Organizational Culture and Strategic Performance

work & stress, 2002, vol. 16, no. 1, 18± 36

An analysis of safety culture attitudes in a highly regulated environment

JOAN HARVEY*, GEORGE ERDOS, HELEN BOLAM,

MICHAEL A. A. COX, JOHN N. P. KENNEDY and

DAVID T. GREGORY

School of Management, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 7RU, UK

Keywords: Safety; Safety culture; Safety attitudes; Nuclear industry.

Culture is a complex construct in organizations, consisting of attitudes, perceptions, values and beliefs, which must necessarily be set in context. Many authors imply that culture is organization-wide and common to all employees. In terms of safety culture, the organizational context may determine its salience and likelihood of aVecting behaviour, especially in a highly regulated environment such as the nuclear industry. This study investigates the components of safety culture and how it varies in a highly-regulated nuclear power plant. A 60-item questionnaire measuring safety attitudes and values was administered anonymously to 1550 employees at two plants in the UK nuclear industry, with a 64.7% mean response rate. Principal components analyses revealed six factors conceptually common to shop ¯ oor and management (supervisor/manager/professional ) groups for both plants, relating to: management style and communication; responsibility and commitment; risk-taking; job satisfaction; complacency; and risk awareness. A canonical discriminant analysis of the items suggested that the data could distinguish three employee groups. The implications of these ® ndings and the proposal that there are two or more safety cultures in the organization is discussed.

1. Introduction

Organizational culture is a complex construct, de® ned in many ways during the last four

decades. Perhaps the characteristics that have endured through most or all of these de® nitions

are that culture is a learned set of values that may take the form of practices interpreted

through rules and norms of behaviour. Hofstede (1994, p. 5) suggests that culture `distin-

guishes one category of peoples from another, and is speci® c to each group’ ; this de® nition

might be contrasted with that of Handy (1985) who implies that there are four culture

types (based on power, role, task or person), each one of which might describe a whole

organization. Harrison (1972) also implies an organization-wide culture when he suggests

that culture provides the prescriptions for the ways in which people should work in those

organizations.

Several writers refer to how critical it is for corporate success of getting the culture

r̀ight’ for the organization; this approach is inherent, for example, in the `excellence’

* Author for correspondence. e-mail: [email protected]

Work & Stress ISSN 0267-8373 print/ISSN 1464-5335 online 2002 Taylor & Francis Ltd http://www.tandf.co.uk/journals/

DOI: 10.1080/02678370110113226

Analysis of safety culture attitudes 19

literature (Peters, and Waterman, 1982), which suggests that, in the pursuit of excellence,

desired behaviours need to be rewarded. Thus, the pursuit of excellence in terms of safety

suggests that safe behaviours be rewarded. Unfortunately, this notion of rewarding appro-

priate behaviours has been lost as organizations have d̀ownsized’ or put employees on

short contracts, changing the whole nature of the psychological contract, with extreme

lowering of morale and commitment in many organizations (Cooper, 1995).

So far, culture has been considered as one set of values associated with practices, rules

and norms of behaviour. The development of separate literatures relevant to diVerent organizational behaviours, such as absence culture, safety culture and the q̀uality’ literature

implies that there are probably several sets of values and that the term culture is an

`umbrella’ one. Some evidence does suggest that it is better to talk about group cultures

rather than organizational cultures (Beck, and Woolfson, 1999; Chute, and Weiner, 1995;

Clarke, 1999; Hofstede, 1991). If we take this notion of many value sets plus the suggestion

that culture is better applied at group than at organizational level, then the culture of an

organization might be viewed as the summation of a (Group×Behaviour type) matrix of values, practices and norms. It is then quite possible for a group of workers to hold many

diVerent cultural values compared, for example, to a group of managers or to those in a diVerent department or plant. This diVers from most of the safety culture literature that talks about t̀he safety culture’ of an organization, which assumes that everybody conceptual-

izes safety in the same manner, but we suggest that this is untrue and that culture diVers for diVerent groups within an organization.

1.1. What is safety culture?

Safety culture is believed to be a key predictor of safety performance (Advisory Committee

for Safety in Nuclear Installations [ACSNI], 1993), but is a concept with no clear de® nition

or measurement (Cox, and Flin, 1998). It has often been used interchangeably with safety

climate, although the two may be distinct, climate re¯ ecting attitudes, perceptions and

beliefs while culture is more complex, re¯ ecting values and norms and being evident in

safety management practices (Mearns, and Flin, 1999). Lee (1993) de® nes safety culture as

all forms of learned behaviour which àdd up to a shared commitment to think safely, to

behave safely and to believe and trust in the safety measures put in place by the organisation’.

However, in a later paper, Booth, and Lee (1995) narrow the learned behaviours to

organizational attitudes, risk perceptions and beliefs in the necessity, practicality and eVect- iveness of controls. Another view of safety culture in a highly-regulated context is t̀hat

assembly of characteristics and attitudes in organizations and individuals which establishes

that, as an overriding priority, nuclear plant safety issues receive the attention warranted

by their signi® cance’ (International Atomic Energy Authority [IAEA], 1991). Thus, in

general, safety culture is viewed as involving perceptions and attitudes as well as the

behaviour of individuals within an organization. Interestingly, these de® nitions all appear

to ignore the transfer of safety culture from one context to another, such as bringing values

and practices from the driving or domestic environments to the work environment.

In terms of absence culture, Nicholson, and Johns (1985) suggest that the main variables

are the level of trust in an organization and the salience of absence behaviour. Taking

these concepts into safety culture, it is suggested that attempting to model safety behaviour

must take the salience of safety into account and that trust in safety policy and managerial

action is critical. Thus, IAEA (1991) describe two major components of safety culture:

(a) the framework determined by organizational policy and by managerial action; and

(b) the response of individuals in working within and bene® ting by the framework, and

J. Harvey et al.20

suggest that s̀uccess depends on commitment and competence, provided in the policy and

managerial context and by individuals themselves’ . IAEA integrate several models from

both nuclear and non-nuclear industries and propose that attention to safety involves:

individual awareness of the importance of safety; knowledge and competence through

training, instruction and self-education; commitment; motivation; supervision; and

responsibility.

The attempts to de® ne the factors that constitute a good safety culture (ACSNI, 1993;

Cox, and Flin, 1998; IAEA, 1991; Institution of Occupational Safety and Health [IOSH],

1994; Mearns, Flin, Gordon, and Fleming, 1998; Ryan, 1991), all imply that eVective provision for health and safety depends as much upon organizational culture generally as

it does upon speci® c attention to health and safety matters. However, the number of groups

of attitudes relevant to safety vary from 2 to 9 and vary in nature although most include

job satisfaction, individual responsibility, management responsibility, leadership style and

communication, commitment, risk awareness and risk taking (Cheyne, Cox, Oliver, and

Tomas, 1998; Cooper, and Phillips, 1994; Cox, and Cox, 1991; Dedobbeleer, and Beland,

1991; Diaz, and Cabrera, 1997). The fact that there is no general consensus about what

groups of attitudes constitute safety culture may be because the area is very complex and

hard to de® ne. This does not mean that because of the complexity it should not be

measured, as the norms that make up the culture can be examined (Ostrom, Wilhelmsen,

and Kaplan, 1993). However, the notion of whether safety culture consists of the same

concepts for everybody within an organization or whether there are diVerent subcultures within an organization still needs to be clari® ed.

When modelling safety behaviour the salience of safety must be taken into account

since this determines the importance placed upon some behaviours, and managers’ roles

are very important in establishing a meaningful work and safety culture (Harshbarger, and

Rose, 1991; Zohar, 1980). Indeed, job satisfaction is deemed to be an important construct

related to safety culture and this is aVected by management style and leadership (Cheyne et al. 1998; Clarke, 1998; Hofmann, and Morgeson, 1999; Lee, 1994; Ryan, 1991;

Williamson, Feyer, Cairns, and Biancotti, 1997).

1.2. Attitudes to safety and behaviour

There are several theories linking attitudes and behaviour (Ajzen, 1991; Fishbein, and

Ajzen, 1975; Hanisch, Hulin, and Roznowski, 1998); attitudes and safety behaviour are

not likely to be exceptional to these theories. Cox, and Cox (1991) suggest that c̀onstruct-

ive’ attitudes are probably the most important single index of the eVectiveness of a safety culture as they result from all other contributory features. This view is reinforced by Lee

(1994) and by Williamson et al. (1997), who argue that attitudes, de® ned as stable predisposi-

tions, are the most comprehensive and useful indicators of a safety culture. Lee (1994)

identi® es groups of attitudes that are relevant to safety, including job satisfaction, stress,

leadership style and communication, respect for safety, risk-taking and sensation-seeking,

locus of control and fatalism. However, some of these can be construed as personality

rather than attitudinal variables, which may have serious implications for safety management

insofar as attitudes may be changed whereas personality factors may not be. For example,

in terms of locus of control, individuals whose locus is external may take less preventive

behaviour, as Lee suggests, and we would argue that this characteristic is relatively unchange-

able and thus training to increase preventive behaviours would be ineVective. However, it is implicit in attitude theory that attitudes can be changed, and thus respect for safety

and communications might be amenable to training interventions.

Analysis of safety culture attitudes 21

Using a safety survey in the nuclear industry, Lee, MacDonald, and Coote (1993)

investigated nine general attitudinal areas derived from focus groups. They found that the

nine areas further divided into 19 factors, 17 of which discriminated between an àccident

group’ (resulting in time oV work for >3 days) and a `non-accident’ group. Interestingly, the attitudinal area which most diVerentiated in this way involved satisfaction and `discon- tentment’ . However, a safety culture measure developed by Cooper, and Phillips (1994)

identi® ed seven factors aVecting employee perceptions of safety climate but did not include satisfaction directly, instead focusing largely on the importance and status of safety.

There is evidence that diVerent groups of employees may hold diVerent attitudes, values and perceptions in relation to safety culture and risk (Chute, and Weiner, 1995;

Marek, Iverson, and Hellesoy, 1985; Mearns et al., 1998; Pidgeon, 1998), although whether

these constitute diVerent conceptions within safety culture or diVerent levels of the same variables is hard to say.

1.3. Accident causation and risk in relation to safety culture

When disasters occur, there are often attempts to ensure that accountability and blame can

be laid at someone’s doorÐ a so-called b̀lame culture’ . Lee (1994) argues that many

disasters result from malpractices that corrupt large parts of the social system, while Reason

(1987) suggests that (organizational ) l̀atent failures’ will always underlie (individual) àctive

errors’ . It is suggested that these both imply that the organization’ s overall culture can be

responsible for accidents, a view which could be derived from other investigations

(Rousseau, 1988, 1990).

Risk awareness and risk taking are highly relevant to safety culture, especially when

risk aVects intentions indirectly through subjective norms and attitudes associated with behaviour (Fleming, Flin, Mearns, and Gordon, 1998; Rundmo, Hestad, and Ulleberg,

1998; Strasson, and Fishbein, 1990). Both risk homeostasis theory and behavioural decision

theory imply that safety may be improved through changing risk-taking behaviours and

habits rather than risk perception or evaluation (Rundmo, 1996; Wagenaar, 1994; Wilde,

1982). However, it remains possible that behaviour change may be localized to àctive

errors’ rather than l̀atent errors’ and therefore be limited in extent (Reason, 1987, 1988).

Dake (1992) argues that most research on risk perception and risk communication

ignores the cultural context in which hazards are framed and debated and in which risk-

taking and risk perception occur; the theory of reasoned action incorporates such non-risk

pressures and other elements of the cultural context as one would expect in the workplace

(Ajzen, 1991). Dake (1992) further argues that disagreements about risk are not mainly

about the probability of dying but about what our world is like while we are constructing

it. Thus, appeals to shop ¯ oor employees on the basis of their own personal safety may be

of little consequence. Even the language used may re¯ ect cultural or contextual diVerences in the organization, e.g. a major incident may variously be described as òperator error’

( implicitly blaming individuals) or as caused by institutional neglect or poor policy decisions

(focusing blame on larger social units).

Social construction of risk theory describes several patterns of social relationships and

the industry sector investigated here clearly ® ts into two of these Ð hierarchical and fatalist,

where compliance with regulations and downward instruction is expected by management

and where the shop ¯ oor may be characterized as being resigned to high levels of

prescriptio n and with minimum collective participation (Dake, 1992). Fatalist shop ¯ oor

workers may accept the world as capricious and desire to stay out of harm’ s way or not,

as they so choose (Dear, 1995). Hierarchical management employees may expect shop

J. Harvey et al.22

¯ oor workers to accept and indeed take ownership of what they have been told to do or

how they have been told to behave. This social construction analysis is paralleled by

attribution theory, which would imply internal attributions by the management and external

attributions by the shop ¯ oor in relation to management safety initiatives (Harvey, and

McGlynn, 1982; Kelley, 1973; Miller, and Ross, 1975; Reason, 1990; Ross, 1977; Weisz,

and Jones, 1993).

In addition to the cultural in¯ uences, many other variables have major in¯ uence upon

individual risk-taking behaviour, such as risk-taking disposition, locus of control, perception

of luck, framing as a threat or opportunity, performance feedback and mood states

(Cherpitel, 1999; Dixon, Hayes, and Ebbs, 1998; Horswill, and McKenna, 1999; Jones,

and Wuebker, 1988; Lee, 1994; Mittal, and Ross, 1998; Niskanen, 1994; Reber, and

Wallin, 1984; Teigen, 1998). Person-environment (P-E) ® t theory may incorporate the

above and therefore oVer an explanation of work performance, accident behaviour and stress (French, Caplan, and Harrison, 1982; Sherry, 1991); it may, in conjunction with

attribution theory, also oVer predictions about safety culture.

1.4. Summary and rationale for this study

Safety culture would seem to be a complex construct that includes values and attitudes

( particularly commitment and job satisfaction), perceptions and communications, most of

which are potentially changeable and relate to actual accident behaviour. However, the

notion of an organization-wide safety culture common to all employees in an organization

has been established as an ideal rather than proved from existing evidence.

This study concerns one area of culture, safety culture, investigated across all employee

levels in a highly regulated work environment. The organization is involved in nuclear

reprocessing and had recently seen changes at senior management level; at the time of this

investigation, the new management was keen to see changes in safety culture and was

about to introduce safety training initiatives to enhance attitudes and behaviours in respect

of both conventional as well as radiological safety. The purpose of this study is to identify

the components of safety culture and to make pre-training comparisons between plants

and between occupational grades.

Several hypotheses were developed, all of which are derived from the literature and

are set in the above context. The ® rst hypothesis is derived from the work of Cheyne

et al. (1998), Cox, and Cox, (1991), IAEA (1991), Lee (1994), suggesting that commitment

and job satisfaction are important aspects in safety culture, especially in forming a positive

culture. The second hypothesis follows on from the ® rst hypothesis where the literature

states that job satisfaction is closely linked and to some extent a function of management

style and commitment (Clarke, 1998; Hofmann, and Morgeson, 1999; Lee, 1994;

Williamson et al., 1997).

(1) Hypothesis 1. Commitment and satisfaction are salient constructs in s̀afety culture’ .

(2) Hypothesis 2. Job satisfaction is a function of management style and commitment.

The third hypothesis comes from the literature surrounding risk, particularly risk

homeostasis and behavioural decision theories, which propose that risk evaluation is not a

frequent occurrence but that risky behaviour and thus attitudes to risk are determined by

routine and habit, rather than by other work attitudes (Dake, 1992; Wagenaar, 1994;

Wilde, 1982).

(3) Hypothesis 3. Perception of and behaviour in relation to risk are salient constructs

in a safety culture but are unrelated to satisfaction and commitment.

Analysis of safety culture attitudes 23

Hypotheses 4 and 5 are derived from the work of Chute, and Weiner (1995); Clarke

(1999); Hofstede (1994); Mearns et al. (1998) and Pidgeon (1998) who have shown that

culture exists at the group rather than at the organization level and thus one would expect

employment grade and plant diVerences to exist within one organization. This research has also shown that culture as a concept diVers widely for diVerent people, and when applied to safety culture one would expect these groups to conceptually think of safety in

a diVerent way to each other.

(4) Hypothesis 4. Perceptions of safety culture will show signi® cant employment grade

and plant diVerences. (5) Hypothesis 5. Factors may diVer in concept, not just in terms of factor score

diVerences, between two diVerent grades of employee.

2. Method

A survey of 60 Likert-style items was constructed in order to measure the attitudes and

perceptions identi® ed in the hypotheses, namely, perception of others’ attitudes to safety;

perceived trust in management in relation to safety; job interest and satisfaction; perceived

judgement of and attitudes to risk; perceived responsibility for and involvement in safety

issues; upward and downward communication with respect to safety; attitudes to conven-

tional and radiological safety; perceived changes in work and safety attitudes over the

previous 2 years. In addition, whether the employee was shop ¯ oor [Sh] or supervisor/

management/professional [Mgt] was identi® ed, as was the working department.

Information about age, gender and length of service was not requested because management

were concerned to avoid information that could potentially identify individuals on this

® rst survey of this type.

These items were piloted on employees from the population who were to be surveyed.

The scale used was of six points, ranging from strongly agree, agree, mildly agree to

strongly disagree; there was no mid-point because it was felt that asking respondents to

actually agree or disagree with statements was preferable in this study to a scale with a

neutral mid-point.

2.1. Sample

The sample consisted of all employees in each of two plants, which were both based on

the same site of a large nuclear power generating company in the UK. The two plants

performed related work in that they were both involved in nuclear operations, but doing

largely diVerent work on the same site, although they occasionally worked together on some projects. Both operated independently of each other and in recent years there had

been restructuring in the organization, which aVected both plants similarly. Both plants also had similar age and length of service distributions. It is suggested that, although these

two plants were not chosen at random, they formed representative strata of the whole

organization.

Plant A works with older equipment where more liability and greater risks are inherent.

They are probably seen in the organization as being a major `money earner’ . Management

here might see themselves as `head of the pack’ in many ways, typi® ed by being particularly

good at refresher training, consultation, and high levels of displayed commitment by

management to various (safety) initiatives. Management are largely drawn from production

or operations backgrounds, which inherently involve a greater emphasis on people

management.

Plant B works with newer equipment, which is therefore inherently less risky. StaV

J. Harvey et al.24

have been trained recently to work in these newer areas. Management are largely drawn

from technical and research and development backgrounds, which inherently involve less

emphasis on people-management skills.

There were 600 employees at the time of the survey in plant A, and 950 in plant B,

all of whom were asked to respond. The responses were distributed by employment grade

and Plant as in table 1.

2.2. Procedure

The survey took place in 1997/98. A covering letter was prepared so that respondents

were aware that the survey was being conducted in complete con® dence by a person from

outside the organization. Respondents were assured that nothing reported could in any

way be traced back to them individually, and were given a pre-addressed envelope to

facilitate an anonymous response. A reminder was sent to all respondents one month after

the questionnaires had been distributed.

3. Results

3.1. Exploratory analysis

All 60 items were ® rst considered separately in a descriptive way. Means ranged from x= 1.75 (item 27, related to s̀howing oV ’ at work) to x=5.47 ( item 58, concerned with the importance of morale). Standard deviations ranged from 1.39 (taking small risks) to 0.62

(all should be responsible for safety). Standard errors were low, providing a high degree of

con® dence in the means (ranging from 0.02 to 0.04).

By way of exploration and to illustrate the many predicte d grade and plant eVects, 2-way ANOVAs were performed by grade and plant on all 60 individual items. (Full data

available on request from the authors.) Although interaction eVects were not predicted, 2-way analyses would reveal them rather better than the larger number of 1-way nonpara-

metric tests necessary to achieve the same thing. (To have conducted non-parametric

equivalents to identify interaction eVects would have necessitated testing 60 grade diVer- ences within each plant and 60 plant diVerences for each grade using Mann-Whitney tests.) Although not strictly justi® ed by the ordinal nature of the data, the total N is large enough

to allow the use of ANOVAs to explore the data. These analyses can be summarized as

follows: out of 60 ANOVAs, 43 yielded signi® cant diVerences between shop ¯ oor and management grades at p<.05, 2-tailed, of which 29 were signi® cant at p<.001 and 40 at p<.01. For plant diVerences, 26 of the 60 ANOVAs yielded signi® cant diVerences at p<.05, of which 3 were signi® cant at p<.001 and 21 at p<.01.

As a statistical comparison, Kruskal-Wallis’ s one-way analyses of variance were con-

ducted on the four plant/grade groups, yielding comparable ® ndings with 46 of the 60

tests signi® cant at p<.05, of which 43 were signi® cant at p<.01 and 27 at p<.001. Of

Table 1. Number of responses for Plants A and B by employee type.

Management Grade % Response Shop¯ oor (supervisor/manager/professional) not given Total rate

Plant A 166 227 24 417 69.5 Plant B 384 191 11 586 61.7

Total 550 418 35 1003 64.7

Analysis of safety culture attitudes 25

the 43 which were signi® cant, 41 showed grade diVerences and 22 showed plant diVerences. These data are available on request from the authors.

The Bonferroni adjustment procedure (Bland, and Altman, 1995) would suggest that

only those ® ndings at p<.001 are considered when conducting 60 tests, in order to minimize the likelihood of type I errors. The large number of extremely signi® cant

diVerences between shop ¯ oor and management employees suggests that these groups can be considered as diVerent populations. However, inspection of the data for plant diVerences showed that it was largely shop ¯ oor diVerences which accounted for the signi® cant ® ndings, the managements of both plants often scoring very similarly. These diVerences did not follow any particular theme other than the shop ¯ oor for Plant B often scoring

less `desirably’ than Plant A, for example in items relating to job satisfaction.

Taken overall, these ® ndings suggest two populations in relation to safety culture, with

perhaps a possibility that the shop ¯ oor culture comprises two sub-cultures. Bearing in

mind the exploratory nature of these ® ndings, two factor analyses were conducted, one

for Sh and one for Mgt employees and the results of these analyses are given in tables 2 and 3.

Two principal components analyses of the items were conducted. In both cases the

scree plots suggested that seven factors be rotated. Varimax rotations with orthogonal

transformations were performed. The ® rst three factors had similar structures for the two

groups, and are shown in table 2. The remaining four factors for both groups are shown

in table 3. Where the factors are discussed below, Cronbach’ s a is given for the total sample except for factor 6, where it is given for each sample separately.

The two factor structures in tables 2 and 3 look similar apart from the sixth factor,

which only accounted for just over 3% of the variance for both grades. In order to test whether these factor structures are signi® cantly diVerent, a con® rmatory factor analysis was performed using SAS/STAT (1993), with the following goodness-of-® t indices [GFI]: the

model based on shop ¯ oor employees yielded GFI=.75 for Shop ¯ oor and GFI=.75 for Management, while the model based on Management yielded GFI=.75 for Shop ¯ oor and GFI=.78 for Management. These GFIs clearly do not diVer signi® cantly from one another, although each model is marginally better for its own group. From this analysis,

we cannot conclude that the two groups yield diVerent factor structures. However, the fact that six factors are so similar would statistically outweigh any diVerence in one factor which only accounted for 3% of the variance, even if that factor represented a real conceptual diVerence in psychological terms. So, although it is diYcult to ascertain from the con® rmatory factor analysis that there are any conceptual diVerences, factor 6 cannot be built as one factor for both groups, and will be considered in two parts, 6a and 6b, for

the purposes of considering plant and grade diVerences. The ® rst factor in table 2, management style and communication, was common to

both Shop ¯ oor and Management employees and accounts for approximately 16% of the variance in both cases. This ® nding supports hypothesis 1. Of the 13 items common to

both grade groups, 11 yielded very highly signi® cant grade diVerences and 9 yielded signi® cant plant diVerences. This factor has some items in common with factor 6b in table 7, although factor 1 is positively loaded and factor 6b is negatively loaded. Further,

6b contains items which suggest that the poor management is conducted proactively by,

for example, withholding information, playing down accidents, etc., whereas many items

in factor 1 relate to the perception of management behaviour in terms of its impact on

others. Cronbach’ s a was calculated as a=.88 for the 13 common items in this factor. Factor 2, including commitment and involvement, shows some items which are highly

loaded for Management but not Shop ¯ oor, particularly those relating to responsibility.

This is re¯ ected in a lower a coeYcient than for factor 1, being a=.61. Of the six items

J. Harvey et al.26

Table 2. Principal components analyses: factor loadings for shop ¯ oor (Sh) and management (Mgt) employees separately; ® rst three factors only.

Factor

Factor and item 1 Sh 1 Mgt 2 Sh 2 Mgt 3 Sh 3 Mgt

1. Management style & communication 1 Manager works on a need-to-know basis ­ .43 2 The safety emphasis is good .59 3 Supervisor has good ideas .62 .73 4 Gives time/energy to safety ­ .59 6 Manager orders without question ­ .38 ­ .46 7 Management concern for safety .67 .63

11 Information reaches me .57 .36 12 Management commitment ­ .48 13 Manager places safety ® rst .67 .65 15 Manager listens to safety ideas .54 .54 16 My ideas ignored ­ .56 ­ .32 17 Manager consults me .63 .68 19 All activities done safely .54 20 Manager does not pass on information ­ .50 ­ .50 24 I am encouraged to make suggestions .65 .40 30 Pressure to get things done ­ .47 ­ .33 32 Respect manager’s attitudes .67 .73 39 Supervisor likely to consult .63 .64 44 More time now to do things .40

2. Commitment, responsibility and involvement 5 Responsible for plant safety .51 .60 9 All contribute to safety .60 .38

10 Discuss safety with others .38 11 Information reaches me .53 14 Safety not my role ­ .58 ­ .33 16 My ideas ignored ­ .38 18 Little safety responsibility ­ .54 21 Safety should take priority .53 22 Not want more responsibility ­ .47 29 All should be responsible .55 .41 35 Tell a person if they were wrong .40 .35 46 Conventional safety not important ­ .34 ­ .34 58 Morale is important .48

3. Risk-taking of self and responses to risk-related behaviour of others 23 Take risk if know consequences .51 .48 27 If no risk, would show oV .55 .37 28 Would take small risks .72 .54 37 Would cut corners .71 .69 45 Occasions when risk taken .63 .60 50 Slow workers are a nuisance .42 .50 53 Not report a near miss .53 .39 59 People raise issues to waste time .50 .43

Percentage of variance 16.3 16.0 6.2 7.8 4.5 5.6 N of items common to both groups 13 6 8

Only factor loadings >.300 are included.

Analysis of safety culture attitudes 27

Table 3. Principal components analyses: factor loadings for shop ¯ oor and management employees separately; last four factors only.

Factor

Item 4 Sh 4 Mgt 5 Sh 5 Mgt 6 Sh 6 Mgt 7 Sh 7 Mgt

4. Job satisfaction 1 Manager works on a need-to-know basis ­ .44 ­ .32

26 I am interested in my job .65 .64 34 Teams are closer than before .45 .47 40 Job satisfaction lower now ­ .57 ­ .65 41 Supervisors trust operators .43 47 I am satis® ed with my job .73 .80 49 Recommend job to a friend .73 .64

5. Complacency 2 The safety emphasis is good .48 8 Operators could be more committed ­ .38 ­ .48

10 I discuss safety with others ­ .53 12 Management commitment ­ .54 19 All activitie s done safely .36 .71 31 Proud of attitudes .35 .50 38 Forget conventional safety ­ .51 41 Supervisors trust operators .31 .46 43 Accidents are played down ­ .41 ­ .41 60 Reasonable safety record .55 .55

6 [a]. Sh² Avoid responsibility 14 Safety not my role .52 16 My ideas ignored .42 18 Little safety responsibility .42 20 Manager does not pass on information .41 22 Not want more responsibility .45 33 Managers care less now .54 48 Less attention to safety .56

6 [b]. Management² Good vs. poor management 1 Manager works on a need-to-know basis .32 6 Manager issues orders .36

20 Manager does not pass on information .33 27 If no risk, I would show oV .35 33 Managers care less now .38 36 Use equipment unskilled .41 43 Accidents are played down .31 44 More time now to do tasks .39 46 Conventional safety not important .39 48 Less attention to safety .36 60 Reasonable safety record .30

7. Risk awareness 21 Safety should take priority .45 29 All should be responsible .41 35 Would tell manager if he was wrong .38 .42 54 I am good judge of risk .70 .32 55 Can foresee hazards .69 56 Would tell a person if unsafe .49 .46 57 Cannot aVord radiological mishaps .53 .47 58 Morale is important .37 .60

Percentage of variance 3.6 3.6 3.3 3.4 3.1 3.2 2.7 2.8 N of items common to both groups 6 6 7 11 5

Only factor loadings >.300 are included. ² Factors 6a and 6b relate to shop ¯ oor and management employees, respectively.

J. Harvey et al.28

common to both grade groups in the factor, 4 signi® cantly diVerentiated grade and 2 diVerentiated plants. It is possible that notions of commitment and involvement amongst shop ¯ oor employees contain less emphasis on responsibility because they perceive that it

is less a part of their job descriptions, whereas this may be inbuilt into job descriptions for

supervisors and management. This factor supports the proposition in hypothesis 1.

Factors 3 and 7 both involve risk, although one relates to behaviour and the other to

perception. Factor 3 contained 5 (out of 8) items which signi® cantly diVerentiated grade and 2 which diVerentiated plant, but factor 7 contained only one item that diVerentiated either grade or plant. The a coeYcients were calculated as a=.70 for factor 3 and a=.62 for factor 7. These factors do not contain items common to other factors, thus supporting

hypothesis 3.

Factor 4 includes, of the items common to both groups, two relating to the closeness

between employees and four to job satisfaction; this factor yielded a=.74. This ® nding supports hypothesis 1.

Factor 5 relates to acceptance of the current situation and the feeling that everything

is ® ne. It involves people feeling that there is already suYcient commitment, that things are done as they should be and that there is trust between shop ¯ oor and management.

The a coeYcient was a=.65. The loadings on this factor for managers are higher, although the factor accounts for the same amount of variance in responses for both groups.

Factor 6 presents very diVerent items for Shop ¯ oor and Management employees with diVerent interpretations of meanings, being the notion of avoiding taking responsibility (6a) on the one hand and distinguishing good and poor management (6b) on the other.

The a coeYcients are a=.70 for Sh and a=.66 for Mgt. The loadings on factor 6a for Sh employees are higher than the loadings on factor 6b for Mgt employees, implying that

the factor has a stronger conceptual level for the former group. These two factor structures

support hypothesis 5 insofar as this ® nding of two diVerent sets of items in factor 6a and factor 6b allow the possibility of a conceptual diVerence between the two groups.

3.2. Hypothesis testing

Factor scores were calculated using only those items that were signi® cant ( i.e. loading

>.3) in the factor for both shop ¯ oor and management employees. Equal weights were assigned to each item for simplicity. For factors 6a (the Shop ¯ oor only factor) and 6b (the

Management only factor), all items with loadings >.3 were included in the calculation of a factor score. Henceforth, all plant and grade comparisons are made using factors rather

than single items.

The distributions of the aggregated factor scores approximate to normal and there are

few tied values; although strictly the data remain ordinal, aggregated scores are closer to

interval measurement. These reasons plus the large N allow the use of parametric tests on

the data. In order to ascertain interactive eVects and to provide an overview, 2-way ANOVAs by grade and plant were conducted for each factor, and these are summarized

in table 4.

In table 4, statistics are given for the factors common to both grade groups of employees

plus the two factors 6a and 6b. Although neither 6a nor 6b are common to both groups

and should strictly not be compared, their existence supports hypothesis 5 and they are

included in order to illustrate the profound grade diVerences obtained where factors which diVer conceptually are compared. It can be seen in the table that the attitudinal factors show considerable grade diVerences and two plant diVerences, thus supporting hypothesis 4.

To examine the data further and in particular to ascertain the nature of the interaction

Analysis of safety culture attitudes 29

Table 4. Two-way analyses of variance of factors by employment grade and plant.

Grade Plant Interaction

Factor F p F p F p

1. Management style/communication 73.36 <.001 2.17 n.s. 1.60 n.s. 2. Commitment/responsibility/involvement 18.82 <.001 2.02 n.s. 1.08 n.s. 3. Risk-taking 0.04 n.s. 0.42 n.s. 5.27 .022 4. Job satisfaction 52.19 <.001 4.92 .027 0.96 n.s. 5. Complacency 0.03 n.s. 1.97 n.s. 2.96 n.s. 7. Risk awareness 0.29 n.s. 3.59 .058 1.72 n.s.

6a. Avoid responsibility 92.50 <.001 0.29 n.s. 2.46 n.s. 6b. Good vs. poor management 195.13 <.001 5.83 .016 1.85 n.s.

The analyses for factors 6a and 6b are included for illustrative purposes only as they are not common to both groups.

Table 5. Plant A: mean scores and standard deviations; z and signi® cance levels for Sh/Mgt comparisons.

Shop¯ oor Management

x SD x SD z p

1. Management style/communication 50.62 10.08 55.68 9.42 ­ 5.17 <.001 2. Commitment/responsibility/involvement 26.22 2.86 26.88 2.77 ­ 2.42 .016 3. Risk-taking 25.12 7.08 24.06 5.71 0.18 n.s. 4. Job satisfaction 22.17 5.53 24.28 4.60 ­ 3.56 .001 5. Complacency 22.37 4.05 21.89 4.38 0.84 n.s.

6a. Avoid responsibility 18.57 5.11 15.88 4.34 5.23 <.001 6b. Good vs. poor management 34.26 6.02 29.13 5.57 7.86 <.001 7. Risk awareness 26.14 2.37 26.01 2.25 0.68 n.s.

N Sh

=166; N M gt

=227.

Table 6. Plant B: mean scores and standard deviations; z and signi® cance levels for Sh/Mgt comparisons.

Shop¯ oor Management

x SD x SD z p

1. Management style/communication 48.73 10.44 55.54 8.97 ­ 7.48 <.001 2. Commitment/responsibility/involvement 25.81 2.92 26.85 2.64 ­ 3.90 <.001 3. Risk-taking 24.42 6.39 25.31 5.15 ­ 1.76 n.s. 4. Job satisfaction 21.09 4.76 23.86 4.87 ­ 5.98 <.001 5. Complacency 21.74 4.05 22.33 4.19 ­ 1.24 n.s.

6a. Avoid responsibility 19.27 5.22 15.54 4.37 8.33 <.001 6b. Good vs. poor management 35.79 5.67 29.56 6.25 10.78 <.001 7. Risk awareness 25.60 2.80 25.91 2.18 ­ 0.61 n.s.

N Sh

=372; N M gt

=189.

eVects, Sh and Mgt employees were compared within Plants for each factor using z-tests; the results are shown in tables 5 and 6. In table 5, it can be seen that factors 1, 2, 4, 6a

and 6b all continue to show grade diVerences.

J. Harvey et al.30

Table 7. Classi® cation results in percentages for canonical discriminant analysis.

Predicted group membership

Actual group membership Group 1 Group 2 Group 3

Group 1: Plant A shop ¯ oor 63.0 22.7 14.3 Group 2: Plant B shop ¯ oor 28.0 56.8 15.2 Group 3: Management [A+B] 12.5 11.7 75.8

In table 6, a similar pattern occurs. It is interesting to observe lower means for factors

1, 2 and 4 for Sh employees in Plant B which have the eVect, when Mgt scores remain more or less the same as in Plant A, of yielding higher z values.

Factor 3 showed an interaction eVect at p=.022, and this is illustrated by the means, which are higher for Sh in Plant A and Mgt in Plant B. Factor 5, although not signi® cant

in this respect, may be explained similarly with higher means for the same two groups.

Plant diVerences were identi® ed in table 4 for factors 4 and 6b. Inspection of the means for factor 4 (shown in tables 6 and 7) shows clearly lower scores for each grade in Plant B

than their counterparts in Plant A, suggesting lower job satisfaction throughout Plant B.

For factor 6b, where a high score re¯ ected perceived poor management, the pattern was

predictably in reverse to that of factor 4, i.e. Plant B had a poorer perception of management

than did Plant A.

In order to test hypothesis 5 and examine whether there are two cultures based on

grade, or whether the shop ¯ oor culture contains two subcultures for each plant, a canonical

discriminant analysis was conducted on the whole 60 items. This analysis yielded x2= 512.5, p<.001 with 120 d.f. and the data for predicted group membership are given in table 7.

In order to test hypothesis 2, the aggregated score for factor 4 was correlated with that

of factors 1 and 2; although the factors generated in tables 2 and 3 are orthogonal, these

correlations are based on the items used in calculating the aggregate scores, which are only

those common to both grade groups. Job satisfaction correlated with factor 1, management

style and communications at r=.54 and with factor 2, commitment, responsibility and involvement at r=.17. Although both of these are signi® cant, as one would expect with such a large value of N, the ® rst is also signi® cantly larger (using the Fisher z transformation)

than the second. Therefore in relation to hypothesis 2 we can say that job satisfaction is

much more a function of management style than it is of commitment.

We can therefore conclude that although management perceptions of themselves are

similar across the two plants, workers’ attitudes and perceptions of their respective manage-

ments clearly diVer in several important areas, thus supporting the hypotheses.

4. D iscus sion

This study suggests that the concept of safety culture in a highly regulated organization

varies according to whether the employee belongs to the shop ¯ oor or management, even

to the extent that one measure of culture may be inappropriate for all employees. Safety

culture not only shows very large grade diVerences in items comprising the factors but also suggests the possibility of a conceptual diVerence between the two employee grade groups. In addition, the study also supports the proposition that safety culture may diVer across diVerent departments or plants within an organization. These ® ndings would point to

Analysis of safety culture attitudes 31

certainly two and possibly three diVering safety cultures within this organization and this accords with the suggestion by Chute, and Weiner (1995), Hofstede (1994), Mearns et al.

(1998) and Pidgeon (1998) that culture as a concept is better applied at the group than at

the organizational level, a view diVering from that of Handy (1985). The three groups here would comprise management as one group and shop ¯ oor as another, the latter group

being subdivided into two groups, one for each plant. Management as a group can be

characterized as having largely positive views of themselves, greater commitment to safety,

more involvement, seeing responsibility for safety as their concern. The shop ¯ oor groups

together can be characterized as having more negative views about management commun-

ication, management commitment to safety, personal responsibility for safety and being

listened to. Additionally, the shop ¯ oor subgroup in Plant A can be characterized as having

more positive views of management and greater risk awareness and risk-taking than the

management there; the Plant B subgroup can be characterized as having more negative

views of management and lower job satisfaction than their management. There is also the

possibility, unexplored here with only two shop ¯ oor groups, that there may be more than

two shop ¯ oor sub-cultures within this, or indeed any, organization.

One explanation of the three safety cultures identi® ed is the almost pervasive shop

¯ oor and management diVerences found in this study. Yet very little safety culture research has addressed this issue, preferring instead to say what should happen (IAEA, 1991; Lee

et al., 1993; Ryan, 1991). The implication perhaps is that there is one desirable safety

culture which should be striven for, and presumably employees possess this desired state

to some degree or another. It is proposed that the safety cultures found here are not simply

more or less versions of the same thing, but that these diVer more fundamentallyÐ that shop ¯ oor employees may have diVering perceptions about what safety behaviours are desirable. If there are diVering perceptions, then these may have their roots in diVerent loci of control or diVerent levels of distractibility (Hansen, 1989; Jones, and Wuebker, 1988; Lee, 1994). The work environment diVers considerably for diVerent levels of employee, and thus person-environment ® t theory may oVer an explanation of diVerences in safety culture and even accident behaviour (French et al., 1982; Sherry, 1991).

The number of factors that comprise safety culture has variously been de® ned as 3 to

19 (Donald, and Canter, 1993; Lee, 1994; Lee et al., 1993). The present study has found

seven, but only six of these apply to all employees in the organization. Most studies agree

on several attitude areas in safety culture, such as satisfaction, leadership style and commun-

ication; the ® ndings here support these areas. Other factors found here include risk taking

and awareness, responsibility and commitment, which are also apparent in some other

studies (ACSNI, 1993; IAEA, 1991; Lee, 1994). Some of the factors in this study, such as

job satisfaction, risk-taking and management style and communications, are very similar

for all employees; others, while being identi® able as the same factors, show some variation

across employee groups. This variation may be worthy of further investigation, perhaps

with repeated measures one or two years later. This study has, however, identi® ed two

factors that are not apparent elsewhere: complacency and avoidance of responsibility. It is

suggested that the latter factor is not simply a negation of a commitment-responsibility

factor but concerns general negative attitudes more akin to alienation.

Factors 1, 6a and 6b contain most of the items which diVerentiate between shop ¯ oor and management employees, and are also the factors containing the items which diVerentiate between plants. These factors are largely attitudinal and therefore potentially can be changed

(Ajzen, 1991) and may indeed relate to accident rates (Donald, and Canter, 1993). It is

likely that attribution theory oVers an explanation for many of the Sh/Mgt diVerences, since one would expect managers to attribute more positive attitudes and behaviours to

J. Harvey et al.32

themselves in a self-serving way than to shop ¯ oor employees (Hewstone, 1989; Miller,

and Ross, 1975). This has important implications for the way that workplace interactions

are conducted: if management believe that the way they are communicating is very eVective but shop ¯ oor employees do not, problems may arise and misunderstandings are highly

probable.

Risk perceptions and risk-taking showed few diVerences across the plants and between grades, factor 3 yielding only a signi® cant interaction eVect which seems to result from management rather than shop ¯ oor diVerences. This may be explained by context diVer- ences and really suggests that there are few if any group cultural diVerences whereby one group of individuals perceives risk diVerently to another group due to the diVerent values and beliefs of the groups (Dake, 1992; Slovic, FischoV, and Lichenstein, 1980). However, it is likely that, given the resounding diVerences between the two employment grades on the other safety culture factors, the social construction of risk diVers in some fundamental ways, such as in attributions relating to taking responsibility for one’ s own personal safety

and participation in reducing risk (Dake, 1992; Harvey, and McGlynn, 1982; Kelley, 1973;

Miller, and Ross 1975; Reason, 1990; Ross, 1977; Weisz, and Jones, 1993).

Job satisfaction is clearly identi® ed here as a factor relevant to perceptions of safety.

This ® nding accords with other research and may diVerentiate those who have accidents at work from those who do not (Cox, and Cox, 1991; Lee, 1994; Lee et al., 1993).

However, this proposition must be treated with caution as variance in accident behaviour

can be accounted for by many variables, such as the hazard exposure of shop ¯ oor

employees, of which job satisfaction is only one. This view is reinforced by the fact that

this factor accounted for only 3% of the variance in the survey for both groups. It is not possible to correlate accident or incident numbers with job satisfaction from this study,

since the survey was conducted anonymously; however, had respondents been able to be

identi® ed, they may have been less honest in their answers and thus render spurious any

relationship found between job satisfaction and accidents. In this study, job satisfaction was

found to be related to management style and communication, a factor accounting for much

more of the variance. There are clear implications here for management, who may be

making wrong assumptions about the actual nature of job satisfaction for shop ¯ oor

employees.

Some 20% of respondents disagreed that they were satis® ed and they would not recommend their job to a friend; these ® ndings suggest that there are issues relating to

morale and the psychological contract following restructuring (Cooper, 1995). The organ-

ization studied here might well view themselves as leading edge and `excellent’ (Peters,

and Waterman, 1982), and this implies that there is a need to get the culture r̀ight’ for

corporate success. This would imply that this organization has got the culture right but it

would be erroneous to jump to this conclusion. However, the ® nding of an avoidance/

alienation factor, coupled with the high variances for the job satisfaction items suggests the

existence of an unhappy minority, perhaps 20% of employees, for whom the culture is not right.

Reber, and Wallin (1984) suggest that feedback about employee performance of safe

behaviours on the job contributes signi® cantly to the occurrence of safety behaviour. This

has important implications within this organization. Communication is a major part of

factors 1 and 6b, whereby shop ¯ oor employees are more likely to feel that insuYcient communication reaches them and that management operate on a need to know basis.

Management employees on the other hand are more likely to believe the reverse and thus

a mismatch of perceptions may arise, which could have important implications for safety

behaviours. If Reber, and Wallin (1984) are correct, the occurrence of safety behaviours

could be signi® cantly improved if better communication was achieved.

Analysis of safety culture attitudes 33

This study has found that most individuals have a high level of personal responsibility

for safety and that they are committed to safety, but has also shown that there are several

areas of discontent, mainly amongst shop ¯ oor employees, in terms of avoidance or

alienation. These areas mostly concern communication, management style and responsibil-

ity. If these areas are viewed as organizational l̀atent failures’ (Reason, 1987) then individual

accidents or àctive errors’ will continue to occur until the latent failures are resolved.

Within this organization, the survey shows communication as being generally more of a

problematic issue than commitment. Commitment to safety was evident in both groups

of employees, whereas communication was seen more negatively by shop ¯ oor employees

than by management.

An interesting factor to arise concerns the acceptance that everything is ® ne and that

there are no major problems; notwithstanding the high levels of safety in the organization,

it seems that there is a complacency factor in both groups of employees such that everything

is conducted as it should be and done as safely as possible. This would accord with the

notion of a role culture (Handy, 1985) and would imply resistance to change amongst the

most complacent employees.

5. Conclus ions

This study was conducted within a highly regulated organization, which clearly and

necessarily has safety at the forefront of its agenda. It is important to consider whether the

® ndings of the study: those of three safety cultures, almost pervasive diVerences between shop ¯ oor and management, plus some diVerences between plants, will generalize to other parts of the organization, other highly regulated industries or elsewhere.

From what has been found here, it is possible that there is some variation in culture

across diVerent plants in the organization, particularly for shop ¯ oor employees, but the most likely situation is that these factors will prevail organization-wide for management.

Despite the restructuring, this organization presents nothing very diVerent to other highly regulated organizations and industries and thus it is proposed that two, three or more safety

cultures exist elsewhere, although the precise details and scores on the factors might diVer. The implications of several safety cultures in an organization may be considerableÐ in

terms of training, teamworking, etc., and there is much potential for further research here.

It is more diYcult to say whether similar numbers or types of safety culture would be found in other organizations that are not highly regulated. The issue here is primarily

whether safety is of salient importance to employees in these organizations. Less salience

may simply mean lower factor s̀cores’ or it may mean diVerent factors altogether. Given that communications, commitment, job satisfaction, etc., are important issues in any

organization and are related in many cases to work performance, then it is likely that they

would appear as part of the factor structures in safety or any other form of culture. Again,

these issues oVer much potential for further research.

A ® nal issue is the stability of the factor structure. The results of any factor analysis are

dependent on the items that are fed into it. Over time, various studies have generated

various factor structures, and although these are beginning to converge and de® ne what

will probably eventually become the de® nitive factors in safety culture, researchers are still

engaged with this issue. The possibility of some conceptual diVerences within safety culture for diVerent sub-groups may be a way forward but it may complicate the issue further; nevertheless it must be addressed. There is also the added complication that factor structures

may change over time, for example as safety increases or decreases in salience. This was

J. Harvey et al.34

not investigated here, but repeated measures over the next few years may shed some light

on this matter.

Acknowled g em ents

The authors would like to thank two anonymous reviewers for their comments, which

have helped to strengthen this paper.

References Advisory Committee for Safety in Nuclear Installations (ACSNI) Study Group on Human

Factors (1993). Third report, Organising for Safety. London: HMSO. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Decision Processes,

50, 179± 211. Beck, M., & Woolfson, C. (1999). Safety culture Ð a concept too many? The Health and Safety

Practitioner, 16 (1), 14± 16. Bland, J. M., & Altman, D. G. (1995). Multiple signi® cance tests. The Bonferroni method. British

Medical Journal, 310, 170. Booth, R. T., & Lee, T. R. (1995). The role of human factors and safety culture in safety

management. Proceedings of the Institution of Mechanical Engineers, Part B. Journal of Engineering Manufacture, 209, 393± 400.

Cherpitel, C. J. (1999). Substance use, injury, and risk-takin g dispositions in the general population. AlcoholismÐ Clinical and Experimental Research, 23 (1), 121± 126.

Cheyne, A., Cox, S., Oliver, A., & Tomas, J. M. (1998). Modelling safety climate in the prediction of levels of safety activity. Work and Stress, 12, 255± 271.

Chute, R., & Weiner, E. L. (1995). Cockpit-cabi n communication. I. A tale of two cultures. International Journal of Aviation Psychology, 5, 257± 276.

Clarke, S. (1998). Safety culture on the UK railway network. Work and Stress, 12, 285± 292. Clarke, S. (1999). Perceptions of organizational safety: implications for the development of safety

culture. Journal of Organizational Behavior, 20, 185± 198. Cooper, C. L. (1995). Papering over the cracks: Individual strategies or organisational interventions in dealing

with stress at work. Address given to the Occupational Psychology Conference of the British Psychological Society, Warwick, January.

Cooper, M. D., & Phillips, R. A. (1994). Validation of a safety climate measure. Paper presented at the Occupational Psychology Conference of the British Psychological Society, Birmingham, January.

Cox, S. J., & Cox, T. (1991). Structure of employee attitudes to safety: A European example. Work and Stress, 5, 93± 106.

Cox, S., & Flin, R. (1998). Safety culture: Philosopher’ s stone or man of straw? Work and Stress, 12, 189± 201.

Dake, K. (1992). Myths of nature: Culture and the social construction of risk. Journal of Social Issues, 48 (4), 21± 37.

Dear, L. (1995). Negotiated safety, what you don’ t know wont hurt you, or will it?, Drug and Alcohol Review, 14, 323± 329.

Dedobbeleer, N., & Beland, F. (1991). A safety climate measure for construction sites. Journal of Safety Research, 22 (2), 97± 103.

Diaz, R. I., & Cabrera, D. D. (1997). Safety climate and attitude as evaluation measures of organizational safety. Accident Analysis and Prevention, 29, 643± 650.

Dixon, M. R., Hayes, L. J., & Ebbs, R. E. (1998). Engaging in ìllusory control’ during repeated risk-taking. Psychological Reports, 83 (Part 1), 959± 962.

Donald, I., & Canter, D. (1993). Psychological factors and the accident plateau. Health, 18, 77± 125. Fishbein, M., & Ajzen, I. (1975). Belief, Attitude, Intention and Behavior: An Introduction to Theory and

Research. Reading, MA: Addison-Wesley. Fleming, M., Flin, R., Mearns, K., & Gordon, R. (1998). Risk perceptions of oVshore workers

on UK oil and gas platforms. Risk Analysis, 18, 103± 110. French, J., Caplan, R., & Harrison, R. V. (1982). The Mechanisms of Job Stress and Strain. New

York: Wiley.

Analysis of safety culture attitudes 35

Handy, C. (1985). Understanding Organizations. Harmondsworth: Penguin. Hanisch, K. A., Hulin C. L., & Roznowski M. (1998). The importance of individuals’ repertoires

of behaviors: The scienti® c appropriateness of studying multiple behaviors and general attitudes. Journal of Organizational Behavior, 19, 463± 480.

Hanson, C. P. (1989). A causal model of the relationship among accidents, biodata, personality and cognitive factors. Journal of Applied Psychology, 74,. 81± 90.

Harrison, R. (1972). Understanding your organization’s character. Harvard Business Review, Vol. 50, May± June, 119± 128.

Harshbarger, D., & Rose, T. (1991). New possibilities in safety performance and the control of workers’ compensation costs. Journal of Occupational Rehabilitation, 1 (2), 133± 143.

Harvey, J. H., & McGlynn, R. P. (1982). Matching words to phenomena: The case of the fundamental attribution error. Journal of Personality and Social Psychology, 432, 345± 346.

Hewstone, M. (1989). Causal Attribution: From Cognitive Processes to Collective Beliefs. Oxford: Blackwell.

Hofmann, D. A., & Morgeson, F. P. (1999). Safety-related behavior as a social exchange: the role of perceived organizational support and leader-member exchange. Journal of Applied Psychology,

84, 286± 296. Hofstede, G. (1994). Cultures and Organizations: Software of the Mind. New York: Harper-Collins. Horswill, M. S., & McKenna, F. P. (1999). The eVect of perceived control on risk taking. Journal

of Applied Social Psychology, 29, 377± 391. Institution of Occupational Safety and Health [IOSH] (1994). Policy Statement on Health and

Safety Culture. IOSH: Leicester. International Atomic Energy Authority [IAEA]. (1991). Safety CultureÐ a report by the

International Nuclear Safety Advisory Group. Safety Series No. 75 INSAG-4. Vienna: IAEA. IAEA (1994). Ascot GuidelinesÐ guidelines for organisational self-assessment of safety culture and for reviews

by the assessment of safety culture in organization teams. IAEA Tecdoc 743, May. Jones, J. W., & Wuebker, L. J. (1988). Accident prevention through personnel selection. Journal of

Business and Psychology, 3, 187± 198. Kelley, H. H. (1973). The process of causal attribution. American Psychologist, 28, 107± 128. Lee, T. R. (1993). Seeking a safety culture. Atom, 429, 20± 23. Lee, T. R. (1994). Employee attitudesÐ the quintessence of safety culture. Paper presented at the Third

European Seminar on Human Factors in OVshore SafetyÐ Their Importance in Safety Case Implementation, Aberdeen, Sept.

Lee, T. R. MacDonald, S. M., & Coote, J. A. (1993). Perceptions of risk and attitudes to safety at a nuclear reprocessing plant. Paper presented at the SRA Fourth Conference, Rome, October.

Marek, J., Iverson, E., & Hellesoy, O. H. (1987). Risk, Organization and Safety on an Oil Platform. In W. T. Singleton, & J. Hovden (Eds.), Risk and Decisions ( pp. 123± 127). Chichester: Wiley

Mearns, K. J., & Flin, R. (1999). Assessing the state of organizational safetyÐ Culture or climate? Current Psychology: Developmental, Learning, Personality, Social, 18 (1), 5± 17.

Mearns, K. J., Flin, R., Gordon, R., & Fleming, M. (1999). Measuring safety climate on oVshore installations. Work and Stress, 12, 238± 254.

Miller, D. T., & Ross, M. (1975). Self-serving biases in the attribution of causality: Fact or ® ction? Psychological Bulletin, 82, 213± 225.

Mittal, V., & Ross W. T. (1998). The impact of positive and negative aVect and issue framing on issue interpretation and risk taking. Organizational Behavior and Human Decision Processes, 76, 298± 324.

Nicholson, N., & Johns, G. (1985). The absence culture and the psychological contractÐ Who’ s in control of absence? Academy of Management Review, 10, 397± 407.

Niskanen, T. (1994). Assessing the safety environment in work organization of road maintenance jobs. Accident Analysis and Prevention, 26 (1), 27± 39.

Ostrom, L., Wilhelmsen, C., & Kaplan, B. (1993). Assessing safety culture. Nuclear Safety, 34, 163± 172.

Peters, T., & Waterman, N. (1982). In Search of Excellence: Lessons from America’s Best Run Companies. New York: Harper & Row.

Pidgeon, N. (1998). Safety culture: key theoretical issues. Work and Stress, 12: 202± 216. Reason, J. R. (1987). A Framework for Classifying Errors, in New Technology and Human Error.

Chichester: Wiley. Reason, J. R. (1990). Human Error. New York: Cambridge University Press.

Analysis of safety culture attitudes36

Reason, J. R. (1998). Achieving a safe culture: theory and practice. Work and Stress, 12, 293± 306. Reber, R. A., & Wallin, J. A. (1984). The eVects of training, goal-setting and knowledge of results

on safe behaviour: A component analysis. Academy of Management Journal, 27, 544± 560. Ross, L. (1977). The intuitive psychologist and his shortcomings. Advances in Experimental Social

Psychology, 10, 174± 220. Rousseau, D. M. (1988). The construction of climate in organization research. In C. L. Cooper,

& I. T. Robertson (Eds.), International Review of Industrial and Organizational Psychology, Vol. 13: 139± 158. Chichester: Wiley.

Rousseau, D. M. (1990). Assessing organizational culture: The case for multiple methods. In B. Schneider: Organizational Climate and Culture, San Francisco: Jossey Bass.

Rundmo, T., Hestad, H., & Ulleberg, P. (1998). Organisational factors, safety attitudes and workload among oVshore oil personnel. Safety Science, 29 (2), 75± 87.

Ryan, T. G. (1991). Organisational factors. Regulatory Research Brie® ng to the ACSNI Study Group on Human Factors and Safety, London, July.

SAS/STAT (1993). SAS/STAT Software: Syntax Version 6. North Carolina: SAS Institute Inc. Sherry, P. (1991). Person-environment ® t and accident prediction. Journal of Business and Psychology,

5, 411± 416. Slovic, P., Fischhoff, B., & Lichtenstein, S. (1980). Facts vs. fears: Understanding perceived risk. In

R. C. Schwing and W. A. Albers (Eds): Societal Risk Assessment: How Safe is Safe Enough, New York: Plenum Press, pp. 181± 216.

Strasson, M., & Fishbein, M. (1990). The relation between perceived risk and preventive action: A within-subject analysis of perceived driving risk and intentions to wear seatbelts. Journal of Applied Social Psychology, 20, 1541± 1557.

Teigen, K. H. (1998). Hazards mean luck: Counterfactual thinking in reports of dangerous situations and careless behavior. Scandinavian Journal of Psychology, 39, 235± 248.

Wagenaar, W. A. (1994). Risk-takin g and accident causation. In J. F. Yates, & ?? Stone. Risk Taking Behaviour. Chichester: Wiley

Weisz, C., & Jones, E. E. (1993). Expectancy discon® rmation and dispositional inferenceÐ latent strength of target-based and category-based expectancies. Personality & Social Psychology Bulletin, 19, 563± 573.

Wilde, G. J. S. (1982). The theory of risk homeostasis: Implications for safety and health. Risk Analysis, 2, 209± 225.

Williamson, A. M., Feyer, A. M., Cairns, D., & Biancotti, D. (1997). The development of a measure of safety climate: The role of safety perceptions and attitudes. Safety Science, 25 (1± 2), 121± 140.

Zohar, D. (1980). Safety climate in industrial organizations: Theoretical and applied implication. Journal of Applied Psychology, 65, 96± 102.