2.1 Human factor for railway safety
In every country with a rail network of any importance, the relevant operational,
regulatory, and government bodies are trying to achieve something similar. This is, to move
more people and goods, on time and safely, to satisfaction of their customers. Rail human
factors is not limited in its contribution in terms of providing dimensional and performance
information for design of equipment, interfaces, and workspaces (e.g., train cab design),
working environments (e.g., signal boxes), minimization of risk from manual handling (e.g.,
track working), and design of information display systems (e.g., lineside signage or train
movement displays). Rail human factors is increasingly working at a systems level, being
central to systems engineering. The drive behind this comes from contributions such as better
understanding of organizational failure in accidents (Reason, 1997), acceptance that cognitive
task performance is situated in a setting that it strongly influences (Nardi, 1993), and is spread
across people, places, and time (Hutchins, 1995).
Elms’ (2001) research certified that 1/3 of railway accidents and personnel casualties are
caused by human factors, which is a greater effect than for technical factor. Approximate 60%
to 90% of accidents are caused by human factors, and the most important factor in railway
operation safety is human reliability (Dickens, 1992). The focus of rail human factors research
includes signalers and controllers; drivers; station and on-train staff; planners, engineers, and
managers; track (maintenance) workers; passengers and the general public. However, the
drivers are in a critical role for railway safety.
Treat’s (1977) study stated that more than 90% of human factor-related accidents are
related to drivers, which is also the leading cause of accidents. Ugajin (1999) investigated the
total number of railway accidents that occurred in Japan in 1997 and found that 40% of these
accidents were due to human factors on the part of drivers. According to Hall (1997), driver
errors accounted for approximately 46% (N = 18) of the registered accidents during period
1970-1997 in the U.K. During the period 1970-1998 in Norway, train drivers error accidents
accounted for 38% of all accidents that occurred (Jernbaneverket, 1999). In addition, driver
responsible accidents made up 31% of the total accidents in Taiwan Railway Administration
(TRA, 2006). It is our crucial on-going mission to keep accidents like these caused by human
factors of train drivers from occurring, and to contribute to establishing methods for
preventing them.
2.2 Human factors research for train drivers
The current research addresses the fundamental elements of the train driver’s role and
performance, including route knowledge and the underlying psychological components of
train driving (Farring-Darby et al., 2005; Jansson et al., 2006; McLeod et al., 2005). Much
research in this area has investigated the potential causes of human error (Reason et al., 1994)
and the extent to which the in-cab environment supports the driver’s ability to maintain
situational awareness (Endsley et al., 2003). A part of reducing the potential for driver error
and increasing effective (on-time) performance lies in the design of jobs and job aids
(Kecklund et al., 2001) as well as understanding and optimizing – neither too high nor too
low–workloads.
There has been, in particular, a broad and relatively well-studied research field in the
area of train driver vigilance and perception, in terms of their recognition of and acting upon
signs and signals. This includes investigations into signals passed at danger (SPAD) and the
appropriate design of signage and signaling systems. Recently in the UK, motivated by the
Ladbroke Grove rail crash and by reports of incidents not leading to injury, there have been
various studies of SPADs (Pasquini et al., 2004; Turner et al., 2003), predictive tools (Wright
et al., 2007), and development of tools to identify the risk of SPADs at different signals
(Holywell, 2005; Lowe and Turner, 2005). Because of the desire to find a technological fix to
such incidents, there has been related research into the use of vigilance devices and reminder
appliances (McLeod et al., 2005; Whitlock et al., 2005). Modern observation techniques, such
as the measurement of eye movements and of direction of gaze, allow interpretation of
drivers’ behavior and of the possible reasons for it (Luke, 2006; Merat et al., 2002).
One use of eye tracking is as a technique to investigate the onset, manifestation, and
consequences of fatigue (e.g., dwell or fixation times will become longer as people get
fatigued). Studies have also examined the prevalence of sleep apnea (Hack et al., 2007).
Research related to fatigue has also examined the effects of driver work shift (Hawarth and
Tapas, 2001), used observation and self-report to study the effects of long (>6 h) journey
times (Gouin et al., 2001), run simulator studies (Dorrian et al., 2005), and developed
checklist tools such as the Fatigue Index (Cotterill and Jones, 2005), as well as prototypical
preferred roster patterns (Ashton and Fowler, 2005). Related to impairment through fatigue is
the incidence and effects of drugs and alcohol use on performance (Ervasti et al., 2007).
2.3 The major factors affecting accident risk by train drivers
Prolonged attention is the most significant factor influencing the occurrence of accidents
in human factors, especially errors related to railway traffic signals (Grahan, 1997). Factors
that result in the reduction of prolonged attention are fatigue, noise (Hancock, 1990),
particularly cold or hot temperatures (Davies and Parasuraman, 1981), time of day (Folkard
and Monk, 1997); alcohol effects (Horne and Gibbons, 1991), stress (Kecklund, 1999) and
differences in experience (Bisseret, 1988). It has been found that railway drivers’ negligence
of traffic signals is due to distraction, which is also the major cause of reductions in prolonged
attention (Haga, 1984). In addition, temperature, humidity, and noise within the driver’s
cabin are environment factors influencing prolonged attention (Smiley, 1990). Finally, most
of the evidence indicates that decreasing prolonged attention has an adverse effect on
performance and safety for train drivers.
Previous studies are confirmed that consecutive driving (Dinges, 1995; Horne and
Reyncr, 1995) and stress (Hockey and Hamilton, 1993; Kecklund et al., 1999; Zakay, 1993)
decrease prolonged attention. For improving railway safety, it is important to identify the
most critical human factors influencing job stressors while train driving. In so doing, it is
hoped the results will aid in mitigating job stress for train drivers. At the same time, it is also
important to understand the relationship between consecutive driving and train accident risk.
2.3.1 Stress and railway safety
The concept of stress is built upon three principle components: the environment or the
demands of the situational context (the environment here is identified as the work situation);
the capacity and resources the individual has for dealing with these demands; the individual’s
physiological, psychological and behavioral reactions (Kalimo et al., 1998). Stress in its
negative sense implies an imbalance between the demands of the environment and the
capacity of the individual to cope, or that the individual’s expectations exceed what is offered
by the environment. If a stress situation cannot be controlled, negative reactions arise (e.g.,
discontent, worry, fear, frustration, and a lack of pleasure or motivation at work (Brown,
1980), or endocrine or physiological responses such as increased heart rate (Robertson,
1988)).
There are numerous factors and situations that can cause stress at work and some of the
most common are difficult social relationships, problems with the organization, poor career
opportunities, strenuous physical conditions, excessive workloads and time pressure, demands
are too low, little decision-making opportunity, no stimulation, lack of control and an inability
to exert influence on the job. But, stress conditions experienced by humans affect
performance and reliability (Dhillon, 1986). Zakay (1993) stated that increased stress can lead
to reduced productivity and performance, and performance is generally shown to deteriorate
under some sources of load (e.g., noise, vibrations and heat) (Hockey and Hamilton, 1993).
One reason why performance is often impaired by stress is that, under favorable
conditions, many people are able to deal with moderate levels of stress by mobilizing extra
(mental) resources and focusing themselves on the task (Hockey and Wastell, 1998;
Schonflug, 1983; Meijman, 1997). There have been a number of studies that link highly
aroused stress states with impaired decision making capabilities (Baddeley, 1972), decreased
situational awareness (Vidulich et al., 1994), and degraded performance which impaired
driving ability (Helmereich et al., 1990). Hudoklin (1996) mentioned that stress levels above a
moderate level cause a decrease in human reliability, and lower driver reliability can cause
lower driver performance, which increases accident rates for railways. In order to improve
railway safety, it is necessary to measure train drivers’ conceptualization of stressors during
the driving process. However, very little is known about the strategies people use for coping
with stress and, in real situations, those coping mechanisms influence performance and safety.
Previous studies have explored different job stressor for train drivers. For example,
Chang (2005) applied Cooper’s Occupational Stress Indicator to explore job stress in train
drivers, and Yu (1998) examined the factors in job stress and strain using the National
Industry Safety and Healthy Research Institute’s generic job stress questionnaire. It should be
noted, those studies utilized only a general industry job stress questionnaire. Because the
railway is a specialized field, a specific questionnaire needs to be designed for measuring job
stress in train drivers.
Table 2-1 Stressors that influence driving safety
Dimension Latent stressor
Fixed equipment: route equipment (including railroad ties, bridges, track or overhead power lines), safety equipment (including ATW/ATS), communication equipment, signal system, signal device position, signing site for construction
Equipment
Mobile equipment: train type, train length, train fault, train equipment, train emergency device
Internal environment: control cabin space, noise, vibration, temperature Environment
External environment: weather conditions (including raining, temperature or visibility), night driving, road signal confusion, route slope, effect of buildings and trees, passing attended (unattended) grade crossing, trespassing by people or animals
Work shift: mixed shifts, length of duty time, duty break Management
Management measures: training (including simple troubleshooting, operating different types of trains, operating skill, learning operation rules), monotonous tasks, work assurance, promotional channel, performance evaluation, single-driver or two-driver duty service, supervisor communication and management style, work requirements (including safety rule examination, call-response evaluation, physical examination specification)
As a first step in that direction, we apply rail system safety theory as proposed by Zhao
et al. (2003), which includes equipment, environment, management and driver as the four
dimensions of a railway safety system. Furthermore, based on relevant literature (e.g., Chang
et al., 2005; Cooper et al., 1988; Dorrian 2006, Yu et al., 1998), those variables fit particularly
well in Taiwan. The general key stressors that influence train driver safety are shown in Table
2-1.
Stressors related to equipment problems in railway systems during operation can be
divided into fixed equipment and mobile equipment. Stressors related to fixed equipment
include whether the railroad ties, track, and overhead power lines are functioning normally,
and whether the signal devices are mounted and functioning properly. Ohlsson (1990) pointed
out that, as drivers rely more on driving safety equipment, such as automatic train warning
(ATW) or automatic train stop (ATS), the availability of such safety equipment would affect a
driver’s response, thus causing huge work stress for a driver. The potential problems with
mobile equipment include locomotive or passenger car conditions. Since the train is operated
via many electronic devices, any malfunctioning equipment would affect the safety of the
train and drivers. Moreover, due to limited control cabin space and lack of toilets, drivers’
physiological needs cannot be met while on duty, thus bringing added stress to driving.
Apart from stress related to equipment, the environment inside and outside the
locomotive can also bring stress to drivers on duty. Internal environment refers to the work
environment in the control cabin. Akerstedt (1980) indicated that the space, noise,
temperature, and vibration of the control cabin could make drivers uncomfortable, and
monotonous train driving distracted drivers’ attention; all these factors resulted in stress on
drivers, and further jeopardized their driving safety (Edkins and Pollock, 1997). The external
environmental factors that bring driving safety stress include: (1) too cold, too hot, foggy, and
snowy days that hamper drivers’ responses and driving safety (Park, 1987); (2) weather or
environmental changes (e.g., growing trees along the railway) make it hard for drivers to
identify fixed signals, thus affecting their response speed; and (3) when a train is passing a
railway grade crossing and there are cars or pedestrians intruding. The external environment
poses a serious threat to train driving safety, and involves unexpected risks that are out of the
control of train drivers, which results in additional sources of stress for train drivers.
Besides tangible driving stressors, such as equipment and environment factors, drivers
also face intangible stress from management. Kolmodin-Hedman (1975), Akerstedt (1980),
and Netterstom (1981) suggested that mixed work shifts and irregular work timed are main
factors contributing to drivers’ work stress and affect their driving safety. Due to limitations
in work time and shift switch sites, train drivers often have to serve prolonged shifts, thus
facing more stress. In addition, as pointed out by Edkins (1997), without enough motivation
or aspiration, train drivers would be stressed about personal uncertainty of the future. Yu
(1998) also mentioned that uncertainty about work is an important source of stress for train
drivers. Therefore, reasonable supervision and promotional channels could cause stress on
drivers to a variable extent. Tsang and Wilson (1997) pointed out that, if work demand is
beyond one’s capability, then higher work stress would be brought to drivers, leading to more
errors. Therefore, the availability of sufficient training so that drivers can drive various types
of locomotives, execute simple locomotive troubleshooting procedures, and master operating
skills to respond to driving needs, could also cause considerable work stress for drivers.
Finally, as drivers have to make immediate responses to various situations when driving (such
as signal identification or responding to obstacles), work stress for drivers in “single-driver
duty” service is naturally higher than for those in “two-driver duty” due to lack of warning
and aids.
2.3.2 Consecutive driving and railway safety
Driving behavior is exemplified as information management; the efficiency of
information management is closely related to the conscious status of driver. Driving and
working for a sustained period of time can generate fatigue (Okogbaa et al., 1994; Smiley,
1998; Sussman and Coplen, 2000) so drivers would not be able to maintain the level of
driving safety under conditions of continuous driving (Dinges, 1995; Horne and Reyncr,
1995). Consecutive driving can decrease vigilance, which is a major factor accounting for
driver error. Moreover, consecutive driving is one of the greatest causes of traffic accidents
(Harris, 1977; Hermann, 2004; Li et al., 2005). The accident rate goes up significantly as the
number of consecutive driving hour increases for highway truck operations (Chang and
Hwang, 1991). According to a National Transportation Safety Board (NTSB) analysis of
railway accident research from 1990 to 1999 in America, it also shows that consecutive
driving is the major cause of train collisions and accident (Sussman and Coplen, 2000).
Therefore, the risk of being involved in a railway accident is expected to increase as the
number of hours of consecutive driving increases.
Few studies have examined the risk of train drivers’ accidents as a function of
consecutive driving time. Wharf (1993) analyzed the frequency of signals passed at danger
(SPAD) per million driving hours for British Rail train drivers and found a distinct peak
during the second and third hours of duty, followed by a relatively low level, which then
subsequently increased again. Based on an investigation of accident records from the Swedish
National Rail Administration during the period 1980–1997, Kecklund et al. (1999) also
indicated that a risk peak existed at the third hour of the shift, followed by a period of low risk,
which then showed an exponential increase in risk over hours on duty. Additionally, a Dutch
study (van der Flier and Schoonman, 1988) explored the relationship between driver errors
(missed signals) and working hours and found that the probability of error is at its peak during
the second and third hours of the shift. Kecklund (2001) and van der Flier (1988) also
discussed possible reasons for the findings and speculated that such mistakes are due to
fatigue accumulated from previous shifts or that drivers might relax too much during the start
of a shift.
However, little has been said about the definition of working time and types of train
drivers in the previous studies. Actually, given the work tasks and missions assigned to TRA
train drivers, a work shift can usually be divided into three sequential stages: pre-starting,
on-board driving, and post-arrival stages (Figure 2-1). Using the start and end times of shifts,
including the static pre-driving and post-arrival times mostly collected for payroll or other
reasons as the input, in most studies (Fletcher et al., 2001) may confound the time effect on
risk of accidents for actual on-board driving.
Figure 2-1 Three stages of one driver’s work shift under TRA operation.
In addition, ignoring differences between passenger and freight train drivers in accident
risk may result in the loss of some important implications due to their different working
environments and requirements. Freight train drivers usually have irregular work schedules,
boredom during operations (Sussman and Coplen, 2000), and a higher proportion of night
operation problems (Jackson, 2005). It is well-documented that irregular shift workers suffer
from restless sleep while undertaking early morning and night-time work (Akerstedt and Pre-starting stage On-board driving stage Post-arrival stage
Folkard, 1996; Pollard, 1996). Furthermore, shunting in marshalling yards by the starting
station is exclusively required for freight trains. Shunting is a notoriously unsafe activity
(Elms, 2001); therefore, freight train driving is expected to have more accident risk than
passenger train driving because it has more wearying and complex work requirements.
CHAPTER 3 METHODOLOGY
3.1 The train driver’s stressors
This study applied rail system safety theory proposed by Zhao et al. (2003) and designed
variables that fit particularly well in Taiwan. The different stressors for train driver as shown
Table 2-1 are all latent variables, which are inferred from subjective judgments by the
respondents. Researchers in transportation have directed their attention to the relationship
between one’s latent consideration and his/her response. After multiple related studies were
conducted, one was left to question: Were convincible and comparable measures on the
related latent constructs obtained? Such a challenge in the measurement is critical, especially
for those latent variables that have no normalized scales (the norms) to serve as a reference of
measurement. In practice, researchers usually measure such latent constructs by collecting
respondents’ opinions, and those opinions are mostly represented by items with ordinal scales
(e.g. the Likert-type scale) in questionnaires. If these ordinal categories in the items are
naively assigned some incremental integers, such integers can only represent the rank among
categories in a single item, which has limitations in statistical inference. For this reason, one
of the aims of the study is to demonstrate approaches in how to measure a newly-specified
latent construct; especially for ensuring the results on the trait level can serve as reasonable
and effective factors for further statistical inference.
3.2 Methods for measuring a latent trait
To provide an objective and valid rating scales for solving such a problem, the item
response model has been developed and improved. Item response theory (IRT), which is a
model-based measurement in which trait level estimates depend on both persons’ responses
and on the properties of the item that were administered, has become the mainstream of
psychological measurement (Hambleton and Swaminathan, 1978). Among the various models
of IRT, the Rasch model is one that is widely applied for exploring psychological constructs.
A review of IRT and the Rasch model are provided in the following parts of this chapter.
3.2.1 Review of Item Response Theory
Psychological constructs are usually conceptualized as latent variables that underlie
behavior. Latent variables are assumed as unobservable entities that influence the manifest
variables (e.g., test scores or item responses). Thus the observation of these manifest variables
can only serve as indicators of a person’s standing on the latent variables. As a result,
measurements of psychological constructs are usually indirect; that is, latent variables are
measured by observing behavior on relevant tasks or items. A measurement theory in
psychology must provide a rationale that both persons and items on a psychological
psychology must provide a rationale that both persons and items on a psychological