In chapter 1 the different concepts that play a role in driver mental workload were introduced and defined. The task that has to be performed or the goal that has to be reached can be described in objective terms. Goal setting (task conception or subjective task interpretation) determines the task goal that has to be reached (in terms of accuracy or speed) and thus affects task demand. Task demand can now be described in terms of operating stages which determine task complexity. How well the task is performed is an objective measure, namely the level of performance achieved. However, how the task is experienced, i.e. task difficulty, is not an objective property. Task difficulty depends upon task complexity, the operator's possibilities (i.e. capacity), his or her state and the applied strategy. Finally mental workload, the central concept in this thesis, is determined directly by task difficulty. On the basis of task difficulty processing resources are allocated and mental workload is reflected by the amount of allocated resources.
A relation between task demand and task performance has been described by Meister (1976, see also O'Donnell & Eggemeier (1986) for an adapted reprint). Meister defined three regions, region A, B and C. Region A is described as low operator workload with high performance. An increase in demands does not lead to performance decrements. In region B the level of performance declines with increased task demands. So, region B is the region where performance decreases with increases in demand, and increases in workload. In region C extreme levels of load have diminished performance to a minimum level, and performance remains at this minimum level with further increases in demand (see figure 1).
Figure 1. Hypothetical relationship between demand and performance (based on Meister, 1976)
According to this model, a primary-task workload measure, i.e. a measure of performance, will only be sensitive
to variations in levels of workload in region B. In region A performance remains stable and is independent of variations
in demand, while in region C performance will remain at a minimum level, independent of demand. Other measures, e.g.,
self-report measures of workload, may be sensitive in region B and may clearly reveal overload in the C-region, while
they need not to be sensitive in region A.
While extreme levels of load resulting in overload can be situated in the C-region, it is not clear where the domain
of underload is. A relation between arousal, task difficulty and performance as was first found almost a century ago,
the so-called `inverted U', could help to complete the region model. In 1908 Yerkes & Dodson (Yerkes & Dodson,
1908) published a famous paper on performance in a learning task under various levels of stress. The original paper
did not describe the performance of human subjects, but of mice. Possibly due to their behaviour as a consequence of
electric shocks that were administered, these mice were even called `dancers'. The major result of the study, which
led to what later became the inverted-U hypothesis, was that with different strengths of stimulation the medium
electrical stimuli were more favourable to the acquisition of a `habit' than stronger or weaker stimuli.
Although the original Yerkes-Dodson paper described a relation between stimulus strength and learning, the law has
implicitly been broadened to account for the effects of arousal level on performance (Hebb, 1955, see Teigen, 1994,
for a discussion of the law's history). To return to the region model, this model could be completed by adding a
deactivation or D-region at the far left end. The effects of monotonous tasks, for example, are situated in the
D-region. These are low demand tasks that can result in increases in task difficulty and workload
by a reduction in capacity. In case of, e.g., boredom a reduction in capacity requires that a larger proportion
of the capacity is used for performance of the same task, thus increasing mental workload (Meijman & O'Hanlon,
1984, O'Hanlon, 1981). It may also be that an affected state impedes the allocation of resources. By means of the
addition of the D-region the complete inverted-U is split into four regions, the D, A, B and C regions.
A question that comes to mind with respect to the region model is "How much workload is too much?". This
issue is usually referred to as the determination of a workload redline (Reid & Colle, 1988,
Wierwille & Eggemeier, 1993). When trying to tackle the determination of a redline there is a need to
first decide upon the context of `too much'. Degraded performance may indicate too much workload, but
affected personal well-being is equally valid. Preliminary work on workload redline puts this line at
the transition from region A to B (Rueb et al., 1992). Reid & Colle (1988) related just detectable
performance decrements to self-report ratings, and this workload rating designated the absolute workload redline.
The point of a just detectable performance decrement is at the transition from region A to B. While it is clear
that performance measures themselves have defined the A-region, it may be useful to split the A-region up into
three parts. In the middle part, region A2, the operator can easily cope with task demands and performance
remains at a stable level with increases in demand without increased effort. In the A3 region, however,
performance measures still do not show a decline, but the operator is only able to maintain the level of
performance by increasing effort. Temporary compensation by the exertion of effort in region A3 is one of
the advantages of human flexibility and is not critical. If, however, continuous effort is required to maintain
performance, or if peak loads occur frequently, this can lead to stress, an unhealthy situation that has to be
avoided (Zijlstra & Mulder, 1989, Meijman, 1989). This is in particular true if the operator has no control
over the situation (e.g., Van Ouwerkerk et al., 1994b). It may therefore be more useful to put a workload redline
at the transition from region A2 to A3 instead of at the transition from region A (A3) to B, as Rueb et al. (1992)
did. In this way, the word workload redline remains related to workload instead of relating it to
primary-task performance breakdown. A similar situation exists at the region that is to the right of the D-region,
region A1. Here for instance monotony starts to affect the operator's state, but by `trying harder', i.e. by the investment of effort, the primary-task performance level is not yet affected. A second workload redline then arises at the transition from region A2 to A1, where the operator is effectively counteracting a reduced operator state. When effort investment is no longer effective, the D-region is entered where performance is affected.
When demand increases, starting from the optimal operator state in region A2, the operator's capability of (effort)
compensation will be exceeded at a certain moment and a transition from the A3 to the B region takes place. In the
B-region performance is affected and at the moment that it has deteriorated to a minimum level the C region is
entered. Task performance and workload as a function of demand are depicted in figure 2. It is important to stress
that demand on the x-axis in figure 2 is not directly linked to region of performance. Task demands are determined
by the goals that have to be reached by task performance and cannot be linked directly to workload, which is
subjective. Region merely indicates the interaction between performance and workload. The same task can result
in performance in region A2 for one individual, and may require effort compensation and thus region A3 performance
for another. Also, in figure 2 the two types of effort compensation (Mulder, 1986, Cnossen, 1994) are split
over two regions. In the A1 region deactivation is counteracted by state-related effort, while in region A3
task-related effort is exerted.
Figure 2. Workload and performance in 6 regions. In region D (D for deactivation) the operator's state is affected.
In region A2 performance is optimal, the operator can easily cope with the task requirements and reach a (self-set)
adequate level of performance. In the regions A1 and A3 performance remains unaffected but the operator has to exert
effort to preserve an undisturbed performance level. In region B this is no longer possible and performance declines,
while in region C performance is at a minimum level: the operator is overloaded.
In the model (figure 2) only one dimension of mental workload is displayed. What is depicted denotes the overall or sum relation between demand, workload and performance. The relation exists in principle for each separate resource. The implication is that auditory task demands, visual task demands and central demands do not necessarily have to be in the same region, which is in accord with Wickens' multiple-resource theory (Wickens, 1984).
With respect to the model the following questions can be asked: `Which measure is sensitive when'? `In order to assess mental workload, is one measure sufficient?' `Do measures dissociate?' `Can we deduce whether state-related effort or task-related effort was exerted, and if we can, how?'
This thesis focuses on how to measure driver workload. Different techniques, their characteristics and their use in applied settings, in particular in traffic research, will be evaluated. The technique's sensitivity in traffic research will be evaluated in the model on the basis of studies that my colleagues and I have performed. Particular attention will be paid to possible differential sensitivity of measures to increases in mental workload by changes in driver state opposed to changes in task complexity. The focus will be on performance measures that are specific for traffic research, a physiological measure (heart rate and its variability) and on two self-report scales. An overview of previously found results with respect to mental workload studies will be attended to before that, in chapter 3. But first the general properties of the measures will be considered.
to chapter 3
back to thesis summary
I like to hear from you, so if you find this information useful, a short message is very much appreciated. For more information you can also contact me.
© Dick de Waard 1996
You may only use (parts) of this thesis if you quote the source:
De Waard, D. (1996). The measurement of drivers' mental workload. PhD thesis, University of Groningen. Haren, The Netherlands: University of Groningen, Traffic Research Centre.
Back to my HomePage