur die Verkehrser-
ziehung vo: Kindern im Vorschul Alter. Zusammenfassender Bericht iiber das Forschungsproject
nr. 751 1 Die Rolle !on Erziehungspersonen in der Verkehrserziehung und Moglichkeiten
gezielter Einfluszname . Tubingen.
MATTSON,K., and LINDENSJ~, S. (1971). The Skandia Report: A Report on Children in Traffic.
Stockholm: Skandia Insurance Company.
MICHALIK, CH. (1975). Verhaltensbeobachtung 5-6,iahrigen Kindern im Straszenverkehr. Kuratorium
fur Verkehrssicherheit. Wien.
MOLEN,H. H. VAN DER (1981a). Child pedestrian’s exposure, accidents and behavior. Accident
Analysis and Prevention, 13 (3), 193-224.
MOLEN,H. H. VAN DER (1981b). Behavior of Child Pedestrians and Accompanying Adults at a Zebra
Crossing. Report VK 79-05, Traffic Research Centre, University of Groningen, Groningen.
MOLEN,H. H. VAN DER (1982a). Child Behavior when Performing Three Pedestrian Tasks. Report
VK 82-01, Traffic Research Centre, University of Groningen, Groningen.
MOLEN,H. H. VAN DER (1982b). Some Determinants of Child Pedestrian Task Performance and the
Effects of a Training Program. Traffic Research Centre report VK 82-02, University of Gron-
ingen, Groningen.
MOLEN,H. H. VAN DER,ROTHENGATTER, J. A., and VINJB,M. (1981). Blueprint of an analysis of the
pedestrian’s task, 1 : method of analysis. Accident Analysis and Prevention, 13 (3), 175-192.
MOLENAAR, W. (1970). Approximations to the Poisson, Binomial and Hypergeometric Distribution
Functions. Mathematical Centre Tracts No. 31, Mathematisch Centrum, Amsterdam.
MOLENAAR, W. (1977). Ik word ziek van de statistiek, of: ervan weten zonder ernaar te handelen.
Mens en Maatschappij, 52, 58-72.
Moou, J. (1980). Her neneren van rood licht door voetzanners.
- - Report WR 80-10, Traffic Research
Centre, University of Groningen, Groningen.
ROTHENGATTER, J. A. (1981). Traffic Education for Young Children; an Empirical Approach. Lisse:
Sets and Zeitlinger BV.
168 Parents and Children as Pedestrians
SANDELS,
S., (1974). Why are Children Injured in Traffic? Stockholm: Skandia Insurance Company.
SANDELS,
S., (1975). Children in Traffic. London: Paul Elek.
SIEGEL,
S. (1956). Nonparametric Statistics for the Behavioral Sciences. Tokyo: McGraw-Hill
Kogakusha.
SNJJDERS,
T. (1980). De X*-toets voor kruistabellen: konditioneren of niet? V.V.S. Bulletin, 2, 14;
16-21.
V I N J ~M.
, (1981). Children as pedestrians: abilities and limitations. Accident Analysis and
Prevention, 13, 225-240.
ZUMKLEY-MONKEL, C. (1976). Imitationslernen. Diisseldod: Schwann.
(Manuscript received 16th July, 1982)