Sustained performance under overload: Personality and individual differences in stress and coping

Abstract

Individuals differ considerably in their vulnerability to task-induced stress, in part because of individual differences in cognitions of task demands. This study investigated the personality and cognitive factors that may control stress vulnerability, using a ‘rapid information processing’ task that was configured to overload attention. Stress response was assessed using the Dundee Stress State Questionnaire (Matthews, G. et al., 2002. Fundamental dimensions of subjective state in performance settings: task engagement, distress and worry. Emotion, 2, 315–340), as well as instruments assessing workload, appraisal and coping. Time pressure was manipulated as a between-subjects stress factor. Higher time pressure tended to elicit decreased effort and task engagement and avoidance coping. However, much of the variance in state response was attributable to individual differences in appraisal and coping. The personality trait of neuroticism related to some of these cognitive processes. Subjective state, appraisal and coping were also predictive of objective performance indices. Consistent with the transactional theory of stress, subjective states appear to correspond to configurations of cognitive processes that signal the participant's mode of adaptation to task demands. The findings underscore the importance of accommodating individual differences in selecting operators for handling overload, for designing interfaces and for training operators to manage overload successfully.

Full text

Theoretical Issues in Ergonomics Science
Vol. 10, No. 5, September–October 2009, 417–442
Sustained performance under overload: personality and individual
differences in stress and coping
Gerald Matthews
a
*
and Sian E. Campbell
b
a
Department of Psychology, University of Cincinnati, Cincinnati, OH 45221, USA;
b
South East Employers, Newfrith House, 21 Hyde Street, Winchester, SO23 7DR, UK
(Received 20 February 2008; final version received 1 June 2009)
Individuals differ considerably in their vulnerability to task-induced stress, in
part because of individual differences in cognitions of task demands. This study
investigated the personality and cognitive factors that may control stress
vulnerability, using a ‘rapid information processing’ task that was configured
to overload attention. Stress response was assessed using the Dundee Stress State
Questionnaire (Matthews, G. et al., 2002. Fundamental dimensions of subjective
state in performance settings: task engagement, distress and worry. Emotion,2,
315–340), as well as instruments assessing workload, appraisal and coping. Time
pressure was manipulated as a between-subjects stress factor. Higher time
pressure tended to elicit decreased effort and task engagement and avoidance
coping. However, much of the variance in state response was attributable to
individual differences in appraisal and coping. The personality trait of
neuroticism related to some of these cognitive processes. Subjective state,
appraisal and coping were also predictive of objective performance indices.
Consistent with the transactional theory of stress, subjective states appear to
correspond to configurations of cognitive processes that signal the participant’s
mode of adaptation to task demands. The findings underscore the importance
of accommodating individual differences in selecting operators for handling
overload, for designing interfaces and for training operators to manage overload
successfully.
Keywords: stress; sustained attention; task engagement; coping; overload
1. Introduction
Task-induced stress is ubiquitous in a range of operational settings, including industry,
transportation and medical practice. Work activities may elicit stress responses that are
governed by the nature of the task performed itself, as opposed to extraneous stressors
such as environmental factors (e.g. noise), social interactions (e.g. poor relationships with
co-workers) and the organisational climate (Matthews et al. 2002, 2006b). High workload,
monotony, frustration and provision of negative feedback are all qualities of tasks that are
potentially stressful. The introduction of new technology into the workplace may also
contribute to stress (Schabracq and Cooper 2000). Individuals differ markedly in their
*Corresponding author. Email: matthegl@email.uc.edu
ISSN 1463–922X print/ISSN 1464–536X online
ß 2009 Taylor & Francis
DOI: 10.1080/14639220903106395
http://www.informaworld.com
Downloaded by [University of Central Florida] at 12:41 15 August 2014

vulnerability to task-induced stress. Some operators appear to remain calm and focused
even under the most difficult circumstances, whereas others are easily upset by even minor
difficulties in task execution. This article will consider some of the theoretical and practical
challenges inherent in predicting individual differences in task-induced stress response and
its impact on operator performance.
Current understanding of individual differences in operator stress is limited. Most
research on instruments that are potentially diagnostic has focused on personality traits
linked to general stress vulnerability. Within the Five Factor Model (McCrae and
Costa 2003), the most important trait is neuroticism, which is reliably linked to a variety
of indices of acute and chronic stress (Matthews et al. 2003). Neuroticism relates also
to measures of occupational stress, such as low job satisfaction (Judge et al. 2002), and to
measures of negative affect in performance settings (Matthews and Gilliland 1999).
However, despite extensive evidence for predictive validity, neuroticism is limited in its
utility as a predictor of task-induced stress and compromised performance. Matthews and
Gilliland (1999) noted that when true state measures are used to assess stress response
(i.e. immediate feelings) correlations between neuroticism and negative affect are robust
but of moderate magnitude (typically 0.2–0.4). Furthermore, high neurotic persons
experience more negative mood even before exposure to any task demand (Matthews et al.
1999), so that correlations observed following task performance may simply carry over
from the pre-task state. Matthews et al. (2006b) investigated neuroticism as a predictor of
distress induced by three stressful tasks. A regression analysis showed that neuroticism
predicted increased distress relative to initial baseline. However, personality factors
explained less than 4% of the variance in post-task distress, with pre-task distress
statistically controlled. The study also showed that the effect of neuroticism was fully
mediated by individual differences in emotion-focused coping. Thus, while neuroticism
may be a useful predictor, it is only modestly related to individual differences in the
distress response to the task.
It is also uncertain whether neuroticism is a robust predictor of the performance
impairment that may accompany stress. Meta-analysis of data on personality and job
performance typically fails to show any robust general association between neuroticism
and poorer performance (Barrick et al. 2001). However, neuroticism and other personality
traits may be more predictive of performance in theory-driven studies that link personality
to specific criteria, such as performance under stress (e.g. Tett and Burnett 2003).
Laboratory studies also provide a very mixed picture of the effects of neuroticism
(Matthews and Gilliland 1999). Relationships between neuroticism and performance on
tasks making high attentional demands vary considerably across studies (see Szymura and
Wodniecka 2003). Consistent with cognitive theories of stress, neuroticism may relate to
performance impairment especially when the operator is taxed by changing task demands
(Cox-Fuenzalida et al. 2004). Even when a link between neuroticism and performance
is found, the mechanism may be unclear. High neurotic individuals may be especially
vulnerable to impairments in attention, working memory or executive control as a con-
sequence of stress, including time pressure (Szymura in press). Alternatively, similar to
trait anxiety, neuroticism may relate to qualitative biases in selective attention, such
as increased selectivity (‘attentional narrowing’), or diversion of attention to potential
sources of threat (Eysenck et al. 2007).
Other personality traits may add to the predictive validity of neuroticism. Extraversion
has also been linked to hardiness under stress and, under some circumstances, to
better performance on high workload tasks, especially those requiring verbal or symbolic
418 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

processing (Matthews 2008). Again, though, extraversion is only weakly related to
job performance (Barrick et al. 2001) and to affect in controlled laboratory settings
(Matthews and Gilliland 1999) and its effects on performance are complex and dependent
on task and environmental factors (Matthews and Gilliland 1999). Predisposition to
negative affectivity may also be unpacked into numerous more narrowly defined traits,
including trait anxiety, pessimism, lack of hardiness, negative meta-cognition and others.
Some of these traits appear useful for predicting operator stress (e.g. Helton et al. 1999),
but, as with neuroticism, effect sizes in relation to task-induced stress and task
performance are typically modest.
There are also compelling theoretical reasons for the limited utility of trait measures.
Traits are multi-levelled constructs that integrate many overlapping components of
personality, of differing relevance to performance. From a cognitive science perspective
traits may be seen as distributed across multiple, independent biases in neurological
functioning, in information processing and in personal understanding of the task
environment (see Matthews (2008) for a formal cognitive science theory). Thus, trait
measures may not be strongly predictive of the specific information-processing routines
controlling a given task. This paper will mention briefly that the traditional psycho-
physiological theories of personality, such as arousal theory (Eysenck and Eysenck 1985),
have not been very successful as a basis for predicting cognitive task performance
(Matthews and Gilliland 1999). Trait effects are also typically moderated by situat-
ional and contextual factors that may vary across different operational tasks (Matthews
et al. 2003).
1.1. Individual differences and transactional theory
Traditional theories of stress saw the construct as either an attribute of external, noxious
stimuli or as a generalised response (Selye 1976). By contrast, contemporary, transactional
theories of stress (Lazarus and Folkman 1984, Lazarus 1999) conceptualise stress as
a relational construct that describes a relationship between external demands and
the person’s active efforts at demand management. The person–task interaction is also at
the core of many contemporary theories of stress and performance (e.g. Hancock
and Warm 1989, Hockey 1997, Hancock and Szalma in press). Stress refers (in the
performance context) to task demands that tax or overload the person’s abilities to cope
with task load (Matthews 2001). It follows that individual differences cannot be under-
stood simply as a broad vulnerability to stressors, or as a scaling factor for a general stress
response. Stress theory also highlights the variability of stress response as a concomitant
of dynamic, changing interactions between operator and task (Szalma et al. 2004).
Thus, to predict performance, it is essential to look at transient mental states, as well
as enduring traits.
Matthews (2001) argues that operator–stressor interactions are multi-levelled transac-
tions that may be variously understood in terms of biophysics, symbolic information-
information processing, or assignment of personal meaning to the task. Often, there are
multiple paths that link the impact of the stressor to observable performance change.
A basic distinction is between ‘biocognitive’ paths mediated by changes in the operations
of components of the neural and/or cognitive architectures and ‘cognitive-adaptive’
changes that reflect the operator’s changing goals and strategies for attaining those goals
(cf. Hockey 1997). (A reviewer of this article correctly noted that strategy choice is itself
Theoretical Issues in Ergonomics Science 419
Downloaded by [University of Central Florida] at 12:41 15 August 2014

supported by a neural architecture, as evidenced by numerous studies of the frontal circuits
supporting executive control. The point here is that understanding voluntary strategy
choice requires an understanding of self-beliefs and the personal meaning the individual
attaches to the performance environment, separate from an understanding of how
voluntary control is implemented via neurocognitive processes).
It follows that there are multiple sources of individual differences (that may not map
simply onto standard personality traits).
This article examines the potential contribution of the transactional model of stress to
understanding individual differences in performance. The transactional model as typically
expressed (e.g. Lazarus 1999) is rather vague about information processing. It offers
a general framework for understanding stress but does not specify in detail how the
self-regulative processes underpinning stress response impinge on processing and task
performance. Progress requires the investigation of the key stress processes identified by
Lazarus – appraisal and coping – in performance settings (Matthews 2001). Appraisal
refers to the evaluation of the personal significance of stimuli (and of personal
competence). Performance may be influenced by the extent to which the operator
appraises the task as threatening, challenging and difficult to control. Perceived lack of
controllability, in particular, has been identified as a source of stress and performance
degradation (Fisher 1984, Siegrist and Marmot 2004). Coping refers to the operator’s
efforts to regulate task demands in relation to personal aspirations. The coping strategies
described by Lazarus (1999) and other stress theorists may be related to performance by
specifying how they operate within a cognitive architecture (Wells and Matthews 1994).
Matthews and Desmond (2002) linked task-focused coping to investment of effort in
performance, avoidance to deliberate withdrawal of effort and lowering performance
standards, and emotion-focus to processing the personal significance of performance
impairment. Task-focused coping is expected to be more adaptive than emotion-focus or
avoidance in the performance setting, provided that the operator has the skill to implement
task-focused strategies effectively.
The transactional approach also informs understanding of temporary mental states
such as anxiety and fatigue that may mediate effects of personality traits on performance.
Matthews et al. (1999, 2002) have proposed a comprehensive model of subjective states
in performance settings that proposes three distinct dimensions of stress response. Task
engagement is defined by energy, motivation and concentration; loss of engagement
corresponds to fatigue. Distress refers to tension, negative mood and lack of confidence,
whereas worry is defined by self-focused attention, low self-esteem and intrusive thoughts
about the task and personal concerns. Matthews et al. (2002) suggest that these
fundamental states index different operator–task transactions. Task engagement signals
commitment to task-directed effort, distress signals management of overload of capacity
and worry is an index of self-evaluation in the performance context. Consistent with this
position, Matthews et al. (2002) found that, in an organisational setting, the state factors
were associated with different patterns of appraisal and coping with work demands.
Because of the multi-levelled nature of transactions, they may also be expressed through
individual differences in parameters of neural and cognitive architectures.
1.2. Application to sustained performance
Potentially, an important application for the transactional model is understanding
individual differences in sustained performance and vigilant attention. People differ
420 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

markedly in their ability to sustain attention in operational environments, including
monitoring displays in industrial, medical, military and transportation settings
(Warm et al. 2008). Individual differences are similarly conspicuous in laboratory settings.
There is a long research tradition of using personality scales to predict vigilance in applied
and laboratory studies, which has had very mixed outcomes. Various traits have been
identified as promising predictors (e.g. Rose et al. 2002), but even the more successful ones,
such as extraversion–introversion, tend to be rather weakly and inconsistently related to
performance, at best (Koelega 1992). One source of inconsistency is that trait effects are
often dependent on a host of moderator factors including, in the case of extraversion, time
of day, arousal and task demands (Matthews et al. 2003). Investigation of such interactive
effects is of prime importance for theory building (e.g. Humphreys and Revelle 1984,
Matthews 2008), but unhelpful for prediction in the uncontrolled environments of real life.
By contrast, research suggests that state factors are more reliably predictive of vigilance
than are traits. Associations between states and performance may be understood within the
influential workload/resource model of vigilance (Warm and Dember (1998)). Several
studies show that the mood dimension of energetic arousal – contrasting vigour with
tiredness – predicts subsequent vigilance performance (see Matthews and Davies 1998 for a
review). Furthermore, energy predicts vigilance only when task demands are high, as
predicted by the theory that vigilance tasks require increasing investment of attentional
resources as task demands increase (Parasuraman et al. 1987). The workload-dependence of
the energetic arousal effect is the opposite to that predicted by traditional arousal theory,
which supposes that easier tasks should be more likely to show beneficial effects of arousal
(Matthews et al . 1990b). Resource theory is also supported by findings that energy is related
to elevated performance on other demanding attentional tasks, such as visual and semantic
search (e.g. Matthews and Margetts 1991). The applied relevance of these results is
demonstrated by generalisation of the energy effect to skilled performance in paradigms
including multi-tasking (Singh et al. 1993) and simulated vehicle driving (Funke et al . 2007).
More recent work has used the Dundee Stress State Questionnaire (DSSQ; Matthews
et al. 1999, 2002) to assess the three broad state factors of task engagement, distress
and worry, together with their various facets. Energetic arousal loads highly on the
task engagement factor. Studies have confirmed that, like energy, task engagement is
a consistent predictor of a range of demanding vigilance tasks, including tasks requiring
both a sensory and a cognitive discrimination (Matthews et al. 1999, 2001, 2006a, Helton
et al. 2009). Vigilance studies using the DSSQ have also confirmed that even relatively
short monitoring tasks elicit a distinctive pattern of distress and task disengagement, which
differs from stress profiles produced by other demanding tasks (Matthews et al. 2006b).
Furthermore, the state response appears to be sensitive to psychophysical properties of
the task including signal salience, sensory modality and task duration (e.g. Szalma et al.
2004, Helton et al. 2009). However, although performance of vigilance tasks may be
distressing, individual differences in distress do not seem to be reliably associated with
performance deficit. The more predictive factor is task engagement. From the applied
standpoint, though, the difficulty is that task engagement is a transient state rather than
a stable trait, so the measure may not be suitable for prediction of typical operational
performance. This critical issue will be discussed later in discussion.
1.3. Aims and hypotheses
The objective of the present study was to investigate predictors of sustained monitoring in
an overload situation. The study employed a ‘rapid information processing’ (RIP) task
Theoretical Issues in Ergonomics Science 421
Downloaded by [University of Central Florida] at 12:41 15 August 2014

(Wesnes and Warburton 1983) configured with a rate of stimulus presentation (event rate)
that was so high that most participants could not perform the task effectively. Wesnes
and Warburton (1983) used an event rate of 100 stimuli/min to tax sustained attention.
The present study compared performance at higher (150 stimuli/min) and lower
(75 stimuli/min) rates to manipulate task load. It was expected that the scope for
performance failure and lack of control of event rate would be stressful for participants,
affording the opportunity to test predictors of stress response and performance under
overload. The aim was to test for correlates both of overall level of performance and
performance change during the 15-min task duration: previous studies have confirmed
that demanding short-duration tasks may elicit both vigilance decrement and stress
response (e.g. Temple et al. 2000).
The task may provide a limited analogue to real-life overload situations, in which
stimuli cannot all be fully processed, including military command and control and
emergencies in industrial process control (e.g. Three Mile Island). The transactional
theory of stress predicts that the experience of overload may elicit maladaptive
strategies such as emotion-focus (brooding on the aversive experience) and avoidance
(setting lower targets for performance), in place of more adaptive task-focus (Wells and
Matthews 1994). The task should elicit both distress, due to overload, and task
disengagement because of the need for sustained attention in a monotonous environment
(Matthews et al. 2002). The manipulation of event rate was expected to accentuate these
general trends.
Because the task was designed to be stressful, it was anticipated that neuroticism would
relate to both subjective response and performance. However, as previously discussed,
neuroticism (and the related trait of anxiety) is not very reliable as a predictor of vigilance
(Davies and Parasuraman 1982, Rose et al. 2002). However, because the task was intended
to be unusually stressful (cf. Cox-Fuenzalida et al. 2004), it was expected that neuroticism
might relate to a lower detection rate, especially at the higher event rate. Extraversion is
a paradoxical trait in the vigilance context. It is typically related to positive affect and
the dopaminergic reward system of the brain (e.g. Wacker et al. 2006), implying that it
should enhance sustained attention. In fact, extraversion is usually negatively correlated
with detection frequency across a range of tasks (Koelega 1992, Beauducel et al. 2006).
Matthews et al. (1990a) suggested that extraversion–vigilance associations are dependent
on task parameters and extraverts may even out-perform introverts when the task requires
high-workload symbolic processing. The traditional analysis of extraversion would predict
poorer performance; Matthews et al.’s (1990a) analysis predicts that extraverts might
out-perform introverts.
Turning to state factors, it was predicted on the basis of prior findings (e.g. Matthews
et al. 2001, 2006a) that task engagement would be the main state correlate of better
performance, assuming the task to be resource-demanding. Pharmacological research
with the RIP task has also shown that it is sensitive to agents including caffeine that may
increase engagement and task interest (e.g. Frewer and Lader 1991). The study aimed
to extend previous research by also investigating appraisal and coping as predictors of
performance. It was expected that the cognitive underpinnings of engagement – challenge
appraisal, task-focused coping, low avoidance coping (Matthews et al. 2002) – would also
support superior performance. Finally, the study aimed to explore the extent to which
trait and state factors predict overlapping or independent portions of the variance in
performance using multivariate analyses.
422 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

2. Method
2.1. Participants
The participants were 144 undergraduate students (60% female) attending a British
university. They were paid a fee of £10 for participation. They ranged in age from 17 years
to 34 years, with a mean age of 20.4 years (SD 3.6 years). All participants had normal or
corrected-to-normal vision.
2.2. Apparatus and measures
2.2.1. Rapid information processing task
Task duration was 15 min, analysed as five continuous 3-min periods of performance.
Participants were presented with a sequence of single digit stimuli (0–9). Each digit was
approximately 8 mm (vertical)  5 mm (horizontal) and was viewed from a distance of 1 m.
Each stimulus was presented for 120 ms on a 15 inch computer monitor. In the higher
event rate condition, 150 stimuli/min were presented; in the lower event rate condition,
75 stimuli/min were presented. The lower event rate nevertheless represents a rapid rate
relative to the typical vigilance study; event rates of greater than 24 stimuli/min are often
classified as ‘high’ (See et al. 1995). Critical signals were the final digit presented in
a sequence of either three odd digits or three even digits. 8% of digits presented were
critical signals. That is, at the higher event rate, there were 36 critical signals in each 3-min
task period, whereas, at the lower event rate, 18 critical signals were presented in each
3-min period. Participants signified their detection of critical signals by pressing the
spacebar on the computer keyboard. No other responses were required. Responses
occurring within a window of 1200 ms after the onset of the critical signal were recorded as
correct detections. All other key presses were recorded as false alarms.
2.2.2. Questionnaire measures
Questionnaires were used to measure five types of construct: personality; subjective state;
workload; appraisal; coping. The Eysenck Personality Questionnaire-Revised (EPQ-R;
Eysenck et al. 1985) was used to measure extraversion and neuroticism traits. Items for
additional psychoticism and lie scales were not used. The EPQ-R has been extensively
employed in personality research (Eysenck and Eysenck 1985).
Subjective state was measured with the DSSQ (Matthews et al. 1999, 2002). The
questionnaire comprises 11 scales relating to mood (energetic arousal, tense arousal,
hedonic tone), motivation (interest motivation, success motivation) and cognitive state
(self-consciousness, self-esteem, concentration, confidence and control, task-related
cognitive interference, task-irrelevant cognitive interference). Scales have been shown to
be internally consistent in a variety of performance contexts. They are also appropriately
sensitive to various task and environmental stressors (Matthews et al. 1999, 2002, 2006b).
In the present study, data analysis focused on the three higher-order factors of the DSSQ,
derived from factor analysis of the 11 first-order scales (Matthews et al. 2002). Factor
scores for task engagement, distress and worry were estimated using regression weights
from a large normative sample (Matthews et al. 1999). Factor scores are distributed with
a mean of 0 and SD 1, so that values calculated for a sample represent a deviation from
normative values in standard deviation units.
The post-task version of the DSSQ, administered following task performance, includes
a modified version of the NASA-Task Load Index (TLX) (Hart and Staveland 1988)
Theoretical Issues in Ergonomics Science 423
Downloaded by [University of Central Florida] at 12:41 15 August 2014

workload measure. The participant is asked to rate six key features of workload on 0–10
scales: mental demand, physical demand, temporal demand, performance difficulty, effort
and frustration/stress. Overall workload was calculated as the unweighted mean of the six
ratings.
Appraisal was measured with the threat and challenge scales (six items per scale) of the
Appraisal of Life Events scale (ALE; Ferguson et al. 1999). Scales are internally consistent
and have been validated in studies of life stressors described by Ferguson et al. (1999). The
ALE is directed towards primary appraisal (Lazarus 1999). An additional eight-item scale
for controllability developed and validated by Falconer (2002) was used to assess this
element of secondary appraisal. Sample items are: ‘The task was a situation ... which you
could deal with effectively’ (positive) and ‘The task was a situation ... in which efforts to
change the situation made it worse’ (negative).
Coping was measured using the Coping Inventory for Task Stress (CITS; Matthews
and Campbell 1998), which comprises three seven-item scales for task-focus, emotion-
focus and avoidance. These scales represent three fundamental categories of coping
identified in studies of life stressors (Endler and Parker 1990). The CITS is modelled on the
Endler and Parker (1990) scale for coping with life stressors, but using items that are
directed towards the performance context. Instructions require respondents to report
the strategies used ‘as a deliberately chosen way of dealing with problems’ during
performance. Matthews and Campbell (1998) reported that the scales were internally
consistent and appropriately sensitive to task stressors.
2.2.3. Design and procedure
Participants completed a battery of questionnaire measures including the EPQ-R prior to
arrival at the laboratory. Participants were tested individually in artificially lit, sound-
attenuated cubicles. First, they completed the pre-task version of the DSSQ to assess initial
subjective state. Participants were then given task instructions and viewed a sequence of
50 example trials, with the word ‘TARGET’ printed above each of four critical signals
presented. Then, they performed a practice session of 9 min without feedback. During
practice, signal probability was the same as in the main task, but a lower event rate of
50 stimuli per min was used. Participants were randomly assigned to either the high event
or low event condition, with 72 participants assigned to each condition. Participants were
also randomly assigned to either a feedback or a no-feedback condition. In the feedback
condition, the message ‘PERFORMANCE RATED POOR’ was printed on the screen on
12 occasions during task performance. Delivery of the message was unrelated to actual
performance. However, feedback did not affect either performance or subjective state,
perhaps because the high event rates used limited participants’ attention to non-task
stimuli. In the present article, data were pooled across the two feedback conditions.
Next, they performed the main RIP task for 15 min, during which performance
was recorded. Following performance, they completed, in succession, the post-DSSQ
(including embedded workload items), the appraisal scales and the CITS. Debriefing
emphasised that the task was designed to be very difficult to perform.
3. Results
3.1. Effects of task parameters
Initially, analyses were run to determine the effects of event rate (ER) and exposure to the
task on task performance, subjective state, workload and coping, prior to subsequent
424 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

analyses of individual differences. Effects of task parameters on detection rate were
analysed using a 2  5 (event rate  task period) mixed-model ANOVA, with repeated
measures on the task period factor. Box’s correction was employed in significance testing,
because of violations of the sphericity assumption. Uncorrected degrees of freedom (dfs)
are reported here. Main effects of event rate (F(1,142) ¼ 92.73, partial 
2
¼ 0.295, p 50.01)
and task period (F(4,568) ¼ 4.76, partial 
2
¼ 0.032, p50.01) were both significant, but the
interaction between the factors was not. As shown in Figure 1, detection rate was much
lower at the higher event rate and detection rate tended to improve over time in both
conditions (contrary to the typical vigilance decrement). False positive responses were rare
throughout – mean rates were 1.4% (lower event rate) and 2.5% (higher event rate) – and
so subsequent analyses used only detection rate as a performance index.
Task effects on subjective state were analysed using three 2  2 (event rate  pre-post)
mixed-model ANOVA, with repeated measures on the pre-post factor, which represented
the contrast between pre-task and post-task administrations of the DSSQ. The dependent
variables for the ANOVA were task engagement, distress and worry. The main effect
of pre-post was significant in each case (F
engagement
(1,142) ¼ 49.16, partial 
2
¼ 0.257;
F
distress
(1,142) ¼ 484.73, partial 
2
¼ 0.773; F
worry
(1,142) ¼ 0.930, partial 
2
¼ 0.061,
p50.01, for all Fs). The only significant effect of event rate was found in the analysis of
task engagement, in which the event rate  pre-post interaction was significant
(F
engagement
(1,142) ¼ 7.05, partial 
2
¼ 0.047, p50.01), indicating that task-induced state
change varied with event rate. Figure 2 shows state changes (post-task state–pre-task
state). Both task versions induced large magnitude increases in distress and smaller
declines in engagement and worry; engagement decreased more sharply in the high event
rate condition.
Effects of event rate on workload ratings were analysed using a series of t-tests,
employing the Bonferroni correction. Figure 3 shows the mean for each rating in each
condition. All aspects of workload except physical demand were high. Significant effects of
event rate were found for two ratings only: performance impairment (t(142) ¼ 4.83,
p50.01) and effort (t(142) ¼ 2.72, p50.05). The higher event rate induced greater loss of
performance, together with reduced effort. t-tests were also employed to test for effects of
0
10
20
30
40
50
60
70
1234 5
Period
Detections (%)
ER: 150/min
ER: 75/min
Figure 1. Detection rate as a function of event rate (ER) and task period.
Theoretical Issues in Ergonomics Science 425
Downloaded by [University of Central Florida] at 12:41 15 August 2014

event rate on appraisal and coping; means for each condition are shown in Figure 4.
Significant effects were found for task-focus (t(142) ¼ 3.22, p50.05), avoidance
(t(142) ¼ 3.06, p50.05) and perceived controllability (t(142) ¼ 2.84, p50.05). Increasing
the event rate lowered task focus and controllability and increased avoidance.
3.2. Predictors of workload, appraisal and coping
The next set of analyses tested the role of personality as a predictor of these constructs.
In bivariate analyses, neuroticism was significantly correlated with threat appraisal
0
1
2
3
4
5
6
7
8
9
10
Workload rating
Mean rating
ER = 75
ER = 150
MD PD TD PF EF FS
Figure 3. NASA-Task Load Index workload ratings as a function of event rate (ER). MD ¼ mental
demand; PD ¼ physical demand; TD ¼ temporal demand; PF ¼ poor performance; EF ¼ effort;
FS ¼ frustration/stress.
–1
–0.5
0
0.5
1
1.5
2
State factor
State change (z)
ER = 75
ER = 150
Engagement Distress Worry
Figure 2. Task-induced changes in three Dundee Stress State Questionnaire factors as a function of
event rate (ER).
426 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

(r ¼ 0.382, p50.01), low controllability ( r ¼0.186, p50.05) and emotion-focused coping
(r ¼ 0.370, p50.01), whereas extraversion correlated only with threat (r ¼0.185,
p50.05). Correlation magnitudes were similar in high and low event rate conditions.
Neither personality factor correlated with overall workload, although neuroticism was
significantly correlated with the rating of frustration and stress (r ¼ 0.273, p50.01).
Multiple regressions were run to compare personality with cognitive factors in predicting
workload and coping. Table 1 gives summary statistics for the prediction of the three coping
scales. Event rate was entered at the first step, to control for task-related variance, followed
by the two personality variables, followed by the three appraisal variables at the final step.
In each case, the final regression equation was significant (R
2
task-focus
(6,137) ¼ 0.297,
p50.01; R
2
emotion-focus
(6,137) ¼ 0.498, p50.01; R
2
avoidance
(6,137) ¼ 0.349, p50.01). The data
showed that, even as the final block of predictors, appraisal added more to the variance
explained (21–36%) than the other sources of variance. Personality was relevant only to the
prediction of emotion-focus, and neuroticism remained modestly significant as a predictor
in the final equation for this criterion variable. The regressions suggest that multiple aspects
of appraisal may bias choice of coping.
To investigate different predictor sets for workload as a criterion, a similar regression
was run, but coping was also included as a further step in the equation, between personality
and appraisal in the sequence of predictor blocks entered (see Table 2). The final regression
equation was significant (R
2
(8,135) ¼ 0.235, p50.01). Both appraisal and coping
contributed to the prediction of workload, jointly adding 23% to the variance, but event
rate and personality did not. An individual perceiving the task as uncontrollable and using
emotion-focused coping but not avoidance is the person likely to rate workload as highest.
3.3. Predictors of subjective state
Bivariate analyses showed that neuroticism correlated with distress both pre-task and
post-task (r ¼ 0.427, 0.383, respectively, p50.01 for each) and also worry pre-task and
5
10
15
20
25
30
Scale
Mean score
ER = 75
ER = 150
Threat Chall Control Task-F Emotion-F Avoid
Figure 4. Appraisal (Appraisal of Life Events) and coping (Coping Inventory for Task Stress) scores
as a function of event rate (ER). Chall ¼ challenge; Control ¼ controllability; Task-F ¼ task-focus;
Emotion-F ¼ emotion-focus; Avoid ¼ avoidance.
Theoretical Issues in Ergonomics Science 427
Downloaded by [University of Central Florida] at 12:41 15 August 2014

Table 1. Summary statistics for regressions of Coping Inventory for Task Stress scales on personality (Eysenck Personality Questionnaire-Revised) and
appraisal (Appraisal of Life Events) predictor sets.
Step: predictor set added
1: Event rate 2: Personality 3: Appraisal
Criterion R
2
F (1,142) R
2
DR
2
F
D
(2,140) R
2
DR
2
F
D
(3,137) Significant predictors (at step 3)
Task-focus 0.068 10.35** 0.083 0.015 1.11 0.297 0.214 13.96** ER ( ¼0.18*)
challenge ( ¼ 0.38**)
Emotion-focus 0.008 1.17 0.142 0.134 10.91** 0.498 0.356 32.46** ER ( ¼0.21**)
neuroticism ( ¼0.16*)
threat ( ¼ 0.41**)
controllability ( ¼0.36**)
Avoidance 0.062 9.36** 0.087 0.025 1.95 0.349 0.262 18.37** challenge ( ¼0.36**)
controllability ( ¼0.32**)
*p50.05. **p50.01.
428 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

Table 2. Summary statistics for regression of NASA-Task Load Index workload on personality (Eysenck Personality Questionnaire-Revised), coping
(Coping Inventory for Task Stress) and appraisal (Appraisal of Life Events) predictor sets.
Step: predictor set added
1: Event rate 2: Personality 3: Coping 4: Appraisal
Criterion R
2
F(1,142) R
2
DR
2
F
D
(2,140) R
2
DR
2
F
D
(3,137) R
2
DR
2
F
D
(3,134)
Significant predictors
(at step 4)
Workload 0.011 1.54 0.043 0.032 2.38 0.209 0.166 9.55** 0.271 0.062 3.84* E ( ¼ 0.16*)
emotion-focus ( ¼ 0.24*)
avoidance ( ¼0.25**)
controllability ( ¼0.31**)
*p50.05; **p50.01.
Theoretical Issues in Ergonomics Science 429
Downloaded by [University of Central Florida] at 12:41 15 August 2014

post-task (r ¼ 0.266, 0.290, p50.01 for each). Extraversion was more weakly related
to state; only its correlation with pre-task distress attained significance (r ¼0.174,
p50.05). Correlation magnitudes were similar at each event rate. Multiple regressions
were run to compare the contributions of personality and cognitive variables in predicting
task-induced state change. Three regressions were run, using each post-task state
factor (engagement, distress, worry) as the criterion, in turn. Each regression included
the appropriate pre-task state to control for individual differences in baseline state prior
to exposure to the task. Pre-task state was entered at the first step, followed by event rate,
personality, coping and appraisal.
Summary statistics are given in Table 3. Each equation was significant at the final step
(R
2
engagement
(10,133) ¼ 0.675, p50.01; R
2
distress
(6,137) ¼ 0.533, p50.01; R
2
worry
(6,137) ¼
0.647, p50.01). In each case, the pre-task state variable was significant initially and
remained significant in the final equation. That is, some of the variance in post-task state
simply reflects individual differences in state prior to performance. The remaining
predictors added substantially to the variance explained, with coping and appraisal
contributing most of the additional variance. Engagement and distress appeared to relate
to both appraisal and coping. Three of these variables remained significant in the final
equation in both cases. The initial entry of personality into the equation at step 3 was
significant only with distress as the criterion. At this step, both neuroticism ( ¼ 0.261,
p50.01) and extraversion ( ¼ 0.162, p50.05) made significant contributions. In the final
equation, only extraversion was significant with a positive , perhaps surprisingly. In the
case of worry, only coping added to the variance explained initially; high emotion-focus
and high avoidance were associated with greater post-task worry, relative to baseline.
Further regressions were run for each of the criteria listed in Tables 1–3 to check
whether personality effects were moderated by event rate. Perhaps neuroticism might be
expected to relate more strongly to negative appraisals, dysfunctional coping and distress
in the more demanding condition. To do so, interaction terms for extraversion  event rate
and neuroticism  event rate were calculated (using centred variables to reduce collinearity
with linear terms). In no case did inclusion of the interaction terms add to the variance
explained by personality and event rate alone, implying that personality effects were
similar in both task conditions.
3.4. Predictors of performance
The final set of analyses investigated predictors of performance (correct detection rates).
Two performance indices were calculated. Overall detection rate was found as the
arithmetic mean of detection rate in each of the five task periods. Detection change was
found by regressing period 5 detection rate against period 1 detection rate and calculating
the residual. A high score indicates that the person detected more signals in period 5 than
would be predicted on the basis of their period 1 performance. Given the overall trend
towards performance improvement, such an individual is learning or adapting to the task
environment. The post-task DSSQ scores were used as predictors of performance, because
the moderate test–retest correlations for the DSSQ factors indicated that pre-task state
was not a reliable index for the state experienced during performance. Bivariate analyses
showed that several state, appraisal and coping scales related to performance in the whole
sample (correlation magnitudes were similar at each event rate). Overall detection rate was
significantly correlated with task engagement (r ¼ 0.306, p50.01), task-focused coping
(r ¼ 0.314, p50.01), challenge (r ¼ 0.235, p50.01), controllability (r ¼ 0.303, p50.01) and
430 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

Table 3. Summary statistics for regression of Dundee Stress State Questionnaire factors on personality (Eysenck Personality Questionnaire-Revised),
coping (Coping Inventory for Task Stress) and appraisal (Appraisal of Life Events) predictor sets.
Step: predictor set added
1: Pre-task state 2: Event rate 3: Personality 4: Coping 5: Appraisal
Criterion R
2
F (1,142) R
2
DR
2
F
D
(1,141) R
2
DR
2
F
D
(2,139) R
2
DR
2
F
D
(3,136) R
2
DR
2
F
D
(3,133)
Significant predictors
(at step 5)
Engagement 0.288 57.44** 0.324 0.036 7.51** 0.324 0.000 0.03 0.610 0.286 33.17** 0.675 0.065 8.87** pre-task state ( ¼ 0.25**)
task-focus ( ¼ 0.27**)
avoidance ( ¼0.27**)
challenge ( ¼ 0.30**)
Distress 0.201 35.87** 0.202 0.001 0.041 0.259 0.057 5.39** 0.475 0.216 18.68** 0.533 0.058 5.43** pre-task state ( ¼ 0.27**)
E( ¼ 0.19**)
emotion-focus ( ¼ 0.34**)
avoidance ( ¼0.23**)
controllability ( ¼0.23**)
Worry 0.416 101.19** 0.416 0.000 0.05 0.436 0.020 2.47 0.642 0.206 25.96** 0.647 0.005 0.72 pre-task state ( ¼ 0.39**)
emotion-focus ( ¼ 0.42**)
avoidance ( ¼0.15*)
*p50.05; **p50.01.
Theoretical Issues in Ergonomics Science 431
Downloaded by [University of Central Florida] at 12:41 15 August 2014

(low) avoidance (r ¼0.343, p50.01). Likewise, the residualised performance change
score correlated with task engagement (r ¼ 0.328, p50.01), task-focused coping (r ¼ 0.196,
p50.05), challenge (r ¼ 0.285, p50.01), controllability (r ¼ 0.170, p50.05) and avoidance
(r ¼0.186, p50.05). Personality correlates of performance did not attain significance.
Initial multiple regressions explored two issues: (1) relative predictive power of state,
appraisal and coping predictor sets; (2) moderator effects of event rate. The bivariate
data showed that state, appraisal and coping variables were all linked to performance.
The study tested whether, if one variable set (e.g. state) was entered into the equation first,
the remaining variables sets (e.g. appraisal, coping) added to the variance explained.
In general, these regressions failed to establish any precedence between predictor sets; the
first set entered typically added a significant variance increment, but additional sets did
not. Because previous research has focused on states as predictors, and because states are
conceptualised as integrations of appraisal and coping processes (Matthews et al. 2002),
it was decided to focus on the state variables in subsequent regressions.
Further regressions investigated whether centred product terms representing inter-
actions between trait and state predictors and event rate added to variance explained.
These analyses showed that personality  event rate terms added to the variance in overall
detection rate, but state  event rate terms were not predictive. Subsequent regressions
included the personality product terms.
On the basis of the initial regressions, predictors of performance were handled as
follows. Event rate was added at the first step, followed by personality, personality  event
rate and the three state factors. Summary statistics are given in Table 4. Each equation
was significant at the final step (R
2
overall
(8,135) ¼ 0.482, p50.01; R
2
change
(8,135) ¼ 0.177,
p50.01). For overall detections, event rate explained a substantial part of the variance
initially, but the personality  event rate and state factors added significantly to the
variance explained. There were no linear personality effects but both interaction terms
were significant in the final equation. The positive sign of the interaction term indicates
that both high extraversion and high neuroticism persons tended to perform relatively
better in the high ER condition and relatively worse in the low ER condition. Task
engagement also added to the variance explained. In the prediction of performance
change, in addition to an ER effect not previously evident, task engagement emerged as
the only significant predictor.
4. Discussion
This study succeeded in producing a highly demanding, stressful task environment. In both
task conditions, workload levels exceeded those typically seen in other demanding tasks
used in the laboratory, such as vigilance. For example, Matthews et al. (2006a) found
means in the range 5–6 for mental and temporal demands for a 36-min sensory vigilance
task, substantially less than the means of around 8 for these scales in the present data. The
lack of a ‘vigilance decrement’ on the task suggests that it was rather different to typical
high-workload vigilance tasks that do show decrement (See et al. 1995). Possibly, subjects
were still learning the task, given its difficulty, including development of strategies for
managing the exceptional task load. Indeed, the transition from the lower event rate
used in practice to the higher event rates of the main task may have required a strategy
shift. Both task versions elicited large magnitude increases in distress, exceeding 1.5 SD,
and worry declined less than is typical in performance settings (Matthews et al. 2006b).
The event rate parameter primarily affected task engagement, with engagement showing
432 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

Table 4. Summary statistics for regression of two indices of vigilance performance on personality (EPQ-R) and state (DSSQ) predictor sets.
Step: predictor set added
1: Event rate 2: Personality 3: Personality  Event rate 4: States
Criterion R
2
F (1,142) R
2
DR
2
F
D
(2,140) R
2
DR
2
F
D
(3,137) R
2
DR
2
F
D
(3,134)
Significant predictors
(at step 4)
Detections 0.395 92.73** 0.412 0.017 2.02 0.445 0.033 4.10* 0.482 0.037 3.18* ER ( ¼0.60*)
extraversion  ER ( ¼ 0.13*)
neuroticism  ER ( ¼ 0.16*)
engagement ( ¼ 0.18**)
Change
(residual)
0.068 10.34** 0.084 0.016 1.26 0.097 0.013 0.97 0.177 0.080 4.36** ER ( ¼0.21**)
engagement ( ¼ 0.27**)
*p50.05; **p50.01.
Theoretical Issues in Ergonomics Science 433
Downloaded by [University of Central Florida] at 12:41 15 August 2014

a more pronounced decline in the high event rate condition. Similarly, NASA-TLX data
showed a lower level of effort at the higher event rate.
Several features of the data support a transactional understanding of individual
differences in subjective state and performance in this especially taxing environment.
Individual differences in coping were substantially dependent on appraisal. Perceiving the
task as challenging and controllable tended to drive more adaptive coping. Individual
differences in both appraisal and coping also related to workload and to subjective state.
Although workload is typically treated as a property of tasks rather than individuals
(e.g. Hart and Staveland 1988), the data showed that variation in the NASA-TLX measure
is sensitive to controllability appraisals and individual differences in coping. The data also
showed that several dimensions of coping and appraisal related to the two performance
indices. Appraising the task as challenging and controllable, and using task-focused but
not avoidance coping, correlated with better performance. The data support Hancock
and Szalma’s (in press) identification of appraisal processes as key to operator stress.
Here, appraisal was linked to coping, workload, stress state and objective performance.
By contrast, personality factors were rather modestly related to subjective state and
performance, demonstrating that these traditional trait measures make only a limited
contribution to understanding individual differences in the transactional process.
The remainder of this discussion addresses the following issues in more depth. First,
the personality data and their significance are reviewed. Next, the role of subjective states
as indices of the interaction between operator and task and their relationship with
performance are discussed. Finally, the applied utility of the transactional model of stress
as a basis for predicting individual differences in stress response and performance is
reviewed. This section will also consider implications for design of systems.
4.1. Personality and task-induced stress
As discussed in Section 1, there is a wealth of evidence that links personality traits,
especially neuroticism, to general stress vulnerability. Given that the RIP task used here
was highly stressful, it might be expected that neuroticism, and perhaps introversion,
would relate to an elevated stress response. Some evidence to support this prediction was
found. Consistent with previous studies (Matthews et al. 1999, 2006b), neuroticism was
correlated with distress, worry, emotion-focused coping and negative appraisals of the task
as being threatening and uncontrollable. Indeed, neuroticism made a unique contribution
to the prediction of emotion-focused coping in the multiple regression (replicating
Matthews et al. 2006b). Neuroticism also related to task-induced change in distress,
although the multivariate analyses suggested that the cognitive variables were more
strongly predictive. Thus, the data confirm that neuroticism predicts operator stress,
perhaps because high neurotic persons tend to have maladaptive cognitions of high-
pressure task environments. By contrast, extraversion did not predict post-task state in the
bivariate data and, surprisingly, emerged as a predictor of higher distress in the multiple
regression.
As also noted previously, neuroticism tends to be a more robust predictor of stress than
of performance. Consistent with other vigilance research (e.g. Rose et al. 2002), the present
study failed to find any general association between neuroticism and detection rate.
The interactive effect of neuroticism and event rate evident in the regression analysis
suggested that high neuroticism was detrimental only at the lower event rate. Thus, more
neurotic individuals were not especially vulnerable to overload.
434 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

The transactional model may offer an explanation for this seemingly anomalous result.
The multiple regression data (Table 1) showed that higher event rate was actually
associated with lower emotion-focus, as well as with lower task-focused coping. Overload
of information may actually make it more difficult for operators to allocate attention to
their feelings about the task. Thus, a high event rate may be paradoxically beneficial for
high neurotic persons in suppressing their typical but potentially distracting tendencies
to engage in emotion-focused coping during task performance. The inconsistency of
associations between neuroticism and sustained performance may reflect the variable
influence of different forms of task demand in accentuating or discouraging emotion-
focused coping.
Extraversion data were similar, in that only the extraversion  event rate interaction
reached significance. Again, the effect was suggestive of extraversion being relatively more
damaging to performance at the lower event rate. The effect was weak but in line with the
view that under high cognitive demands the typical impairment in vigilance of extraverts
may disappear or even reverse (Matthews et al. 1990a). More generally, the data confirm
that standard personality traits are rather weak predictors of sustained performance that
are dependent on the configuration of the task. Such personality effects may be
insufficiently robust to be useful in predicting operational performance in real-life
contexts.
4.2. Subjective states and cognitive processes
The data confirm the imperfect correspondence between external reality and inner
experience that is typical of stress (Lazarus 1999). Although event rate had a strong effect
on performance, its effects on subjective state were of smaller magnitude and confined
to task engagement. Notably, loss of performance was not accompanied by increased
distress. Conversely, the multiple regressions showed that appraisal and coping together
explained about 20–30% of the change in post-task state relative to pre-task baseline
levels. Event rate added only 3.6% to the variance in task engagement and was non-
significant in the final regression equation. The present research did not set out to test
causal (mediation) models, but it is plausible that the event-rate effect was mediated by
the lower task-focused coping produced by the higher event rate. State change appears
to be linked more directly to cognitions of the task environment than to event rate per se.
Similarly, varying temporal trends in subjective stress response seen during vigilance tasks
may be attributed to the role of coping (Szalma et al. 2004).
Matthews et al. (2002) propose that the broad state factors assessed by the DSSQ
represent integrated descriptors of the transaction between operator and task environ-
ment. Thus, task engagement represents commitment to effort, distress represents damage
control under overload and avoidance signals personal reflection and self-regulation.
The data are consistent with this perspective. There were no one-to-one mappings between
states and cognitive processes. Instead, task-induced state change was sensitive to multiple
processes. For example, task engagement was maintained to the extent that the task was
perceived as challenging and the person employed task-focus rather than avoidance.
Each cognitive factor contributed independently to the regression equation. In relation to
coping, the present study almost exactly replicates the findings of Matthews et al. (2006b),
regarding independent predictors of each state factor. The only difference is that Matthews
et al. (2006b) did not find any association between avoidance and post-task distress in the
regression analysis. Perhaps avoidance is actually adaptive in dealing with task stress when
Theoretical Issues in Ergonomics Science 435
Downloaded by [University of Central Florida] at 12:41 15 August 2014

task demands are so high that effective performance is precluded. Thus, to understand
individual differences in subjective stress response, it is essential to investigate the
individual’s fine-grained cognitions of the task and choice of coping strategies.
Turning to individual differences in performance, the study replicated the positive
association between task engagement and signal detection found in several previous
studies (Matthews et al. 1999, 2001, 2006a, Helton et al. 2009), including studies of
energetic arousal (e.g. Matthews et al. 1990a,b). Correlation magnitudes are typically in
the range 0.2–0.4. Furthermore, task engagement correlates modestly with the increment
in cerebral blood flow induced by performing short, demanding tasks, which may be a
biological marker for resource availability (Matthews et al. 2006a). Typically, formal tests
of resource theory predictions (e.g. Matthews et al. 1990a, Matthews and Margetts 1991)
have supported the hypothesis that engagement/energy is a marker for attentional resource
availability, in line with resource models of vigilance (Parasuraman et al. 1987, Warm et al.
2008). The resource model is also supported by the generalisation of the effect across tasks
of differing characteristics, including simultaneous and successive tasks (Matthews et al.
1999) and sensory and cognitive tasks (Matthews et al. 2006a, 2007).
The present data extend previous findings by showing also that the cognitive processes
supporting engagement themselves correlate with performance indices; specifically,
challenge and controllability appraisals and high task-focus and low avoidance coping.
An alternative hypothesis is that task engagement is associated with greater task-directed
effort and allocation of those resources available. This position is consistent with Hockey’s
(1997) view that fatigue (corresponding to low engagement) results in lower targets for
performance and reduced effort. That is, one may see task engagement as a ‘biocognitive’
factor that relates to individual differences in basic efficiency of attentional processing,
or as a ‘cognitive-adaptive’ factor that relates to the effectiveness of strategic deployment
of the functionality provided by the cognitive architecture (Matthews 2001). The two
hypotheses are not differentiated by the current data. It is possible too that availability of
resources encourages effort allocation, so that the two mechanisms work synergistically
(Helton et al. 2009).
4.3. Applications
The data confirm the limited utility of two standard personality measures for prediction
of sustained performance. Neuroticism and extraversion were only weakly related to
performance and the associations were moderated by event rate. This finding does not
mean that personality assessment is of no value. Assuming that personality–performance
associations are moderated by various task and contextual factors (Matthews et al. 2003),
there may be circumstances under which the moderators are aligned so as to produce
practically useful effect sizes. Furthermore, even small effect sizes may support improve-
ments in real-life personnel selection, especially when the selection ratio (ratio of
applicants:applicants hired) is low (Anastasi and Urbina 1998). The data also confirmed
that neuroticism is a reliable predictor of distress, worry and dysfunctional coping, in
keeping with other stress research (e.g. Suls 2001). Even in circumstances in which distress
and emotion-focused coping have little impact on performance, vulnerability to task-
induced stress may have other undesirable consequences, such as absenteeism, poor
morale and health problems (e.g. Siegrist and Marmot 2004). Other types of task may be
more strongly impacted by distress. Some evidence has been found that multiple-task
performance is vulnerable to distress (Matthews and Campbell 1998), perhaps because
436 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

distress disrupts the executive processes that coordinate multiple processing streams
(cf. Eysenck et al. 2007).
The data also suggest that transactional stress process measures, and their expression
in the state of task engagement, are potentially more predictive of sustained ‘overload’
performance than is personality. Unfortunately, states are by their nature more volatile
than stable personality traits, so it is not feasible simply to use state measures in place
of states for personnel selection and assessment. There are two possible solutions to the
problem. The first is to investigate dispositional measures that may predict the task-
induced engagement response. Similar to previous studies (e.g. Matthews et al. 1999,
2006b), the data confirm that extraversion, despite its link to positive affect, is not a good
predictor of engagement. Other studies (e.g. Matthews et al. 1999) have shown that the
additional traits of the Five Factor Model are also only modestly predictive of task
engagement, at best. Conscientiousness and agreeableness show correlations with post-
task engagement up to 0.3 or so (Matthews et al. 1999). The association between
conscientiousness (C) and engagement may be mediated by greater use of task-focused
coping by high C individuals (Matthews et al. 2006b). Recent research (Finomore et al.
2009) suggests that more narrowly defined traits related to fatigue-proneness, abnormal
personality and cognitive disorganisation may also be predictive of task engagement.
Correlations in the 0.2–0.3 range were found between post-task engagement and scales
including those for adult attention deficit hyperactivity disorder traits, schizotypy, chronic
daytime sleepiness, mind-wandering and cognitive failures. However, as in the present
study, these scales failed to predict the extent of task disengagement directly elicited by the
task – in this case, a version of Temple et al.’s (2000) brief vigilance task. Furthermore,
both pre- and post-task engagement correlated significantly with performance on the
vigilance task, but these traits failed to do so. The search for reliable trait predictors of
vigilance remains frustrating.
A second solution to the selection problem is to investigate the stability of the task-
induced response in context. Perhaps there are some individuals who consistently react to
a particular task setting with engagement or with fatigue. Matthews and Falconer (2000)
investigated state responses to a simulation of customer service work. They showed some
stability of the state response over a 7-month interval; the test–retest correlation for
engagement was 0.45. If there is some consistency in response, a work sample task might
be used to predict characteristic task engagement in the operational setting. An assessment
centre setting, affording multiple assessments of engagement to improve reliability through
aggregation, might be even more successful.
Thus far, this paper has discussed the relevance of the results to the traditional problem
of personnel selection. Findings may also be relevant to design and training issues. The
challenge in this instance is to configure systems to maintain task engagement, even when
there is a potential for a level of overload that may drive avoidance coping and withdrawal
of effort. In recent work using a driving simulator, it was found that enhancing
background scenery to reduce monotony serves to maintain task engagement during
a potentially fatiguing drive (Saxby et al. 2007), in line with other findings on monotony
(Thiffault and Bergeron 2003). Given that real-life overload situations, e.g. in military and
industrial settings, are liable to be emergency situations that are highly motivating, it may
seem odd to focus on monotony as a problem. However, as classic vigilance research
showed (Mackworth 1948), high motivation is not necessarily a panacea for the difficulties
intrinsic to sustained performance. Thus, in line with Hancock’s (1997) recommendation
for increasing the enjoyment of work activities, it is important to design systems to
maintain task interest and challenge, even in emergencies. Furthermore, given that Richter
Theoretical Issues in Ergonomics Science 437
Downloaded by [University of Central Florida] at 12:41 15 August 2014

and Salvendy (1995) demonstrated that interfaces may need to be differently configured
for different personalities, it may be important to investigate what design features serve
to maintain engagement in different individuals.
In the specific context of vigilance and sustained attention, designers may also need to
enhance the operator’s sense of control of the task environment. Allowing some scope for
personal configuration of the system may be valuable. Parsons et al. (2007) compared two
versions of a vigilance task, presented in a military context. Participants were required
either to simply register detection of symbols representing enemy missile launchers or to
additionally acquire and destroy the target using a mouse response, following detection.
The opportunity to follow detection with this action elevated task engagement and
performance. Szalma and Hancock (2006) demonstrated that task engagement was higher
in observers who believed (falsely) that they were able to choose their level of task demand,
relative to a no-choice condition. The rather passive quality that attaches to the detection
response in vigilance may encourage dysfunctional cognitions, such as low challenge
appraisal; designers may usefully attend to the response requirements of the task.
A final application is in the realm of training. Advances in simulation make it possible
to train operators to handle low-probability but critical overload situations. The danger
suggested by studies of task-induced stress is that the overload factor is likely to induce
task disengagement, especially if the contingency is perceived as unlikely, which may
adversely impact training. Again, designing to increase perceived challenge and affor-
dances for task-focused coping is likely to maintain engagement with the training exercise.
By contrast with the typical vigilance task, with less time pressure, the present data showed
an overall increment in detection rate; an overall trend that masks a variety of different
temporal changes in individuals. Task engagement and cognitive stress processes predicted
performance improvement, implying that these may be adaptive qualities during training.
Research is needed to test how engagement during training relates to subsequent
operational stress and performance.
5. Conclusions
The present study demonstrates the importance of a transactional understanding of
individual differences in the stress and fatigue that may be induced by sustained
performance. Multivariate assessment of individual differences in stress is essential for
understanding the interaction between operator and task environment. Task engagement
(energy, motivation and concentration) supports effective performance. Individual
differences in engagement reflect multiple aspects of the operator’s cognitions of the
task, including challenge appraisal and choosing to employ task-focused coping rather
than avoidance. Personality factors play a role in biasing the transactional process. In the
present data, neuroticism related to appraisals and cognitions linked to distress, although
the performance of the RIP task was insensitive to distress. The challenge for application
of these findings is to find reliable predictors of individual differences in task engagement;
refinement of personality models and/or use of work sample tests provide possible
strategies. The data also reinforce the importance of designing tasks to be challenging and
supportive of task-focused coping.
Acknowledgements
The first author gratefully acknowledges support from the US Army Medical Research and Materiel
Command under Contract No. DAMD17–04-C-0002, and from the Army Research Institute under
438 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

SBIR Contract No. W74V8H-06-C-0049 to JXT Applications, and Subcontract No. JXT-06-S-1003
to the University of Cincinnati. The views, opinions and/or findings contained in this report are
those of the author(s) and should not be construed as an official Department of the Army position,
policy or decision unless so designated by other documentation. In the conduct of research where
humans are the subjects, the investigator(s) adhered to the policies regarding the protection of
human subjects as prescribed by 45 CFR 46 and 32 CFR 219 (Protection of Human Subjects).
References
Anastasi, A. and Urbina, S., 1998. Psychological testing. 7th ed. Englewood Cliffs, NJ: Prentice-Hall.
Barrick, M.R., Mount, M.K., and Judge, T.A., 2001. Personality and performance at the beginning
of the new millennium: what do we know and where do we go next? International Journal of
Selection and Assessment, 9, 9–30.
Beauducel, A., Brocke, B., and Leue, A., 2006. Energetical bases of extraversion: effort, arousal,
EEG, and performance. International Journal of Psychophysiology, 62, 212–223.
Cox-Fuenzalida, L., Swickert, R., and Hittner, B., 2004. Effects of neuroticism and workload history
on performance. Personality and Individual Differences, 36, 447–456.
Davies, D.R. and Parasuraman, R., 1982. The psychology of vigilance. London: Academic Press.
Endler, N. and Parker, J., 1990. Multidimensional assessment of coping: a critical review. Journal
of Personality and Social Psychology, 58, 844–854.
Eysenck, H.J. and Eysenck, M.W., 1985. Personality and individual differences: a natural science
approach. New York: Plenum.
Eysenck, M.W., et al. , 2007. Anxiety and cognitive performance: attentional control theory.
Emotion, 7, 336–353.
Eysenck, S.B.G., Eysenck, H.J., and Barrett, P.T., 1985. A revised version of the Psychoticism scale.
Personality and Individual Differences, 6, 21–29.
Falconer, S., 2002. Predictors of occupational stress. Unpublished doctoral dissertation. University of
Dundee, UK.
Ferguson, E., Matthews, G., and Cox, T., 1999. The Appraisal of Life Events (ALE) scale: reliability
and validity. British Journal of Health Psychology, 4, 97–116.
Finomore, V., et al., 2009. Predicting vigilance: a fresh look at an old problem . Ergonomics, 52,
791–808.
Fisher, S., 1984. Stress and the perception of control. Hillsdale, NJ: Erlbaum.
Frewer, L.J. and Lader, M., 1991. The effects of caffeine on two computerized tests of attention
and vigilance. Human Psychopharmacology: Clinical and Experimental, 6, 119–128.
Funke, G.J., et al., 2007. Vehicle automation: a remedy for driver stress? Ergonomics, 50, 1302–1323.
Hancock, P.A., 1997. Essays on the future of human-machine systems. Minneapolis, MN: Peter
A. Hancock.
Hancock, P.A. and Szalma, J.L., in press. Stress and performance. In: P.A. Hancock and
J.L. Szalma, eds. Performance under stress. Aldershot, UK: Ashgate Publishing.
Hancock, P.A. and Warm, J.S., 1989. A dynamic model of stress and sustained attention. Human
Factors, 31, 519–537.
Hart, S.G. and Staveland, L.E., 1988. Development of NASA-TLX (task load index): results of
empirical, & theoretical research. In: P.A. Hancock and N. Meshkati, eds. Human mental
workload. Oxford, UK: North-Holland, 139–183.
Helton, W.S., Matthews, G., and Warm, J.S., 2009. Stress state mediation between environmental
variables and performance: the case of noise and vigilance. Acta Psychologica, 130, 204–213.
Helton, W.S., et al., 1999. Optimism-pessimism and false failure feedback: effects on vigilance
performance. Current Psychology: Research and Review, 18, 311–325.
Hockey, G.R.J., 1997. Compensatory control in the regulation of human performance under stress
and high workload: a cognitive-energetic framework. Biological Psychology, 45, 73–93.
Theoretical Issues in Ergonomics Science 439
Downloaded by [University of Central Florida] at 12:41 15 August 2014

Humphreys, M.S. and Revelle, W., 1984. Personality, motivation and performance: a theory of the
relationship between individual differences and information processing. Psychological Review,
91, 153–184.
Judge, T.A., Heller, D., and Mount, M.K., 2002. Five-factor model of personality and job
satisfaction: a meta-analysis. Journal of Applied Psychology, 87 (3), 530–541.
Koelega, H.S., 1992. Extraversion and vigilance performance: 30 years of inconsistencies.
Psychological Bulletin, 112, 239–258.
Lazarus, R.S., 1999. Stress and emotions: a new synthesis . New York: Springer.
Lazarus, R.S. and Folkman, S., 1984. Stress, appraisal, and coping. New York: Springer.
Mackworth, N.H., 1948. The breakdown of vigilance during prolonged visual search. Quarterly
Journal of Experimental Psychology, 1, 6–21.
Matthews, G., 2001. Levels of transaction: a cognitive sciences framework for operator stress.
In: P.A. Hancock and P.A. Desmond, eds. Stress, workload, and fatigue. Mahwah, NJ:
Erlbaum, 5–33.
Matthews, G., 2008. Personality and information processing: a cognitive-adaptive theory.
In: G.J. Boyle, G. Matthews and D.H. Saklofske, eds. Handbook of personality theory and
assessment. Volume 1: Personality theories and models. Thousand Oaks, CA: Sage.
Matthews, G. and Campbell, S.E., 1998. Task-induced stress and individual differences in coping.
Proceedings of the Human Factors and Ergonomics Society, 42, 821–825.
Matthews, G. and Davies, D.R., 1998. Arousal and vigilance: the role of task demands.
In: R.R. Hoffman, M.F. Sherrick and J.S. Warm, eds. Viewing psychology as a whole: the
integrative science of William N. Dember. Washington, DC: APA, 113–144.
Matthews, G., Davies, D.R., and Holley, P.J., 1990a. Extraversion, arousal, and visual sustained
attention: the role of resource availability. Personality and Individual Differences, 11,
1159–1173.
Matthews, G., Davies, D.R., and Lees, J.L., 1990b. Arousal, extraversion, and individual differences
in resource availability. Journal of Personality and Social Psychology, 59, 150–168.
Matthews, G., Deary, I.J., and Whiteman, M.C., 2003. Personality traits. 2nd ed. Cambridge:
Cambridge University Press.
Matthews, G. and Desmond, P.A., 2002. Task-induced fatigue states and simulated driving
performance. Quarterly Journal of Experimental Psychology, 55A, 659–686.
Matthews, G. and Falconer, S., 2000. Individual differences in task-induced stress in customer
service personnel. Proceedings of the Human Factors and Ergonomics Society, 44, 145–148.
Matthews, G. and Gilliland, K., 1999. The personality theories of H.J. Eysenck and J.A. Gray:
a comparative review. Personality and Individual Differences, 26, 583–626.
Matthews, G. and Margetts, I., 1991. Self-report arousal and divided attention: a study of
performance operating characteristics. Human Performance, 4, 107–125.
Matthews, G., Warm, J.S. and Washburn, D., 2006a. Diagnostic methods for predicting performance
impairment associated with combat stress. Report for US Army Medical Research and Materiel
Command, contract # W23RYX-3106-N605.
Matthews, G., Warm, J.S. and Washburn, D., 2007. Diagnostic methods for predicting performance
impairment associated with combat stress. Report for U.S. Army Medical Research and
Materiel Command, contract # W23RYX-3106-N605.
Matthews, G., et al., 1999. Validation of a comprehensive stress state questionnaire: towards a state
‘Big Three’. In: I. Mervielde, et al., eds. Personality psychology in Europe. Vol. 7, Tilburg:
Netherlands Tilburg University Press, 335–350.
Matthews, G., et al., 2001a. The common cold impairs visual attention, psychomotor performance,
and task engagement. Proceedings of the Human Factors and Ergonomics Society, 45,
1377–1381.
Matthews, G., et al., 2002. Fundamental dimensions of subjective state in performance settings:
task engagement, distress and worry. Emotion, 2, 315–340.
Matthews, G., et al., 2006b. Emotional intelligence, personality, and task-induced stress. Journal
of Experimental Psychology: Applied, 12, 96–107.
440 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014

McCrae, R.R. and Costa, P.T., 2003. Personality in adulthood: a five-factor theory
perspective. 2nd ed. New York, NY: Guilford Press.
Parasuraman, R., Warm, J.S., and Dember, W.N., 1987. Vigilance: taxonomy and utility.
In: L.S. Mark, J.S. Warm and R.L. Huston, eds. Ergonomics and human factors: recent
research. New York: Springer-Verlag, 11–32.
Parsons, K.S., et al., 2007. Detection-action linkage in vigilance: effects on workload and stress.
Proceedings of the Human Factors and Ergonomics Society 51st annual meeting . Santa Monica,
CA: Human Factors and Ergonomics Society.
Richter, L.A. and Salvendy, G., 1995. Effects of personality and task strength on performance in
computerized tasks. Ergonomics, 38, 281–291.
Rose, C.L., et al. , 2002. The role of the Big Five personality factors in vigilance performance and
workload. European Journal of Personality, 16, 185–200.
Saxby, D.J., Matthews, G., and Hitchcock, T., 2007. Fatigue states are multidimensional: evidence
from studies of simulated driving. Proceedings of the Driving Simulation Conference – North
America 2007. Iowa City, IA: University of Iowa.
Schabracq, M.J. and Cooper, L., 2000. The changing nature of work and stress. Journal of
Managerial Psychology, 15, 227–241.
See, J.E., et al., 1995. Meta-analysis of the sensitivity decrement in vigilance. Psychological Bulletin,
117, 230–249.
Selye, H., 1976. Stress in health and disease . Reading, MA: Butterworth.
Siegrist, J. and Marmot, M., 2004. Health inequalities and the psychosocial environment – two
scientific challenges. Social Science and Medicine, 58, 1463–1473.
Singh, I.L., Molloy, R., and Parasuraman, R., 1993. Individual differences in monitoring failures of
automation. Journal of General Psychology, 120, 357–373.
Suls, J., 2001. Affect, stress, and personality. In: J.P. Forgas, ed. Handbook of affect and social
cognition. Mahwah, NJ: Lawrence Erlbaum, 392–409.
Szalma, J.L. and Hancock, P.A., 2006. Performance, workload, and stress in vigilance:
the power of choice. Proceedings of the Human Factors and Ergonomics Society, 50, 1609–1613.
Szalma, J.L., et al., 2004. Effects of sensory modality and task duration on performance, workload,
and stress in sustained attention. Human Factors, 45, 349–359.
Szymura, B. and Wodniecka, Z., 2003. What really bothers neurotics? In search for factors impairing
attentional performance. Personality and Individual Differences, 34, 109–126.
Szymura, B., in press. Individual differences in resource allocation policy. In: A. Gruszka,
G. Matthews and B. Szymura, eds. Handbook of individual differences in cognition: Attention,
memory and executive control. New York: Springer.
Temple, J.G., et al., 2000. The effects of signal salience and caffeine on performance, workload and
stress in an abbreviated vigilance task. Human Factors, 42, 183–194.
Tett, R.P. and Burnett, D., 2003. A personality trait-based interactionist model of job performance .
Journal of Applied Psychology, 88, 500–517.
Thiffault, P. and Bergeron, J., 2003. Monotony of road environment and driver fatigue: a simulator
study. Accident Analysis and Prevention, 35, 381–391.
Wacker, J., Chavanon, M., and Stemmler, G., 2006. Investigating the dopaminergic basis of
extraversion in humans: a multilevel approach. Journal of Personality and Social Psychology,
91, 171–187.
Warm, J.S. and Dember, W.N., 1998. Tests of a vigilance taxonomy. In: R.R. Hoffman,
M.F. Sherrick and J.S. Warm, eds. Viewing psychology as a whole: the integrative science of
William N. Dember. Washington, DC: American Psychological Association, 87–112.
Warm, J.S., Matthews, G., and Finomore, V.S., 2008. Workload and stress in sustained attention.
In: P.A. Hancock and J.L. Szalma, eds. Performance under stress. Aldershot, UK: Ashgate
Publishing, 115–141.
Wesnes, K. and Warburton, M., 1983. Effects of smoking on rapid information processing
performance. Neuropsychobiology, 9, 223–229.
Wells, A. and Matthews, G., 1994. Attention and emotion: a clinical perspective. Hove: Erlbaum.
Theoretical Issues in Ergonomics Science 441
Downloaded by [University of Central Florida] at 12:41 15 August 2014

About the authors
Gerald Matthews is Professor of Psychology at the University of Cincinnati. He received a PhD in
Experimental Psychology in 1984 from the University of Cambridge. His research interests include
cognitive science models of personality and emotion, cognitive fatigue, emotional intelligence and
traffic psychology. He is currently Secretary-Treasurer of the International Society for the Study of
Individual Differences and President-Elect of Division 13 (Traffic Psychology) of the International
Association for Applied Psychology.
Sian Campbell is Improvement and Development Consultant with South East Employers in
Winchester, UK. She holds a Masters degree in IT from Aston University, UK. She was formerly
a graduate research assistant in psychology at the University of Dundee. Sian is currently carrying
out work in areas of improvement and development for employees, managers, elected members and
organisations in the South East of England, using psychometric tests to underpin her work in
providing mentoring support and guidance.
442 G. Matthews and S.E. Campbell
Downloaded by [University of Central Florida] at 12:41 15 August 2014