See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/262357754

Pupillometry as a method for measuring mental

workload within a simulated driving task
Dataset January 2008

CITATIONS

READS

609

3 authors:
Maximilian Schwalm

Andreas Keinath

RWTH Aachen University

BMW

16 PUBLICATIONS 179 CITATIONS

26 PUBLICATIONS 186 CITATIONS

SEE PROFILE

SEE PROFILE

Hubert D. Zimmer
Universitt des Saarlandes
128 PUBLICATIONS 2,200 CITATIONS
SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate,

letting you access and read them immediately.

Available from: Maximilian Schwalm

Retrieved on: 14 November 2016

Pupillometry as a method for measuring mental

workload within a simulated driving task
Maximilian Schwalm1, Andreas Keinath1, & Hubert D. Zimmer2
1
BMW Group, Mnchen
2
Saarland University, Saarbrcken
Germany

Abstract
In this contribution a method of pupillometry is discussed to identify high mental
demands on the driver in a simulated driving task. A new method allows to identify
the effect of mental demand by measuring changes in size of the drivers pupil and to
display the actual demand through an index called Index of Cognitive Activity
(ICA, Marshall et al., 2004). A study will be discussed where a simulated driving
task was used in combination with the method of pupillometry. This study shows that
the ICA increases in situations with a higher mental demand on the driver when
performing lane change manoeuvres or an additional secondary task. Hence the
index of cognitive activity seems to be a suitable method for continuously measuring
mental demands while driving.
Introduction
The main issues when developing new automotive human machine interaction
concepts are to provide an optimum of functionality and to ensure the safe and
efficient use of these systems while driving. Additional systems such as driver
information and communication systems that can be used while driving may
significantly increase the operational demands on the driver. Using these systems
while driving, forces the driver to divide his or her attention between the additional
system and the driving task itself (De Waard, 1996). If the resulting demand for the
driver through such a dual task situation reaches a critical level, the risk of driving
errors increases. Therefore when developing new driver-information or drivercommunication systems the goal is to design these systems in such a way as to
reduce the additional demand on the driver when interacting with these systems to a
minimum.
Because of the apparent necessity to identify the amount of mental demand while
driving, several different methods have been proposed to identify this effect on a
driver. ODonnell & Eggemeier (1986) specify three groups of workload measures:
(1) Subjective measures (or self-report measures, see De Waard, 1996). These
measures directly investigate the subjectively perceived mental workload in a given
situation (e.g. questionnaires, interviews). (2) Performance measures. Performance
In D. de Waard, F.O. Flemisch, B. Lorenz, H. Oberheid, and K.A. Brookhuis (Eds.) (2008), Human
Factors for assistance and automation (pp. 1 - 13). Maastricht, the Netherlands: Shaker Publishing.

Schwalm, Keinath, & Zimmer

measures indirectly infer from the performance in a given task to the actual workload
of the driver, based on the idea that an increased demand on the driver results in a
decrease in performance (e.g. reaction time, accuracy). (3) Physiological measures.
These measures use physiological reactions as correlates for the actual demand on
the driver (e.g. diameter of the pupil, heart rate).
These three measurement groups that are discussed in the literature each have
different methodological advantages as well as disadvantages.
Subjective as well as performance measures have the advantage of a very fast and
efficient application during a product development process. The disadvantage of
these measures lies in a reduced temporal resolution of the measurement as well as
the insensitivity caused by individual compensating strategies while performing a
task. Performance measures reflect how a task is performed, but performance may
also be protected by the amount of invested effort for a task (Hockey, 1997).
Especially when performing a realistic dual task situation such as driving, the driving
performance may be deliberately protected by a reduction of the invested effort for
the secondary task, leading to a reduced performance in the secondary task but also a
reduced demand for the driver.
Physiological measures however offer an objective and direct approach to the
measurement of mental workload and do not require an overt response by the subject
(De Waard, 1996). Moreover these measures may offer a continuous, interferencefree measurement of the actual mental demand with a high temporal resolution. The
disadvantage lies in the increased technical effort for the detection of the signal as
well as the separation of the relevant to the interfering signals (signal-to-noise ratio,
Kramer, 1991 cited in De Waard, 1996).
The current contribution will discuss a specific method of pupillometry as a
physiological measurement of mental workload. Especially the objectivity, the high
temporal resolution and the online availability of the signal are the main advantages
that make the method of pupillometry interesting for an application oriented research
in the automotive context.
The method of pupillometry as a physiological measurement
The positive relation between an increase in the pupil diameter and mental demand
has been discussed in the literature for decades (Beatty & Lucero-Wagoner, 2000;
Just, Carpenter & Miyake, 2003). Empirical evidence could be found for a whole
range of tasks such as mental arithmetic (Hess & Polt, 1964; Ahern & Beatty 1979,
1982; Breadshaw, 1968), signal detection tasks (Hakerem & Sutton, 1966 Beatty &
Lucero-Wagoner, 1975; Kahnemann & Beatty, 1967), memory tasks (Kahnemann &
Beatty, 1966; Graholm, Asarnow, Sarkin & Dykes, 1996) and language processing
(Schlurhoff, 1982; Beatty, 1982; Just & Carpenter, 1993). The reason why
pupillometry as a measure for mental demand has only been playing a minor role in
applied research lies mainly in the fact that the size of the human pupil is dependent
on a wide range of different variables such as light density, distance accommodation,
heartbeat, breathing frequency, etc, factors that are not directly related to mental

pupillometry as a method for measuring mental workload

demands. The challenge when developing a new measure of mental demand using
the size of the human pupil therefore lies in resolving the raw pupil signal from the
influence of these other influences.
In previous studies in that the pupillometric measures were used it was attempted to
hold irrelevant influences to the size of the pupil constant throughout the data
collection. Such an approach, however, drastically limits the application of the
method to those laboratory based situations where these variables can indeed be
experimentally controlled. But such a procedure could certainly not be applied to
realistic and application oriented situations, as for example a driving situation based
on the continuously changing environment (lightning condition, distances etc).
An additional problem when measuring the size of the human pupil results from the
fact that the absolute size of the human pupil varies between individuals. A
comparison of absolute differences in size is therefore not practicable. A meaningful
measurement of mental demand through the size of the human pupil should therefore
be based on relative rather than absolute changes in size (Marshall, Davis, & Knust,
2004).
The Index of Cognitive Activity
In order to cope with these problems, Marshall and colleagues developed the Index
of Cognitive Activity (ICA) (US Patent Marshall, 2000). More detailed discussions
of this index can be found in Marshall (2000, 2005, 2007; Marshall, Davis & Knust,
2004). Therefore only a short overview on its rational should be given here instead.
The ICA was developed to resolve the previously discussed methodical problems
when assessing mental demands through changes in the size of the human pupil. The
Index is calculated from high-frequency components of changes in the pupil size
while an individual performs a specified task that requires significant cognitive
processing. This effect has previously been described as the dilation reflex
(Loewenfeld 1993) and is based on the fact that in this case the two muscle groups
namely the dilator as well as sphincter muscles that cause the pupil to dilate and to
constrict are working together simultaneously. While the dilator muscles are
activated (causing the pupil to dilate) the sphincter muscles are inhibited (also
causing the pupil to dilate). The result is a brief dilation that is greater than either
muscle group could effect, subserved by both sympathetic as well as parasympathetic
inputs (Beatty & Lucero-Wagoner, 2000). These changes in the size of the pupil
based on the dilation reflex are irregular and sharp, often exhibiting large jumps
followed by rapid declines (Marshall, 2000). This response is quite different to the
one caused by the light reflex. Therefore the ICA is a measure of relative changes of
pupil size that reflects the number of times each second that those large and abrupt
increases occur which exceed a specified threshold in the amplitude of the signal. A
higher ICA value indicates more changes per second and therefore a higher amount
of cognitive activity.
For a more detailed discussion about the calculation of the index from the raw pupil
data please refer to the works of Marshall (2000), Marshall, Davis & Knust (2004),
Marshall (2007).

Schwalm, Keinath, & Zimmer

The ICA can be computed for short time periods such as single trials in a multi-trial
study as well as for long periods of time. While the index is a relative measure of the
fast and abrupt changes in the high frequency components of the pupil signal, and as
such it is independent of changes in lightning conditions as shown in Marshall
(2000) as well as Marshall, Davis, & Knust (2004).
The Index of Cognitive Activity has been shown to be sensitive to different levels of
mental load in a whole range of different cognitive tasks. Examples are mental
arithmetic and sequence tasks (Marshall, Davis & Knust, 2004), gauge monitoring
and the detection of strategic shifts (Marshall, Pleydell-Pearce, 2002; Marshall,
2002), and military tactical decision making (Marshall, 2007). Higher ICA values
were observed in situations or with tasks that high demands on cognitive processing.
Hence, the index has delivered a suitable measure of cognitive activity in practice.
Higher mental load was associated with an increase of the ICA value and the
measure should therefore also be appropriate to estimate cognitive load in a driving
situation.
Experiment
The aim of the present study is to apply the ICA methodology to the driving
situation. It should be used as a method to identify high mental workload situations
in the context of a simulated driving task, by measuring the ICA in the course of
driving. There are three situations when mental workload is expected to be
increased: (1) at the moment the driver is instructed visually via signs to change
lanes, work load should increase; (2) if workload is a consequence of processing of
the command and not of perceptual processing, the increase should also occur when
the command to change is given as an auditory instruction, and (3) when the driver
has to perform a secondary task while driving this enhanced load should be visible in
a higher ICA value. In order to investigate this dependency we required our
participants to perform a simulated driving task and while driving lane change
commands were given.
Driving simulation
For the simulated driving task, software called the Lane Change Test (LCT) (Mattes
2003) has been used. The LCT is a PC based driving simulation designed for
measuring driver distraction. Drivers have to repeatedly perform lane changes while
driving on a simulated roadway that consists of a three-lane road with a total track
length of about 3000 m. While participants are instructed to drive with the gas pedal
pressed to its maximum, speed is limited to 60 km/h resulting in a total driving
duration of about 180 seconds per track. In the original LCT setup the participants
are instructed to change lanes according to visual signs situated at the roadside. Each
sign indicates the to-be-taken lane (see Figure 1). There are 18 signs along a track
and the mean distance between the signs is about 150 m, resulting in a lane change
approximately every nine seconds. Distances between signs show minor temporal
differences of a few seconds differing between tracks in order to prevent learning
effects (ISO Draft TC22/SC13/WG8). In the present study the LCT has additionally
been used in a slightly modified form, so that instructions were given via voice

pupillometry as a method for measuring mental workload

instructions through headphones instead of the signs next to the road. In this
condition the signs showed no information. Between the manoeuvres the driver has
been instructed to keep the lane as precisely as possible.
In the lane change task, driving performance is measured through the deviation
between a normative driving path this is the optimal lane change path prescribed
by the model when a change command is presented and the actual course of the
subject along the track averaged over time points. The LCT has been shown to be
sensitive to both visual and cognitive distraction while driving (Mattes, 2003, 2006;
ISO Draft TC22/SC13/WG8).

Figure 1. The drivers view of the simulated three lane road scene with the signs indicating
the correct lane

Secondary task
To manipulate the amount of mental demand for the driver, the LCT is usually
combined with different secondary tasks. In the present study the Surrogate
Reference Task (SURT) (Mattes, 2006) has been used as a secondary task. The
Surrogate Reference Task is a visual search task, where subjects have to identify the
position of a target (a white ring) among 50 distractors (white rings with a smaller
diameter than the target) on a screen placed next to the LCT screen (distances and
viewing angles according to the LCT ISO standards, ISO Draft TC22/SC13/WG8).
The area in that the target circle was located had to be marked by a grey bar that
could be moved by pressing the cursor keys. The task could be interrupted and
started again at any time during execution.

Schwalm, Keinath, & Zimmer

Figure 2. An example of the secondary task display showing the target and distractors in the
easy condition

Eye-tracking apparatus
The data reported here was collected using the EyeLink 2 Eye-Tracking System from
SR Research, Ltd., with binocular tracking at a sampling rate of 250 Hz. The
EyeLink 2 system consists of small video cameras mounted on a lightweight
headband. The system offers a pupil size resolution of 0.1% of diameter at an
averaged pupil diameter of about 3-7 mm.
Procedure
On arrival at the laboratory the aim of the current study as well as the general
procedure was briefly explained to each participant. At the beginning participants
were given the opportunity to practice the Lane Change Task while being instructed
to perform each of the lane change manoeuvres as fast but as exactly as possible.
After having become familiar with the task, participants had to drive one of the
tracks until the criteria of a mean deviation less than 1 could be reached (ISO Draft
TC22/SC13/WG8).
After calibrating the eye-tracker it was made sure that the quality of the obtained
pupil data was sufficient in order to calculate the ICA throughout the whole
experimental procedure. Each participant performed each of the three experimental
conditions through one complete track on the Lane Change Task: LCT visual, visual
plus SURT, LCT auditory. The sequence was interindividually counterbalanced.
When performing the LCT plus SURT, participants were instructed not to prioritise
one of the tasks but to try to perform as well as possible in both of the tasks.
At the end of each experimental condition the subjectively perceived mental demand
was rated by the participants via a German version of the NASA TLX questionnaire
(Hart & Staveland, 1988). The NASA TLX includes ratings on six subscales of

pupillometry as a method for measuring mental workload

workload such as mental demands, physical demands, temporal demands, own

performance as well as effort and frustration.
Sample
Twenty participants (10 men and 10 women) took part in the current study. The
participants ranged in age from 23 to 33 years, with a median age of 26.6. All
participants were holding a valid drivers licence.
Results
Driving performance
To determine the performance on the driving task, the mean deviation in the lane
change path (determined by the difference between the actual driven path and the
optimal path) was calculated for the three experimental conditions (see Figure 3). An
analysis of variances (ANOVA) with repeated measures showed a significant main
effect for the factor of experimental condition F(2,28) = 55.9, p < .001. Mean
deviation in the dual task condition (LCT plus SURT) was significantly higher than
in the two pure driving conditions, F(1,14) = 70.85, p < 0.001. In contrast, the
driving performances in with visual and auditory change commands were not
different, F(1,14) < 1. Hence, the secondary task caused the intended additional load,
but processing the visual signals was not harder than processing the auditory signals
although the visual signal puts a strain on the same system as driving whereas the
auditory condition did not.

Figure 3. Mean deviation in the lane change path for the condition LCT visual, LCT auditory
and LCT plus SURT (error bars indicate +- 1 SD)

Subjective ratings
The subjective ratings of the perceived mental demands as measured by the NASA
TLX questionnaire at the end of each experimental condition showed the same
pattern of results. Significant differences occurred in the perceived mental demands

Schwalm, Keinath, & Zimmer

between the two conditions with and without a secondary task on each of the six
dimensions, F(1, 14) = 46.95, p < 0.001 and no significant difference between the
visual and auditory conditions. Here the condition with an additional secondary task
was rated to be much more demanding than without such an additional task while
driving.
Index of Cognitive Activity
The ICA was calculated for each participant and the period of each experimental
condition (see Figure 4). An ANOVA with repeated measures over the three
conditions showed a significant main effect for the mean ICA values, F(2,
28) = 217.3, p < 0.001. A planned contrast revealed a significant increase in the
mean ICA values in the LCT plus SURT compared to the two mere driving
conditions, F(1, 14) = 259.17, p < .001. No significant difference was observed
between the visual and auditory LCT, F(1, 14) < 1. To check the relationship
between ICA and task performances we additionally correlation the number of
correctly solved tasks (correctly identified targets) and the mean ICA values. Both
measures significantly correlated, r = .55, p<.05. That means, when in the same
amount of time more additional tasks were completed while driving, the mental
workload of the driver indicated by the ICA increased as well and vice versa.

Figure 4. Mean Index of Cognitive Activity (ICA) for the condition LCT visual, LCT audi-tory
and LCT plus SURT (error bars indicate +- 1 SD)

The ICA has the advantage that it offers not only a global measure of load, but a
continuous measurement of mental workload in a given situation with a high time
resolution. We therefore additionally conducted a closer analysis of the changes of
ICA in the course of driving. For that purpose we synchronised the starting point of
each track and then averaged the ICA values over the 20 participants. Figure 5
depicts the grand averages of ICA during driving separately for the visual and
auditory LCT condition. The grey bars indicate the time interval in that participants
processed the signal and changed lane. The temporal distribution of ICA revealed a
very clear synchrony between the ICA amplitudes and the time periods of the
different task components. ICA increased and had a local maximum during any of

pupillometry as a method for measuring mental workload

the 18 lane change manoeuvres (see Figure 5). The local maximum of ICA was more
than twice as large (around 10) as during driving without lane change (around 4). In
light of the standard deviation of ICA at the group level and of the complete
regularity of the effect this can be considered as a highly reliable result.

Figure 5. Grand average of ICA (a) in the visual LCT and (b) in the auditory LCT. Grey bars
indicate the 18 lane change manoeuvres

10

Schwalm, Keinath, & Zimmer

Additionally, we analyzed ICA in the LCT plus SURT (see Figure 6). Contrary to
out expectation, the due to the secondary task condition generally high ICA values
significantly decreased around each of the 18 signs. An ANOVA was carried out
to compare the mean ICA values during the lane change manoeuvres to the situations
where drivers only had to keep the lane while driving. The analysis confirmed this
decrease around each of the 18 signs compared to the situations without a sign,
F(1,155) = 92.65; p < .001.

Figure 6. Grand average of ICA in the LCT plus SURT condition. Gray bars indicate lane
change manoeuvres

Discussion
The Lane Change Task was developed to manipulate mental demands in driving
performance. Correspondingly, we observed the expected results in combination
with the secondary task. The mean deviation clearly increased when an additional
secondary task had to be performed while driving. The higher demand was also
reflected in the subjective measure of mental effort and in the mean ICA. The two
instruction modalities did not differ. Obviously, perceptual processing of the simple
signs cause only minimal additional costs, so that the driving condition is not
impaired by modality-specific interference.
One advantage of ICA is that one can additionally identify the time course of
cognitive demands during driving. The ICA dramatically changed during driving and
lance change. It significantly increased at the moment the command was presented
and the driver had to understand the instruction and had to change lanes. Hence, by
the ICA a higher temporal resolution could be reached when measuring mental

pupillometry as a method for measuring mental workload

11

demands while driving than could ever be obtained by using traditional global
performance measures such as the mean deviation from the optimal driving path or
subjective ratings. Because this effect was unaffected by the modality of instruction,
we conclude that the enhanced load is due to additional load on higher cognitive
processes and not perception.
The unexpectedly observed decrease of ICA in the secondary task condition,
demonstrates the potential of the ICA for a task analysis and for estimating the online demands of a task. When driving together with the secondary task, ICA
significantly decreased during the lane change manoeuvres. Such a decrease is
counterintuitive. It is plausible, however, when we analyse participants driving
behaviour together with their processing of the secondary task. They stopped
working on the SURT when the instruction was given and they focussed on driving.
In other words, they switch from a dual task (driving and performing the secondary
task) to a single task situation (only driving). This clearly demonstrates that a micro
analysis of task is necessary and it is not sufficient to evaluate the task as whole. We
have to consider such strategic serial scheduling of two tasks if the discrete
characteristic of the secondary task (or the primary task) allows such behaviour.
Applied to the driving situation, this follows the idea of a driver who actively
manages his workload when performing secondary tasks while driving.
If one follows this interpretation, the ICA has the ability to detect strategic shifts of
attention by the driver, as produced by a switch between a dual-task to a single-task
situation. ICA would therefore allow insights into the internal strategies of the driver
and his workload managing behaviour, independently of the type of system the driver
is interacting with. That the ICA is useful for a detection of strategic shifts has
previously been shown in experimental settings (Marshall, Pleydell-Pearce, Dickson,
2002, Marshall, 2002, 2007). But this is the first time it has been shown in a driving
context.
In summary, the current experimental results show that the Index of Cognitive
Activity could be a suitable method for measuring mental workload while driving
and to detect strategic shifts of attention by the driver. The index could therefore be a
valuable instrument when optimising new HMI concepts to the needs of the driver,
ensuring a safe and efficient use while driving. One can think on many other
applications. Examples are exploring driver learning, the acquisition of new skills,
the efficiency of training, the influence on drugs on cognitive load, effects of
advanced aging, etc. In all these cases, the Index of Cognitive Activity gives us the
chance to get insights into the time course of mental workload during driving. This
measure therefore is a valuable enrichment of the instruments of cognitive
ergonomics.
References
Ahern, S., & Beatty, J. (1979). Pupillary responses during information processing
vary with scholastic aptitude test scores. Science, 205, 1289-1292.

12

Schwalm, Keinath, & Zimmer

Ahern, S., & Beatty, J. (1981). Physiological evidence that demand of processing
capacity varies with intelligence. In M. Friedman, J. & Das J. & OConnor, O.
(Eds.), Intelligence and learning (pp. 121-128). New York: Plenum.
Beatty, J. (1982). Task-evoked pupillary responses, processing load, and the
structure of processing resources. Psychological Bulletin, 91, 276-292.
Beatty, J., & Kahneman, D. (1966). Pupillary changes in two memory tasks.
Psychonomic Science, 5, 371- 372.
Beatty, J., & Lucero-Wagoner, B. (1975). Pupillary measurement of sensory and
decision processes in a signal-detection task. Paper presented at the Meeting of
Psychonomic Society, Denver.
Beatty, J., Lucero-Wagoner, B., Cacioppo, J.T., Tassinary, L.G., & Berntson, G.G.
(2000). The pupillary system. New York, NY, US: Cambridge University Press.
Ben-Nun, Y. (1986). The use of pupillometry in the study of on-line verbal
processing: Evidence for depths of processing. Brain and Language, 28, 1-11.
Bradshaw, J.L. (1968). Pupil size and problem solving. The Quarterly Journal of
Experimental Psychology, 20, 116- 122.
De Waard, D. (1996). The measurement of drivers mental workload.PhD Thesis,
University of Groningen. Haren, The Netherlands: University of Groningen,
Traffic Research Centre.
Hakerem, G.A.D., & Sutton, S. (1966). Pupillary response at visual threshold.
Nature, 212, 485-486.
Hart, S.G., Staveland, L.E., Hancock, P.A., & Meshkati, N. (1988). Development of
nasa-tlx (task load index): Results of empirical and theoretical research.
Oxford, England: North-Holland.
Hess, E. H., & Polt, J. M. (1964). Pupil size in relation to mental activity during
simple problem-solving. Science, 143, 1190-1192.
Hockey, G.R.J. (1997). Compensatory control in the regulation of human
performance under stress and high workload: A cognitive-energetical
framework. Biological Psychology, 45, 73-93.
Just, M.A., & Carpenter, P.A. (1993). The intensity dimension of thought:
Pupillometric indices of sentence processing. Canadian Journal of
Experimental Psychology, 47, 310-339.
Just, M.A., Carpenter, P.A., & Miyake, A. (2003). Neuroindices of cognitive
workload: Neuroimaging, pupillometric and event-related potential studies of
brain work. Theoretical Issues in Ergonomics Science, 4, 56-88.
Kahneman, D., & Beatty, J. (1966). Pupil diameter and load on memory. Science
(New York, N.Y.), 154, 1583-1585.
Kahneman, D., & Beatty, J. (1967). Pupillary responses in a pitch-discrimination
task. Perception & Psychophysics, 2, 101-105.
Loewenfeld, I.E. (1993). The pupil: Anatomy, Physiology and Clinical Applications,
vols. I, II. Ames: Iowa State University Press / Detroit: Wayne State University
Press.
Marshall, S.P. (2000). U.S. Patent No. 6,090,051. Washington, DC: U.S. Patent &
Trademark Office.
Marshall, S.P. (2002). The index of cognitive activity: Measuring cognitive
workload. Paper presented at the 7th IEEE Conference on Human Factors and
Power Plants, New York.

pupillometry as a method for measuring mental workload

13

Marshall, S.P. (2005). Assesing cognitive engagement and cognitive state from eye
metrics. Paper presented at the 1st International Conference on Augemted
Cognition, Las Vegas.
Marshall, S.P. (2007). Identifying cognitive state from eye metrics. Aviation, Space,
And Environmental Medicine, 78, B165.
Marshall, S.P., Davis, C., & Knust, S. (2004). The index of cognitive activity:
Estimating cognitive effort from pupil dilation.San Diego, CA: Eyetracking Inc.
Technical Report ETI-0401.
Marshall, S.P., Pleydell-Pearce, C., & Dickson, B. (2002). Integrating
psychophysiological measures of cognitive workload and eye movements to
detetect strategy shifts. Paper presented at the 36th Hawaii International
conference on System Sciences.
Mattes, S. (2003). The lane-change-task as a tool for driver distraction evaluation.
Paper presented at the Annual Spring Conference of the GfA/17th Annual
Conference of International-Society-for-Occupational-Ergonomics-and-Safety
(ISOES).
Mattes, S. (2006). Messung der Fahrerablenkung in der Fahrzeugentwicklung.
Paper presented at the Vehicle Interaction Summit 3, Fraunhofer IAO,
Stuttgart, Germany.
ODonnell, R.D., Eggemeier, F.T., Boff, K.R., Kaufman, L., & Thomas, J.P. (1986).
Workload assessment methodology.Oxford, England: John Wiley & Sons.
Schluroff, M. (1982). Pupil responses to grammatical complexity of sentences. Brain
and Language, 17, 133-145.
Schluroff, M., Zimmermann, T.E., Freeman, R.B., Jr., Hofmeister, K., Lorscheid, T.,
& Weber, A. (1986). Pupillary responses to syntactic ambiguity of sentences.
Brain and Language, 27, 322-344.