Full text of DOI:10.1080/02678370120688

This article was downloaded by: [Universidad Autonoma de Barcelona]
On: 16 October 2014, At: 03:37
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,
UK
Work & Stress: An International
Journal of Work, Health &
Organisations
Publication details, including instructions for authors
and subscription information:
http://www.tandfonline.com/loi/twst20
Nurses' representations of
the perceived causes of work-
related stress: A network
drawing approach
Steven Muncer , Shirley Taylor , D. W. Green & I. C.
Mcmanus
Published online: 26 Nov 2010.
To cite this article: Steven Muncer , Shirley Taylor , D. W. Green & I. C. Mcmanus
(2001) Nurses' representations of the perceived causes of work-related stress: A
network drawing approach, Work & Stress: An International Journal of Work, Health &
Organisations, 15:1, 40-52
To link to this article: http://dx.doi.org/10.1080/02678370120688
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the
information (the “Content”) contained in the publications on our platform.
However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or
suitability for any purpose of the Content. Any opinions and views expressed
in this publication are the opinions and views of the authors, and are not the
views of or endorsed by Taylor & Francis. The accuracy of the Content should
not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions,
claims, proceedings, demands, costs, expenses, damages, and other liabilities
whatsoever or howsoever caused arising directly or indirectly in connection
with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan, sub-
licensing, systematic supply, or distribution in any form to anyone is expressly
forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

&
, 2001,
15,
1, 40±52
no.
work stress
vol.
Nurses’ representations of the perceived causes of
work-related stress: a network drawing approach
STEVEN MUNCER*
Department of Applied Psychology, University of Durham, University Boulevard,
Thornaby, Stockton-on-Tees TS17 6BH, UK
SHIRLEY TAYLOR
School of Health, University of Teesside, Middlesbrough TS1 3BA, UK
D. W. GREEN and I. C. McMANUS
Psychology Department, University College, Gower Street, London WC1E 6BT, UK
Keywords: Stress; Nursing; Network drawing; Causal paths.
This paper reports a study, involving a network drawing approach, that examined how nurses
perceive the interrelationship between causes of workplace stress. Network analysis originated in
sociology as a method of examining the relationship between people, objects or events. It has recently
been adapted to examine participants’ perceptions of the relationships between causes of a phenom-
enon, either by asking participants to complete a grid rating the strength of all the possible links
between causes or by getting them to draw a diagram of the links that they think are important.
The network drawing technique, in which participants are asked to draw a diagram indicating
perceived causal links between nominated causes of stress and also to indicate the strength of these
links, was employed in this study. The causes of stress were taken from a previous study in which
nurses kept a diary for one week detailing stressful events and their causes. There were 48 participants
in the present study and the main results con®rmed the importance of staYng levels and inadequate
support as perceived direct causes of stress. The study also revealed the importance of indirect links
between staYng levels and other causes of stress. The networks illustrate how direct and mediating
causes of stress are connected and lie largely outside nurses’ control. The results are discussed in
relation to other recent work on the causes and experience of stress by nursing staV.
1. Introduction
In the past few years there has been an increasing recognition of the importance and impact
of job-related stress (Cox, and GriYths, 1995). It has been recognized that work-related
stress and depression are among the most important factors aVecting the health of UK
employees (Hodgson, Jones, Elliott, and Osman, 1993). There is also evidence that those
employees who work in the health-related professions are suVering more than other
*Author for correspondence. e-mail: SJMuncer@durham.ac.uk
Work & Stress ISSN 0267-8373 print/ISSN 1464-5335 online 2001 Taylor & Francis Ltd
http://www.tandf.co.uk/journals/
DOI: 10.1080/02678370110063402

A network drawing approach to work-related stress
41
workers (Charlton, Kelly, Dunnell, Evans, and Jenkins, 1993). A report commissioned by
the NuYeld Trust has suggested that National Health Service (NHS) staV suVer more
stress-related psychological morbidity than other professional sectors (Williams, 1989). A
study of the mental health of National Health Service staV from 19 hospital trusts (Institute
of Work Psychology, 1996) showed greater levels of stress than amongst workers generally.
While it is true that doctors suVer higher levels of stress, there can be no doubt that nurses
are also suVering (Wall, Bolden, Borrill, Carter, Golya, Hardy, Haynes, Rick, Shapiro, and
West, 1997). Data on early retirements from the NHS (Moore, 1996) and the increase in
nursing staV turnover (Snell, 1998) support this view. Although there have been few
attempts to quantify the levels of stress, a clear picture of widespread negative experiences
in nursing emerges (Kennedy, and Grey, 1997; Potts, Halliday, Plimley, Wright, and
Cuthbertson, 1995; Power, and Sharp, 1988). It is also clear that stress has a strong negative
relationship with nurses’ job satisfaction (Blegen, 1993), and that this pattern of results is
not limited to Britain and the USA (Piko, 1999; Trucco, Valenzuela, and Trucco, 1999).
The eVects of stress on those in the health professions are important for three main
reasons. First, the NHS is one of the larger employers in Britain. Second, there is evidence
that stress can eVect the eYcacy of treatment and that medical personnel are less likely to
acknowledge the eVects of stress than other professionals (Sexton, Thomas, and Helmreich,
2000). Third, the nature of their jobs make it likely that nurses will experience factors that
have been implicated in causing stress (role con¯ ict, role ambiguity and work demands)
more than other occupations, which makes them theoretically interesting (Haynes, Wall,
Bolden, Stride, and Rick, 1999).
Recent studies have identi®ed the following causes of nursing stress: inappropriate
advice from junior and inexperienced staV; con¯ icts within the multi-disciplinary team;
bureaucracy; inadequacies of nursing care by others; verbal abuse from patients and relatives;
physical abuse from patients; dealing with death and dying; shift work; lack of emotional
support; insuYcient resources; responsibility without power; con¯ ict with doctors and
uncertainty due to political issues (Callaghan, Tak-Ting, and Wyatt, 2000; Farrington,
1997; Prosser, Johnson, Kuipers, Szmulker, Bebbington, and Thornicroft, 1997; Tyler, and
Cushway, 1992, 1995; Tyler, and Ellison, 1994). Changes in the management of the health
service have also been implicated. Some nurses, for example, have expanded their roles via
The Scope of Professional Practice (United Kingdom Central Council for nursing, midwifery
and health visiting, 1992) and the outcomes of this on patient care and nurses’ stress
experiences are unknown. Initiatives such as this, intended to reduce the stress on junior
doctors, can thus be perceived as merely redistributing the pressure from one group of
workers to another. Such added responsibilities may serve to exacerbate the multiple roles
that are evident in contemporary nursing (Adomat, and Killingworth, 1994; Taylor, White,
and Muncer, 1999). The eVects of health policy on the working patterns of nurses,
changing roles associated with reduction in junior doctors’ working hours (Fish, 1995),
the scope of professional practice and continuous changes in health policy are all potential
causes of stress. However, it is worth remembering that nurses in systems other than the
National Health Service also experience stress.
While many studies have established various factors as causally implicated in stress, few
have looked at how these causes interact. Wheeler (1997) has argued that a model of stress
in nursing that accounts for its complex nature has not so far been developed, and criticizes
the lack of attention in research to the nurses’ own constructions of stress (Wheeler, 1998a,
1998b). Others have argued more generally for the importance of understanding lay health
beliefs (Furnham, 1988; Furnham, and Henley, 1988; Furnham, Wardley, and Lillie, 1992).
More speci®cally Furnham (1997) has examined lay people’s beliefs about the causes,

42
S. Muncer et al.
manifestations, consequences and alleviation of occupational stress using factor analysis. In
that study, participants were presented with 27 items that dealt with proposed causes of
stress and asked to rate each item for agreement on a scale of 1±7. For example, two of
the items were; `People who aren’t very busy or challenged by their work cannot really
experience stress’ and `Stress only aVects people who aren’t their own boss, i.e. have to
take orders from others’. Factor analysis with varimax rotation identi®ed ®ve factors that
were perceived to be causing stress: con¯ ict and satisfaction; career development; demo-
graphic subgroups; danger and intimidation; and authority. The two items given as examples
both loaded on the factor identi®ed as career development.
We believe that the approach of Furnham (1997) to establishing lay beliefs is a useful
and important contribution to the literature, but would argue that there may be other
more appropriate methods for examining the perceived interrelationships between causes.
Factor analysis can reveal consistent patterns of response or rating but it does not tell us
about the importance of a given cause. In a recent study examining the perceived causes
of health, for example, participants tended to rate the importance of the impact of
supernatural powers consistently, so it came out as a factor; however the majority of the
participants rated it as unimportant (Taylor, and Muncer, 1999). In Furnham’s (1997) own
study the majority of the items with the highest levels of agreement or disagreement, that
is those with scores above 5 or below 2, had low loadings on the ®ve factors that were
identi®ed. For example, the following items `The risk of redundancy is a very stressful
factor’ and `Lack of consultation creates an uncommunicative climate and this may lead to
stress’ were the most strongly endorsed items but both had loadings of less than 0.3 (which
is the conventional cut-oV point for the importance of factor loading (Kline, 1999)) on
any of the identi®ed factors. In other words, the factors that were identi®ed do not include
the most important perceived causes of stress.
In the past few years network analysisÐa technique that involves the analysis of either
a network grid or drawingÐhas been modi®ed from its more frequent use as a tool for
studying social relationships in sociology and anthropology, to investigating the causal
models of various social phenomena that lay people have in their minds. It has been
adapted to try to explain the process by which causal attributions are made, a process that
has held a central place in social psychology (Kelley, 1967). In the early studies using this
method, participants were presented with a grid with selected causes of the phenomena in
question down the side and along the top. They were then asked to indicate whether there
was a causal link between each of the causes, by putting a `1’ in the appropriate cell if
there was a link and a `0’ if there was no causal link. This binary grid data collection
technique has been used to investigate the perceived causal structure of examination failure
(Lunt, 1988), personal debt (Lunt, and Livingstone, 1991; Muncer, and Gillen, 1992),
crime (Campbell, and Muncer, 1990; Muncer, Gillen, and Campbell, 1996), drug use
(Muncer, Sidorowicz, Epro, and Campbell, 1992), date rape (Gillen, and Muncer, 1995)
and poverty (Heaven, 1994; Muncer, 1995).
These early network studies have been criticized for ignoring the importance of direct
links between a nominated cause and the target social phenomena, and consequently over-
estimating the importance of indirect links (Green, and McManus, 1995; Green, McManus,
and Derrick, 1998). In the previously mentioned studies only indirect links between causes
have been examined, as the target (for example, loneliness) does not appear on the grid.
These studies have also been criticized for using a binary `yes’ or `no’ scoring system rather
than a more sensitive system in which participants can say how strongly they perceive the
link to be between two causes (Lunt, 1991; Muncer, and Gillen, 1997). There is now
clear evidence that if participants are given the opportunity to include direct links, these

A network drawing approach to work-related stress
43
play an important part in the network (HeVernan, Green, McManus, and Muncer, 1998;
Taylor, White, and Muncer, 1999).
Taylor et al. (1999) provided participants with the opportunity to include direct links
in their network grid study of nurses’ perceived causes of stress and also asked participants
to rate the strength of perceived connections on a scale of 1 to 5 (1=impossible to 5=
highly likely). In that study, nurses were given grids with the following nominated causes
down the side and along the top: inadequate support; multiple roles; patients and relatives’
behaviour; patient suVering; powerlessness; interruptions; attitude and ability of staV;
behaviour of managers; behaviour of doctors; and shift patterns. These causes of stress had
been the most frequently nominated causes by the participants (Taylor et al., 1999). They
were also provided with a brief sentence explaining each cause. For example, multiple
roles were described as `con¯ icting demands of diVerent roles’ and powerlessness as `feeling
that one has no in¯ uence on decisions’. Stress also appeared on the grid as a possible eVect
of all these causes so that perceived direct links and also a possible reciprocal cause could
be investigated. For example, it is possible that stress could cause the behaviour of doctors
or any of the other nominated causes.
The network that was produced showed three relatively clear systems in which there
were perceived to be direct links to stress from patient suVering, manager’s behaviour and
staYng levels. StaYng levels were the distal cause in a system of causal links including
multiple roles and inadequate support, which in turn caused stress and were also perceived
as causing the attitude and ability of staV and feelings of powerlessness, which also led to
stress. (These seven most important causes were those used in the current study.)
Green, and McManus (1995) used both a diVerent method of data collection, in which
they asked participants to directly draw the network of causes in a diagram, and use
diVerent kinds of analysis when appropriate (e.g. Item Response Theory). They have
argued that one of the problems with the grid method is that there are theoretical problems
about the threshold for deciding whether or not to include a causal link (which they call
a path). They have also argued that the method confounds the issue of presence and
strength of a path, as it is the mean strength of a path that determines whether it should
be included, and this depends on both the number of participants including it and how
strongly they rate it. Clearly it is possible to have a path which most people think exists
but at low strength, or to have a path that few people think exists but that is believed to
be very important by those who do include it. However, the grid method does not allow
disentanglement of the presence versus strength of a path (Green, and McManus, 1995;
Green et al., 1998). There is also evidence that the direct drawing method is less in¯ uenced
by response factors than the grid method; this is because in the grid method causes appear
in ®xed positions and participants are asked to put a number in each cell (Wall, Gillen,
Muncer, and Holmes, 1998; HeVernan et al., 1998), whereas in the drawing method they
are asked to draw a network diagram. The drawing and understanding of such diagrams
appears to be relatively easy for both academic (Hoc, 1989) and lay people (Green, and
McManus, 1995). Indeed, Argyle, Furnham, and Graham (1981) have used similar diagrams
to represent the goal con¯ icts experienced by nurses in diVerent situations.
The present study thus ensured that participants were able both to nominate direct
links to stress and to use the direct diagram drawing method advocated by Green and
colleagues. This approach was employed as a method to corroborate previous studies and,
in particular, to see if similar networks would be produced by using a diVerent method of
data collection. The causes that participants could put in the diagram were the seven most
important of those that appeared on the Taylor et al., (1999) network. Participants were
also given the same short de®nitions of each cause.

44
S. Muncer et al.
2. Method
2.1. Participants
There were 48 participants in this study. They were state registered nurses who were
enrolled at an English university on a part-time post-registration degree programme. They
participated voluntarily in the study during a research methods class. A total of 48 out of
a possible 56 nurses completed the networks.
2.2. Procedures
Participating nurses ®rst took part in an introductory session, which included a discussion
of stress among nurses. They were then given a sheet of paper with the following
instructions printed on it.
The following seven factors have been identi®ed as being the most important in causing stress:
patient suVering; inadequate support; attitude and ability of staV; feelings of powerlessness; staYng levels;
multiple roles; and behaviour of managers. We would like you to draw a diagram (on the sheet below)
of how you think these causes are linked to stress and each other, using arrows to indicate the
direction of the eVect. We would also like you to rate the strength of the link on a scale of 1 to
100 and write the appropriate number onto the diagram. If you do not think a cause has an
important eVect leave it oV your diagram. Only include those causes that you think are important.
In the diagram method the strength of a link has always been rated on a 100-point
scale rather than a 5-point scale; this is to increase the sensitivity of the measure. If there
are a large number of possible paths it enables individuals to express their discriminations
between causal links more easily.
The word `Stress’ was printed in the middle of the page so that participants were able
to draw direct links to it if they wished. Participants could, therefore, include in their
drawing both direct links such as that between Inadequate support and Stress, and also
indirect links such as that between Behaviour of managers and Inadequate support.
Participants were provided with the same de®nitions of the causes as were used in the
study by Taylor et al. (1999). Participants completed their diagrams and handed them in
anonymously.
3. Results
A composite diagram was prepared by examining the separate paths between any two
factors for each of the 48 participants. Most participants represented both direct paths from
a possible causal factor to the target factor (stress) and indirect paths that interconnected
these possible causal factors. Four participants included only direct paths from causes to
stress, with no indirect paths. Figure 1 shows the composite diagram for all of the paths
that were drawn by at least 5 (10%) of the participants. The various proportions of
individuals (10±39%, 40±69%, 70±100%) who included a particular path are shown by
lines of diVerent types.
We also calculated for each path its mean strength, using the scores given by those
who included it. The mean strength of each path is represented by diVerent types of
arrowhead. Thus we are able to separate path frequency, which is the proportion of
participants who included a particular path, and path strength. For simplicity’s sake (see
also below) we excluded seven paths that were drawn backward from the target factor (i.e.
stress) to the possible causal factors. It is, however, worth noting that participants who
included backward paths from stress usually had a higher number of paths on their diagrams;
this was signi®cantly the case for 6 out of 7 paths on two-tailed t tests.

A network drawing approach to work-related stress
45
Figure 1. Composite network diagram of the perceived causes of stress in nurses, derived from
causal paths drawn between up to seven possible causes and directly from those causes to
stress. The diagram shows directions of each path, the mean strength of each path (as scored
by the nurses) and frequency of inclusion of that path. Only paths drawn by more than 5
%
(10 ) of nurses are included.
Path inclusion frequency. Proportion of nurses who included a particular path in their drawing
(out of 48).
%
%
%
High (70±100
)
Medium (30±69
)
Low (10±39
)
Mean path strength. Scored by nurses out of a possible 100:
%
%
%
Low (0±39 )
High (70±100
)
Medium (40±69
)
Arrow type:
High frequency+High path strength
High frequency+Medium path strength
Medium frequency+Medium path strength
Low frequency+High path strength
The proportion of participants who drew a particular line is indicated by line type, as above. The
mean path strengths are depicted by type of arrowhead. For the sake of clarity, paths included by a
low proportion of participants were only included if they had a high path strength.

46
S. Muncer et al.
The composite diagram indicates that the majority of nurses considered staYng levels
and lack of support to be important direct contributors to stress. Both of these paths were
present on over 70% of participants’ diagrams and had an average path strength of over
70. Five other direct paths were also included by over 70% of participants, but with a
somewhat lower (i.e. medium) mean strength (behaviour of management, multiple roles,
attitude and ability of staV, feelings of powerlessness and patient suVering). Two indirect
paths were included by a medium proportion of nurses, between staYng levels and multiple
roles, and behaviour of management and inadequate support; these paths had a medium
mean strength. It is noteworthy too that there were a number of indirect paths that were
included by relatively few individuals but that were considered to have a strong impact.
For example, the paths from powerlessness to management, from attitude to management,
and from patient suVering to powerlessness were included by a low percentage of participants
but had a high path strength. These are the links that are likely to be missed by the grid
method using inductive eliminative analysis (Taylor et al., 1999).
Previous work has already gone some way towards establishing both the reliability and
validity of models derived from both types of network analysis (Muncer, 1995; Muncer,
and Gillen, 1997; Green, and McManus, 1995; Green et al., 1998; HeVernan et al., 1998).
We can provide a weak check of the reliability of the ®ndings by randomly splitting the
sample into two and comparing the scores of the two groups on all of the possible causal
paths. If the sample is performing consistently we should not expect signi®cant diVerences
between the two groups. In this case two randomly selected groups were not signi®cantly
diVerent ( p<.05) for scores on any of the paths. Furthermore, for those paths that appear
on the network they showed no signi®cant diVerences at p<.2.
In this study it is also possible to calculate an agreement index (Muncer, 1995) between
the present results and those produced by a diVerent sample using a diVerent method. The
agreement index is calculated by dividing the number of common paths between networks
by the total number of paths. In this case the directional agreement index, which takes the
direction of the causal path into account, is 0.78 (a score of 1 would indicate total agreement
and 0 total disagreement). The agreement index for this present study is considerably better
than that found in some previous studies where similar groups were compared using the
same grid data collection method for both groups. For example, the agreement index for
right-wing voters on networks of the causes of poverty in Heaven’s (1994) research was
only 0.43 and for left-wing voters only 0.38 (Muncer, 1995). The non-directional agree-
ment index in the present study was extremely high at 0.87, which is important, as previous
work has suggested that the direction of the path may be less signi®cant than the fact that
a path exists (Green et al., 1998; Wall et al., 1998).
4. Discussion
How should this composite diagram be interpreted? Our view is that inclusion of any path
in the diagram is a probabilistic process. The likelihood of an individual including a path
that nurses endorsed with relatively low frequency will tend to re¯ ect the tendency of that
individual to include paths in his or her diagram. We have shown elsewhere by using item
response theory (Green et al., 1998) that people tend to have a single representation of the
causes of social phenomena and sample probabilistically from it. That is, there is a shared
representation of paths that will unfold in sequence, and the more paths an individual
includes in his or her diagram then the more of the shared representation will be revealed.
This view is also supported in this case, by the ®nding that participants that included the
backwards paths from stress included more paths in their diagrams overall.

A network drawing approach to work-related stress
47
The agreement between the present drawing study and the network grid study (Muncer,
1995) suggests that a reliable and transferable explanation of the perceived causes and
relations between the causes of workplace stress in nursing, may be emerging. This present
study allows a more holistic explanation to emerge by including paths that may not have
been revealed by the grid method of data collection and analysis. This explanation is one
that is more able to account for the complex relationships between the potential causal
factors of stress. This methodological triangulation between studies has enhanced the degree
to which we feel that the causes and the relationships between the causes of stress have
been validly represented. This is manifested via the participants’ nominated causes of stress;
their network ratings of causes of stress and the personally drawn diagrams of the causes
of stress, many of which probably show the consequences of imposed working practices
and organizational constraints.
The ®ndings of the present study con®rms most of the perceived causes of stress as
being related to issues that impede nurses in the performance of their role; these are
primarily inadequate staYng levels and lack of support. StaYng levels were perceived as a
direct cause of stress, with high frequency of inclusion and also high path strength. They
were also perceived of as leading to the adoption of multiple roles, which was viewed as
causing stress, with a high inclusion frequency and medium path strength. This situation
had also been seen in the study using the grid network method of data collection (Taylor
et al., 1999). It seems that nurses perceive increased job demands to be an important stressor
(Karasek, and Theorell, 1990).
It was hoped that this study, by focusing on the nurses’ personal drawings of the causal
networks of stress would clarify and con®rm the links previously found between managers’
behaviour and stress. Taylor et al. (1999) found that the behaviour of managers was initially
identi®ed as one of the three distal causes of stress. In the present study the behaviour of
managers is viewed as both a direct and an indirect stressor. The indirect path is to
inadequate support. The impact of inadequate support on stress is however placed as one
of two key causes of stress, the other being staYng levels, both with high frequency and
strength levels. The signi®cant issue here is that the behaviour of managers is the only
nominated cause of inadequate support. We consider this to be a crucial antecedent in the
causal process.
The qualitative data from which were derived the list of possible causes of stress used
in the present study (Taylor et al., 1999) presented a strong image of nurses working with
a lack of support. In that study, the more comprehensive diary entries highlighted how
the interplay of excessive demands on nurses, to perform their roles without adequate staV
and support from management and other staV, created feelings of isolation as they had to
cope alone (Taylor et al., 1999). The present study shifts the nominated causes of stress to
causes that are overtly present in nurses’ day-to-day work: staYng and inadequate support.
While behaviour of managers has a substantial degree of endorsement from nurses as a
cause of stress, it remains on a day-to-day basis distant yet integral to their working
practices. When asked to represent what they considered to be the causes of stress, these
nurses focused on visible and direct causes rather than their antecedents.
It is likely that the way in which the changes to nurses’ working lives have been
pushed through, taking only a management perspective into account, have exacerbated
stress levels (Obholzer, and Roberts, 1994). It has been suggested elsewhere that diVerent
perspectives on the management of waiting lists, staV and skill mix and the management
of patient throughput should have been sought (Firth-Cozens, and Moss, 1998). It has also
been shown that taking into account the wishes and beliefs of staV members can signi®cantly
reduce stress levels and increase performance (Sonnetag, 1996).

48
S. Muncer et al.
The other direct paths to stress not thus far discussed included patient suVering, attitude
and ability of staV, and powerlessness, all of which had high frequency of inclusion and
medium path strength. These ®ndings ®t with other work performed in this area that
suggests that it is not the nature of caring work that is stressful but the barriers that impede
nurses in this role (Adomat, and Killingworth, 1994). The dangers of `controlled care’
have been described by Benner, and Wrubel (1989). Such care is delivered from a distance,
in a mechanistic mode rather than in a manner that focuses on what is important to the
person receiving care. The barriers to such care have been outlined in the literature on
stress and burnout in nursing. Adomat, and Killingworth (1994) suggest that it is not
`caring’ that is stressful but the organizational pressures of the caring environment. The
need to be able to oVer safe care at a high standard may explain the links between
powerlessness and stress, patient suVering and stress, and multiple roles and stress. One
cause of concern emerging from this study is that external pressures and organizational
issues were key causes of stress.
In this study, the paths with low frequency of inclusion need particular consideration
as some of these paths were not on the network produced in the earlier study using the
grid method (Taylor et al., 1999). These paths indicate that the nurses in the present study
were perhaps somewhat more understanding or insightful of a manager’s predicament.
Powerlessness was considered to impact on a manager’s behaviour, as was the attitude and
ability of staV, staYng levels, and multiple roles. These links, although not included with
high frequency, were rated strongly. This positions managers in the middle of a complex
network of causes of stress; they can be considered either as causing stress or as having to
respond to situations that cause stress. Thus, while simplistically managers can be viewed
as causing stress, this is perhaps somewhat naõve. The direct antecedents of stress indicated
È
by the respondents were also considered to impact on the behaviour of managers. Those
respondents who drew complex paths were expressing, via their networks, an awareness
of a complex work situation. They acknowledged the external issues that impact on their
stress experience (including the behaviour of managers) yet they were simultaneously aware
that the behaviour of managers shares some of the same antecedents. Thus a main antecedent
in the experience of work-based stress is also an eVect of the working practices.
Some support for our model (as shown in ®gure 1) can be found in a recent study that
attempted to evaluate and quantify work characteristics as causal factors of stress within the
National Health Service. Haynes et al. (1999) surveyed 9000 staV from the seven major
occupational groups within the National Health Service. The sample included 3441 nurses
who were asked to ®ll out a questionnaire designed to measure eight constructs that were
deemed from prior research and theory to be implicated in stress (Warr, 1987; Karasek,
and Theorell, 1990; Arnold et al., 1991). These were autonomy/control, feedback on work
performance, in¯ uence over decisions, leader support, role clarity, role con¯ ict, peer
support and work demands. Each of these constructs was measured by a minimum of four
or a maximum of six items. For example, in the work demands section participants were
asked `How often do you ®nd yourself meeting the following problems in carrying out
your job?’ and then given six items to respond to such as `I am asked to do work without
adequate resources to complete it’. The importance of support, which is clearly highlighted
by our research, is acknowledged in the questionnaire of Haynes et al. (1999), which
includes items dealing with both peer support and managerial support. Interestingly, as our
model would predict, in that study nurses believed that they received signi®cantly more
peer support (M=3.77 and SD=0.87) than managerial support (M=3.32 and SD=1.02;
t=19.69, p<.001).
Although, Haynes et al. (1999) argue that the nine dimensions they are trying to

A network drawing approach to work-related stress
49
measure are separate, they acknowledge that there will be some overlap. Indeed, some of
the items appear to ®t into more than one category. For example, `Lack of clarity/
agreement about the diVerent responsibilities of doctors and nurses’ is in the professional
compromise category but could clearly be in the role clarity or role con¯ ict sections, which
contain items such as `I know what my responsibilities are’ and `Professionals make
con¯ icting demands on me’. However, the importance of multiple roles suggested by our
network is supported by this research. It is particularly noteworthy that individuals with
higher role clarity show greater psychological well-being in the study by Haynes et al.
(1999) and also in other studies (Fang, and Baba, 1993; Revicki et al., 1993).
As one may expect, in the study by Haynes et al. (1999) there were signi®cant
correlations between many of the nine domains. From our model of nursing stress we
would expect there to be links between staYng levels and multiple roles, which are most
nearly represented as work demands involving role clarity, role con¯ ict and professional
compromise. Work demands in the study by Haynes et al. (1999) were indeed signi®cantly
positively associated with role con¯ ict (r=.48) and professional compromise (r=.48) and
signi®cantly negatively associated with role clarity (r=Õ.24). We would also expect, as
the study by Haynes et al. (1999) found, multiple roles as measured by professional
compromise (r=.29), role clarity (r=Õ.3) and role con¯ ict (r=.39) to be signi®cantly
related to stress as measured by job-related anxiety. Job anxiety had its highest correlation
with work demands (r=.44). While the pattern of correlations show some similarities with
our model, it should be recognized that their correlations come from the whole sample
and not speci®cally from the nursing group.
It is also noteworthy that while our participants believed that powerlessness is a cause
of stress, in the study by Haynes et al. (1999) autonomy and control, which are most like
powerlessness, were not signi®cantly associated with job-related anxiety (r<.005). This is
surprising, given the continued evidence in support of the job-demand-control model of
stress (Van der Doef, and Maes, 1999). This could be because autonomy as measured by
Haynes et al. (1999) was concerned with the amount of choice a person has on the job,
rather than their ability to control events. Given that some of our participants drew a path
from patient suVering to powerlessness it seems likely that they were using it in a wider
sense. Having said that, we believe that it is important to acknowledge the eVect that
patient suVering may have on nurses, and would argue that it is a weakness of the approach
by Haynes et al. (1999) that it was not included in the questionnaire.
It is worth noting in this respect that other scales designed to measure the stress
experienced by nurses have included items on dealing with patients (Gray-Toft, and
Anderson, 1981; Harris, 1989). The Nurse Stress Index, for example, includes items on
dealing with patients and relatives, managing workload, organizational support, home and
work con¯ icts and competence in the role (Harris, 1989).
Green, and McManus (1995) have argued before that their diagram method reveals
cognitive structural models with `formal similarity to covariance structural models’ but that
`their origins are in individual cognitions rather than in data per se’. Clearly, it would be
possible to use structural equation modelling (Bentler, 1988) on the sort of data provided
by Haynes et al. (1999) to compare it to the individual cognitions of our participants. We
would argue that the examination of correlations mentioned suggests that such a procedure
may well produce a similar picture to the network produced by our participants. We
would also argue, however, that network analysis as described in the present paper can
provide important explanations of perceived causes of stress and relationships between
potential causes as identi®ed from individual cognitions.

50
S. Muncer et al.
References
Killingworth
, A. (1994). Care of the critically ill patient: The impact of stress on
Adomat
, R., &
touch in intensive therapy unit. Journal of Advanced Nursing, 19, 912±922.
Graham
A., &
Argyle Furnham,
, M.,
University Press.
, G. A. (1981). Social Situations. Cambridge: Cambridge
Arnold Robertsion, I. T.,
Cooper
&
, J.,
Behaviour in the Workplace. London: Pitman.
Benner Wrubel
, C. L. (1991). Work Psychology, Understanding Human
, P., & , J. (1989). Caring is the candle that lights the dark, that permits us to ®nd
answers where others see none. American Journal of Nursing, 88, 1073±1074.
Bentler
, P. M. (1988). Causal modelling via structural equation systems. In J. R. Nesselroade, &
R. B. Cattell (Eds.), Handbook of Multivariate Experimental Psychology, 2nd edn. (pp. 317±335).
New York: Plenum Press.
Blegen,
M. (1993). Nurses’ job satisfaction: A meta-analysis of related variables. Nursing Research,
42, 36±41.
Callaghan, P., Tak-Ting, S
Wyatt,
P. A. (2000). Factors related to stress and coping among
., &
Chinese nurses in Hong Kong. Journal of Advanced Nursing, 31, 1518±1527.
Campbell, Muncer
A., &
, S. (1990). Causes of crime: Uncovering a lay model. Criminal Justice and
Behavior, 17, 410±420.
Charlton Kelly, Dunnell
Evans
, B., &
Jenkins
, R. (1993). Suicide deaths in England
, J., , K.,
S.,
and Wales: Trends in factors associated with suicide deaths. Population Trends, 69, 34±42.
Cox
, T., &
Griffiths
, A. (1995). Guidance for UK employers on managing work related stress.
Work and Stress, 9, 1±3.
Fang, Y. Q Baba
., &
, V. (1993). Stress and turnover intention: A comparative study among nurses.
International Journal of Comparative Sociology, 34, 24±38.
Farrington
, A. (1997). Strategies for reducing stress and burnout in nursing. British Journal of Nursing,
6(1), 44±50.
Firth-Cozens
Moss
, J., &
1335±1337.
, F. (1998). Hours, sleep, teamwork and stress. British Medical Journal, 317,
Fish,
J. (1995). The impact of reducing junior doctors’ hours on nursing. British Journal of Nursing,
4, 306±307.
Furnham
, A. (1988). Lay Theories. Oxford: Pergamon.
Furnham,
A. (1997). Lay theories of work stress. Work and Stress, 11, 68±78.
Furnham
Henley
, A., &
, S. (1988). Lay beliefs about overcoming psychological problems. Journal
of Social and Clinical Psychology, 6, 423±438.
Furnham Wardley Lillie
, A., , Z., & , F. (1992). Lay theories of psychotherapy. III. Comparing
the ratings of lay persons and clinical psychologists. Human Relations, 45, 839±858.
Gillen, Muncer
K., &
, S. (1995). Sex diVerences in the perceived causal structure of date rape.
Aggressive Behavior, 21, 101±112.
Gray-Toft Anderson
, P., & , J. (1981). The nursing stress scale: Development of an instrument.
Journal of Behavioral Assessment, 3, 11±23.
Green
Green
McManus,
, D., &
prevention in coronary heart disease. British Journal of Psychology, 86, 321±336.
McManus, Derrick,
B. J. (1998). Cognitive structural models of unemployment
I. (1995). Cognitive structural models: The perception of risk and
, D.,
and employment. British Journal of Social Psychology, 37, 415±438.
Harris
I., &
, P. E. (1989). The nurse stress index. Work and Stress, 3, 335±346.
Haynes, Wall Bolden, Stride Rick,
J. (1999). Measures of perceived work
C.,
, T., , C., &
R.,
characteristics for health services research: Test of a measurement model and normative data.
British Journal of Health Psychology, 4, 257±275.
Heaven,
P. (1994). The perceived causal structure of poverty: A network analysis approach. British
Journal of Social Psychology, 33, 259±271.
Heffernan,
Green
, D.,
McManus
, I., &
Muncer
, S. (1998). Comments on network analysis
and loneliness: Some issues of consensus and reliability by Muncer and Gillen. British Journal
T.,
of Social Psychology, 37, 253±254.
, J. (1989). Cognitive approaches to control. In G. Tiberghein (Ed.), Advances in Cognitive Science,
vol. 2, Theory and Application (pp. 178±202). Place: Publisher.
Hoc
Hodgson
, J. T.,
Jones
, J. R.,
Elliott Osman,
, R. C., & J. (1993). Self-Reported Work Related Illness.
Sudbury: HSE.
Institute of Work Psychology (
1996). Mental health of the work force in NHS trusts. Phase 1
Report (Summary), SheYeld University, SheYeld.

A network drawing approach to work-related stress
51
Karasek
, R., &
Theorell,
T. (1990). Healthy Work: Stress, Productivity and the Reconstruction of
Working Life. New York: Basic Books.
Kelley
, H. H. (1967). Attribution theory in social psychology. In D. Levine (Ed.), Nebraska
Symposium on Motivation, vol. 15, pp. 343±369. Lincoln, NB: University of Nebraska Press.
Kennedy,
Grey
, N. (1997). High pressure areas. Nursing Times, 93 (29), 16 July, pp. 26±30.
P., &
Kline,
Lunt,
P. (1999). The Handbook of Psychological Testing. London: Routledge.
P. (1991). The perceived causal structure of loneliness. Journal of Personality and Social Psychology,
30, 309±323.
Lunt
Lunt
, P. (1988). The perceived causal structure of examination failure. British Journal of Social
Psychology, 27, 171±179.
Livingstone
, P., &
British Journal of Social Psychology, 30, 309±323.
Moore
, S. (1991). Everyday explanations of personal debt: A network approach.
, W. (1996). All stressed up and nowhere to go. Health Service Journal, 5 (Sept.), 23±25.
Muncer
, S. (1995). A disagreement index for network analytic studies of perceived causes. Psychological
Reports, 77, 650.
Muncer, Gillen
S., &
, K. (1992). Network analysis and analytic induction: A research note. British
Psychological Society, London, December.
Muncer Gillen
, S., & , K. (1997). Network analysis and lay interpretation: Some issues of consensus
and representation. British Journal of Social Psychology, 36, 537±551.
Muncer, Gillen, Campbell,
A. (1996). Social representations and comparative network
S. J.,
analysis: A preliminary report. Textes sur les Representation Sociales, 5, 1±11.
Muncer Sidorowicz Epro Campbell,
K., &
, S., , L., , R., &
A. (1992). Causes of drug use: What do
college students think? Criminal Justice and Behavior, 19, 363±371.
Obholzer, Roberts
A., &
in the Human Services. New York: Routledge.
, V. (1994). The Unconscious at Work: Individual and Organizational Stress
Piko,
B. (1999). Work-related stress among nurses: A challenge for health care institutions. Journal
of the Royal Society for the Promotion of Health, 119, 156±162.
Potts,
Halliday
, S.,
satisfaction in a small staVed house in the community. British Journal of Nursing, 4, 452±458.
Plimley
, C.,
Wright Cuthbertson
, E., & , A. (1995). StaV stress and
M.,
Power, Sharp
K., &
, G. (1988). A comparison of sources of nursing stress and job satisfaction among
mental handicap and hospice nursing staV. Journal of Advanced Nursing, 13, 726±732.
Prosser Johnson, Kuipers Szmulker Bebbington, Thornicroft
, G. (1997).
, D., , E., , G.,
S.,
P., &
Perceived sources of work stress and satisfaction among hospital and community mental health
staV, and their relation to mental health, burnout and job satisfaction. Journal of Psychosomatic
Research, 43 (1), 51±59.
Revicki,
Whitley,
stress and depression in emergency-medicing residents. Behavioral Medicine, 19(2), 74±81.
Gallery
T., & , M. (1993). Organizational characteristics, perceived work
D.,
Sexton Thomas, Helmreich
, J. B., J., & , R. L. (2000). Error, stress, and teamwork in medicine
and aviation; cross-sectional surveys. British Medical Journal, 320, 745±749.
, J. (1998). When the going gets tough. Health Services Journal, 108 (26 February), 28±30.
Snell
Sonnetag,
S. (1996). Work group factors and individual well-being. In M. West (Ed.), Handbook of
Work Group Psychology. Chichester: Wiley.
Taylor Muncer
, S., & , S. J. (1999). Social representations of health from health professionals. 6th
European Congress of Psychology Abstracts, 1, 310±311.
Taylor, White, Muncer
S.,
R., &
, S. (1999). Nurses’ cognitive structural models of stress. Journal
of Advanced Nursing, 29, 974±983.
Trucco Valenzuela Trucco,
, M., , P., &
D. (1999). Occupational stress in health care personnel.
Revista Medica de Chile, 127, 1453±1461.
Tyler
Tyler
Cushway,
, P., &
Medicine, 8, 91±98.
Cushway
D. (1992). Stress coping and mental well being in hospital nurses. Stress
, P., &
Medicine, 11, 243±251.
Tyler, P., & Ellison,
, D. (1995). Stress in nursesÐThe eVects of coping and social support. Stress
R. (1994). Sources of stress and psychological well-being in high dependency
nursing. Journal of Advanced Nursing, 19, 469±476.
United Kingdom Central Council for nursing, midwifery and health visiting
Scope of Professional Practice. London: UKCC.
(1992). The
Van Der Doef
, M., &
Maes
, S. (1999). The job demand-control (-support) model and psychological
well-being. Work and Stress, 13, 87±114.

52
A network drawing approach to work-related stress
Wall, Gillen
I., , K.,
Muncer Holmes,
, S., & M. (1998). Taking health risks: Adolescent perceptions
about pregnancy and smoking. Proceedings of the APA Annual Congress, San Francisco, CA,
August.
Wall
Bolden
, R. I.,
Borrill
, C. S.,
Carter
, A. J.,
West
Golya
, D. A.,
Hardy Haynes
, G. E., ,
, M. A. (1997). Minor psychiatric disorder in
, T. D.,
C. E.,
Rick
Shapiro
, J. E.,
, D. A., &
NHS trust staV: Occupational and gender diVerences. British Journal of Psychology,, 519±523.
P. B. (1987). Work,Unemployment and Mental Health. Oxford: Oxford University Press.
Warr,
Wheeler
, H. (1997). Nurse occupational stress research. 2: De®nition and conceptualization. British
Journal of Nursing, 6, 1256±1260.
Wheeler
, H. (1998a). Nurse occupational stress research. 5: Sources and determinants of stress.
British Journal of Nursing, 7, 40±43.
Wheeler
, H. (1998b). Nurse occupational stress research. 6: Methodological approaches. British
Journal of Nursing, 7, 226±229.
Williams
, A. C. (1989). Empathy and burnout in male and female helping professionals. Research in
Nursing and Health, 12, 169±178.