- Email: [email protected]
Railway signals passed at danger
Applied Ergonomics 1988, 19.2, 135-141
Railway signals passed at danger Situational and personal factors underlying stop signal abuse H. van der Flier and W. Schoonman Department of Industrial Psychology, Dutch Railways, PO Box 2025, 3500 HA, Utrecht, The Netherlands
Based upon a systematic registration of cases of unauthorised passing through signals at danger (SPD cases) during the period 1 January 1983 to 1 January 1985, the contribution of situational and personal factors to the occurrence of this kind of safety error was studied. Exposural deviances were taken into account by presenting the relevant exposural details as well as the SPD frequencies. To study the effect of personal factors, the group of drivers involved was compared with an individually matched control group. Data analysis indicated that SPD cases tended to be fairly evenly dispersed over the months of the year. The distribution of incidents throughout the days of the week corresponded rather closely to the numbers of train driver shifts. Neither technical nor mechanical faults nor weather nor visibility conditions seemed to be significant contributory factors. SPD cases were relatively many during the morning hours (midnight till 6 am and 8 am till noon) and at the start of duty periods. At a number of (black) spots two or more cases were recorded accounting for a relatively high percentage of all cases. The most frequently mentioned hazard was that of the signal being situated behind a bend. About 90% of the cases took place at or near stations and most were with arriving trains. The direct cause of an SPD case often seemed to be that a signal was overlooked or not anticipated. Personal factors (age, time on duty, length of service or track and rolling-stock experience) did not seem to be important variables. However, it was proven that previous incidents, worse performance with multiple choice reaction tests, and less job satisfaction are predictive of a future SPD case. In this study the consequences of SPD cases were not as serious as they could be. Most damage was to points (forced open). Other consequences include level crossings not closed and a very small number of crashes. There were no fatalities and only one person was injured as a result of stop signal abusage.
Keywords: Train driving, errors, signals Introduction SPD Of all the visual signals that an NS (Nederlandse Spoorwegen (Dutch Railways)) engine-driver has to contend with while doing his work, the one of supreme importance is the red stop signal. Only in exceptional circumstances - and then only in situations explicitly set out in the regulations - is an engine-driver authorised to take his train through a stop signal at red. At the NS it has become usual to speak in terms of the 'Signal At Stop' (STS). In this paper, however, an instance where the stop signal is ignored will be referred to as an SPD (Signal Passed at Danger) case. Many an SPD case owes its origin to problems in the section leading up to the signal in question. Almost invariably a red signal is preceded by a yellow one which in turn may or may not be preceded by a 'reduce speed now' warning sign, generally to be interpreted as meaning: reduce speed and continue braking so that the train will be brought to a standstill at the stop signal. As a rule trains will automatically be brought
to a halt in cases where the trains and track sections are equipped with ATC (Automatic Train Control) when signals are not observed. When braking is too light this sort of back-up system offers no guarantee. All ATC can do is instigate the minimum amount of braking. Moreover, below a certain speed limit (40 km/h) it is feasible to keep travelling, without braking, provided that a so-called 'vigilance' button is pressed at 20 s intervals. SPD cases may also arise from: signals being reversed (returned to the stop position) by traffic control, brakes not functioning properly, or the track surface being too slippery Then, of course, there always remain the trains and railway line sections that are not protected by ATC where the guarantees that engine-driver errors will be without repercussions are considerably lower. Fortunately, SPD cases rarely lead to accidents but they do create situations of increased risk and may therefore serve as indicators of the degree to which the company is successful in maintaining safe transport.
0003 6870/88/02 0135-07 $03.00 © 1988 Butterworth & Co (Publishers) Ltd
Applied Ergonomics
June 1988
135
Literature Various disciplines (mechanics, medicine, psychology, ergonomics) have contributed to research into the causes of SPD cases. Study has especially centred upon driver errors, the most recurrent kinds being: incorrect expectations, responding to the wrong signal, preoccupation and distraction, inattentiveness and timing errors (Buck, 1963; Davis, 1966). The problem of a general lowering of driver vigilance in a monotonous work environment is also frequently mentioned (Sharp Grant, 1971; Endo and Kogi, 1975; Andrews, 1979) particularly in connection with shift work and irregular working hours (Kogi and Ohta, 1975; Hildebrandt et al, 1974). The possible influence of personality linked factors has also been explored. Davis (1966) found some differences between drivers who had been involved in SPD incidents and those in a control group where complaints of a psychiatric and a psychosomatic nature were concerned. Verhaegen and Rijkaert (1986) found significant positive correlations. between error frequencies (delayed reactions to yellow signals and speeding) on the one hand and the personality dimensions of extroversion and neuroticism or emotional instability on the other hand. Thus it would appear that the literature available does provide us with certain points relevant to the research into SPD cases. However, one can only conclude that all in all there is very little empirical material available. Despite the fact that for most railway companies it is common practice to carry out investigations when safety has been breached and have the findings put down in a report, it remains a fact that on the whole no centralised statistical collation of results takes place. Hence the fact that the publication of findings also remains undone.
The SPD project The objective of the SPD project was to find out what situational and personal factors accompany the unauthorised passing through a signal at danger. Knowledge of these factors and about how they interact is an essential prerequisite to forming an effective policy aimed at cutting back the number of SPD incidents. When examining situational and personal factors one must allow for exposural deviances. If at given times of the day the incidence of SPD cases is relatively high, this tells us little if, at these same times, relatively many trains arrive and depart. Likewise, if engine-drivers have in certain periods had few or relatively few intensive on-duty stints, they naturally have run less risk of passing a red signal than other colleagues. When analysing the situational factors, the exposural differences were accounted for - where desirable and possible - by presenting the exposural details as well as the SPD frequencies. Unfortunately, individual exposural data were not available. To study the effect of individual factors it was decided that a scheme be set up in which the SPD group would be compared with a control group matching the SPD group on a number of background variables thought to be of relevance.
Design of the study Collecting the infecmation: the SPD group The systematic registration of SPD cases commenced on 1 January 1983 with the aid of specially designed forms. In
136
AppliedErgonomics June 1988
the months leading up to that date the forms had been introduced in the various railway districts. During the period 1 January 1983 to 1 January 1985, the period covered by the study, 224 SPD cases were registered. Besides the SPD details for 1983 and 1984 ascertained from the forms that were circulated, some information on SPD incidents or previous years (1980 to 1982) was also collected. This involved the dates of the incidents and the registration numbers of the signals. For the cases in 1983 and 1984, the information drawn from the last psychological examination (PE) before the SPD incident of the engine-driver concerned was included in the data file. More often than not, these particulars were therefore derived from the last periodical examination (generally, NS engine-drivers are tested once in every five years, those above 50 years of age more frequently, to verify that they are still fit to perform their job in safety). Where files were incomplete (e g, because the engine-driver concerned had not been employed long enough to have received a periodical examination), the intake data (PE dating from recruitment time) was referred to instead.
PE variables The periodical check-up of engine drivers consists of a medical and a psychological" part. During the psychological examination three different tests are used. The tests are aimed at personal characteristics like 'concentration', 'fieldattention' and 'reaction'. One of the tests is the so-called 'Wiener Determinations Geriit' (DTG), a multi-choice reaction test. The DTG-unit has a display on which five different colours can be shown at different places. Two different tones can be given on a headphone. On the display are also two fixed lamps. The examinee reacts by pushing on coloured buttons or by pressing pedals with his feet. The test has machine and a self-paced series of stimuli. For a description of the other tests (Attention Diagnostic Method, Bourdon concentration test) see Visser et al (1983). The four possible outcomes of the psychological examination are: Suitable, Suitable with some reservations, Doubtful suitability, and Unsuitable. The categories Doubtful suitability and Unsuitable arose very seldom among this group of selected employees so in effect one was only dealing with two outcome categories. Apart from this overall outcome the scores on the separate sections of the test programme were also kept -- i e, the reaction, attention and concentration tests. Evaluations derived from an interview conducted by the psychologist or his co-worker at the time were also included. Finally, the results of a medical questionnaire were also entered into the data file.
Composition of the control group To discover which of the PE variables are predictive with respect to the incidence of SPD errors, it was decided that a control group should be formed. The following method was used. For each of the 210 engine-drivers (in the 1983 to 1984 period 14 were involved in two registered SPD cases) a 'comparison' engine-driver was sought. This comparison driver, or 'twin brother', had to possess similar characteristics to the SPD engine-driver in several respects, namely: railway district, qualifications/promotional status within the group, years of service and age. The selection of 'twin brothers' was based upon these four criteria since it was thought that in this way a number of exposural differences could be cancelled out. This matching also served to ensure that variation in age (which is
known to be related to reaction speed and concentration see, for example, Heinze (1983)) - w o u l d not affect the comparison of test achievements. For the control group the same PE (and medical examination) variables were put on file.
Table 1: Distribution of SPD cases (in 1983 and 1984 and for 1980 to 1984) and engine-driver on duty periods (person-duties) over days of the week
Comparisons between the SPD group and the control group were carried out by means of the t-test.
Day of the week
Total
%
Total
%
Total
%
Monday
34
15.2
80
16-3
1943
16"3
Tuesday
32
14.3
83
16-9
....
Results
The most important results of this study may be summed up in the following points: 1.
For the purposes of the SPD study the following cases of trains or parts of trains passing stop displaying signals were registered: 115 in 1983 and 109 in 1984. These numbers should be seen in relation to the 109 and 111 million train-kilometres for passenger and freight trains (taken together) registered in 1983 and 1984 respectively. It is not very clear whether, in comparison with previous years, the number of SPD cases had increased.
2.
The occurrence of SPD cases is fairly evenly distributed over the months of the year (see Fig. 1). The deviations between the SPD distribution and the distribution of numbers of days per month are not significant (×2(11 ) = 2.76). Contrary to what one might expect no winter month peak period exists. Taking the statistics for 1980 to 1984, the months of December, January and February produced a mere 21-2% of all the SPD cases while the months of June, July and August were responsible for 28"9% of SPD cases.
3.
The way in which SPD incidents are dispersed throughout the week does not deviate substantially from what one would expect when looking at the numbers of engine-driver duties (see Table 1).
4.
SPD case percentages are high for the morning hours in comparison with the numbers of arriving and departing trains during the same hours (especially from 10 o'clock till noon). During the evening rush hour and late evening the number of SPD incidents is relatively low. The deviations between the distributions of SPD cases and numbers of arriving and departing trains are significant at the 1% level (X2(23~= 42-83). The results (for 1983 and 1984) are giv~n'graphically in Fig. 2.
10 9
8
i
7 6 5 I 4 I 3 2 I
I
I
I
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Months Fig. 1
SPD cases by time of year for the period 1980/1984 (percentage of the total number)
SPD cases 1983/1984
SPD cases 1980 to 84
Duties *
Wednesday
34
15"2
73
14"9
....
Thursday
37
16"5
79
16-1
....
Friday
40
17.9
87
17"7
....
Saturday
24
10.7
44
9.0
1177
9.9
Sunday
23
10"3
45
9.2
1040
8.7
224
100.0
491
Total
100-0 11 930 100.0
Testing deviations STS cases 1983/1984 X2(6) = 1"98; ns from numbers of duties STS cases 1980 to 1984 X2t.~ = 1.80; ns *Based on the duty roster as of September 1984 (including shunting schedules) which is representative for the total period.
16 [-
'iF ,~
• L I
• SPD-cases
•_ =_, n Art.and r~ trains dep.
4
1
2 o
,~I l l I O0- 02- 0q- ~ i - 08- 10- 12- lq- 16- 18- 20- 220002 0q 06 08 I0 12 lq 16 18 20 22 24 Hour of the day
Fig. 2
SPD cases and a sample of arrival and departure trains for each hour of the day
5.
Where the years 1983 and 1984 are concerned, the direct effects of SPD cases were fortunately not as bad as one might imagine. The two most frequently mentioned after.effects were: points having been switched open by trains (43 times) and level crossings not having been closed on time (18 times). Six cases led to a crash or collision and five were the cause of derailment. There were no fatalities and only in one case were people wounded. Seven instances involved damage to rolling-stock and two gave rise to damage of the superstructure.
6.
Between 1980 and 1984, two or more SPD cases arose at each of 51 different signal spots. In all, 147 SPD cases - almost 30% of the grand total - took place at these 51 locations. As a point of interest, the Dutch railway network possesses 7000 light and arm signals
Applied Ergonomics
June 1988
137
that are able to display a 'stop' signal. Only two of the 51 signals were semaphore signals. The reason most often given as accounting for the reduced conspicuotrsness of a signal was that of being positioned after a bend (in 52 cases). Unfortunately, there is no information on the total number of signals positioned after a bend. 7.
8.
9.
10.
At least 90% of all registered SPD cases happened at marshalling yards and/or stations, half of these being with arrival trains. Departures proved to produce considerably fewer SPD cases. Less than 20% of the incidents at yards or stations involved departing trains. Slightly more than 20% of all SPD mishaps occurred during shunting manoeuvres. The commonest cause of SPD cases would appear to be that stop signals are either overlooked or are noticed too late. Other, more indirect, causes mentioned are: having incorrectly anticipated what the signal will be and having somehow been distracted. In approximately 10% of cases, faulty braking was indicated as one of the causes. Technical factors such as lack of braking power and track slipperiness only played a part in relatively few cases. In 16 instances (7' 1%) the conclusion was that the conductor had given an incorrect departure order. Finally, it appears that in rather a large proportion of the cases (29.0% of the total), compulsory use of the vigilance button was reported (at a maximum speed of 40 km/h see introduction). The results are produced in Table 2.
Table 2: Causes of SPD cases (1983/1984)
Causes/particular circumstances Engine-drivers: •
stop signal not noticed
71!
31.7
•
stop signal noticed too late
57
25.4
•
previous signal not noticed
15
6"7
•
incorrect anticipation of signal
24
10"7
•
having been distracted
25
11'2
•
faulty braking
24
10'7
Rolling stock: •
train braked too lightly
5
2-2
•
brake test had not been done
2
0-9
12
5"4
Track: •
Particular circumstances: second person in cabin
15
6"7
•
compulsory use of vig. button
65
29"0
•
wrong departure order from conductor
16
7' 1
*SPD incidents can be classified under more than one cause/ circumstance
ATC. If we confine our thoughts to 'ATC sections' we have grounds enough to demonstrate that ATC is advantageous. However, it remains a fact that the present ATC system cannot perfectly guarantee the prevention of SPD cases. 12.
The distribution of SPD cases over various types of traction units significantly deviates from the distribution of total numbers of units = 42'80; p < "01). But again, although no precise eklgosural data are available, the deviations largely reflect the way in which the units are utilised.
(×:fi2I
11.
138
Automatic Train Control apparatus will only function when both track stretches and rolling-stock are equipped with ATC. Going on the reports one may conclude that nearly half of all SPD cases (43.8%) occurred in situations where ATC was actually in operation (applying brakes). Because of the lack of company statistics on the frequency of the use of lines and trains with and without ATC (in combination), it is hard to establish whether with ATC there is decidedly less chance of an SPD mistake than without
Applied Ergonomics
June 1988
slippery rails
•
Adverse weather and adverse conditions such as twilight/darkness, poor visibility and the sudden onset of rain or fog did not contribute significantly to SPD incidents. As found in the enquiry, only 25.5% of cases took piace in twilight hours or hours of darkness. In 11-2% cases it was raining at the time (NB: the rainfall average in the Netherlands for 1983 and 1984 was 7-4% of the time) and in 1-3% cases it was loggy. Considerably more SPD incidents are caused by slow trains than by inter-cities (44"6% against 9.4%). Metropolitan trains in particular seem to be responsible for a relatively high number of SPD cases. However, when differences in numbers of trains and train schedules are taken into account the discrepancies can be explained by exposural deviances. Based upon a random sample from the timetable, the numbers of planned stops for slow trains and inter-cities stand in the relation of 6 to 1.
Frequency % *
Comparisons between SPD cases where ATC was and was not in use prove that, in ATC situations, there is relatively more often evidence of slightly overshooting (brakes having been applied before the stop signal was reached) and in non-ATC instances of completely shooting past (X20) = 4'64; p < .05). On the whole the distance overstiot in ATC situations was therefore also less (Differences between distributions of distances significant at I% level). Differences in the seriousness of the implications of SPD incidents where ATC was or was not in operation are hard to show because most of the consequences listed do not crop up often enough. The only definite difference is that in ATC situations the points are not pushed open so often 3"95; p < "05). Of course 'not having noticed tt16 previous signal' is less often mentioned as causing an SPD case where ATC is in operation.
(X2h)
13.
=
On the whole there is no clear relationship between SPD incidents, age and the engine-driver's length of service. Comparison of the distribution of age and years of experience as a driver within the SPD group and the total group of drivers results in a X2tR~value of 11.64 for age and a X2fio-ivalue of 16.64 for'length of service. Neither value i~ glgnificant. It is striking, however, that in the 40 to 45 age range the proportion of SPD cases is relatively high - see Fig. 3). Although no comparison data are available, factors such as
20 18 16 14 12 ~o 10
accounted for 31.7% of all SPD cases. The distribution of SPD cases and scheduled duties (Fig. 4) differ significantly at the 1% level (X2(19) = 41-09).
8 6 4 2 0
The dispersal of SPD cases throughout the hours of duty (see Fig. 5) also differs from what one would expect on the basis of a random sample of scheduled engine-driver duties (×2(s] = 28"54; p < "01). What is most striking is the fact'fhat relatively many SPD cases occur at the start of duty periods. At a first glance this percentage would only seem to relate to the 2nd and 3rd hours of the shift. However, one ought to bear in mind that much of the first hour on duty consists of waiting (pre-travelling time) and that the situation of most risk in terms of SPD, the arrival at the station or yard, often only arises for the first -time in the second half-hour of that initial hour on duty. The lunch-hour probably accounts for the low number of SPD instances in the 4th hour on duty. If one makes a breakdown of the distribution of SPD cases over duty hours by commencement of shifts, it becomes obvious that the relatively high proportion of SPD cases towards the start of shifts is recurrent with various types of shifts, especially the late ones. Another rather striking finding is that with the very early shifts (starting between 4.00 am and 6.00 am), SPD cases tend to increase towards the end of the shift. One must not rule out fatigue (insufficient night's sleep) as being at the root of this problem.
• .-25 25-30 30-35 35-40 40-45 45-50 50-55 55-60'60-.. Age group
SPD cases per age group and total numbers of engine-drivers per age group (percentages)
Fig. 3
30
• SPD-cases I-I Scheduled duties
25 20 ~o 15 10
04-06 06-0808-1010-1212-14 14-1616-18 18-20 20-22 22-24 Shift commencement
Fig. 4
15.
The engine-drivers who were involved in SPD cases in the 1983/1984 period had, relatively often, already committed such an offence in the two years prior to that period. Taking into account that in 1981/1982 there were 100 SPD cases a year and that about 3500 engine-drivers were engaged by NS, the average enginedriver, roughly speaking, would have had a chance of 200 : 3500 = 0.057 of being involved in an SPD case in this period. For the 210 engine-drivers (of which a small portion were still not driving unaccompanied in 1981/1982) this would amount to an expected total of 11 to 12 SPD cases. In reality this same group was involved in 21 SPD incidents in 1981/1982. To what degree this outcome links with exposural differences or individual personal factors is hard to
SPD cases and a random selection of scheduled engine-driver shifts according to shift commencement time (in percentages)
insufficient knowledge of the route or too little experience with the rolling stock in question do not seem to explain SPD mistakes either; 86-6% of the drivers in the SPD group had more than a year of experience on the route and the same percentage took the route at least weekly. As to the experience with rolling stock, the same picture arises; 91.5% of the drivers indicated that they used the kind of material at least once a week. 14.
It seems that SPD cases cannot, in the first place, be put down to long working periods on the part of the engine-driver (having had too many successive duty periods or shifts that were too long). How SPD cases are distributed over the accumulated number of onduty periods since the last day off differs little from what, taking a random selection of scheduled enginedriver shifts, one would expect (X2(7)= 9"89; ns). Engine-drivers working on early shifts were involved in relatively many SPD incidents in proportion to the number of scheduled duties (based on random samples). Engine-drivers going on duty between 4.00 am and 9.00 am (42.8% of the total) produced 50"9% of the SPD cases. For engine-drivers who did not go on duty until the afternoon (between 1.00 pm and 6.00 pm) a relatively low number of SPD cases was found. These later shifts (38"7% of all the scheduled shifts) only
25 20
I i /
~
--0- SPD-cases ~ Scheduled duties
NN~
15
Fig. 5
I
i
l
~
2
3
4
5 6 7 On d u t y hours
8
9
1
1
10
11
SPD cases and a random sample of scheduled engine-driver shifts by duty hours
Applied Ergonomics
June 1988
139
establish from these results. In any case, the information gathered does not exclude the possibility that personal factors do play some part. 16.
A comparison between the test results of the SPD group derived from the last periodical examination and the control group did bring to light certain differences. These differences gain significance and, in quite a few cases, become statistically meaningful when we put two particular SPD groups alongside the control group. The groups concerned are: 'the second time offenders' (engine-drivers who had already been involved in an earlier case) and the group of those who, as has been determined, 'saw the stop signal too late or not at all'. Both these SPD groups had lower ratings on the multi-choice reaction test. The DTG-test has three machine-paced series (inter-stimulus intervals 1-1, 1 "0 and 0"8 s) and two self-paced series. The machine-paced series are scored in number correct, late, wrong or no responses. In both groups significant t-values were found in the number of wrong responges in the last two conditions. In the two self-paced series, differences in number of wrong responses were in the expected direction (p-values between -10 and -29). When the final conclusions of the periodical examinations were compared differences also became apparent. For the SPD case group the category 'suitable with some reservations' came up more frequently and the category 'suitable' less frequently. The percentages for both groups with results other than 'suitable' were: 21 "0% (SPD group) and 15'7% (control group). This tendency is stronger in the two sub-groups just mentioned. Within the groups of those who noticed the stop signal too late or not at all the percentages with a classification other than 'suitable' crept up to 29-1%, and within the groups of second time offenders it was 26"4%. However, the general conclusion to be drawn must be that in tightening up the standards of the present periodical check-up only a limited preventative effect can be expected.
17.
On comparing the data from interviews several slight differences emerge. Individuals in the SPD group gain slightly less job satisfaction. They give rise to more reports of previous irregularities and/or calamities (41 vs 34%). With this group mention is rather more often made of exceptional circumstances (38 vs 31%). The groups did not differ on the medical questionnaire completed during periodical company medical checkups.
Conclusion There was a rather negative side to the results of the investigation, which was that no substantial correlation was found between certain personal and situational variables and the occurrence of SPD incidents. This would imply that from the steps and measures taken relating to the different variables little effect is to be expected. In addition to this a number o f results relate to variables that are difficult to influence within the limitations created by the implementation of the timetable. Here, too, the possibilities for reducing the number of SPD cases are very few. In the first place it would seem that the answer to the SPD problem must be sought in the technical side of matters, in particular
140
AppliedErgonomics June 1988
in refining the ATC system still further. In its present Iorm this system leaves responding to a stop signal in time entirely up to the engine-driver even though it does provide a safe alternative for not reacting to the yellow signal and, with long-lasting use of the vigilance button, it can even lead to an SPD incident. in the immediate future the priority should lie in rectifying hazardous signals. Where signals have been the cause of more than two SPD cases in recent years and have not in any way been adjusted since then it would seem logical to recommend that these be investigated. Even when alterations appear to be impossible or undesirable (and only too often the concern that the problem will merely be transferred to elsewhere is not ungrounded) it is still possible to devote extra attention to these signals .... for example, in training sessions. At any rate it is important that, as was done for the purposes of this study, the SPD cases per actual signal continue to be registered in the tuture. How SPD cases were found to be distributed between the beginning of shifts and engine-driver's duty hours would also be reason enough to prompt further research. In conjunction with this it would be vital to gather more detailed information on preceding duty periods. If certain shift patterns prove to brifig with them added risk, this could be taken into consideration by personal supervisors to the engine-drivers, or those involved in drawing up the timetables. Person-linked factors were found to be of relevance: a previous SPD record tended to warn of more SPD mistakes to come and the SPD case chances (in particular, because of stop signals not having been seen or having been seen too late) do tend to correlate with the outcome of a person's last periodical psychological check-up (especially the part of the multi-choice reaction test). However, the relationships in this selected group were rather weak and, apart from where medical unfitness was involved, only combinations of factors such as: a previous SPD record, a PE result other than 'suitable', and negative information about job performance could provide reason enough for taking drastic steps. Other than improving upon standards, starting with the present psychological examination, one could consider concentrating more upon personal characteristics during the check-up which, up until now, seem to have received too little attention. The points particularly being thought of are the capacity for 'quick and accurate observation' and the ability to remain alert for long periods in relatively monotonous situations (vigilance). Looking at the results of this study it would certainly seem worthwhile to recommend further research into the predictive nature of these two personal characteristics with respect to SPD mistakes. The intention is to begin such a study in the near future.
References Andrews, M.J.
1979, Human factors in train operation. Proceedings of the Institute of Railway Signalling Engineers, London.
Buck, L. 1963, Errors in the perception of railway signals. Ergonomics, 6, 181- 192. Davis, R., 1966, Railway signals passed at danger: the drivers, circumstances and psychological processes. Ergonomics, 9, 2 1 1 - 2 2 2 .
Endo, T., and Kogi, K. 1975, Monotony effects of the work of motormen during high speed train operation. Journal o/Human Ergology, 4, 129-140. Heinze, E.E. 1983, Vigilanzforschung bei LokfLihrern, aueh hinsichtlich des Alters. Wissenschaftliche Konferenz der UIMC, Mtlnchen. Hfldebrandt, G., Rohmert, W., and Rutenfranz, J. 1974, Twelve and twenty four hour rhythms in error frequency of locomotive drivers and the influence of tiredness. In ternational Journal of Chronobiology, 2, 175-180. Kogi, K., and Ohta, T. 1975, Incidence of near accidental drowsiness in locomotive driving during a period of rotation. Journal o/Human Ergology, 4,65-76.
Sharp Grant, J. 1971, Concepts of fatigue and vigilance in relation to railway operation. In: K. Hashimoto, K. Kogi and E. Grandjean (Eds), Methodology in human fatigue performance. Taylor and Francis, London. Verhaegen, P.K., and Rijkaert, R.W. 1986, Vigilance of train engineers. Proceedings of the Human Factors Society, 30th annual meeting
Visser, R.S.H., van Vliet-Mulder, J.C., Evers, A., and ter Laak, J. 1983, Documentatie van tests en testresearch in Nederland - 1982 (Documentation on tests and testresearch in the Netherlands - 1982). NIP, Amsterdam.
Applied Ergonomics
June 1988
141
Railway signals passed at danger Situational and personal factors underlying stop signal abuse H. van der Flier and W. Schoonman Department of Industrial Psychology, Dutch Railways, PO Box 2025, 3500 HA, Utrecht, The Netherlands
Based upon a systematic registration of cases of unauthorised passing through signals at danger (SPD cases) during the period 1 January 1983 to 1 January 1985, the contribution of situational and personal factors to the occurrence of this kind of safety error was studied. Exposural deviances were taken into account by presenting the relevant exposural details as well as the SPD frequencies. To study the effect of personal factors, the group of drivers involved was compared with an individually matched control group. Data analysis indicated that SPD cases tended to be fairly evenly dispersed over the months of the year. The distribution of incidents throughout the days of the week corresponded rather closely to the numbers of train driver shifts. Neither technical nor mechanical faults nor weather nor visibility conditions seemed to be significant contributory factors. SPD cases were relatively many during the morning hours (midnight till 6 am and 8 am till noon) and at the start of duty periods. At a number of (black) spots two or more cases were recorded accounting for a relatively high percentage of all cases. The most frequently mentioned hazard was that of the signal being situated behind a bend. About 90% of the cases took place at or near stations and most were with arriving trains. The direct cause of an SPD case often seemed to be that a signal was overlooked or not anticipated. Personal factors (age, time on duty, length of service or track and rolling-stock experience) did not seem to be important variables. However, it was proven that previous incidents, worse performance with multiple choice reaction tests, and less job satisfaction are predictive of a future SPD case. In this study the consequences of SPD cases were not as serious as they could be. Most damage was to points (forced open). Other consequences include level crossings not closed and a very small number of crashes. There were no fatalities and only one person was injured as a result of stop signal abusage.
Keywords: Train driving, errors, signals Introduction SPD Of all the visual signals that an NS (Nederlandse Spoorwegen (Dutch Railways)) engine-driver has to contend with while doing his work, the one of supreme importance is the red stop signal. Only in exceptional circumstances - and then only in situations explicitly set out in the regulations - is an engine-driver authorised to take his train through a stop signal at red. At the NS it has become usual to speak in terms of the 'Signal At Stop' (STS). In this paper, however, an instance where the stop signal is ignored will be referred to as an SPD (Signal Passed at Danger) case. Many an SPD case owes its origin to problems in the section leading up to the signal in question. Almost invariably a red signal is preceded by a yellow one which in turn may or may not be preceded by a 'reduce speed now' warning sign, generally to be interpreted as meaning: reduce speed and continue braking so that the train will be brought to a standstill at the stop signal. As a rule trains will automatically be brought
to a halt in cases where the trains and track sections are equipped with ATC (Automatic Train Control) when signals are not observed. When braking is too light this sort of back-up system offers no guarantee. All ATC can do is instigate the minimum amount of braking. Moreover, below a certain speed limit (40 km/h) it is feasible to keep travelling, without braking, provided that a so-called 'vigilance' button is pressed at 20 s intervals. SPD cases may also arise from: signals being reversed (returned to the stop position) by traffic control, brakes not functioning properly, or the track surface being too slippery Then, of course, there always remain the trains and railway line sections that are not protected by ATC where the guarantees that engine-driver errors will be without repercussions are considerably lower. Fortunately, SPD cases rarely lead to accidents but they do create situations of increased risk and may therefore serve as indicators of the degree to which the company is successful in maintaining safe transport.
0003 6870/88/02 0135-07 $03.00 © 1988 Butterworth & Co (Publishers) Ltd
Applied Ergonomics
June 1988
135
Literature Various disciplines (mechanics, medicine, psychology, ergonomics) have contributed to research into the causes of SPD cases. Study has especially centred upon driver errors, the most recurrent kinds being: incorrect expectations, responding to the wrong signal, preoccupation and distraction, inattentiveness and timing errors (Buck, 1963; Davis, 1966). The problem of a general lowering of driver vigilance in a monotonous work environment is also frequently mentioned (Sharp Grant, 1971; Endo and Kogi, 1975; Andrews, 1979) particularly in connection with shift work and irregular working hours (Kogi and Ohta, 1975; Hildebrandt et al, 1974). The possible influence of personality linked factors has also been explored. Davis (1966) found some differences between drivers who had been involved in SPD incidents and those in a control group where complaints of a psychiatric and a psychosomatic nature were concerned. Verhaegen and Rijkaert (1986) found significant positive correlations. between error frequencies (delayed reactions to yellow signals and speeding) on the one hand and the personality dimensions of extroversion and neuroticism or emotional instability on the other hand. Thus it would appear that the literature available does provide us with certain points relevant to the research into SPD cases. However, one can only conclude that all in all there is very little empirical material available. Despite the fact that for most railway companies it is common practice to carry out investigations when safety has been breached and have the findings put down in a report, it remains a fact that on the whole no centralised statistical collation of results takes place. Hence the fact that the publication of findings also remains undone.
The SPD project The objective of the SPD project was to find out what situational and personal factors accompany the unauthorised passing through a signal at danger. Knowledge of these factors and about how they interact is an essential prerequisite to forming an effective policy aimed at cutting back the number of SPD incidents. When examining situational and personal factors one must allow for exposural deviances. If at given times of the day the incidence of SPD cases is relatively high, this tells us little if, at these same times, relatively many trains arrive and depart. Likewise, if engine-drivers have in certain periods had few or relatively few intensive on-duty stints, they naturally have run less risk of passing a red signal than other colleagues. When analysing the situational factors, the exposural differences were accounted for - where desirable and possible - by presenting the exposural details as well as the SPD frequencies. Unfortunately, individual exposural data were not available. To study the effect of individual factors it was decided that a scheme be set up in which the SPD group would be compared with a control group matching the SPD group on a number of background variables thought to be of relevance.
Design of the study Collecting the infecmation: the SPD group The systematic registration of SPD cases commenced on 1 January 1983 with the aid of specially designed forms. In
136
AppliedErgonomics June 1988
the months leading up to that date the forms had been introduced in the various railway districts. During the period 1 January 1983 to 1 January 1985, the period covered by the study, 224 SPD cases were registered. Besides the SPD details for 1983 and 1984 ascertained from the forms that were circulated, some information on SPD incidents or previous years (1980 to 1982) was also collected. This involved the dates of the incidents and the registration numbers of the signals. For the cases in 1983 and 1984, the information drawn from the last psychological examination (PE) before the SPD incident of the engine-driver concerned was included in the data file. More often than not, these particulars were therefore derived from the last periodical examination (generally, NS engine-drivers are tested once in every five years, those above 50 years of age more frequently, to verify that they are still fit to perform their job in safety). Where files were incomplete (e g, because the engine-driver concerned had not been employed long enough to have received a periodical examination), the intake data (PE dating from recruitment time) was referred to instead.
PE variables The periodical check-up of engine drivers consists of a medical and a psychological" part. During the psychological examination three different tests are used. The tests are aimed at personal characteristics like 'concentration', 'fieldattention' and 'reaction'. One of the tests is the so-called 'Wiener Determinations Geriit' (DTG), a multi-choice reaction test. The DTG-unit has a display on which five different colours can be shown at different places. Two different tones can be given on a headphone. On the display are also two fixed lamps. The examinee reacts by pushing on coloured buttons or by pressing pedals with his feet. The test has machine and a self-paced series of stimuli. For a description of the other tests (Attention Diagnostic Method, Bourdon concentration test) see Visser et al (1983). The four possible outcomes of the psychological examination are: Suitable, Suitable with some reservations, Doubtful suitability, and Unsuitable. The categories Doubtful suitability and Unsuitable arose very seldom among this group of selected employees so in effect one was only dealing with two outcome categories. Apart from this overall outcome the scores on the separate sections of the test programme were also kept -- i e, the reaction, attention and concentration tests. Evaluations derived from an interview conducted by the psychologist or his co-worker at the time were also included. Finally, the results of a medical questionnaire were also entered into the data file.
Composition of the control group To discover which of the PE variables are predictive with respect to the incidence of SPD errors, it was decided that a control group should be formed. The following method was used. For each of the 210 engine-drivers (in the 1983 to 1984 period 14 were involved in two registered SPD cases) a 'comparison' engine-driver was sought. This comparison driver, or 'twin brother', had to possess similar characteristics to the SPD engine-driver in several respects, namely: railway district, qualifications/promotional status within the group, years of service and age. The selection of 'twin brothers' was based upon these four criteria since it was thought that in this way a number of exposural differences could be cancelled out. This matching also served to ensure that variation in age (which is
known to be related to reaction speed and concentration see, for example, Heinze (1983)) - w o u l d not affect the comparison of test achievements. For the control group the same PE (and medical examination) variables were put on file.
Table 1: Distribution of SPD cases (in 1983 and 1984 and for 1980 to 1984) and engine-driver on duty periods (person-duties) over days of the week
Comparisons between the SPD group and the control group were carried out by means of the t-test.
Day of the week
Total
%
Total
%
Total
%
Monday
34
15.2
80
16-3
1943
16"3
Tuesday
32
14.3
83
16-9
....
Results
The most important results of this study may be summed up in the following points: 1.
For the purposes of the SPD study the following cases of trains or parts of trains passing stop displaying signals were registered: 115 in 1983 and 109 in 1984. These numbers should be seen in relation to the 109 and 111 million train-kilometres for passenger and freight trains (taken together) registered in 1983 and 1984 respectively. It is not very clear whether, in comparison with previous years, the number of SPD cases had increased.
2.
The occurrence of SPD cases is fairly evenly distributed over the months of the year (see Fig. 1). The deviations between the SPD distribution and the distribution of numbers of days per month are not significant (×2(11 ) = 2.76). Contrary to what one might expect no winter month peak period exists. Taking the statistics for 1980 to 1984, the months of December, January and February produced a mere 21-2% of all the SPD cases while the months of June, July and August were responsible for 28"9% of SPD cases.
3.
The way in which SPD incidents are dispersed throughout the week does not deviate substantially from what one would expect when looking at the numbers of engine-driver duties (see Table 1).
4.
SPD case percentages are high for the morning hours in comparison with the numbers of arriving and departing trains during the same hours (especially from 10 o'clock till noon). During the evening rush hour and late evening the number of SPD incidents is relatively low. The deviations between the distributions of SPD cases and numbers of arriving and departing trains are significant at the 1% level (X2(23~= 42-83). The results (for 1983 and 1984) are giv~n'graphically in Fig. 2.
10 9
8
i
7 6 5 I 4 I 3 2 I
I
I
I
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Months Fig. 1
SPD cases by time of year for the period 1980/1984 (percentage of the total number)
SPD cases 1983/1984
SPD cases 1980 to 84
Duties *
Wednesday
34
15"2
73
14"9
....
Thursday
37
16"5
79
16-1
....
Friday
40
17.9
87
17"7
....
Saturday
24
10.7
44
9.0
1177
9.9
Sunday
23
10"3
45
9.2
1040
8.7
224
100.0
491
Total
100-0 11 930 100.0
Testing deviations STS cases 1983/1984 X2(6) = 1"98; ns from numbers of duties STS cases 1980 to 1984 X2t.~ = 1.80; ns *Based on the duty roster as of September 1984 (including shunting schedules) which is representative for the total period.
16 [-
'iF ,~
• L I
• SPD-cases
•_ =_, n Art.and r~ trains dep.
4
1
2 o
,~I l l I O0- 02- 0q- ~ i - 08- 10- 12- lq- 16- 18- 20- 220002 0q 06 08 I0 12 lq 16 18 20 22 24 Hour of the day
Fig. 2
SPD cases and a sample of arrival and departure trains for each hour of the day
5.
Where the years 1983 and 1984 are concerned, the direct effects of SPD cases were fortunately not as bad as one might imagine. The two most frequently mentioned after.effects were: points having been switched open by trains (43 times) and level crossings not having been closed on time (18 times). Six cases led to a crash or collision and five were the cause of derailment. There were no fatalities and only in one case were people wounded. Seven instances involved damage to rolling-stock and two gave rise to damage of the superstructure.
6.
Between 1980 and 1984, two or more SPD cases arose at each of 51 different signal spots. In all, 147 SPD cases - almost 30% of the grand total - took place at these 51 locations. As a point of interest, the Dutch railway network possesses 7000 light and arm signals
Applied Ergonomics
June 1988
137
that are able to display a 'stop' signal. Only two of the 51 signals were semaphore signals. The reason most often given as accounting for the reduced conspicuotrsness of a signal was that of being positioned after a bend (in 52 cases). Unfortunately, there is no information on the total number of signals positioned after a bend. 7.
8.
9.
10.
At least 90% of all registered SPD cases happened at marshalling yards and/or stations, half of these being with arrival trains. Departures proved to produce considerably fewer SPD cases. Less than 20% of the incidents at yards or stations involved departing trains. Slightly more than 20% of all SPD mishaps occurred during shunting manoeuvres. The commonest cause of SPD cases would appear to be that stop signals are either overlooked or are noticed too late. Other, more indirect, causes mentioned are: having incorrectly anticipated what the signal will be and having somehow been distracted. In approximately 10% of cases, faulty braking was indicated as one of the causes. Technical factors such as lack of braking power and track slipperiness only played a part in relatively few cases. In 16 instances (7' 1%) the conclusion was that the conductor had given an incorrect departure order. Finally, it appears that in rather a large proportion of the cases (29.0% of the total), compulsory use of the vigilance button was reported (at a maximum speed of 40 km/h see introduction). The results are produced in Table 2.
Table 2: Causes of SPD cases (1983/1984)
Causes/particular circumstances Engine-drivers: •
stop signal not noticed
71!
31.7
•
stop signal noticed too late
57
25.4
•
previous signal not noticed
15
6"7
•
incorrect anticipation of signal
24
10"7
•
having been distracted
25
11'2
•
faulty braking
24
10'7
Rolling stock: •
train braked too lightly
5
2-2
•
brake test had not been done
2
0-9
12
5"4
Track: •
Particular circumstances: second person in cabin
15
6"7
•
compulsory use of vig. button
65
29"0
•
wrong departure order from conductor
16
7' 1
*SPD incidents can be classified under more than one cause/ circumstance
ATC. If we confine our thoughts to 'ATC sections' we have grounds enough to demonstrate that ATC is advantageous. However, it remains a fact that the present ATC system cannot perfectly guarantee the prevention of SPD cases. 12.
The distribution of SPD cases over various types of traction units significantly deviates from the distribution of total numbers of units = 42'80; p < "01). But again, although no precise eklgosural data are available, the deviations largely reflect the way in which the units are utilised.
(×:fi2I
11.
138
Automatic Train Control apparatus will only function when both track stretches and rolling-stock are equipped with ATC. Going on the reports one may conclude that nearly half of all SPD cases (43.8%) occurred in situations where ATC was actually in operation (applying brakes). Because of the lack of company statistics on the frequency of the use of lines and trains with and without ATC (in combination), it is hard to establish whether with ATC there is decidedly less chance of an SPD mistake than without
Applied Ergonomics
June 1988
slippery rails
•
Adverse weather and adverse conditions such as twilight/darkness, poor visibility and the sudden onset of rain or fog did not contribute significantly to SPD incidents. As found in the enquiry, only 25.5% of cases took piace in twilight hours or hours of darkness. In 11-2% cases it was raining at the time (NB: the rainfall average in the Netherlands for 1983 and 1984 was 7-4% of the time) and in 1-3% cases it was loggy. Considerably more SPD incidents are caused by slow trains than by inter-cities (44"6% against 9.4%). Metropolitan trains in particular seem to be responsible for a relatively high number of SPD cases. However, when differences in numbers of trains and train schedules are taken into account the discrepancies can be explained by exposural deviances. Based upon a random sample from the timetable, the numbers of planned stops for slow trains and inter-cities stand in the relation of 6 to 1.
Frequency % *
Comparisons between SPD cases where ATC was and was not in use prove that, in ATC situations, there is relatively more often evidence of slightly overshooting (brakes having been applied before the stop signal was reached) and in non-ATC instances of completely shooting past (X20) = 4'64; p < .05). On the whole the distance overstiot in ATC situations was therefore also less (Differences between distributions of distances significant at I% level). Differences in the seriousness of the implications of SPD incidents where ATC was or was not in operation are hard to show because most of the consequences listed do not crop up often enough. The only definite difference is that in ATC situations the points are not pushed open so often 3"95; p < "05). Of course 'not having noticed tt16 previous signal' is less often mentioned as causing an SPD case where ATC is in operation.
(X2h)
13.
=
On the whole there is no clear relationship between SPD incidents, age and the engine-driver's length of service. Comparison of the distribution of age and years of experience as a driver within the SPD group and the total group of drivers results in a X2tR~value of 11.64 for age and a X2fio-ivalue of 16.64 for'length of service. Neither value i~ glgnificant. It is striking, however, that in the 40 to 45 age range the proportion of SPD cases is relatively high - see Fig. 3). Although no comparison data are available, factors such as
20 18 16 14 12 ~o 10
accounted for 31.7% of all SPD cases. The distribution of SPD cases and scheduled duties (Fig. 4) differ significantly at the 1% level (X2(19) = 41-09).
8 6 4 2 0
The dispersal of SPD cases throughout the hours of duty (see Fig. 5) also differs from what one would expect on the basis of a random sample of scheduled engine-driver duties (×2(s] = 28"54; p < "01). What is most striking is the fact'fhat relatively many SPD cases occur at the start of duty periods. At a first glance this percentage would only seem to relate to the 2nd and 3rd hours of the shift. However, one ought to bear in mind that much of the first hour on duty consists of waiting (pre-travelling time) and that the situation of most risk in terms of SPD, the arrival at the station or yard, often only arises for the first -time in the second half-hour of that initial hour on duty. The lunch-hour probably accounts for the low number of SPD instances in the 4th hour on duty. If one makes a breakdown of the distribution of SPD cases over duty hours by commencement of shifts, it becomes obvious that the relatively high proportion of SPD cases towards the start of shifts is recurrent with various types of shifts, especially the late ones. Another rather striking finding is that with the very early shifts (starting between 4.00 am and 6.00 am), SPD cases tend to increase towards the end of the shift. One must not rule out fatigue (insufficient night's sleep) as being at the root of this problem.
• .-25 25-30 30-35 35-40 40-45 45-50 50-55 55-60'60-.. Age group
SPD cases per age group and total numbers of engine-drivers per age group (percentages)
Fig. 3
30
• SPD-cases I-I Scheduled duties
25 20 ~o 15 10
04-06 06-0808-1010-1212-14 14-1616-18 18-20 20-22 22-24 Shift commencement
Fig. 4
15.
The engine-drivers who were involved in SPD cases in the 1983/1984 period had, relatively often, already committed such an offence in the two years prior to that period. Taking into account that in 1981/1982 there were 100 SPD cases a year and that about 3500 engine-drivers were engaged by NS, the average enginedriver, roughly speaking, would have had a chance of 200 : 3500 = 0.057 of being involved in an SPD case in this period. For the 210 engine-drivers (of which a small portion were still not driving unaccompanied in 1981/1982) this would amount to an expected total of 11 to 12 SPD cases. In reality this same group was involved in 21 SPD incidents in 1981/1982. To what degree this outcome links with exposural differences or individual personal factors is hard to
SPD cases and a random selection of scheduled engine-driver shifts according to shift commencement time (in percentages)
insufficient knowledge of the route or too little experience with the rolling stock in question do not seem to explain SPD mistakes either; 86-6% of the drivers in the SPD group had more than a year of experience on the route and the same percentage took the route at least weekly. As to the experience with rolling stock, the same picture arises; 91.5% of the drivers indicated that they used the kind of material at least once a week. 14.
It seems that SPD cases cannot, in the first place, be put down to long working periods on the part of the engine-driver (having had too many successive duty periods or shifts that were too long). How SPD cases are distributed over the accumulated number of onduty periods since the last day off differs little from what, taking a random selection of scheduled enginedriver shifts, one would expect (X2(7)= 9"89; ns). Engine-drivers working on early shifts were involved in relatively many SPD incidents in proportion to the number of scheduled duties (based on random samples). Engine-drivers going on duty between 4.00 am and 9.00 am (42.8% of the total) produced 50"9% of the SPD cases. For engine-drivers who did not go on duty until the afternoon (between 1.00 pm and 6.00 pm) a relatively low number of SPD cases was found. These later shifts (38"7% of all the scheduled shifts) only
25 20
I i /
~
--0- SPD-cases ~ Scheduled duties
NN~
15
Fig. 5
I
i
l
~
2
3
4
5 6 7 On d u t y hours
8
9
1
1
10
11
SPD cases and a random sample of scheduled engine-driver shifts by duty hours
Applied Ergonomics
June 1988
139
establish from these results. In any case, the information gathered does not exclude the possibility that personal factors do play some part. 16.
A comparison between the test results of the SPD group derived from the last periodical examination and the control group did bring to light certain differences. These differences gain significance and, in quite a few cases, become statistically meaningful when we put two particular SPD groups alongside the control group. The groups concerned are: 'the second time offenders' (engine-drivers who had already been involved in an earlier case) and the group of those who, as has been determined, 'saw the stop signal too late or not at all'. Both these SPD groups had lower ratings on the multi-choice reaction test. The DTG-test has three machine-paced series (inter-stimulus intervals 1-1, 1 "0 and 0"8 s) and two self-paced series. The machine-paced series are scored in number correct, late, wrong or no responses. In both groups significant t-values were found in the number of wrong responges in the last two conditions. In the two self-paced series, differences in number of wrong responses were in the expected direction (p-values between -10 and -29). When the final conclusions of the periodical examinations were compared differences also became apparent. For the SPD case group the category 'suitable with some reservations' came up more frequently and the category 'suitable' less frequently. The percentages for both groups with results other than 'suitable' were: 21 "0% (SPD group) and 15'7% (control group). This tendency is stronger in the two sub-groups just mentioned. Within the groups of those who noticed the stop signal too late or not at all the percentages with a classification other than 'suitable' crept up to 29-1%, and within the groups of second time offenders it was 26"4%. However, the general conclusion to be drawn must be that in tightening up the standards of the present periodical check-up only a limited preventative effect can be expected.
17.
On comparing the data from interviews several slight differences emerge. Individuals in the SPD group gain slightly less job satisfaction. They give rise to more reports of previous irregularities and/or calamities (41 vs 34%). With this group mention is rather more often made of exceptional circumstances (38 vs 31%). The groups did not differ on the medical questionnaire completed during periodical company medical checkups.
Conclusion There was a rather negative side to the results of the investigation, which was that no substantial correlation was found between certain personal and situational variables and the occurrence of SPD incidents. This would imply that from the steps and measures taken relating to the different variables little effect is to be expected. In addition to this a number o f results relate to variables that are difficult to influence within the limitations created by the implementation of the timetable. Here, too, the possibilities for reducing the number of SPD cases are very few. In the first place it would seem that the answer to the SPD problem must be sought in the technical side of matters, in particular
140
AppliedErgonomics June 1988
in refining the ATC system still further. In its present Iorm this system leaves responding to a stop signal in time entirely up to the engine-driver even though it does provide a safe alternative for not reacting to the yellow signal and, with long-lasting use of the vigilance button, it can even lead to an SPD incident. in the immediate future the priority should lie in rectifying hazardous signals. Where signals have been the cause of more than two SPD cases in recent years and have not in any way been adjusted since then it would seem logical to recommend that these be investigated. Even when alterations appear to be impossible or undesirable (and only too often the concern that the problem will merely be transferred to elsewhere is not ungrounded) it is still possible to devote extra attention to these signals .... for example, in training sessions. At any rate it is important that, as was done for the purposes of this study, the SPD cases per actual signal continue to be registered in the tuture. How SPD cases were found to be distributed between the beginning of shifts and engine-driver's duty hours would also be reason enough to prompt further research. In conjunction with this it would be vital to gather more detailed information on preceding duty periods. If certain shift patterns prove to brifig with them added risk, this could be taken into consideration by personal supervisors to the engine-drivers, or those involved in drawing up the timetables. Person-linked factors were found to be of relevance: a previous SPD record tended to warn of more SPD mistakes to come and the SPD case chances (in particular, because of stop signals not having been seen or having been seen too late) do tend to correlate with the outcome of a person's last periodical psychological check-up (especially the part of the multi-choice reaction test). However, the relationships in this selected group were rather weak and, apart from where medical unfitness was involved, only combinations of factors such as: a previous SPD record, a PE result other than 'suitable', and negative information about job performance could provide reason enough for taking drastic steps. Other than improving upon standards, starting with the present psychological examination, one could consider concentrating more upon personal characteristics during the check-up which, up until now, seem to have received too little attention. The points particularly being thought of are the capacity for 'quick and accurate observation' and the ability to remain alert for long periods in relatively monotonous situations (vigilance). Looking at the results of this study it would certainly seem worthwhile to recommend further research into the predictive nature of these two personal characteristics with respect to SPD mistakes. The intention is to begin such a study in the near future.
References Andrews, M.J.
1979, Human factors in train operation. Proceedings of the Institute of Railway Signalling Engineers, London.
Buck, L. 1963, Errors in the perception of railway signals. Ergonomics, 6, 181- 192. Davis, R., 1966, Railway signals passed at danger: the drivers, circumstances and psychological processes. Ergonomics, 9, 2 1 1 - 2 2 2 .
Endo, T., and Kogi, K. 1975, Monotony effects of the work of motormen during high speed train operation. Journal o/Human Ergology, 4, 129-140. Heinze, E.E. 1983, Vigilanzforschung bei LokfLihrern, aueh hinsichtlich des Alters. Wissenschaftliche Konferenz der UIMC, Mtlnchen. Hfldebrandt, G., Rohmert, W., and Rutenfranz, J. 1974, Twelve and twenty four hour rhythms in error frequency of locomotive drivers and the influence of tiredness. In ternational Journal of Chronobiology, 2, 175-180. Kogi, K., and Ohta, T. 1975, Incidence of near accidental drowsiness in locomotive driving during a period of rotation. Journal o/Human Ergology, 4,65-76.
Sharp Grant, J. 1971, Concepts of fatigue and vigilance in relation to railway operation. In: K. Hashimoto, K. Kogi and E. Grandjean (Eds), Methodology in human fatigue performance. Taylor and Francis, London. Verhaegen, P.K., and Rijkaert, R.W. 1986, Vigilance of train engineers. Proceedings of the Human Factors Society, 30th annual meeting
Visser, R.S.H., van Vliet-Mulder, J.C., Evers, A., and ter Laak, J. 1983, Documentatie van tests en testresearch in Nederland - 1982 (Documentation on tests and testresearch in the Netherlands - 1982). NIP, Amsterdam.
Applied Ergonomics
June 1988
141