Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
net/publication/228742286
CITATIONS READS
223 866
1 author:
Raja Parasuraman
George Mason University
295 PUBLICATIONS 18,327 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Raja Parasuraman on 02 June 2014.
Raja Parasuraman*
Keywords: Brain and behaviour; cognitive science; neuroscience; ergonomics; human factors.
1. Introduction
1.1. Neuroergonomics: definitions and scope
‘Neuroergonomics’ is the study of brain and behaviour at work. As the name
implies, this emerging area comprises two disciplines that are themselves interdisci-
plinary, neuroscience and ergonomics. Neuroscience is the study of brain structure
and function. Ergonomics (also known as human factors) examines human use of
technology at work and in other real-world settings. As the intersection of these two
fields, neuroergonomics is concerned both with the brain and with humans at work,
but more precisely with their dynamic interaction—with human brains at work
(Parasuraman 1998a).
Neuroergonomics focuses on investigations of the neural bases of such percep-
tual and cognitive functions as seeing, hearing, attending, remembering, deciding
and planning in relation to technologies and settings in the real world. Because the
human brain interacts with the world via a physical body, neuroergonomics is also
concerned with the neural basis of physical performance—grasping, moving or lift-
ing objects and one’s limbs. The real-world environments that neuroergonomics
deals with are many and diverse. They include, for example: working with computers
and various other machines at home, in the workplace, or when engaged in leisure
activities, using consumer products and operating vehicles such as aircraft, cars,
trains and ships.
* e-mail: parasuraman@cua.edu
Theoretical Issues in Ergonomics Science ISSN 1463–922X print/ISSN 1464–536X online # 2003 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
DOI 10.1080/14639220210199753
6 R. Parasuraman
tially more profitable questions about attention allocation than before. Measures of
brain function that reflect visual attention and oculomotor control can help deter-
mine the impact of the new display on the pilot’s visual scanning and attentional
performance. Finally, neuroergonomic evaluation of the manual and physical
demands involved in interacting with the information panels and controls of the
new traffic monitoring system would also be required for this system to be used
effectively and safely by pilots. Several other examples of neuroergonomic research
and practice are discussed in a later section of this article. More specifically, three
topic areas relevant to this example—assessment of cognitive workload, attention
and oculomotor control, and manual performance—are each discussed from a neu-
roergonomic perspective in other papers in the two special issues on neuroergo-
nomics in this journal, by Just et al. (2003), Kramer and McCarley (2003) and
Karwowski, Siemionow and Gielo-Perczak (2003), respectively.
ers from artificial intelligence and computer science. Their rallying cry was that the
mind is software. The task of cognitive science, in their view, was to determine the
characteristics and rules of the software, irrespective of the hardware that imple-
ments the software. Accordingly, the rules of mind could as easily be studied in a
computer as they could in a human. The actual hardware, the structures and
mechanisms of the brain, were deemed unimportant. In contrast, the recently devel-
oped field of cognitive neuroscience proposes that neural structure and function
constrain and, in some cases, determine theories of human mental processes (Gaz-
zaniga 2000). The influence of neuroscience has carried over not only to the other
cognitive sciences (Pinker 1997), but also to philosophy (Churchland 2002).
If neuroscience has freed cognitive science from rigid functionalism, then ergo-
nomics may serve to liberate it from a disembodied existence devoid of context and
provide it an anchor in the real world. Even though researchers are aware of the
importance of ecological validity, modern cognitive psychology (with a few excep-
tions) tends to study mental processes in isolation, apart from the artifacts and
technologies of the world that require the use of those processes. However, it may
be more useful to see technology, particularly computers, as representing an exten-
sion of human cognitive capability, a view that emerged as psychology escaped from
the grip of behaviourism in the late 1940s (Craik 1947). A modern version of this
doctrine is the view espoused by the new field of cognitive engineering that humans
and intelligent computer systems constitute ‘joint cognitive systems’ (Roth et al.
1987, Hutchins 1995). Furthermore, much human behaviour is situated and context
dependent. Context is often defined and even driven by technological change. How
humans design, interact with and use technology—the essence of ergonomics—
should, therefore, also be central to cognitive science.
The idea that cognition should be considered in relation to action in the world
has many antecedents. Piaget’s (1952) work on cognitive development in the infant
and its dependence on exploration of the environment anticipated the concept of
situated or embodied cognition. In a recent philosophical work, Clark (1997) provides
a modern statement of this thesis. Going beyond the old distinction between mind
and matter, Clark examines the characteristics of an embodied mind that is shaped
by and helps shape action in a physical world. If cognitive science should, therefore,
study mind not in isolation but in interaction with the physical world, then it is a
natural second step to ask how to design artifacts in the world that best facilitate that
interaction. This is the domain of ergonomics. Neuroergonomics goes one, critical,
step further. It postulates that the human brain that implements cognition and is
itself shaped by the physical environment must also be examined in order to under-
stand fully the inter-relationships of cognition, action and the world of artifacts.
Currently, a coherent body of concepts and empirical evidence that constitutes
neuroergonomics theory does not exist. Of course, broad theories in the human
sciences are also sparse, whether in ergonomics (Hancock and Chignell 1995) or in
neuroscience (Albright et al. 2001). What one finds are small-scale theories that
could be integrated into a macrotheory but which would still apply only to a specific
domain of human functioning. For example, neural theories of attention are becom-
ing increasingly well specified, both at the macroscopic level of large-scale neural
networks (Posner and Dehaene 1994, Parasuraman 1998b), at the level of neuronal
function (Sarter et al. 2001), and, in the near future, at the more fundamental level of
individual genes and their protein products (Fosella et al. 2002, Parasuraman et al.
2002). At the same time, psychological theories of attention have informed human
Neuroergonomics 9
factors research and design (Wickens and Hollands 2000). Difficult though the task
may be, one can envisage amalgamation of these respective theories into a neuroer-
gonomic theory of attention. Integration across a broader range of functional
domains, however, is as yet premature.
neuroergonomics given that this group is already established? After all, the two fields
share one goal—the design of safe and efficient technologies and systems for human
work. Nevertheless, the two represent distinct endeavours for the following reasons.
First, whereas psychophysiology focuses on physiological measures and their
psychological correlates, neuroergonomics is more directly concerned with brain
function and not with physiological measures per se. The physiological measure is
primarily of interest to the extent that it provides an index of neural activity related
to the perceptual, cognitive, affective or motor functioning of the human. The meas-
ure is used principally to evaluate a theory of brain function or to see how brain
function is linked to human performance or use of technology. These characteristics
apply to the brain imaging methods discussed previously, so that even if one were to
re-label and include fRMI or optical imaging as psychophysiological measures, the
distinction between neuroergonomics and psychophysiology would remain.
Secondly, if the physiological measure is more ‘distant’ from brain function, as
with measures such as heart rate and skin conductance that are typically used in
psychophysiological research, then the link to neuroergonomics is less direct or
relevant. Thirdly, psychophysiology also has a major focus on autonomic nervous
system (ANS) measures in relation to somatic factors, emotion and stress. Although
these topics are also of interest to neuroergonomics, especially as they are linked to
brain function, a dominant theme in neuroergonomics is neural activity in inter-
action with the ANS and in relation to human performance. According to this
view, stress and emotion are certainly valid topics for neuroergonomics research,
but only if they are grounded in theories of brain function.
Another major reason for distinguishing neuroergonomics from psychophysiol-
ogy is that one can easily envisage a neuroergonomic study in which no physiological
measure at all is used as a dependent variable, which would rule out a classification
of the work as falling within the realm of psychophysiology. Yet, such a study could
clearly qualify as a neuroergonomic investigation. Consider the following example.
Suppose that as a result of the manipulation of some factor, performance on a target
discrimination task (e.g. detection of an ‘intruder’ aircraft in the cockpit traffic
monitoring example discussed previously) in which location cues are provided
prior to the target yields the following results: reaction time (RT) to the target
when preceded by an invalid location cue is disproportionately increased while
that to a valid cue is not. This might happen, for example, if the cue is derived
from the output of an automated detection system, which is not perfectly reliable
(Hitchcock et al. 2003). In simple laboratory tasks using such a cueing procedure,
there is good evidence linking this performance pattern to a basic attentional opera-
tion and to activation of a specific distributed network of cortical and sub-cortical
regions—on the basis of previous research using non-invasive brain imaging in
humans, invasive recordings in animals and performance data from individuals
who have suffered damage to these brain regions (Posner and Dehaene 1994). One
could then conduct a study using the same cueing procedure and performance meas-
ures as a behavioural ‘assay’ of the activation of the neural network in relation to
performance of a more complex task in which the same basic cognitive operation is
used. (See Parasuraman et al. (2002) for a discussion of the logic of behavioural
assays of brain function as used in a quite different area of study—the genetic
contribution to individual differences in cognition.) If the characteristic performance
pattern was observed—a disproportionate increase in RT following an invalid loca-
tion cue, with a normal decrease in RT following a valid cue, then one could argue
Neuroergonomics 11
times (Wiener 1988, Parasuraman and Riley 1997). This suggests the need for linking
the provision of automated support to the level of operator workload or performance.
Automated support is needed only when the operator’s mental workload is relatively
high, so that the operator is likely to benefit by the freeing up of cognitive resources
that automation provides. At other times, automated support need not be given, or, if
operator workload is too low and there is a danger of the operator becoming disen-
gaged from the system, a task previously carried out by automation can be re-allocated
to the operator. Parasuraman et al. (1996) showed that such temporary re-allocation of
an automated task to human control promotes better monitoring of the automation
and improved fault detection when anomalies occur.
Either of these adaptive strategies—automated support when workload is high,
or task re-allocation when it is low—have been found to benefit performance when
their implementation is closely matched to operator workload (Parasuraman et al.
1999). However, accomplishing such workload-matched adaptation is dependent on
the ability to measure the operator’s mental workload in real time. Although mental
workload can be measured using a number of different techniques (Byrne and
Parasuraman 1996), measures of brain function offer particular benefits for use in
adaptive systems (Scerbo et al. 2001, Parasuraman and Caggiano 2002). First, meas-
ures of cognitive-related brain activity, unlike most behavioural measures (with the
exception of continuous motor tasks) can be obtained continuously. In many
systems where the operator is placed in a supervisory role over automated sub-
systems, very few overt responses (e.g. button presses or cursor movements) may
be made even as the operator is engaged in considerable cognitive activity. In such a
situation, a behavioural measure (e.g. RT or cursor movement error) provides a
relatively impoverished sample of the mental activity of the operator.
Physiological measures, on the other hand, may be recorded continuously without
respect to overt responses and may provide a measure of the covert activities of the
human operator. Moreover, if the measures can be directly linked to brain activity,
as neuroergonomic measures should, then sensitive, real-time monitoring of cogni-
tive workload may be possible. In other words, brain measures have a higher band-
width than behavioural or performance measures. As discussed previously, therefore,
automated systems in which little overt behaviour can be recorded provide a par-
ticularly compelling case for the neuroergonomic approach (see also Kramer and
Weber 2000).
Secondly, in some instances, measures of brain function may provide more infor-
mation when coupled with behavioural measures than behavioural measures alone.
For example, changes in RT may reflect contributions of both central processing
(working memory) and response-related processing to workload. However, when
coupled with the amplitude and latency of the P300 component of the ERP, such
changes may be more precisely localized to central processing stages than to
response-related processing (Donchin et al. 1986, Wickens 1990, Kramer and
Weber 2000). In addition, measures of brain function can indicate not only when
an operator is overloaded, drowsy, or fatigued, but also which brain networks and
circuits may be affected. In short, neuroergonomic measures offer new avenues for
adaptive interventions aimed at enhancing system performance.
reflecting how hard one’s mind is working at any given moment. Accordingly, men-
tal workload can be associated with brain work. Various candidate measures for
brain work have been proposed over the years, all of which owe their existence to the
brilliant early investigations of Sherrington, who studied the regulation of the blood
supply of the brain (Roy and Sherrington 1890). Sherrington demonstrated that
there is a close coupling between the electrical activity of neuronal cells, the energy
demands of the associated cellular processes and regional blood flow in the brain.
His pioneering investigations suggested that, if mental activity results in increased
neuronal response in localized regions of the brain, then in principle it should be
possible to measure mental workload by assessing regional cerebral metabolism and
blood flow.
It took several years before sensitive techniques were developed for measuring
regional brain blood flow in humans. An early development was the invention of the
Xenon-133 (Xe-133) method for assessing regional cortical changes in brain blood
flow and glucose metabolism. Injection in human patient volunteers of the radio-
actively tagged Xenon gas, which passes freely across the blood–brain barrier,
showed that performance of various mental tasks led to increased metabolic activity
in specific cortical regions. However, the Xe-133 technique was too invasive to be
used routinely in normal human participants. The development of PET paved the
way for less invasive measurement of regional cerebral metabolism and blood flow.
Regional cerebral glucose metabolism can be non-invasively determined using PET
and radioactively labelled glucose (18-fluoro-deoxyglucose), while regional cerebral
blood flow may be assessed with PET and radioactively-labelled oxygen (O-15) in
water. A disadvantage of PET is the need for ionizing radiation, which, although
safe when used within exposure limits, prevents frequent use in studies with normal
human participants. The recent development of fMRI has overcome this limitation.
fMRI provides non-invasive, high-resolution assessments of regional cerebral blood
flow that can be repeatedly obtained in the same participant.
There is now an emerging literature indicating that different PET and fMRI
measures of brain activity, as well as electromagnetic measures such as EEG and
ERPs, provide sensitive indexes of moment-to-moment variations in mental work-
load (Parasuraman and Caggiano 2002). The potential use of these and other phy-
siological measures for real-time assessment of mental workload in adaptive human–
machine systems has also been documented (Parasuraman 1990, Kramer et al. 1996,
Scerbo et al. 2001). Prinzel et al. (2000) have also specifically demonstrated the
feasibility of an adaptive system based on EEG measures (see also the article by
Scerbo et al. (2003) in the second special issue on neuroergonomics in this journal).
Technical developments will determine the further utility of measures of brain
function in neuroergonomic research on mental workload and adaptive systems. For
example, PET and fMRI are currently expensive to acquire and to operate, cumber-
some, unduly restrictive of participant movement, and are not portable. These char-
acteristics limit their routine use in neuroergonomic studies. Despite the limitations,
Peres et al. (2000) recently reported an fMRI study examining changes in regional
cerebral blood flow in pilots performing a flight control task. Although it has much
poorer spatial resolution than PET or fMRI, transcranial Doppler sonography
(TCD) has high temporal resolution, thus allowing for continuous, real-time mon-
itoring of cerebral blood flow, which facilitates neuroergonomic applications (see the
TCD study by Hitchcock et al. (2003) in this special issue). The newer optical
imaging technologies that are currently being developed will also be cheaper and
14 R. Parasuraman
more portable (Gratton and Fabiani 2001) and will lead to more ergonomic applica-
tions.
6. Conclusions
Neuroergonomics represents a deliberate merger of neuroscience and ergonomics
with the goal of advancing understanding of brain function underlying human per-
formance of complex, real-world tasks. A second major goal is to use existing and
emerging knowledge of human performance and brain function to design tech-
nologies and work environments for safer and more efficient operation. More pro-
gress has been made on the first goal than on the second, but both neuroergonomic
research and practice should flourish in the future, as the value of the approach is
appreciated. The basic enterprise of ergonomics—how humans design, interact with
and use technology—can be considerably enriched if we also consider the human
brain that makes such activities possible.
References
Albright, T. D., Jessell, T. M., Kandel, E. R. and Posner M. I. 2001, Progress in the
neural sciences in the century after Cajal (and the mysteries that remain), Annals of the
New York Academy of Sciences, 929, 11–40.
Bainbridge, L. 1983, Ironies of automation, Automatica, 19, 775–779.
Neuroergonomics 17
Brain Potentials: Basic and Applied Issues (New York: Oxford University Press), 301–
309.
Wickens, C. D. and Hollands, J. G. 2000, Engineering Psychology and Human Performance
(New York, NY: Longman).
Wiener, E. L. 1988, Cockpit automation, in E. L. Wiener and D. C. Nagel (eds), Human
Factors in Aviation (San Diego, CA: Academic Press), 433–461.
Wiener, E. L. and Curry, R. E. 1980. Flight-deck automation: promises and problems,
Ergonomics, 23, 995–1011.