Transcript
PERSONALITY TRAIT A N D COGNITIVE ABILITY CORRELATES OF UNSAFE BEHAVIOURS
by
ZEHRA PIRANI LEROY
B. Sc., The University of Victoria, 1998 B. A. Hons., The University of British Columbia, 2003
A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ARTS
in
THE F A C U L T Y OF GRADUATE STUDIES
(Psychology)
August 2005
11
Abstract
Unsafe behaviours were presumed to be a primary precursor to accident involvement, related to
personality, attention and memory. In Study 1, 633 undergraduates completed a personality
inventory and a hazardous-behaviours questionnaire. A trait-based scale was empirically
developed to assess safety-oriented tendencies. The scale is suitable for applied use, and draws
on traits related to the Big Five, risk-taking, counterproductivity, and impulsivity. In Study 2,
140 undergraduates completed the Study 1 measures and a battery of cognitive ability tests
assessing attention and memory. Two common-factors—Cognitive Errors and Performance
Speed—were correlated with the Study-1 Safety-Orientation scale, but not with unsafe
behaviours. Individual-differences variables may have a more complex role in the safety system
than previously thought, and could be used to improve various Human Resources interventions
to reduce accidents in the workplace, such as through selection, placement, training, and job
redesign. Recommendations for future research are discussed.
Il l
The Role of Unsafe Behaviours in the Safety-Systems Approach 4
Individual Difference Contributors to Unsafe Behaviours 4
Personality Traits and Unsafe Behaviours 5
Cognitive Abilities and Unsafe Behaviours 6
Methodological Issues Associated with Previous Research 8
Current Directions 9
Operationalization of Unsafe Behaviours Criterion 11
Assessment of Safety-oriented Tendencies by means of a Personality Inventory. 12
Procedures 12
Assessment of Safety-oriented tendencies by means of a personality inventory ..21
Relating Unsafe Behaviours to Accident Involvement Differential 22
Relating Unsafe Behaviours to Cognitive Failures 22
Cognitive Ability Measures used in Sample 1 22
Cognitive Ability Measures used in Sample 2 23
Procedures 26
Individual-Differences Variables and the Safety System 34
Limitations and Future Research 35
References 37
Appendix B. Normative data and reliability estimates for other BIODATA-250 scales 65
Appendix C. Questions from Accident-Related Events Scale 66
Appendix D. Additional Details for Study 2 Measures 67
Cognitive Failures Questionnaire (CFQ) 67
Work Skills Assessment (WSA), Part 4 67
V
Comprehensive Ability Battery - Memory Span (CAB-Ms) 68
Comprehensive Ability Battery - Associative Memory (CAB-Ma) 68
Colour-Word Stroop Test 69
Digit Span Backward Test (DSB) 70
Test of Variables of Attention (TOVA) 71
vi
List of Tables
Table 1. Oblique primary-factor matrix for the 16 items from the Criterion Hazardous-Behaviour
Scale 45
Table 2. Factor scales derived from the commom-factor analysis of the 16-item Criterion
Hazardous-Behaviour Scale, including the weighted by gender internal consistency (a)
reliability estimate for each subscale 46
Table 3. The 51 BIODATA-250 items selected for the BIODATA-250 Safety-Orientation
scale 47
Table 4. Oblique primary-factor matrix for the 51 items from the Safety-Orientation scale, with
R's indicating reverse-scored items 50
Table 5. Primary-factor intercorrelation matrix for the 51-item Safety-Orientation scale 53
Table 6. Factor scales derived from the commom-factor analysis of the 51-item Safety-
Orientation scale, including the weighted by gender internal consistency (a) reliability
estimate for each subscale, and R's indicating reverse-scored items 54
Table 7. Correlations between the Safety-Orientation scale and factor scales, and the other
BIODATA-250 scales 57
BIODATA-250 scales 58
Table 9. Oblique primary-factor matrix for the 13 cognitive ability variables from Sample 2 59
Table 10. Factor scales derived from the commom-factor analysis of the cognitive ability
variables from Sample 2, including the internal consistency (a) reliability estimate for
each subscale 60
Table 11. Correlations between the cognitive ability subscales and variables, and the Criterion
Hazardous-Behaviour Scale and subscales, and the Accident-Related Events Scale 61
V l l
Table 12. Correlations between the Safety-Orientation scale and subscales, and the cognitive
ability subscales 62
ix
Acknowledgements
There are many people that I would like to thank who have supported, guided, and
assisted me throughout the course of my Master's degree. I would like to acknowledge all the
participants who have graciously given their time for my research. I would also like to thank my
research assistants —Yasmin Ahamed, Rosemarie Ong, Vil i ja Petrauskas, Gwen Montgomery,
Shirley Sarkodee-Adoo, and Joan Ewasiw— who have been both professional and diligent in
maintaining high standards while collecting and entering data.
I am grateful to Peter Graf for providing me with some of the cognitive ability measures
that were included in the research, and to Carrie Cuttler for sharing her knowledge of
neuropsychological assessment measures and prospective memory, and for coaching me on how
to administer the cognitive ability tests for the second study. I also want to thank Ekin
Blackwell, Darcy Hallett, and Kevin Williams, from the departmental stats consulting office,
who helped me with the factor analyses.
I want to thank the friends and colleagues who have followed my progress and
tribulations, and helped keep a smile on my face. My family has always supported my desire to
study; I am thankful for their continuing support and understanding during my time as a student.
My husband Sean has been my biggest supporter; I am thankful for his patience and tenderness,
and for the light he brings into my life everyday.
I am grateful to the Elizabeth Young Lacey Foundation, the Social Sciences and
Humanities Research Council of Canada, and the Workers' Compensation Board of British
Columbia, and the T.O.V.A. Research Foundation for supporting my research and providing me
with funding throughout the course of my Masters program.
I would like to thank my committee members, Todd Handy, Sandra Robinson, and Linda
Scratchley, for their time and expertise, and careful consideration of my thesis. And last, I would
X
like to thank my supervisor, Ralph Hakstian, for his time in helping me sort through my ideas
and statistical analyses, for his continuing belief in me, and for his understanding of my trials this
past year. Through his mentorship and expertise I have become a better researcher and writer,
and am thankful for all that he has taught me.
1
Introduction
Personnel selection practices have evolved in the last decade in response to increasing
empirical evidence that personality traits and cognitive abilities successfully predict job
performance (Barrick & Mount, 1991; Schmidt & Hunter, 1998). Notably, the prediction of
workplace counterproductive behaviours has lead to increasing demands for integrity testing
before hiring decisions are made (Hakstian, Farrell, & Tweed, 2002; Ones, Viswesvaran, &
Schmidt, 1993).
behaviours: deviant (theft and substance use at work), absentee (including tardiness), and unsafe
(accidents and injuries). The overall purpose of this thesis is to explain variability in unsafe
behaviours with personality traits and cognitive abilities, specifically attentional and memory
abilities. If individuals who frequently engage in unsafe behaviours can be identified,
organizations can use this information to reduce workplace accidents through Human Resource
interventions such as selection, placement, training, and job design (Hale & Glendon, 1987;
Jones & Wuebker, 1988; Lawton & Parker, 1998).
Each year in Canada there are over 700 fatal occupational injuries and over 400,000 non­
fatal occupational injuries requiring time off (Stewart, 2002). These statistics and the associated
costs to society underscore the need to continue to research ways to reduce workplace accidents.
Although the focus of this thesis is on predicting unsafe behaviours from an individual-
differences approach, this does not suggest that workers are to blame for accidents or that other
contributing factors are less important. The focus on individual differences is meant to add to
the existing body of knowledge in accident research.
2
Structure of Thesis
This thesis is the outcome of two empirical studies. Study 1 presents a trait-based scale
to predict unsafe behaviours, and examines its properties in relation to unsafe behaviours.
Study 2 explores the cognitive-ability correlates of unsafe behaviours. The general discussion
and conclusions section provides an overall discussion as well as recommendations for future
research.
The remaining sections of the introduction review the theoretical underpinnings of this
thesis, beginning with a brief review of the literature on unsafe behaviours and a discussion of
how individual-differences variables relate to unsafe behaviours and differential accident
involvement. This is followed by an examination of the methodological problems associated
with past research, an overview of current research directions, and a summary of the objectives
of this research. It should be noted at the outset that unsafe behaviours are also considered an
individual-differences variable. To avoid confusion, the term "individual-differences variables"
in this thesis will refer only to the enduring, measurable traits of individuals that relate to unsafe
behaviours (e.g. personality traits, cognitive abilities).
Conceptualizing Unsafe Behaviours
Occupational accidents and injuries usually have multiple causes and contributing
factors. Incidents may involve equipment, risk exposure, the actions of other workers,
management decisions, leadership, and the overall safety climate. For this reason, accidents and
accident involvement have not been predicted effectively. Several researchers have proposed a
shift towards predicting the precursors of accidents, such as worker behaviour, because these are
more easily studied than accidents (Bradley, 1997; Hale & Glendon, 1987; Lawton & Parker,
1998).
Unsafe Behaviours Defined
For the purpose of this thesis, unsafe behaviours are described as actions related to risk-
taking, absentmindedness and carelessness, and include both intentional acts (e.g. conscious risk-
taking) and unintentional acts (e.g. automatic behaviours). Safe behaviours include wearing
safety equipment, following safety rules, and having a positive attitude towards safety (Neal &
Griffin, 2004).
Unsafe and safe behaviours can be understood to be either independent or on opposite
ends of a continuum. If they are considered independent, an individual could engage in both
types of behaviour simultaneously. If they are considered on a continuum, an individual who
engages more in one of the behaviours would engage less in the other.
As in Bradley (1997), the latter approach was adopted for this thesis because it is unlikely
for individuals to engage in safe and unsafe behaviours simultaneously unless the environment
imposes a strong safety structure (e.g. enforcement of safety-equipment rules), and for safety
behaviours to generalize to the home once the safety structure has been removed (Lund &
Hovden, 2003).
Unsafe behaviours increase an individual's chance of being involved in an accident-
related event (Bradley, 1997), a likelihood that McKenna (1983) referred to as differential
accident involvement. The presumed positive relationship between unsafe behaviours and
differential accident involvement is based on the idea that certain actions—such as taking risks,
not following rules, and being careless—make an individual more likely to be in a situation that
leads to an accident-related event (Hale & Glendon, 1987; Hofmann & Stetzer, 1996).
Conceptual Model
The conceptual model for this thesis appears in Figure 1. In this model, unsafe
behaviours are neither necessary nor sufficient for accident-involvement to occur, whereas
4
organizational factors are. Unsafe behaviours are a mediating factor between individual-
differences variables and the accident-involvement outcome. Organizational variables are also
moderators of the relationship between individual-differences variables and unsafe behaviours.
The Role of Unsafe Behaviours in the Safety-Systems Approach
The safety-systems approach encompasses the entire organizational system and stresses
the safety climate, attitudes, management initiatives, and the elimination of physical hazards
(Petersen, 1996; Stewart, 2002). Various models of accident causation have been proposed
within this approach, some of which have given prominence to the role of unsafe behaviours
(Bradley, 1997; Hale & Glendon, 1987; Lund & Aaro, 2004). Individual-differences variables
do not exist in isolation when influencing behaviours, but interact with the other components of
the system.
Most of the safety-systems research has focused on the role of organizational variables
(Lawton & Parker, 1998). Hansen (1989) was one of the first to present an empirical accident-
causation model that included individual-difference variables while controlling for risk exposure.
In this study, social maladjustment and distractibility were found to be significant predictors of
accidents. Studies of the relationship between everyday cognitive failures and accidents found
that both conscientiousness (Wallace & Vodanovich, 2003b) and level of stress (Reason, 1990)
each acted as moderator variables of this relationship.
Individual Difference Contributors to Unsafe Behaviours
Individual-differences variables have been studied more extensively outside the safety-
system approach. As the focus of this thesis is on exploring the relationship between unsafe
behaviours, personality traits, and variables of attention and memory, only theories and research
related to these constructs are reviewed, starting with personality and then proceeding through
the cognitive abilities.
Behaviours related to accident involvement may include acting without planning, seeking
thrilling activities, taking short-cuts, not taking responsibility for actions, working carelessly, or
not following rules (Hale & Glendon, 1987). The propensity to engage in one or more of these
behaviours on a regular basis can be characterized as a behavioural tendency. Personality traits
are the characteristics that differentiate individuals according to relatively stable thought patterns
and behavioural tendencies. The behaviours listed above describe aspects related to the
following personality traits—impulsivity, risk-taking, locus of control, and conscientiousness.
The dominant theory in accident research related to personality traits has been the
accident-proneness model, which suggested that individuals are predisposed to having an
unequal accident liability. Multiple accident-involved individuals were assumed to have
inherited a stable personality trait, or set of traits, thus predisposing them to experience more
accidents, regardless of other external variables, including risk exposure (Hale & Hale, 1972;
McKenna, 1983). This model, based on Greenwood and Woods research (1919, reproduced in
Haddon, Suchman, & Klein, 1964), was discredited by McKenna, and replaced by the safety-
systems approach, which avoids blaming the victim and provides incentives for organizations to
eliminate workplace hazards (Hale & Glendon, 1987; Hansen, 1988; Lawton & Parker, 1998).
Even though the accident-proneness model is no longer at the forefront of accident
research, researchers have continued to study the characteristics of accident-involved employees
(e.g. Cellar, Nelson, Yorke, & Bauer, 2001; Forcier, Walters, Brasher, & Jones, 2001). These
and other studies contribute to our understanding of how personality traits relate to accident-
involvement, including from a safety-systems perspective.
Personality correlates of accident involvement. Research on the personality differences
between people who experience more versus fewer accidents has produced inconsistent results.
Much of this research has used the five-factor model (Costa & McCrae, 1992), a
conceptualization of personality that is currently the most widely used in research, and includes
the Big Five traits: neuroticism, extraversion, openness, conscientiousness, and agreeableness.
Within the five-factor model, conscientiousness stands out as a trait negatively correlated with
accident involvement (Arthur & Doverspike, 2001; Arthur & Graziano, 1996; Cellar et a l , 2001)
and unsafe work behaviours (Wallace & Vodanovich, 2003b). Other Big Five traits—
neuroticism, extraversion, and agreeableness—have been positively correlated with accident
involvement (Cellar, Nelson, & Yorke, 2000; Hansen, 1988; Lawton & Parker, 1998; Sutherland
& Cooper, 1991). In a meta-analytic study of accidents using the Big Five personality measures,
Salgado (2002) found that none of the traits was related to workplace accident involvement.
Other traits positively correlated with accident involvement include external locus of
control, aggression, sensation seeking, risk taking, social maladjustment (Hansen, 1988; Lawton
& Parker, 1998; Salminen, Klen, & Ojanen, 1999; Thiffault & Bergeron, 2003), Type A
personality (Magnavita, Narda, Sani, Carbone, De Lorenzo, & Sacco, 1997; Sutherland &
Cooper, 1991), and negative affectivity (Iverson & Erwin, 1997). Additional traits found only
among high-risk drivers include low altruism, and either very low or very high anxiety (Iverson
& Rundmo, 2002; Ulleberg, 2002).
Cognitive Abilities and Unsafe Behaviours
Unsafe behaviours also have a strong cognitive component. Employees may have to:
focus on tasks; evaluate risks and their abilities; make decisions; be vigilant for abnormalities
and hazards; balance speed with accuracy, inhibit dominant responses when a different action is
required (Hale & Glendon, 1987); remember past actions; and remember to complete tasks at a
future time (Dornheim, 2000). These behaviours become more complex and challenging when
stress, fatigue, and the social context play a role (Reason, 1990).
In Reason's (1990) theory of human error, cognitive errors are understood as failures in
different components of the cognitive system. Cognitive failures have been found to predict
(using the self-report Cognitive Failures Questionnaire, CFQ; Broadbent, Cooper, FitzGerald, &
Parkes, 1982) safety-related behaviours and accidents at work (Larson & Merritt, 1991; Wallace
& Vodanovich, 2003a; Wallace & Vodanovich, 2003b). Reason divided errors into slips and
lapses, considered failures in execution and memory, and mistakes, which are considered failures
in judgement and decision-making. Slips and lapses describe functions related to attention,
memory, and information processing (Bradley, 1997), of which attention and memory will be
explored further. Mistakes describe functions related to executive functioning, which are beyond
the scope of this thesis. Past research on variables of attention and memory (when available) are
reviewed in relation to accident involvement.
Attentional ability correlates of accident involvement. Attention is the primary cognitive
resource required to monitor internal processes and to avoid making errors that can lead to
accidents (Hale & Glendon, 1987; Reason, 1990). The Sohlberg and Mateer clinical model of
attention is a common typology used to operationalize this complex function (Kerns, 1996). The
model includes: focused and selective attention (often used interchangeably; Ponsford, 2000), the
ability to focus on a stimulus while ignoring an irrelevant one; sustained attention (also known as
vigilance or continuous attention; Warm, 1984), the ability to maintain focus over time;
alternating attention, which involves shifting focus between stimuli; and divided attention, the
ability to respond to multiple stimuli or making multiple responses.
Selective attention, operationalized with auditory selective and dichotic listening tasks,
was the most promising predictor of motor vehicle accidents in the Arthur et al. (1991) meta-
analytical study. Lawton and Parker (1998) described mixed results with measures of visual
attention, and Edkins and Pollock (1997) found a negative relationship between sustained
attention variables and railway accidents. Measures of attention have also been compared with
the CFQ. Cognitive failures were negatively correlated with sustained attention performance
(Manly, Robertson, Galloway, & Hawkins, 1999; Robertson, Manly, Andrade, Baddeley, &
Yiend, 1997; Wallace, Kass, & Stanny, 2001), but were found to have no relationship with
selective attention or dichotic listening variables (Martin, 1983).
Memory-related abUities and accident involvement. Both retrospective and prospective
memory failures are presumed to have a significant impact on accident involvement (Dornheim,
2000). Failures to retrieve previously stored information are related to retrospective memory,
whereas failures to remember to complete a task at a future time are related to prospective
memory (Graf & Uttl, 2001).
Methodological Issues Associated with Previous Research
Most of the reviewed research has been inconsistent in predicting accidents because the
accidents are infrequent (McKenna, 1983; Salgado, 2002), have multiple causes (Hale &
Glendon, 1987), have been conceptualized and measured differently, or the studies have not
controlled for variables that could moderate the effects of individual-differences (Lawton &
Parker, 1998).
personality traits, accident-involvement and unsafe behaviours has in part been attributed to a
lack of special-purpose inventories designed to predict accidents or safety-related tendencies.
An exception is the Employee Safety Inventory (ESI) by Pearson Performance Solutions (Forcier
et a l , 2001; Jones & Wuebker, 1988), which is a measure of safety consciousness. However, a
common problem with special-purpose inventories is that they are usually administered as a
supplement to other tests, which can lead to motivational distortion (Hakstian et al., 2002). To
avoid this issue, Hakstian et al. derived a trait-based scale from the California Psychological
Inventory (CPI; Gough & Bradley, 1996) on the basis of item-criterion correlations with a self-
report counterproductive behaviours questionnaire. This method produced a reliable and valid
scale that could predict the behaviours of interest.
Measurement of cognitive abilities. Another difficulty in identifying stable relationships
between cognitive abilities, accident involvement and unsafe behaviours derives from the fact
that the bulk of research has focused on motor vehicle accidents, rather than a wider range of
jobs and behaviours. Furthermore, researchers have had mixed results determining the extent to
which cognitive ability test results can be generalized to actual behaviours (Sbordone, 2001;
Sbordone & Guilmette, 1999), and developing tests that can isolate and assess the cognitive
ability of interest (Lawton & Parker, 1998).
Current Directions
In order to move beyond the difficulties in predicting differential accident involvement, a
shift to predicting one of the central precursors to accidents, unsafe behaviours, is needed. Little
research has explored the relationship between personality traits, attention and memory
variables, and unsafe behaviours. Understanding these relationships is a necessary first step to
learning their role in the safety system, as they should be comprehended before the varying
functions of organizational variables are considered. Personality traits should be evaluated with
an unobtrusive measure suitable for applied use. Attention and memory variables should be
appraised with accessible, reliable and established tests of those abilities.
This thesis focuses on everyday unsafe behaviours relevant to a wide variety of
individuals and tasks. This was done in order to account for the wide variability in workplace
safety behaviours, to control for the presence of a safety structure at work, and to be able to
generalize the results to individuals without prior experience in high risk-exposure occupations.
The participants for both studies were university undergraduate students. Although this raises
10
the issue of generalization to the workplace, this sample was preferred at this early stage of
research because responses are less likely to be subject to motivational distortion (Hakstian et al.,
2002), and because the participation time required would have posed logistical difficulties with
industrial samples.
Research Objectives
The overall objective of this research was to explain a person's engagement in unsafe
behaviours using individual-differences variables related to personality traits, and attention and
memory. This knowledge was expected to contribute to the existing accident research literature
and further theoretical understanding of the underlying constructs involved. A broadened
understanding of these relationships can be used to improve job design, personnel selection, job
placement and training decisions, with the intent of reducing injuries and accidents in the
workplace.
A specific research objective was to derive a personality trait-based predictor scale that
predicts unsafe behaviours, and that has suitable psychometric properties for applied industrial
use. It was hypothesized that the predictor scale would correlate negatively with
conscientiousness, internal locus of control, and impulse control, and positively with risk-taking,
extraversion, neuroticism, and agreeableness.
A second research objective related to the relationship between cognitive ability variables
and unsafe behaviours. Unsafe behaviours and differential accident involvement were
hypothesized to correlate positively with sustained, selective/focused, and alternating attention
variables (high scores indicating poor attentional abilities), and to correlate negatively with
memory variables (high scores indicating good memory abilities).
11
Study 1
The purpose of this study was to develop a trait-based scale to predict unsafe behaviours
by measuring safety-oriented tendencies using the BIODATA-250 personality inventory
(Hakstian, 2002). The intent was to produce a scale that had internal consistency, test-retest
reliability, and cross-validity estimates adequate for use in organizations. Other objectives for
this study were to correlate the scale with other psychological constructs related to unsafe
behaviours and to contribute to the existing literature on the personality correlates of unsafe
behaviours.
Method
Participants
Data were collected from 646 undergraduate students at the University of British
Columbia (UBC) from September 2002 to December 2004. Thirteen cases were removed
because participants had too much difficulty completing the measures in the time allotted. Of the
remaining 633 (129 males, 504 females), 408 participated in a two-session design in which 16
did not return for the second session. There were a total of 617 participants with both predictor
and criterion data (128 males and 489 females), and 390 participants with test-retest predictor
data (78 males and 312 females).
In this sample, the median age was 20 years, 56% spoke English as a second language,
and the self-defined ethnic profile included 54% East Asians, 31% Canadians / Western
Europeans / Americans, 6% South Asians, and 9% from other ethnicities.
Operationalization of Unsafe Behaviours Criterion
Unsafe behaviours were measured using the Criterion Hazardous-Behaviour Scale
(Hakstian & Woolley, 1995), which is the second part of the criterion scale used in the article by
Hakstian et al. (2002). Items are reproduced in Appendix A. Participants responded to
12
16 questions on the frequency of their engagement in unsafe behaviours over the last five years,
targeting the constructs of absentmindedness, carelessness, and forgetfulness. Responses were
made on a 6-point scale ranging from "Absolutely never did this or had it happen" to "Did this or
had it happen frequently (more than 4 times)". Scores result from the simple summation of all
16 items; a higher score indicated more frequent engagement in unsafe behaviours.
Assessment of Safety-oriented Tendencies by means of a Personality Inventory
The Biographical Information about Occupationally Descriptive Attitudes, Traits, and
Abilities inventory (BIODATA-250; Hakstian, 2002) is a personality inventory that consists of
250 statements to which participants respond on a 4-point scale ranging from "strongly disagree"
to "strongly agree". The BIODATA-250 inventory yields scores on 31 trait scales (A. R.
Hakstian, personal communication, February 5, 2003).
Procedures
The research sessions were administered by the author and five undergraduate
psychology students following a strict protocol. Experimental sessions were administered to
groups of 1 to 20 persons in one of three rooms in the UBC Psychology Department.
At the start of the session, participants completed a demographics form. The first 227
students participated in one 1.5-hour session where they completed the Criterion Hazardous-
Behaviour Scale, two perceptual speed and accuracy tests (see Study 2), and the BIODATA-250.
The other 406 students participated in two 1-hour sessions separated by a two-week interval.
Session 1 involved one of three procedures — (1) a forward digit span and an associative
memory test (see Study 2), (2) the Cognitive Failures Questionnaire (see Study 2), or (3) no
additional measure— followed by the BIODATA-250. Session 2 consisted of the Criterion
Hazardous-Behaviour Scale followed by the BIODATA-250.
13
Data preparation. Missing item data on the Criterion Hazardous-Behaviour Scale were
not replaced. Missing data on the BIODATA-250 were replaced with a neutral value, and any
multiple responses were averaged. The 16 cases of Session-2 missing data constituted less than
3% of the data set, and were therefore not a concern (Cohen, Cohen, West, & Aiken, 2003).
When appropriate, the following analyses used item-level data and scale scores in zero-centred
form to prevent distortion from gender effects.
Psychometric properties of the Criterion Hazardous-Behaviour Scale. Descriptive
statistics and internal consistency estimates using Cronbach's alpha were computed by gender
for the criterion scale.
Exploratory factor analysis of the Criterion Hazardous-Behaviour Scale. The 16 items
on the criterion scale were factor analyzed to enhance understanding of the underlying
psychological constructs of everyday unsafe behaviours. Preparatory analyses indicated that the
16 items manifested significant gender mean differences (Hotelling's T2 procedure: F(16, 598) =
4.34, p < .001), and that the separate-gender covariance matrices were significantly different
(Bartlett-Box test of homogeneity of dispersion: F(136, 171,204.5) = 130, p < .05). These
results indicated that caution should be applied before pooling the data at the covariance matrix
level, and that separate-gender factor analyses should follow.
In the separate-gender analysis, the standard eigen-decomposition of R, along with the
application of the Scree test, maximum likelihood factor analysis, and transformation of obtained
factors, suggested a two-factor solution for both groups. The separate-gender factor pattern
matrices were judged similar enough to allow pooling of the data. These same procedures were
applied to a pooled sample with equal numbers of males and females. The Scree test provided
the best indication of which factor solution to choose, and again a two-factor solution was used
14
(Unweighted Least Squares—ULS—solutions followed by oblique transformation using direct
oblimin). Subscale scores for these factors were obtained by simple summation of the items that
loaded on similar factors in each of the separate-gender analyses conducted earlier.
Development of the Safety-Orientation predictor scale. Each of the 250 items from the
BIODATA-250 inventory was correlated with the Criterion Hazardous-Behaviour Scale total
scores. To control for capitalization on chance, only the item-criterion correlations that were
significant at or beyond the .001 level were considered further for scale development. Each of
the selected items was also required to demonstrate a conceptually meaningful link with the
criterion scale. Scale scores were obtained by simple summation of the selected items. High
scores indicated a higher tendency towards safety orientation.
Psychometric properties of the Safety-Orientation predictor scale. Descriptive statistics,
internal consistency estimates using Cronbach's alpha, and test-retest reliability estimates were
computed by gender for the Safety-Orientation predictor scale.
Exploratory factor analysis of the Safety-Orientation predictor scale. The 51 items on
the empirically derived Safety-Orientation predictor scale were factor analyzed to explore the
individual-differences predictors of unsafe behaviours and the underlying constructs of this new
predictor scale. Preparatory analyses indicated that the 51 items manifested significant gender
mean differences (Hotelling's T2 procedure: F(5l, 581) = 3.12,/? < .001), and that the separate-
gender covariance matrices were significantly different (Bartlett-Box test of homogeneity of
dispersion: F(l,326, 153,090) = 1.12,p < .005). These results indicated that caution should be
applied before pooling the data at the covariance matrix level, and that separate-gender factor
analyses should follow.
In the separate-gender analysis, the standard eigen-decomposition of R, along with the
application of the Scree test, maximum likelihood factor analysis, and transformation of obtained
15
factors, suggested a six-factor solution for both groups. The separate-gender factor pattern
matrices were judged similar enough to allow pooling of the data. These same procedures were
applied to a pooled sample with equal numbers of males and females. The Scree test provided
the best indication of which factor solution to choose, and again a six-factor solution was used
(ULS solutions followed by oblique transformation using direct oblimin). Factor-scale scores for
these factors were obtained by simple item summation (the terms subscales and factor scales are
equivalent and are used interchangeably).
Evidence of construct validity for the Safety-Orientation predictor scale. To further
understand the underlying constructs of the Safety-Orientation predictor and factor scales, these
scales were correlated with the following BIODATA-250 scales: the Big Five (Emotional
Stability, Extraversion, Openness, Agreeableness, and Conscientiousness), Risk-taking, Impulse
Control, and Internal Locus of Control.
Evidence of criterion-related validity for the Safety-Orientation predictor scale. To
establish the usefulness of the Safety-Orientation predictor scale in predicting unsafe behaviours,
the predictor scale and subscales were correlated with the criterion scale and subscales. These
criterion correlations were then compared with those involving the other BIODATA-250 scales.
Double cross-validation of the Safety-Orientation predictor scale. The criterion-related
validity of the predictor scale was inflated because of capitalization on chance. To evaluate what
the criterion-related validity estimate of the predictor scale would be in a different sample, a
double cross-validation procedure (see, e.g. Hakstian et al., 2002) was used to obtain an average
estimate of the cross-validity. The sample of 633 participants was randomly divided into two
equal subsamples, each with an equal proportion of males and females. Within each of the two
development subsamples, the same scale development procedures were followed as above, but
all items (regardless of their content) yielding a correlation significant at the .005 level were
16
included on each scale. This resulted in two similar but different empirical predictor scales,
Scale 1 and Scale 2. Scale-criterion correlations were then computed for each development
subsample, i.e. Scale 1 in Subsample 1 and Scale 2 in Subsample 2. The scales were then
calculated in each cross-sample to obtain the cross-sample scale-criterion correlations for each
scale, i.e. Scale 1 in Subsample 2 and Scale 2 in Subsample 1. The reduction in the scale-
criterion correlations, and the cross-validity estimates were averaged for the two scales.
Results and Discussion
Criterion Hazardous-Behaviour Scale. As in previous research (Frone, 1998), males
( M = 47.24, SD = 11.95) were found to engage in significantly more unsafe behaviours than
females ( M = 41.61, SD = 11.32; t (613) = 4.93, p < .001; A = .49), justifying the use of zero-
centred scores in the analyses. The reasonably high internal consistency estimate of .77
(weighted by gender) suggested that items were tapping the same constructs.
Exploratory factor analysis of the Criterion Hazardous-Behaviour Scale. The common-
factor analysis of the Criterion Hazardous-Behaviour Scale items produced a clear oblique
primary-factor pattern with two factors; this appears in Table 1. The two common-factors were
named according to their content: (1) Forgetfulness, and (2) Injury-involvement. The subscales
and the items contributing to these subscales are listed in Table 2, along with the internal
consistency estimates.
The construct of forgetfulness is conceptually related to conscientiousness. This factor
arose from the analysis because 7 of the 16 items on the criterion scale included memory-related
content. Unfortunately, there were too few of the other types of items—such as self-control and
focus—to yield other conceptually meaningful factors. The other factor was named Injury-
involvement, indicating that individuals do tend to be differentially involved in injury-related
17
events. The correlation (r = .36) between the two factors confirms that forgetfulness and injury-
related behaviours are positively related.
Psychometric properties of the empirically derived Safety-Orientation predictor scale.
The final Safety-Orientation predictor scale included 51 BIODATA-250 items; these are
included in Table 3. Females (M=128.26, SD = 14.12) were found to have significantly higher
safety-orientation tendencies than males (M= 123.45, SD = 16.67; t (631) = -3.32,p < .001;
A = .33). The high internal consistency estimate of .87 (weighted by gender) suggested that
items were drawing on related constructs. The test-retest reliability estimate over a two-week
interval was .93 (weighted by gender), demonstrating the stability of the scale. Both reliability
estimates indicate that the predictor scale is suitable for use in an industrial setting (Guion,
1998).
Exploratory factor analysis of the Safety-Orientation predictor scale. The common-
factor analysis of the 51 items on the Safety-Orientation predictor scale produced a clear oblique
primary-factor pattern with six factors. The obliquely transformed primary-factor matrix appears
in Table 4, and the intercorrelations among the six common-factors appear in Table 5. The
degree of association among the six factors ranged from .02 to .34, with a mean absolute
correlation of .16.
The subscales and the items that loaded most highly on each of the above factors are
listed in Table 6, along with the internal consistency estimates. The six common factors were
named according to their salient item content:
1. Risk-taking: tendency towards sensation-seeking and dangerous activities;
2. Absentmindedness: tendency towards not focusing and being careless;
3. Assertiveness: tendency towards being outspoken and dominant;
4. Gregariousness: tendency towards being extraverted and impulsive;
18
and being organized; and
integrity.
The presence of the Counterproductive factor suggests that unsafe behaviours are indeed related
to counterproductive behaviours. The construct validity and usefulness of the factor scales is
discussed in the following sections.
Evidence of construct validity for the Safety-Orientation predictor scale. The expected
correlations between the personality constructs and safety-oriented tendencies were found. The
correlations between the BIODATA-250 scales—Big Five, Impulse Control, Risk-taking, and
Internal Locus of Control—and the Safety-Orientation predictor scale and subscales appear in
Table 7. Normative data and reliability estimates for the other BIODATA-250 scales appear in
Appendix B. In general, individuals who were more oriented towards safety tended to be more
conscientious, better able to control their impulses, avoided taking risks, and were less
extraverted and open to change. Since the above scales were all based on the BIODATA-250,
the correlations presented are appreciably inflated (e.g. r's > .50) by common method variance.
Although the scale names are similar and the scales include some of the same items, they are
different and were developed through different means (details are included in Appendix B).
Therefore all the correlations should be interpreted with caution. Superior construct validity
evidence could be obtained by having students complete both the BIODATA-250 along with a
more established personality inventory, such as the CPI or the Revised NEO Personality
Inventory (NEO PI-R; Costa & McCrae, 1992).
One unexpected result was a small significant positive correlation between the Risk-
taking predictor-subscale and the Emotional Stability scale. As the direction expected was
19
negative, one explanation could be that the relationship is truly curvilinear, where individuals
who are either very low or very high in emotional stability both engage in risk-taking behaviours
(see Ulleberg, 2002). Although the relationship is not curvilinear in this restricted sample, such a
relationship in a broader population may explain the change in direction of the correlation, and
would be stronger than the linear correlational results obtained here and in past research.
Assertiveness has not been previously studied in relation to unsafe behaviours. Its
negative relationship with safety-oriented tendencies is likely acting through its association with
high impulsivity or low impulse-control, and high risk-taking. Internal locus of control does not
correlate highly with the six safety-orientation factor scales because the criterion scale did not
include item content on perceptions of control.
Evidence of criterion-related validity for the Safety-Orientation predictor scale. The
criterion-related validity estimates of the Safety-Orientation predictor scale and subscales with
the Criterion Hazardous-Behaviour Scale and subscales appear in Table 8. These validity
estimates should be interpreted with caution as they are inflated because of capitalization on
chance (cross-validity estimates are given and discussed below). The criterion-related validity of
the overall Safety-Orientation predictor scale with the overall Criterion Hazardous-Behaviour
Scale was r = - .48 (p < .001).
The total predictor-scale score was the best predictor of both the total score on the
criterion scale, as well as the Injury-involvement criterion-subscale. The Absentmindedness
predictor-subscale outperformed the total predictor-scale score in predicting the Forgetfulness
criterion-subscale. The most useful predictor subscales correlated with unsafe behaviours were
Risk-taking, Planfulness/Orderliness and Counterproductivity; Assertiveness and Gregariousness
were also significantly correlated with the criteria. These results reflect the salience of the
impulsivity and conscientiousness traits in relation to unsafe behaviours.
20
Similar results were found with the other scales derived from the BIODATA-250:
Extraversion, Conscientiousness, Risk-taking, and Impulse Control. The Safety-Orientation
predictor scale and subscales were more highly correlated with the criteria than any of the other
BIODATA-250 scales. However, this was expected given that the scale was developed using
item-criterion correlations with the Criterion Hazardous-Behaviour Scale.
Double cross-validation of the Safety-Orientation predictor scale. The double cross-
validation procedure provided an estimate of what the criterion-related validity would be if the
scale were applied in a different sample. The same-sample scale-criterion correlations were .51
and .54 for Scales 1 and 2, respectively. These estimates are higher than the .48 obtained for the
Safety-Orientation predictor scale in the total sample, because of capitalization on chance and the
fact that no items were excluded based on content. When Scale 1 and Scale 2 were computed in
their cross-samples, the cross-sample criterion-correlations decreased to .47 (Scale 1 in
Subsample 2) and .42 (Scale 2 in Subsample 1).
The cross-validities are lower bound values of what we would expect the criterion-related
validity to be in a different sample. The average cross-validity reduction was .08, and the
average estimate of cross-validity was .44. These estimates suggest that if the Safety-Orientation
predictor scale were applied to a different sample (but drawn from the same population), we
would expect the criterion-related validity estimate to equal or exceed .44.
Study 2
The aim of this second study was to evaluate the relationship between everyday unsafe
behaviours, and attention and memory-related cognitive ability variables. Furthering knowledge
on the relationship between cognitive-domain variables and unsafe behaviours contributes to
accident research and presents new approaches for minimizing accident-related events. A
21
secondary objective was to explore the relationship between cognitive abilities and personality
traits to further understand the role of individual-difference variables in unsafe behaviours.
Method
Participants
Sample 1: This sample consisted of 408 UBC undergraduate students from Study 1,
88 males and 320 females. The characteristics of this sample were similar to those of the larger
sample (n = 633) used in Study 1. See Study 1 for further details.
Sample 2: Data were collected on 140 UBC undergraduate students from February to
April 2005. Two cases were excluded from the analyses because of too many disruptions during
the study, and another two were removed because the student never returned the criterion scale.
The sample of 136 included 28 males and 108 females, 50% of which spoke English as a second
language. The self-defined ethnic profile included 46% East Asians, 39% Canadians / Western
Europeans / Americans, 5% South Asians, and 10% from other ethnic groups.
Operationalization of Unsafe Behaviours Criterion
Unsafe behaviours were measured using the Criterion Hazardous-Behaviour Scale, as
well as the Forgetfulness and Injury-involvement subscales derived in Study 1 (see Study 1 for
details). Scale scores were the simple sum of all relevant items. High scores indicated more
frequent engagement in unsafe behaviours.
Assessment of Safety-oriented tendencies by means of a personality inventory
Safety-orientated tendencies were measured using the Study-1 Safety-Orientation
predictor scale and six subscales: Risk-taking, Absentmindedness, Assertiveness,
Gregariousness, Planfulness/Orderliness, and Counterproductivity. (The whole BIODATA-250
inventory was given; see Study 1 for more details). Scale scores were the simple sum of all
relevant items, with high scores indicating a higher display of the given personality trait.
22
Relating Unsafe Behaviours to Accident Involvement Differential
The Accident-Related Events Scale was developed by the author (items appear in
Appendix C). Study participants responded to seven questions on the frequency of major and
minor near-accident and accident-related events in the past two years. Responses were made on
a 6-point scale ranging from "Absolutely never did this or had it happen" to "Did this or had it
happen frequently (more than 4 times)". Scale scores were the simple sum of all 7 items; high
scores indicated a higher frequency in accident involvement.
Relating Unsafe Behaviours to Cognitive Failures
The Cognitive Failures Questionnaire (CFQ; Broadbent et al., 1982) was included to
establish a link between unsafe behaviours and past research that focused on cognitive failures.
The CFQ consists of 25 questions on the frequency of cognitive mishaps that occur every day.
Responses were made on a 5-point scale ranging from "never" to "very often". Additional
details about the CFQ appear in Appendix D. CFQ scores were the simple sum of all 25 items;
high scores indicated a higher frequency of cognitive failures.
Cognitive Ability Measures used in Sample 1
The following cognitive ability measures were selected on the basis of their internal
consistency and test-retest reliability estimates, content validity, and accessibility for use in
research. The exception was Part 4 of the Work Skills Assessment, which was a new test
undergoing development.
Work Skills Assessment (WSA), Part 4 (Hakstian, 2004). This test measures perceptual
speed and accuracy, direct visual shifting, as well as focused and sustained attention.
Participants matched 54 words or letter-digit sequences to one of five options following a key
printed at the top of the page. Additional details about this measure appear in Appendix D. The
23
number of correctly completed trials during the prescribed time was recorded; higher values
indicated better accuracy and faster performance.
Comprehensive Ability Battery - Perceptual Speed and Accuracy (CAB-P; Hakstian &
Cattell, 1975). The CAB-P measures perceptual speed and accuracy, as well as focused and
sustained attention. Participants discriminated between 72 pairs of 8-letter and 8-digit
sequences, noting whether they were the same or different. Additional details about this measure
appear in Appendix D. The number correctly completed during the prescribed time was
recorded; higher values indicated better accuracy and faster performance.
Comprehensive Ability Battery - Memory Span (CAB-Ms; Hakstian & Cattell, 1975).
The CAB-Ms measures working memory and attentional capacity (Lezak, 1995). An audiotape
presented ten series of digits, increasing from five to ten digits in length. Participants recalled
the series by writing down their responses after a tone. Additional details about this measure
appear in Appendix D. The number of digits correctly recalled was recorded; higher values
indicated better attention and memory.
Comprehensive Ability Battery - Associative Memory (CAB-Ma; Hakstian & Cattell,
1975). The CAB-Ma measures associative memory and focused attention. Participants were
given a list of 14 symbol-number pairs to memorize, and then matched the symbol with the
correct number from a list of five options. Both parts were completed under timed conditions.
Additional details about this measure appear in Appendix D. The number of correct responses
was recorded; higher values indicated better attention and memory.
Cognitive Ability Measures used in Sample 2
The cognitive ability measures of attention were selected on the basis of their test-retest
reliabilities, their established ability to measure various types of attention, and their accessibility
for research use. Given that cognitive tasks generally draw upon several cognitive abilities or
24
types of attention, most of the tests included assess more than one type of attention (see Lezak,
1995; Spreen & Strauss, 1998).
Colour-WordStroop Test (Stroop, 1935). The Stroop test measures cognitive flexibility,
inhibition of dominant response, and selective/focused attention (Lezak, 1995). The Graf, Uttl,
and Tuoko (1995) version was used. Participants read a list of colour-words printed in black ink
(Part 1), a list of coloured X X X s (Part 2), and a list of colour-words in incongruent colours
(Part 3). Additional details about this measure appear in Appendix D. Reading time, and the
number of corrected and uncorrected errors made on each list were recorded. Only the
interference score (Part-2 time subtracted from Part-3 time, Spreen & Strauss, 1998), and Part-3
uncorrected errors were included as variables. A high score on both indicated lower attention,
more errors, and greater difficulty in inhibiting a dominant response.
Trail Making Tests (TMT; Reitan, 1992). The TMT measures performance speed,
alternating and selective attention, cognitive flexibility, and visual search (Lezak, 1995). In Part
A of the test, participants drew a line joining consecutively numbered circles (1 to 25) randomly
organized on a page. In Part B, participants drew a line alternating between randomly ordered
numbered and lettered circles (i.e. 1-A-2-B... 13-L). Additional details about this measure
appear in Appendix D. The time taken and the number of errors for each part were recorded.
Only the difference score (Part-A time subtracted from Part-B time, Spreen & Strauss, 1998) and
Part-B errors were included as variables. A high score on both indicated poor attention and
greater difficulty in cognitive flexibility.
Cancel H Test (CHT; Graf, 2000). A modified version of the standardized letter
cancellation test by Diller, Ben-Yishay and Gerstman (1974) was used. Letter cancellation tests
measure perceptual speed and accuracy, sustained and selective attention, as well as shifts
between activation and inhibition of quick responses (Lezak, 1995; Ponsford, 2000).
25
Participants scanned 12 rows of capital letters and crossed out all the H's (randomly interspersed
10 times in each row). Additional details about this measure appear in Appendix D. Completion
time, and the number of errors and omissions were recorded. Only completion time and number
of omissions were included as variables; a low completion time indicated faster performance,
and more numerous omissions indicated poor attention.
Digit Symbol Test (DST; Wechsler, 1981). The DST measures sustained and selective
attention, direct visual shifting, response speed, and psychomotor performance (Lezak, 1995).
Following a key printed at the top of the page, participants drew symbols into empty boxes
below numbers. Additional details about this measure appear in Appendix D. The numbers of
correct and incorrect substitutions made during a prescribed time were recorded. Only the
number of correct substitutions was included as a variable; a higher score indicated faster
performance and better visual shifting.
Digit Span Backward Test (DSB; Wechsler, 1981). The DSB test measures working
memory, attentional capacity and focused attention (Lezak, 1995; Spreen & Strauss, 1998). The
experimenter read up to 14 series of digits that increased in length from two to eight digits. The
participant was asked to verbally recall each series in reverse order. Additional details about this
measure appear in Appendix D. The number of series correctly recalled was recorded; a higher
score indicated better attention and working memory.
Test of Variables of Attention (TOVA; Greenberg, 1999). The T O V A is a 22-minute
computerized test that measures selective and sustained attention, as well as impulsivity. Using a
microswitch, participants responded to a target stimulus while ignoring a non-target stimulus.
Additional details about this measure appear in Appendix D. The percentage of omissions (a
measure of inattention), commissions (a measure of impulsivity or disinhibition), reaction time,
and reaction time variability were recorded (Leark, Duypuy, Greenberg, Corman, & Kindschi,
26
1999). Only the percentage of omissions and commissions from test-halves one (infrequent
condition) and two (frequent condition) were included as variables. Higher scores on the
omission percentages indicated attentional errors, while the commission percentages indicated
cognitive errors made due to impulsive behaviour.
Record date-of-completion task (ProM-Date). A modification of the Dobbs and Rule
(1987) behavioural remember-the-date prospective-memory task was used. Participants were
asked to write down the date they completed the take-home questionnaires (see Procedures,
below), and were assigned a score on a 3-point scale based on whether they remembered the task
with or without the aid of various retrieval cues. Additional details about this measure appear in
Appendix D. A higher PRoM-Date score indicated better prospective memory.
Personal information form. This form was given to participants to control for factors that
could potentially influence the interpretation of results. Participants were asked whether they
were taking any medication, had suffered from an injury, or were suffering from an illness or
psychological disorder that could have affected their performance during the session.
Procedures
Sample 1 procedures. A l l participants completed a short demographics form at the start
of the session. The first 227 participants completed one 1.5-hour session involving the Criterion
Hazardous-Behaviour Scale, WSA-Part 4, CAB-P, and BIODATA-250 inventory. The other
182 students participated in two 1-hour sessions separated by a two-week interval. Session 1
included either: (a) the CAB-Ms, CAB-Ma and BIODATA-250 (n = 96); or (b) the CFQ and
BIODATA-250 (n = 85). Session 2 always consisted of the Criterion Hazardous-Behaviour
Scale followed by the BIODATA-250. Additional details are included as part of Study 1.
Sample 2 procedures. Participants completed the same demographics form at the start of
the 1-hour session, and then completed the Stroop test, TMT, CHT, DST and DSB in random
27
order, followed by the TOVA and personal information form. Students were then given the
Criterion Hazardous-Behaviour Scale, Accident-Related Events Scale, and BIODATA-250 to
complete at home, and were verbally reminded to write down the date they completed each
questionnaire. Those who failed to write down the dates were cued for recall when they returned
the questionnaires.
Statistical Analyses
Sample 1 data preparation. See Study 1 for details regarding the Criterion Hazardous-
Behaviours Scale and BIODATA-250. Item-level data that were missing (for various procedural
and non-systematic reasons) on cognitive ability measures were not replaced. Univariate outlier
detection revealed one outlier beyond ±3 SD that was replaced with the nearest non-outlying
value. When appropriate, the analyses used variable and scale-level data in zero-centred form to
prevent distortion due to gender effects.
Sample 2 data preparation. See Study 1 for details regarding the Criterion Hazardous-
Behaviours Scale and BIODATA-250. The same procedures used for the Criterion Hazardous-
Behaviour Scale in Study 1, were also used for the Accident-Related Events Scale for this
sample. Item-level data that were missing on cognitive-ability measures were not replaced.
Eighteen cases were excluded to prevent confounding based on responses in the personal
information questionnaire (n = 14), or for other experimental reasons (n = 4). Univariate outlier
detection on the 13 cognitive ability variables revealed 18 outliers beyond ±3 SD that were
replaced with the nearest non-outlying value. A multivariate outlier analysis revealed three
further outliers that were excluded (Mahalanobis distance statistic: % (13) = 22.36, and p < .05).
Analyses for Sample 2 were based on a sample of 119 (23 males and 96 females), 59 of which
had complete data (12 males and 47 females). When appropriate, the analyses that follow used
variable and scale-level data in zero-centred form to prevent distortion due to gender effects.
28
involvement. The Criterion Hazardous-Behaviour Scale and subscales were correlated with the
Accident-Related Events Scale to evaluate the relationship between unsafe behaviours and
accident-involvement likelihood.
Relationship between Reason's model of human error and unsafe behaviours. The total
scores on the CFQ and the Criterion Hazardous-Behaviours Scale and subscales were correlated
to establish a relationship between cognitive failures and unsafe behaviours.
Exploratory factor analysis of the cognitive ability variables from Sample 2. The
13 cognitive ability variables from Study 2 were factor analyzed in anticipation that the types of
attention common to the measures would be revealed as factors. Preparatory analyses prior to
the factor analysis indicated that the 13 variables yielded a non-significant F-ratio (Hotelling's
T2) for gender differences. There were too few males to compute the Bartlett-Box test of
homogeneity of dispersion matrices, but since the Hotelling's T2 was non-significant, it was
judged safe to pool the data and proceed using pairwise deletion to maximize the sample size.
The standard eigen-decomposition of R, along with the application of the Scree test, maximum
likelihood factor analysis, and transformation of obtained factors, suggested that a two-factor
solution should be used (ULS solutions followed by oblique transformation using direct
oblimin). Subscale scores for these factors were obtained by simple summation of the variables
in z-score form.
Cognitive ability correlates of unsafe behaviours and differential accident involvement.
The cognitive-ability factor scales (Sample 2) and variables (Sample 1) were correlated with the
Criterion Hazardous-Behaviour Scale and subscales (Samples 1 and 2), and the Accident-Related
Events Scale (Sample 2) to explore the relationship between cognitive abilities and the given
criteria.
29
Safety-Orientation predictor scale and subscales were correlated with the cognitive ability
subscales to explore the relationship between these individual-differences precursors of unsafe
behaviours.
Hazardous-Behaviour Scale (r = .51,p < .001), the Forgetfulness criterion-subscale (r = .37,
p < .001), and the Injury-involvement criterion-subscale (r = .43, p < .001). Each of the
correlations with the total criterion scale and Injury-involvement criterion-subscale are inflated
because they share similar accident-related item content and the same response format. The
correlation with the Forgetfulness criterion-subscale is less inflated (method variance is still
operating), and thus more meaningful. These correlations are not a test of the model described in
the introduction, but do suggest that a positive relationship exists between engaging in unsafe
behaviours and a person's level of accident-involvement.
Relationship between Reason's model of human error and unsafe behaviours. The CFQ
was positively correlated with the Criterion Hazardous-Behaviours Scale (r = .39, p < .001) and
the Forgetfulness criterion-subscale (r = .38,p < .001), but not with the Injury-involvement
criterion-subscale. These correlations indicate that, even when accounting for method variance
(both the CFQ and criterion scale used self-report formats and focused on everyday behaviours),
there is indeed a relationship between cognitive failures and unsafe behaviours, particularly
forgetfulness. According to Reason's model of human error (Reason, 1990), lapses in memory
are a fundamental type of human error and have a strong potential to be associated with accident-
related events. This is supported by the Forgetfulness criterion-subscale correlation with the
30
measure of cognitive errors related to memory failure.
Exploratory factor analysis of the cognitive ability variables from Sample 2. The
common-factor analysis of the 13 cognitive-ability variables produced a clear oblique primary-
factor pattern with two factors; this appears in Table 9. The factors were named according to
their content: (1) Cognitive Errors, and (2) Performance Speed. The subscales and the items that
loaded most highly on each of the above factors are listed in Table 10, along with the internal
consistency estimates. Scale scores were computed by simple summation of the salient items.
As the common factors above did not reflect various types of attention, the hypotheses
regarding the relationship between unsafe behaviours and focused/selective, continuous, and
alternating attention could not be tested. The most likely reasons for this outcome are that the
aspects of attention the cognitive ability measures were intended to draw upon were not salient
when placed together, or that these aspects were not accessed by the study participants.
However, the factors that were obtained are meaningful and relevant to unsafe behaviours, and
therefore warrant further exploration.
The Cognitive Errors subscale is composed of variables tabulating different kinds of
errors related to cognitive flexibility, omissions, commissions, and difficulty in inhibiting
dominant responses. This subscale suggests a more global construct of cognitive errors that
should be positively correlated with the CFQ, and unsafe behaviours. This subscale would also
be a substantive improvement over the CFQ because it is based on objective performance
measures of cognitive ability. The Performance Speed subscale includes two items related to
speed of performance and is therefore conceptually clear. This subscale should also be
positively correlated with unsafe behaviours, on the assumption that the monitoring of internal
31
processes is reduced when attentional resources are centred on one task, making other cognitive
processes vulnerable to failures.
involvement. None of the cognitive ability subscales was meaningfully correlated with the
Criterion Hazardous-Behaviour Scale and subscales, or with the Accident-Related Events Scale.
These correlations appear in Table 11. A significant negative correlation was noted between the
Cognitive Errors subscale and Injury-involvement criterion-subscale (r = —.24,p < .05). This
relationship is counterintuitive, though it could be explained by gender differences (this
relationship was only found among females) in the responses to the three injury-related items.
However, the correlation is not likely to be a true representation of the relationship, because the
Cognitive Errors subscale and the Accident-Related Events Scale—a longer and better measure
of accident involvement (a = .79, weighted by gender)—were uncorrected.
According to past research on cognitive failures and accident involvement (Wallace,
Kass, & Stanny, 2002; Wallace & Vodanovich, 2003a), there should be a relationship between
the Cognitive Error subscale and the unsafe-behaviour criteria. The failure to find such a
relationship can be attributed the gap between the broad everyday behaviours measured by the
Criterion Hazardous-Behaviour Scale, and the narrowly focused cognitive ability variables.
Another contributing factor is the range restriction operating on all the variables, reduced
variability decreases the likelihood of obtaining a significant correlation.
The absence of a correlation between the unsafe-behaviour criteria and the Performance
Speed subscale could be attributed to the role of risk-taking. Partial correlations—holding the
Risk-taking predictor-subscale constant—resulted in a positive correlation between the
Performance Speed subscale and both the total criterion (r = .23,p < .05) and the Forgetfulness
criterion-subscale (r = .32, p < .01). This suggests that performance speed was correlated with
32
the part of unsafe behaviours and forgetfulness that are not correlated with risk-taking
tendencies.
Cognitive-ability variable correlates of unsafe behaviours and differential accident
involvement. The Criterion Hazardous-Behaviours Scale and the Accident-Related Events Scale
were uncorrelated with the cognitive ability variables from Sample 1 and the ProM-Date task
from Sample 2. These results appear in the lower portion of Table 11. No relationships were
found because the cognitive ability variables were too narrow in scope to correlate independently
with the broader criteria. A study design that combines similar measures into a more global
variable (as in the Sample 2 procedures) would have a better chance of bridging the criterion and
cognitive ability measures.
The lack of significant correlations with the cognitive ability variables and subscales may
also stem from issues related to the ecological validity of the measures. It is not known to what
extent the measures of cognitive ability relate to actual unsafe behaviours, and little research has
explored the extent to which laboratory conditions can be generalized to a work or home
environment (Sbordone, 2001; Sbordone & Guilmette, 1999). If the measures used in this study
do not generalize to actual behaviours, then it is also difficult to build evidence of a relationship
between the cognitive ability subscales and measures of unsafe behaviours.
The relationship among individual-difference precursors of unsafe behaviours. The
correlations between the Safety-Orientation predictor scale and subscales, and the cognitive
ability subscales appear in Table 12. The Cognitive Errors subscale was negatively correlated
with the Safety-Orientation predictor scale (r = -.30,p < .005), and positively correlated with the
Absentmindedness predictor-subscale (r = .38,p < .001), therefore individuals who make more
cognitive errors are less likely to be safety oriented and more likely to be absentminded. Non­
significant correlations also suggest that individuals who make more cognitive errors tend to be
33
line with past research (Wallace & Vodanovich, 2003b), and support further exploration of the
relationship between risk-taking and cognitive errors. Furthermore, these relationships are not
inflated by method variance, and thus are a better representation of how the constructs relate to
one another than the relationships between the predictor scale and subscales, and the criteria.
The Performance Speed subscale was positively correlated with the Assertiveness
predictor-subscale (r = A6,p < .05). The correlations with the Performance Speed subscale
describe individuals who tend to work more quickly as more assertive and counterproductive.
As performance speed was not correlated with risk-taking, the part of assertiveness more relevant
to performance speed may be impulsivity. The significant correlations obtained for both
cognitive ability subscales indicate that the traits and tendencies assessed by the predictor and
factor scales more closely match the range of tasks assessed by the cognitive ability measures
than the unsafe behaviour criteria.
General Discussion and Conclusions
Contributions to Accident Research
The Safety-Orientation predictor scale is most likely the first scale developed from a
personality inventory that is specifically designed to assess safety-oriented tendencies. This
scale draws on most of the constructs previously related to unsafe behaviours and accident
involvement, and has six subscales—Risk-taking, Absentmindedness, Assertiveness,
Gregariousness, Planfulness/Orderliness, and Counterproductivity. Assertiveness has not been
previously studied in accident research and should be explored further. The predictor scale has
suitable psychometric properties for use in an industrial setting. Future research should use a
34
workplace sample to compare this scale with safety behaviours and accident involvement at
work, while controlling for exposure to risk.
The common-factor analysis of the cognitive ability variables from Study 2 did not reveal
factors related to types of attention; therefore, the hypotheses related to attention remain
untested. The common-factor analysis did however produce two meaningful subscales:
Cognitive Errors (or errors in attention in a very broad sense) and Performance Speed. Neither
of these subscales has been previously studied in relation to unsafe behaviours. The subscales
were not correlated with the criteria, but they were correlated with the trait-based Safety-
Orientation predictor and factor scales. These results suggest that further research should
explore the interrelationships among the individual-differences variables related to unsafe
behaviours. These results also support the use of cognitive ability tests in accident research to
study unsafe behaviours.
In the safety-system approach, individual-differences variables are understood to be
related to unsafe behaviours, and these relationships are moderated by organizational variables.
The present results suggest that there are more complex relationships between personality traits,
cognitive abilities, and unsafe behaviours than previously understood that could be investigated
by exploring the partial relationships between variables. New findings, such as the one found
between performance speed and unsafe behaviours while controlling for risk-taking, could
account for past inconsistencies in accident research.
Results with the Safety-Orientation predictor scale are encouraging but will require
additional testing with industrial samples to obtain more realistic cross-validity estimates and
utility analyses. Following this research, the scale could be used by organizations to reduce
accidents through improved hiring decisions and job placement, helping to ensure a better fit
35
between the individual and position within the company. Continued research with attention and
memory variables, and other cognitive abilities—such as executive functioning—could
eventually contribute to accident prevention through the above interventions, and by improving
job design and safety training programmes.
Limitations and Future Research
There were three major limitations across the two studies in this thesis. The first relates
to the operationalization of unsafe behaviours. The Criterion Hazardous-Behaviour Scale was
effective at tapping relevant everyday unsafe behaviours, but the range of items was restricted,
limited to absentmindedness, carelessness, and injury involvement. A longer scale—including
items such as following rules, taking short-cuts and risks, and past experience with safety-related
situations—would shed more light on the underlying psychological constructs of unsafe
behaviours, and would more closely relate these behaviours to those in the workplace. The
unsafe behaviours assessed could also be more specific, resulting in a smaller gap between them
and the cognitive abilities, and thus a larger overlap in the variance with the cognitive-ability
variables.
Risk exposure was not controlled for in this study and could confound the results,
particularly with respect to accident-related events. Although the criterion scale focuses on
everyday unsafe behaviours, there are some items—such as those related to driving—that
individuals are not equally exposed to. Future research with everyday unsafe behaviours should
control for exposure to risk, as well as ensure that all the questions are both relevant to and
similarly interpreted by all the participants.
An undergraduate student sample was suitable for this stage of the research, but does
limit the extent to which the results can be generalized to other populations. Since the majority
of university students are unlikely to work in a high-risk environment, future research should be
36
based on an industrial sample from relevant occupations, or from students in a trade school. The
age of the sample was not a disadvantage, because it remains important to understand the
characteristics of young workers entering the workplace. However, the gender composition was
dominated by females; future research might focus on males, which are usually considered a
higher-risk group (Frone, 1998), or a more gender-balanced group.
37
References
Arthur, W. J. , Barrett, G. V., & Alexander, R. A. (1991). Prediction of vehicular accident
involvement: A meta-analysis. Human Performance, 4, 89-105.
Arthur, W. J., & Doverspike, D. (2001). Predicting motor vehicle crash involvement from a
personality measure and a driving knowledge test. Journal of Prevention and Intervention
in the Community, 22, 35-42.
Arthur, W. J. , & Graziano, W. G. (1996). The five-factor model, conscientiousness, and driving
accident involvement. Journal of Personality, 64, 593-618.
Barrick, M. B., & Mount, M. K. (1991). The Big Five personality dimensions and job
performance: A meta-analysis. Personnel Psychology, 44, 1-26.
Bradley, G. (1997). Safe people and safe places: Psychological contributions to industrial
accident prevention. Journal of Applied Social Behaviour, 3, 1-14.
Broadbent, D. E., Cooper, P. F., FitzGerald, P., & Parkes, K. R. (1982). The Cognitive Failures
Questionnaire (CFQ) and its correlates. British Journal of Clinical Psychology, 21, 1-16.
Cellar, D. F., Nelson, Z. C , & Yorke, C. M. (2000). The five-factor model and driving behavior:
Personality and involvement in vehicular accidents. Psychological Reports, 86, 454-456.
Cellar, D. F., Nelson, Z. C , Yorke, C. M., & Bauer, C. (2001). The five-factor model and safety
in the workplace: Investigating the relationships between personality and accident
involvement. Journal of Prevention and Intervention in the Community, 22, 43-52.
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation
analysis for the behavioral sciences (3rd ed.).'Mahwah, New Jersey: Lawrence Erlbaum
Associates.
Costa, P. T. J., & McCrae, R. R. (1992). The revised NEO Personality Inventory (NEO-PI-R)
and NEO Five-Factor Inventory (NEO-FFI) professional manual. Odessa, FL:
38
Psychological Assessment Resources.
Diller, L., Ben-Yishay, Y., & Gerstman, L. J. (1974). Studies in cognition and rehabilitation in
hemiplegia (Rehabilitation Monograph No. 50). New York: New York University
Medical Center Institute of Rehabilitation Medicine.
Dobbs, A. R., & Rule, B. G. (1987). Prospective memory and self-reports of memory abilities in
older adults. Canadian Journal of Psychology, 41, 209-222.
Dornheim, M. A. (2000). Crew distractions emerge as new safety focus; to reduce accidents,
researchers are targeting human vulnerability to preoccupation and distraction. Aviation
Week & Space Technology, 153, 58-60.
Edkins, G. D., & Pollock, C. M. (1997). The influence of sustained attention on railway
accidents. Accident Analysis & Prevention, 29, 533-539.
Forcier, B. H., Walters, A. E., Brasher, E. E., & Jones, J. W. (2001). Creating a safer working
environment through psychological assessment: A review of a measure of safety
consciousness. Journal of Prevention and Intervention in the Community, 22, 53-65.
Frone, M. R. (1998). Predictors of work injuries among employed adolescents. Journal of
Applied Psychology, 83, 565-576.
Gough, H. G., & Bradley, P. (1996). California Psychological Inventory manual (3rd ed.). Palo
Alto, CA: Consulting Psychologists Press.
Graf, P. (2000). Cancel H Test (CHT). Unpublished test, University of British Columbia,
Vancouver.
Graf, P., & Uttl, B. (2001). Prospective memory: A new focus for research. Consciousness and
Cognition, 10, 437-450.
Graf, P., Uttl, B., & Tuokko, H. (1995). Color- and picture- word Stroop tests: Performance
changes in old age. Journal of Clinical and Experimental Neuropsychology, 17, 390-415.
Greenberg, L. M. (1999). Test of Variables of Attention (TOVA). Los Alamitos, CA : Universal
Attention Disorders.
Greenwood, M, & Woods, H. M. (1919) A report on the incidence of industrial accidents upon
individuals with special reference to multiple accidents. Reproduced in W. Haddon, E. A.
Suchman, & D. Klein (Eds.), Accident research (1964). New York: Harper & Row.
Guion, R. M. (1998). Assessment, measurement, and prediction for personnel decisions.
Mahwah, New Jersey: Lawrence Erlbaum Associates.
Hakstian, A. R. (2002). Biographical Information about Occupationally Descriptive Attitudes,
Traits, and Abilities (BIODATA-250). Unpublished inventory, University of British
Columbia, Vancouver.
Hakstian, A. R. (2004). Work Skills Assessment (WSA), Part 4. Unpublished test, University of
British Columbia, Vancouver.
Hakstian, A. R., & Cattell, R. B. (1975). The Comprehensive Ability Battery (CAB). Champaign,
IL: Institute for Personality and Ability Testing.
Hakstian, A. R., Farrell, S., & Tweed, R. G. (2002). The assessment of counterproductive
tendencies by means of the California Psychological Inventory. International Journal of
Selection and Assessment, 10, 58-86.
Hakstian, A. R., & Woolley, R. M. (1995). Criterion Hazardous-Behaviour Scale. Unpublished
scale, University of British Columbia, Vancouver.
Hale, A. R., & Glendon, A. I. (1987). Individual behaviour in the control of danger (Vol. 2).
Amsterdam: Elsevier Science.
Hale, A. R., & Hale, M. (1972). A review of the industrial accident research literature. London:
Her Majesty's Stationary Office.
Hansen, C. P. (1988). Personality characteristics of the accident involved employee. Journal of
Business and Psychology, 2, 346-365.
Hansen, C. P. (1989). A causal model of the relationship among accidents, biodata, personality,
and cognitive factors. Journal of Applied Psychology, 74, 81-90.
Hofmann, D. A., & Stetzer, A. (1996). A cross-level investigation of factors influencing unsafe
behaviours and accidents. Personnel Psychology, 49, 307-339.
Iverson, H., & Rundmo, T. (2002). Personality, risky driving and accident involvement among
Norwegian drivers. Personality and Individual Differences, 33, 1251-1263.
Iverson, R. D., & Erwin, P. J. (1997). Predicting occupational injury: The role of affectivity.
Journal of Occupational and Organizational Psychology, 70, 113-128.
Jones, J. W., & Wuebker, L. J. (1988). Accident prevention through personnel selection. Journal
of Business and Psychology, 3, 187-198.
Kerns, K. A. (1996). Walking and chewing gum: The impact of attentional capacity on everyday
activities. In R. J. Sbordone & C. J. Long (Eds.), Ecological validity of
neuropsychological testing (pp. 147-169). Delray Beach, FL: GR Press/St. Lucie Press.
Larson, G. E., & Merritt, C. R. (1991). Can accidents be predicted? An empirical test of the
Cognitive Failures Questionnaire. Applied Psychology: An International Review, 40, 37-
45.
Lawton, R., & Parker, D. (1998). Individual differences in accident liability: A review and
integrative approach. Human Factors, 40, 655-671.
Leark, R. A., Duypuy, T. R., Greenberg, L. M., Corman, C. L., & Kindschi, C. (1999). Test of
Variables of Attention (TOVA) professional manual. Los Alamitos, CA: Universal
Attention Disorder.
41
Lezak, M. D. (1995). Neuropsychological assessment (3rd ed.). New York: Oxford University
Press.
Lund, J., & Aaro, L. E. (2004). Accident Prevention. Presentation of a model placing emphasis
on human, structural and cultural factors. Safety Science, 42, 271-324.
Lund, J., & Hovden, J. (2003). The influence of safety at work on safety at home and during
leisure time. Safety Science, 41, 739-757'.
Magnavita, N., Narda, R., Sani, L., Carbone, A., De Lorenzo, G., & Sacco, A. (1997). Type A
behaviour pattern and traffic accidents. British Journal of Medical Psychology, 70, 103-
107.
Manly, T., Robertson, I. H., Galloway, M., & Hawkins, K. (1999). The absent mind: Further
investigations of sustained attention to response. Neuropsychologia, 37, 661-670.
Martin, M. (1983). Cognitive failure: Everyday and laboratory performance. Bulletin of the
Psychonomic Society, 21, 97-100.
McKenna, F. P. (1983). Accident proneness: A conceptual analysis. Accident Analysis &
Prevention, 15, 65-71.
Neal, A., & Griffin, M. A. (2004). Safety climate and safety at work. In J. Barling & M. R. Frone
(Eds.), The psychology of workplace safety (pp. 15-34). Washington, DC: American
Psychological Association.
Ones, D. S., Viswesvaran, C , & Schmidt, F. L. (1993). Comprehensive meta-analysis of
integrity test validities: Findings and implications for personnel selection and theories of
job performance. Journal of Applied Psychology Monograph,

Recommended