ArticlePDF Available

Measuring the Impacts of Safety Knowledge on Construction Workers’ Attentional Allocation and Hazard Detection Using Remote Eye-Tracking Technology

April 2017

April 2017

Authors:

Sogand Hasanzadeh

Purdue University

Behzad Esmaeili

Purdue University

Michael D Dodd

University of Nebraska at Lincoln

Although several studies have highlighted the importance of attention in reducing the number of injuries in the construction industry, few have attempted to empirically measure the attention of construction workers. One technique that can be used to measure worker attention is eye tracking, which is widely accepted as the most direct and continuous measure of attention because where one looks is highly correlated with where one is focusing his or her attention. Thus, with the fundamental objective of measuring the impacts of safety knowledge (specifically, training, work experience, and injury exposure) on construction workers’ attentional allocation, this study demonstrates the application of eye tracking to the realm of construction safety practices. To achieve this objective, a laboratory experiment was designed in which participants identified safety hazards presented in 35 construction site images ordered randomly, each of which showed multiple hazards varying in safety risk. During the experiment, the eye movements of 27 construction workers were recorded using a head-mounted EyeLink II system. The impact of worker safety knowledge in terms of training, work experience, and injury exposure (independent variables) on eye-tracking metrics (dependent variables) was then assessed by implementing numerous permutation simulations. The results show that tacit safety knowledge acquired from work experience and injury exposure can significantly improve construction workers’ hazard detection and visual search strategies. The results also demonstrate that (1) there is minimal difference, with or without the Occupational Safety and Health Administration 10-h certificate, in workers’ search strategies and attentional patterns while exposed to or seeing hazardous situations; (2) relative to less experienced workers (<5 years), more experienced workers (>10 years) need less processing time and deploy more frequent short fixations on hazardous areas to maintain situational awareness of the environment; and (3) injury exposure significantly impacts a worker’s visual search strategy and attentional allocation. In sum, practical safety knowledge and judgment on a jobsite requires the interaction of both tacit and explicit knowledge gained through work experience, injury exposure, and interactive safety training. This study significantly contributes to the literature by demonstrating the potential application of eye-tracking technology in studying the attentional allocation of construction workers. Regarding practice, the results of the study show that eye tracking can be used to improve worker training and preparedness, which will yield safer working conditions, detect at-risk workers, and improve the effectiveness of safety-training programs.

Examples of scenario images (images courtesy of David Ausmus, with permission)

…

. Impact of Work Experience on Workers' Attentional Allocation across Different Types of Hazards

…

. Impact of Safety Training on Workers' Attentional Allocation across Different Types of Hazards

…

Differences in cognitive process and attentional allocation (images courtesy of David Ausmus, with permission): (a and d) original picture; (b) attentional distribution for the group of workers who had received safety training and fewer than 5 years of experience; (c) attentional distribution for the group of workers who had received safety training and more than 10 years of experience; (e) visual search strategy for a worker who had received safety training and fewer than 5 years of experience; (f) attentional distribution and visual search strategy for the group of workers who had received safety training and more than 10 years of experience

…

. Impact of Being Injured on Workers' Attentional Allocation across Different Types of Hazards

…

Figures - uploaded by Sogand Hasanzadeh

Content may be subject to copyright.

Content uploaded by Sogand Hasanzadeh

Content may be subject to copyright.

Measuring the Impacts of Safety Knowledge on Construction

Workers’Attentional Allocation and Hazard Detection Using

Remote Eye-Tracking Technology

Sogand Hasanzadeh, A.M.ASCE

; Behzad Esmaeili, A.M.ASCE

; and Michael D. Dodd

Abstract: Although several studies have highlighted the importance of attention in reducing the number of injuries in the construction indus-

try, few have attempted to empirically measure the attention of construction workers. One technique that can be used to measure worker atten-

tion is eye tracking, which is widely accepted as the most direct and continuous measure of attention because where one looks is highly

correlated with where one is focusing his or her attention. Thus, with the fundamental objective of measuring the impacts of safety knowledge

(speciﬁcally, training, work experience, and injury exposure) on construction workers’attentional allocation, this study demonstrates the

application of eye tracking to the realm of construction safety practices. To achieve this objective, a laboratory experiment was designed in

which participants identiﬁed safety hazards presented in 35 construction site images ordered randomly, each of which showed multiple hazards

varying in safety risk. During the experiment, the eye movements of 27 construction workers were recorded using a head-mounted EyeLink II

system. The impact of worker safety knowledge in terms of training, work experience, and injury exposure (independent variables) on eye-

tracking metrics (dependent variables) was then assessed by implementing numerous permutation simulations. The results show that tacit

safety knowledge acquired from work experience and injury exposure can signiﬁcantly improve construction workers’hazard detection and

visual search strategies. The results also demonstrate that (1) there is minimal difference, with or without the Occupational Safety and Health

Administration 10-h certiﬁcate, in workers’search strategies and attentional patterns while exposed to or seeing hazardous situations; (2) rela-

tive to less experienced workers (<5 years), more experienced workers (>10 years) need less processing time and deploy more frequent short

ﬁxations on hazardous areas to maintain situational awareness of the environment; and (3) injury exposure signiﬁcantly impacts a worker’s

visual search strategy and attentional allocation. In sum, practical safety knowledge and judgment on a jobsite requires the interaction of both

tacit and explicit knowledge gained through work experience, injury exposure, and interactive safety training. This study signiﬁcantly contrib-

utes to the literature by demonstrating the potential application of eye-tracking technology in studying the attentional allocation of construction

workers. Regarding practice, the results of the study show that eye tracking can be used to improve worker training and preparedness, which

will yield safer working conditions, detect at-risk workers, and improve the effectiveness of safety-training programs. DOI: 10.1061/(ASCE)

Introduction

Human error is a main contributing factor to accidents in the work-

place, especially among construction workers (Sanders and

McCormick 1998;Garrett and Teizer 2009). The main human error

that leads to accidents and injuries on a construction site is a work-

er’s lack of attention when detecting potential or active hazards,

which subsequently results in the worker’s failure to react properly

(Garrett and Teizer 2009;Rozenfeld et al. 2010). Accordingly, the

identiﬁcation of variables that impact a worker’s attention and vis-

ual search strategy can have a transformational impact on construc-

tion safety performance.

To study attention, one needs a reliable means of measuring it.

Because visual cues have a direct impact on attentional allocation,

especially in the very ﬁrst phase of exploration (Hallett 1986), one

scientiﬁc way of studying attention is to detect eye-movement pat-

terns. Eye tracking is widely accepted as the most direct and contin-

uous measure of attention, because where one looks is highly corre-

lated with where one is focusing his or her attention (Shepherd et al.

1986;Hoffman and Subramaniam 1995). This strong relationship

between eye movements and cognitive performance manifests in a

variety of domains and explains why eye tracking has been widely

implemented in the ﬁelds of neuroscience, psychology, and behav-

ioral research (Richardson and Spivey 2004).

Several studies have highlighted the importance of attention in

reducing the number of injuries in the construction industry,

although few have attempted to empirically measure the attention

of construction workers (Garrett and Teizer 2009;Lopez et al.

2010). Because eye tracking exhibits the immense potential of provid-

ing deeper insights into construction workers’hazard-identiﬁcation

patterns, researchers have also recently begun to examine potential

applications of this novel technology in studying the attention and

hazard-identiﬁcation skills of construction workers (Bhoir et al.

2015;Hasanzadeh et al. 2016;Dzeng et al. 2016). Although these

past studies have demonstrated the potential application of eye-

tracking technology in construction safety, they face some limita-

tions. For example, although Bhoir et al. (2015) and Hasanzadeh et

Ph.D. Student, Durham School of Architectural Engineering and

Construction, Univ. of Nebraska–Lincoln, 116 Nebraska Hall, Lincoln,

NE 68588. E-mail: smohammadhasanzadeh@huskers.unl.edu

Assistant Professor, Durham School of Architectural Engineering and

Construction, Univ. of Nebraska–Lincoln, 113 Nebraska Hall, Lincoln,

NE 68588 (corresponding author). E-mail: besmaeili2@unl.edu

Associate Professor, Dept. of Psychology, Univ. of Nebraska–Lincoln,

B82 East Stadium, Lincoln, NE 68588. E-mail: mdodd2@unl.edu

Note. This manuscript was submitted on September 6, 2016; approved

on December 28, 2016; published online on April 12, 2017. Discussion pe-

riod open until September 12, 2017; separate discussions must be submit-

ted for individual papers. This paper is part of the Journal of

Management in Engineering, © ASCE, ISSN 0742-597X.

J. Manage. Eng., 2017, 33(5): 04017024

al. (2016) did investigate the feasibility of using eye-tracking tech-

nology in studying hazard-detection skills of construction workers,

they did not investigate variables that might impact the attentional

allocation of construction workers at construction sites. Dzeng et

al. (2016) studied the impact of work experience on worker atten-

tion; however, their study was limited because stimuli consisted of

virtual images of a limited number of hypothetical construction

scenarios and not real images taken from constructions sites.

Furthermore, the images in Dzeng et al. (2016) did not control for

other variables at play in eye tracking (e.g., color, contrast, viewing

time), which is a choice that could confound the outcomes of the

study. Although these studies manifest the promise of applying eye

tracking to construction safety, there remain ongoing gaps in

knowledge about the variables that impact the attentional alloca-

tion of construction workers. Identifying such factors will facilitate

identifying workers who are at greater risk and will thereby further

improve construction safety.

To respond to these past limitations, it is vital to understand the

three major group factors that affect attention, (1) task parameters,

(2) environmental variables, and (3) individual subject characteris-

tics. The ﬁrst group includes several variables that deﬁne the cogni-

tive load required to implement a task, such as difﬁculty, detectabil-

ity, complexity, and administrative conditions. Task parameters can

also interact with other groups of factors to impact attentional allo-

cation, because environmental and subject characteristics will often

inﬂuence the worker’s experience of the task. The second group

includes environmental factors that can affect attentional allocation

that are either extraneous to the task itself (such as noise, tempera-

ture, and situational variables) or intrinsic to the situation (such as

crowded conditions and time of day). It is important to note that the

effects of noise and other environmental factors on attention vary

across studies due to their respective contexts (Ballard 1996); this

reality marks the need for isolating factors and subfactors as much

as possible when studying attention in safety contexts. Finally, the

third group of factors includes subject characteristics, such as demo-

graphics (e.g., gender, work experience, intelligence variations,

past experience), personality traits, and arousal level (e.g., medica-

tion, fatigue, stress) (Ballard 1996;McBride and Cutting 2015).

These subject traits will often impact workers’safety across days

and times, which make them valuable discussion points in the con-

text of construction safety. Combined, these different factors will

interact to impact construction workers’attentional allocation.

Consequently, to reduce the complexity of studying attention in the

context of construction safety, it is important to isolate speciﬁc

processes of interest while controlling for others.

In the context of this paper, the focus is on one of the individual

subject characteristics that plays a crucial role in detecting a hazard,

assessing a hazard’s risk, and selecting proper action: safety knowl-

edge (Chua and Goh 2004). Knowledge manifests in the cognition

and/or behaviors of individuals (Argote and Miron-Spektor 2011)

and has been divided into two categories: tacit and explicit (Zhang

and He 2016). Tacit safety knowledge can be acquired by learning

on the job, gaining experience, and being exposed to an injury

(Koskinen et al. 2003;Podgórski 2010;Hallowell 2012). These past

experiences and exposures will be stored as patterns of knowledge

in the brain, and when a worker is subsequently exposed to related

situations he or she will have a memory-based intuition to respond

correctly (Zhang and He 2016). On the other hand, explicit knowl-

edge can be delivered through safety training (Aboagye-Nimo et al.

2012) or captured from existing theories in books, safety records,

and guidelines (Hadikusumo and Rowlinson 2004). Regardless of

the type of safety knowledge, there is a consensus in the literature

that safety knowledge can be acquired via training, work

experience, and exposure (Podgórski 2010;Argote and Miron-

Spektor 2011;Aboagye-Nimo et al. 2012). Therefore, it is likely to

play a crucial role in guiding worker attention, and it underpins the

hazard recognition and responsiveness of workers.

To determine the importance of safety knowledge on con-

struction workers’safety, this paper investigates the relation-

ship between safety knowledge, attention, and hazard recogni-

tion. More speciﬁcally, the main objective of this study is to

track workers’eye movements and visual search strategies to

assess the impact of safety knowledge acquired from training,

work experience, and past injury exposure on construction

workers’attentional allocation toward hazards. To control for

other inﬂuences on attention (i.e., task parameters and environ-

mental variables), the research team conducted a laboratory-

based eye-tracking experiment using images taken from real-

life scenarios that manifested a variety of construction hazards.

Consequently, this study empirically establishes the relationship

between worker’s subjective safety knowledge characteristics,

their attention, and their safe behavior, ﬁlling an ongoing gap in

knowledge about the impact of safety knowledge on workers’

cognitive processes, and, speciﬁcally, attention.

Attention

The cognitive process of attention is a vital component of this study

and one that has been advanced by the expansion of eye-tracking

research. Unfortunately, a complete review of eye-tracking research

is beyond the scope of the present paper because eye tracking (1) has

been used in numerous domains, (2) has been used in a variety of

ways, and (3) affords the researcher the ability to choose from doz-

ens of different measures dependent on their critical question of in-

terest. Because of the variety of concepts available when examining

attention via eye tracking, it is important to provide an inclusive

overview of these topics so that the reader has the appropriate con-

text needed to understand this paper’s approach and conclusions.

Cognitive processing comprises how the mind receives, stores,

and uses information. The process of understanding and using in-

formation is highly related to many cognitive processes, includ-

ing attention, perception, memory, language, imagination, and

decision making (Anderson 2005). Before the twentieth century,

the roots of this process traced back to philosophy and physiol-

ogy. However, in the early- to mid-twentieth century, with the de-

velopment of information-processing approaches, the study of

cognition extended beyond psychology and behaviorist perspec-

tives. The ﬁndings of Chomsky’sstudy(Chomsky 1959) revealed

that understanding cognitive processes is critical to understand-

ing and assessing behavior. Cognitive processes select which

pieces of information to attend and which to ignore, and this, in

turn, manifests in human behavior. Because the main objective of

this research is to assess hazard detection and to study the unsafe

attentional distribution of construction workers with various

training and experience levels, it was necessary to measure work-

ers’attention as a primary cognitive process.

Attention, an important element in many cognitive tasks, has

been studied for over a century in many different domains. Deﬁning

attention is difﬁcult because it overlaps with many other cognitive

processes (McBride and Cutting 2015). One early proposed deﬁni-

tion of attention is the focus of consciousness to a particular stimu-

lus while ignoring other distracting objects in the environment

(James 1913). Because humans are ﬁnite beings, their capacity for

information processing is limited, which inﬂuences their attentional

abilities. Moreover, emotional arousal, task difﬁculty, and an

observer’s interest/motivation in a task can signiﬁcantly affect the

J. Manage. Eng., 2017, 33(5): 04017024

capacity of mental resources and attention abilities (Duchowski

2007). Because of the limited capacity of human information proc-

essing as well as the various environmental and individual charac-

teristics at play, this study examines attention in construction safety

according to how workers select to attend to or ignore different

types of hazards while working in complex and dynamic construc-

tion sites.

The authors conducted a comprehensive literature review of

studies related to cognition and attention and found that attention,

as a prominent process of cognition, can be classiﬁed into different

levels and that there are various approaches to studying attention.

Salient results of this review are provided in the following sections.

Levels of Attention

Sohlberg and Mateer (1989) identiﬁed four levels of attention: sus-

tained, selective, divided, and alternating. Sustained attention, or

vigilance, is the ability to devote one’s whole attention to a single

but complex stimulus set over extended periods, without being dis-

tracted, to detect infrequent targets (Parasuraman et al. 1998).

Sustained attention represents a basic attentional function that speci-

ﬁes the efﬁcacy of one’s information-processing capacity and the

higher functions of attention (e.g., selective and divided attention).

Because the eyes need to retrieve informationfrom the visual ﬁeld to

perceive it, selective attention is considered the gateway to con-

sciousness (Theeuwes et al. 2009). The ﬁrst step in the selective pro-

cess is to choose what to attend. Eyes and minds do not have enough

capacity to focus on every object in the ﬁeld; therefore, they must

select what is most important to process. Observers concentrate on a

stimulus using selective attention (focus on only one stimulus) or di-

vided attention (splitting attention between two or more tasks/

objects) (Rehder and Hoffman 2005). Finally, alternating attention

represents one’scognitiveﬂexibility, or the ability to shift attention

among different tasks having different cognitive requirements

sequentially (Sohlberg and Mateer 1989). Originally, the intent of

this study was to assess workers’selective attention, but because

construction is such a complex environment with simultaneous

ongoing activities, the focus extended to assess both selective and di-

vided attention across various hazards in a scene.

Modeling Attention

Generally, visual attention plays a crucial role in the control of eye

movements, perception, learning, memory, and other interactions

with the visual world, although people are not always aware of it

(Bisley 2011). A major distinction that has guided all research in

the area of cognition and information processing is whether atten-

tion is goal driven (top down) or stimulus driven (bottom up). In

goal-driven visual attention, observers form deliberate strategies

and intentions for controlling their attention. The top-down model

is used when the task has a strong inﬂuence on where the observer

attends and looks (Yantis 1998). In contrast, stimulus-driven atten-

tion occurs when salient items in the environment capture attention

independent of the observers’intent. For example, people have

been shown to initially ﬁxate on salient attributes of the image

related to color or contrast, and the initial scanning process is not

necessarily relevant to the observer’s perceptual goals (Todd and

Kramer 1993). More generally, top-down and bottom-up attentional

models interact with each other, and it is assumed that any observer

viewing a scene has a complex combination of expectations and

goals related to processing that scene (Chun and Wolfe 2001).

Although current conceptualizations of visual attention acknowl-

edge the strength of bottom-up processing (Torralba et al. 2006), it

is important to consider top-down inﬂuences in estimating attention

allocation, because in the real world, attention tends to be more task

relevant and goal driven than just free viewing of scenes (Jie and

Clark 2008). Moreover, a recent study by Vecera et al. (2014) has

demonstrated that goal-driven attentional control emerges with

increased experience with a task, whereas with relatively little expe-

rience attentional control tends to be more stimulus driven. In this

study, the research team focused on top-down inﬂuences by design-

ing an experiment in which workers were asked to purposefully

search for hazards in each image. It was anticipated that workers

would attend to different parts of the image with a level of intention-

ality that corresponded to their safety knowledge, so this type of

approach allowed for the study of the impacts of their safety knowl-

edge on their attentional modeling.

Attention and Eye Tracking

At any given time an environment contains much more perceptual

information than can be processed. The previously mentioned

evidence illustrates that what a person sees determines to what

that person is going to attend. If a hazard grabs a worker’s atten-

tion, the worker becomes an active seeker and processor of infor-

mation that is able to interact intelligently with the environment

and consider protective measures. A growing number of psycho-

logical and neuropsychological studies have demonstrated the

close relationship between attention and eye movements (Sun

et al. 2008). In one of the seminal studies, Yarbus (1967)showed

that records of observers’eye movements reﬂect human attention

and thought processes. The simple assumption in most attention

studies using eye tracking is that by tracking someone’seye

movements, one is able to follow the observer’s general path of

attention. This process will reveal both what simply drew the

observer’s attention and what fully captures their attention, which

in turn will give researchers clues about how the person perceived

the scene. Such insights help reveal perhaps one of the most im-

portant functions of attention—to guide ﬁxations toward events

that are relevant to ongoing behavior (Duchowski 2007).

In the last decade, the use of eye tracking has been ﬂourishing

in studies related to human-computer interaction and usability

research (Jacob and Karn 2003). On a somewhat smaller scale, eye-

tracking research also has been applied to transportation (Suh et al.

2006), driving (Palinko et al. 2010), aviation (Lavine et al. 2002),

marketing (Wedel and Pieters 2008), nuclear power control rooms

(Ha and Seong 2009), medicine (Zheng et al. 2011), and petrochem-

ical control rooms (Ikuma et al. 2014). Eye tracking in these ﬁelds

has been used mostly to estimate cognitive load, analyze user

behavior, reveal differences in aptitude and expertise, diagnose neu-

rological disorders, and analyze gaze control.

The existing literature on hazard detection and attentional alloca-

tion using eye-tracking technology has mainly focused on automo-

tive driving. These studies use visual behavior as an index of hazard-

detection skill when drivers are exposed to potential hazards using

static pictures or scenarios (Underwood et al. 2005), a movie-based

simulator (Underwood et al. 2011;Borowsky et al. 2012;Mackenzie

and Harris 2015), or ﬁeld experiment (Sun et al. 2016). Eye tracking

has also been used to study the impact of age (Romoser et al. 2005),

experience (Underwood 2007), and cognitive load (such as cell

phone conversations) (Muttart et al. 2007) on drivers’safety per-

formance. Furthermore, to improve the hazard-detection skills of

drivers, eye-tracking technology has been used to evaluate the effec-

tiveness of interactive training processes (Underwood 2007;Fisher

et al. 2007;Pradhan et al. 2009). Although these studies show the

potential relationship between eye movement, hazard detection, and

safety knowledge, the application of this technology remains

J. Manage. Eng., 2017, 33(5): 04017024

unexplored in the ﬁeld of construction safety. Thus, the study of

attention using eye-tracking technology may elucidate the safety be-

havioral paradigms of workers in dangerous situations on construc-

tion sites.

Eye-Tracking Metrics

Each individual research project requires the choice of eye-tracking

metrics that are relevant to the tasks and appropriate for attentional

allocation and cognitive process analysis. The process used for

identifying ﬁxations and monitoring saccades (the rapid movements

of the eye between ﬁxation points) in an eye-tracking protocol is the

most essential part of eye-movement data analysis (Salvucci and

Goldberg 2000;Duchowski 2007). Minimizing the complexity of

eye-tracking data by identifying ﬁxations and removing raw sac-

cade data can improve researchers’understanding of attentional

allocation and visual processing behavior, because little visual proc-

essing is achieved during saccades, and ﬁndings related to saccades

have not been used in many research applications (Salvucci and

Goldberg 2000). To explore the determinants of ocular behavior,

many derived metrics stemming from these basic measures (ﬁxation

and saccade) have been incorporated in different eye-tracking stud-

ies as dependent variables, including ﬁxation-derived and saccade-

derived metrics.

Areas of Interest

The ﬁrst step in assessing eye-movement parameters and establish-

ing dependent measures is to deﬁne an area of interest (AOI) in

each image used in the study. AOIs are objects and locations of in-

terest within a scene as deﬁned by the research team (Jacob and

Karn 2003). For example, in marketing-related studies, specialists

might be interested in knowing the total time it takes each observer

to view the desired target (brand logo) on a company’s home page

(Goldberg et al. 2002). Further, in the study of Weibel et al. (2012),

to understand dynamic allocation of pilot attention, the AOIs were

ﬂight instruments in the cockpits of commercial airplanes. In the

present study, the AOIs are deﬁned as the different types of hazards

that appeared in the construction scenario images, including fall

hazards, struck-by hazards, and housekeeping hazards.

Fixation-Derived Metrics

Fixation is deﬁned as a relatively stationary eye position with a rela-

tively short minimum duration (often 100–200 ms, although longer

ﬁxations can also be observed as a function of an individual’s proc-

essing goals). Although ﬁxations can be interpreted differently in

various studies depending on their context, ﬁxation-count and gaze-

duration (consecutive ﬁxations) metrics have generally been con-

sidered to assess the depth of cognitive processing and the distribu-

tion of attention (Zhao et al. 2014). For example, Hauland (2003)

indicated that a longer gaze that falls into the speciﬁc AOI before an

incident happens can be used as a situation awareness anticipation

measure. Three commonly used ﬁxation-derived metrics in eye-

tracking studies (Bhoir et al. 2015) are (1) ﬁrst ﬁxation time (the

amount of time, i.e., in milliseconds, that passes following the

image’sﬁrst appearance on the screen until the observer ﬁrst ﬁxates

on an AOI), (2) dwell percentage (relative to the amount of time

spent viewing an image, the proportion of time in which gaze was

ﬁxated on an AOI), and (3) run count (the average number of times

that each participant returns their attention to an AOI).

Saccade-Derived Metrics

Saccades, deﬁned as quick eye movements from one location to

another, vary in duration but often take about 25–150 ms, depend-

ing on the amplitude of the saccade. In most studies, researchers

remove the raw saccade data (Salvucci and Goldberg 2000),

whereas in other studies they measure saccade-derived metrics to

understand the observer’s searching duration and to evaluate the

design of the interface (Fuchs 1971;Pan et al. 2004). For example,

in computer interface and usability studies, larger saccade ampli-

tude indicates a well-designed interface with sufﬁcient cues to ena-

ble users to rapidly ﬁnd the desired targets (Goldberg et al. 2002).

Another study showed that regressive saccades during reading are

associated with comprehension difﬁculties (Just and Carpenter

1980).

Visualization of Eye Movement Data

Eye-tracking provides a large amount of data, and the visualization

of eye-movement patterns can provide insight through a compre-

hensive statistical analysis. Visualization techniques commonly

used for representing eye-tracking data are heat maps and scan

paths (Raschke et al. 2014). A heat map is a two-dimensional visu-

alization in which all ﬁxation values that were analyzed are repre-

sented in colors (Bojko 2009). A heat map can be created for an

individual or for a group of people. In this study, the ﬁxation count

heat map was used to deﬁne AOIs and to better compare visual

attention of construction workers across the scene. On the other

hand, a scan path is a compelling visualization of eye movements

deﬁned as a spatial arrangement of a sequence of saccade-ﬁxation-

saccade (Poole and Ball 2006). It has been widely argued in previ-

ous studies that scan paths reveal considerable information about

visual attention and other underlying cognitive processes involved

in eye movements (Laeng and Teodorescu 2002;Foerster and

Schneider 2013). In this study, in addition to statistical analysis out-

puts, the research team used heat maps and scan paths to study vis-

ual search strategies and cognitive processes of workers in each

scene.

Point of Departure

Construction workers need attention to not only complete their

activities, but also to enrich their conscious awareness of the

entire dynamic environment at a construction site. Training can

make a person phenomenally conscious of certain hazards, but

attention is also critical to detect, perceive, and react to hazards

appropriately. Because attending to something elevates conscious

perception of it (De Brigard 2012), workers’lack of attention

may expose them to hazardous situations more frequently.

Therefore, investigating variables that might impact a worker’s

attention can be of great importance.

This study departs from the current body of knowledge because

it is one of the earliest studies to empirically measure construction

workers’attention using eye-tracking technology. Although some

previous studies have been conducted using eye-movement data to

measure the situation awareness of air-trafﬁc controllers (Hauland

2003), a limited number of studies have measured construction

workers’attention empirically using eye-tracking technology

(Bhoir et al. 2015;Dzeng et al. 2016). Additionally, this project

focused on controlling for other attentional factors (i.e., task param-

eters and environmental variables) to more thoroughly examine the

chosen factor, speciﬁcally safety knowledge acquired from training,

work experience, and past injury exposure. A primary objective of

J. Manage. Eng., 2017, 33(5): 04017024

this study is to investigate how workers with different levels of

work experience, training, and injury exposure (independent varia-

bles) attend to a complex construction scene. Thus, to achieve this

overarching objective, four null hypotheses were deﬁned and tested

(Table 1).

Research Methods

The research hypotheses were tested by collecting information on

two types of variables: independent variables (safety knowledge,

i.e., training, work experience, injury exposure) and dependent

variables (eye-tracking metrics). The independent variable data

were collected via a questionnaire, which, in addition to general

questions related to the participants’background and years of ex-

perience in the construction industry, asked respondents whether

they had received the Occupational Safety and Health Administra-

tion (OSHA) 10-h training (formal) or any on-site safety training

(informal), and whether they had previous injury exposure. The de-

pendent variable data were collected in a laboratory experiment using

eye-tracking equipment, described in the following sections.

Apparatus

An EyeLink II (manufactured by SR Research Ltd., Kanata, ON,

Canada), with a high spatial resolution and a sampling rate of 500

Hz, tracked and recorded the participants’eye movements to deter-

mine where they attended. The EyeLink II is a video-based eye-

tracking system that uses cameras mounted on the headset to docu-

ment the path of a viewer’s focus. Participants completed the

experiment seated approximately 45 cm from the computer screen

on which they observed images (the general procedure can be seen

in the left half of Fig. 1).

Participants

Participants were all construction workers who were invited to the

study in one of the following three ways: (1) an invitation ﬂyer was

posted at construction sites in Lincoln and Omaha, Nebraska; (2)

researchers extended invitations by stopping by construction compa-

nies’main ofﬁces and contacting facility managers at the University

of Nebraska-Lincoln; and (3) a ﬂyer with a one-page summary of

the research project was sent to Associated Builders and Contractors

members and department advisory boards. As a result, a total of 31

Table 1. Tested Null Hypotheses

Hypothesis Tested null hypothesis

Gaining more experience improves workers’attentiveness to

hazards

Null hypothesis 1 (H

): Workers’work experience (years of experience) has no impact on

their attentiveness to hazards on a construction site

Receiving training (formal or informal) improves workers’

attentiveness to hazards

Null hypothesis 2 (H

): Workers’safety training (formal or informal) has no impact on their

attentiveness to hazards on a construction site

Exposure to injury impacts workers’attentiveness to hazards Null hypothesis 3 (H

): Workers’past injury exposure (either personal injury or seeing

somebody else injured) has no impact on their attentiveness to hazards on a construction site

By keeping training as a control variable, gaining more

experience improves trained workers’attentiveness to hazards

Null hypothesis 4 (H

): Work experience (years of experience) of trained workers has no

impact on their attentiveness to hazards on a construction site

•Years of experience

•Safety training

•Past injury exposure

•First fixation time

•Dwell percent

•Fixation count

•Run count

ET Metrics

•Fall hazards

•Ladder hazards

•Struck-by hazards

•Housekeeping

hazards

Areas of Interest

•Heatmap (attentional

distribution)

•Scanpath (cognitive

process)

•Real-time map of

worker eye

movement

Visualization

Statistical analyses

and permutation

simulations

Pool of construction

scenario images

Eye-Tracking Lab Experiment

Professional

safety

managers

focus group

Overt attentional

shift

Controlling

exogenous factors

Goal-driven

attentional

modeling

Experimental Design

Analyses and Evaluation Implications

Data Collection

Real-time

tracking of eye

movements

45cm

Quantitative Analyses

Qualitative Analyses

Contribution to Academia

Contribution to Practice

Develop advanced

safety training

Measure effectiveness

of the training

Identifying hidden

hazards and at-risk

workers

Identify limitations and

potential future

research areas

Characterize variation

in construction

workers’ attention

Verify the role of

cognitive process in

construction safety

Setting up and

calibrating

apparatus

(remote

eye-tracker)

Question:

Can you

identify any

hazard?

Recruit

construction workers

Demographic Survey

35 scenario images

Divided

level of attention

Independent Variables

Fig. 1. Research framework (images courtesyof David Ausmus, with permission)

J. Manage. Eng., 2017, 33(5): 04017024

construction laborers (30 males, 1 female) were recruited for the

study. Because eye tracking is experimental research, the sample

size cannot be as large as it would be for survey-style studies.

However, the size of the sample met the parameters of Pernice and

Nielsen (2009), who stipulated that sample sizes for eye-tracking

studies vary greatly,ranging from 6 for a qualitative study to 30 fora

quantitative study. Given the number of subjects used in this study

and the number of trials performed by each, the sample is robust

enough to match the size recommended for eye-tracking studies.

The participants were mostly young to middle-aged (93% were

between the ages of 20 and 55), 35.5% of the participants had fewer

than 5 years of experience, 25.8% had between 5 and 10 years of ex-

perience, and 38.7% had more than 10 years of experience in the

construction ﬁeld. Sixty percent of the participants had received the

OSHA 10-h training, 16% had received informal or on-site safety

training, and the remaining 24% had not received any safety train-

ing. A total of 69.9% of participants reported that they had been

exposed to an injury on the jobsite. All participants had normal or

corrected-to-normal vision. The experiment was conducted in a sin-

gle 30-min session for each worker. The data for four participants

were omitted from the analysis because acceptable levels of calibra-

tion on the eye tracker could not be achieved. The ﬁnal analyses

were based on the data from the remaining 27 workers. All partici-

pants received gift cards as compensation.

Design and Procedure

The eye-tracking experiment consisted of the presentation of 35

construction site images ordered randomly, each of which showed

multiple hazards varying in safety risk (four examples of different

scenario images are shown in Fig. 2). Construction site images

belonged to different private residential and commercial projects

taken from real construction sites across the United States. Scenario

images covered different types of activities, including but not lim-

ited to site work, rooﬁng, lifting materials, ﬁnishing, erecting struc-

tures, and painting. Each image appeared on the screen for a maxi-

mum of 20 s. The participants donned the eye-tracking headset,

which was calibrated before each session. Subjects were to scan

each image and look for potential hazards. Using a video-game

remote control, they then reported whether or not they found any

hazards by pushing Button A for Yes or Button B for No. The sys-

tem gathered eye-tracking data from the moment the image initially

appeared through to the moment the participant pushed A or B.

Once the participant made a selection, he or she was asked about the

number of hazards identiﬁed, and then the screen changed to reveal

the next image.

In addition to the eye-tracking experiment, participants took a

short survey to collect demographic information, including age,

gender, nationality, years of experience, obtained certiﬁcations,

training, and injury exposure. All procedures were approved by

the University of Nebraska-Lincoln Institutional Review Board.

Defining AOIs

To extract eye-tracking metrics, an AOI or an overlap or near-

overlap of the stimulus (a hazardous situation) and the ﬁxation

points needed to be deﬁned. To identify the AOIs, the research

team ﬁrst studied participants’scan paths and heat maps for each

picture to understand where they were looking. In the second step,

a focus group consisting of ﬁve safety managers independently

reviewed and discussed the original pictures without the heat maps

to identify hazards in each scenario image. All safety mangers

were certiﬁed and had at least 10 years of experience in residential

and commercial projects. The focus group of safety managers was

also asked to express the risks they perceived in each scenario.

Thus, for each construction scenario image, they indicated the

expected frequency of each injury type (categorized based on

injury severity). Overlaying and comparing the results of Steps

One and Two enabled the research team to ﬁnalize the AOIs. In total,

177 hazardous situations (AOIs) in 35 pictures were identiﬁed.

Various types of hazards manifested in the scenario images: lad-

der related, fall to lower level related, fall protection system related,

housekeeping, struck by, caught-in/between, and electrical hazards.

However, due to sample and visibility limitations, the research team

had to remove caught-in/between and electrical hazards from the

analysis. First, there was only a single image in which a caught-in/

between hazard existed, which made the sample too small for this

type of hazard. Concurrently, because the main type of electrical

hazard in the images was powerlines, which are very thin objects,

the task of locating ﬁxations on such small and narrow AOIs proved

very challenging. Accordingly, the research team decided to remove

electrical hazards from the analysis.

Therefore, the research team decided to focus on the following

AOIs: fall to lower level (i.e., a worker is in the proximity of an

unprotected building edge or roof, unguarded roof and ﬂoor open-

ings, scaffolding, skylights); fall-protection systems (i.e., misuse of

lanyard and other fall-protection systems); ladder related (i.e.,

improper use of ladders, such as inappropriate type and length of

ladder, ignoring ladder extension rules, unstabilized ladder, unse-

cured straight ladders, unsafe behavior of workers who are working

on the ladder), struck-by hazards (i.e., having the probability of

being struck by heavy equipment or falling objects like tools, or

collapsing masonry or concrete walls); and housekeeping related

(i.e., slippery conditions of working and walking surfaces, unsafe

material storage, unsanitary conditions of work environment).

These types of hazards are among the most typical safety risks that

lead to accidents.

Analysis Procedures

To test the research hypotheses, the research team separated the par-

ticipants into groups based on the independent variables (training,

work experience, and injury exposure). Inferential statistical

Fig. 2. Examples of scenario images (images courtesy of David

Ausmus, with permission)

J. Manage. Eng., 2017, 33(5): 04017024

analyses illustrated how different levels of these variables impacted

the eye-tracking metrics (dependent variables), which indicated the

attentional distribution of construction workers. To study the impact

of training on eye-tracking metrics, participants were grouped based

on whether they had previously received safety training. Regarding

work experience, 5 and 10 years of experience in the construction

industry were considered the cutoffs that determined low- and high-

experienced workers. In terms of injury exposure, people were clas-

siﬁed based on whether they had been injured and/or had seen

someone else injured on a construction site.

The research team linked the eye movement data of participants

to individual AOIs for each hazard to enable analysis. The choice of

an appropriate eye-tracking metric depends on the research context.

According to the results of a 2015 pilot study conducted by Bhoir

et al. (2015) three ﬁxation-related measures were preferred to serve

as indicators of gaze, because they provide unique insight into vis-

ual attention across AOIs. First, the time of ﬁrst ﬁxation on the

deﬁned AOIs was calculated. This metric reveals which hazards

captured workers’attention more quickly than others (e.g., how

quickly was the AOI ﬁxated). Second, because the time each partic-

ipant spent scanning each picture was different, the gaze (consecu-

tive ﬁxations) on the target AOI was divided by the total duration of

all gazes, yielding the dwell-percentage ratio. The average percent-

age of all workers in the same group indicated that the dwell per-

centage on each AOI was a good indicator of which AOIs (or types

of hazard) captured workers’attention more than others. Finally,

the run-count metric revealed the mean number of times workers

returned their attention to each AOI to further investigate the situa-

tion, indicating the extent to which workers perceived the AOI to be

dangerous. The EyeLink Data Viewer was used to analyze the two-

dimensional eye-movement patterns of participants. After grouping

participants based on independent variables and extracting the eye-

tracking data, the research team faced the challenge of selecting an

appropriate statistical test.

Statistical Method

To compare the average of two or more groups, parametric tests,

such as the t-test and ANOVA, are the most commonly used techni-

ques. To use parametric tests, one needs to satisfy several require-

ments, such as testing a random sample from a population, normality,

and equality of variances (Anderson 2001;Field 2013). However, the

majority of laboratory experiments in psychology and physiology are

not based on a random sample from a population, have a small num-

ber of subjects, and often do not meet the requirement of distribu-

tional assumptions in the parametric test (Ludbrook 1994;LaFleur

and Greevy 2009). To overcome these challenges, statisticians have

suggested the use of randomization techniques (Edgington 1995;

Manly 1997;Ludbrook and Dudley 1998;Gleason 2013).

Randomization provides a reliable alternative test when sample

sizes are small, when the sample distributions are nonnormal (either

because of outliers or skewed data), when the data have mixed dis-

tributions, or when the data were not collected at random (Adams

and Anthony 1996;Eudey et al. 2010). Randomization statistical

methods (bootstrap and permutation simulation) are computer-

intensive techniques using reshufﬂing and resampling to build large

samples (e.g., 1,000 samples) from original data and obtain p-values

based on created distributions (Adams and Anthony 1996;

Anderson 2001). Therefore, randomization statistical techniques

can provide a higher power than other nonparametric techniques,

because randomization statistical techniques use the actual data

rather than ranks, which are used in nonparametric techniques

(Edgington 1995;Adams and Anthony 1996;Ludbrook and Dudley

1998;Drummond and Vowler 2012;Gleason 2013).

The permutation test is a randomization technique that was intro-

duced in the early twentieth century by R. A. Fisher, but it only

recently became more popular and practical due to decreases in

computational costs (LaFleur and Greevy 2009;Gleason 2013).

The fundamental idea behind permutation simulation is to generate

a reference distribution by recalculating data statistics using resam-

pling (Berger 2000). In other words, the permutation tests calculate

the probability of obtaining a value equal or more than the observed

value of a test statistic after randomly shufﬂing data several times

(Anderson 2001;Ernst 2004). Using permutation simulation omits

the randomized resampling bias found in the bootstrapping

approach, which means that, in most cases, the permutation simula-

tion of actual data is more powerful than the bootstrapping approach

(LaFleur and Greevy 2009).

As in other laboratory physiological experiments (Ludbrook

1994;LaFleur and Greevy 2009), this study faced some limitations

that inhibited the application of parametric statistical techniques.

First, it was impractical to randomly select a sample of participants

from a broad population because participants needed to be present

in a laboratory at the University of Nebraska’s campus; thus, the

construction workers who were selected for participation were from

a population of workers near Lincoln and Omaha. Second, the sam-

ple size of the study was small, which impacts the power of the

parametric techniques. Third, there was a mixed-normality across

group distributions. Because parametric statistical methods are sen-

sitive to the violation of assumptions, each of these characteristics

obviated the use of traditional parametric statistical techniques.

In response, the research team decided to use a randomization

technique. Among available randomization techniques, permutation

simulations were used (1) to provide higher power even given the

nonnormality and mixed-normality across group distributions and

the small sample size and (2) to obtain more robust outcomes in the

presence of outliers and missing data. Because the permutation sim-

ulations are based on permutations of actual data, the inﬂuence of

extreme data points is reduced compared with parametric counter-

parts, which are affected considerably by the presence of outliers.

To perform the permutation statistical simulations, the research

team used the Deducer package in Java Graphical User Interface

(GUI) for R 1.7-9 of the open-source statistical package R version

R2.15.0 (R Development Core Team). The permutation simulations

were performed at a level of 95% statistical signiﬁcance (p<0.05)

and 90% moderately statistical signiﬁcance (p<0.1). The results of

these statistical analyses appear in the following sections.

Results and Analysis

Table 2describes the various groupings of participants. After

extracting the eye-movement metrics (run count, dwell percentage,

and ﬁrst ﬁxation time) for each participant on AOIs, the research

team then performed the statistical analyses described earlier.

Because inferential statistics are important, permutation simulations

were run to compare groups’means in terms of eye-movement met-

rics across different types of hazards. Signiﬁcant results at 0.05 and

0.1 alpha levels are discussed in more detail.

Work Experience and Worker Attentiveness

Regarding work experience, 38.7% of the respondents had more

than 10 years of experience, 25.8% had between 5 and 10 years of

experience, and 35.5% had fewer than 5 years of experience in the

construction ﬁeld. To magnify the potential difference between the

groups, the research team compared the group with fewer than 5

years of work experience with the group possessing more than 10

J. Manage. Eng., 2017, 33(5): 04017024

years of experience. The descriptive statistics relating eye move-

ment metrics to work experience are provided in Table 3. Workers

with less experience (Group A

) had lower ﬁrst ﬁxation times on

hazardous areas; however, workers with more experience (Group

) had higher run counts, meaning they returned their attention to

hazardous areas in images more often. In addition, more experi-

enced workers (Group B

) needed less processing time for ladder

and fall-to-lower-level hazards, as demonstrated by their shorter

dwell time percentage; experienced workers (Group B

) also gazed

for a longer duration on fall-protection systems and housekeeping

hazards.

In total, 18 (three eye movement metrics for six types of hazards)

permutation simulations were each run 1,000 times to compare eye-

tracking metrics between groups with different levels of experience

across different types of hazards (signiﬁcant differences are indi-

cated on Table 3). Contrary to the null hypothesis H

, results indi-

cated that work experience has a signiﬁcant impact on worker atten-

tiveness to hazards on a construction site. The results showed that

more experienced workers (Group B

) returned their attention more

frequently to hazardous areas related to fall-protection systems

(Welch’st=−2.862; p= 0.015 <0.05), ladders (Welch’st=

−2.336; p= 0.03 <0.05), struck-by hazards (Welch’st=−3.906;

p= 0.002 <0.05), and housekeeping hazards (Welch’st=−1.845;

p= 0.088 <0.1). In addition, the results indicated that ﬁrst ﬁxation

times for hazardous areas did not differ signiﬁcantly among less

experienced workers (Group A

) and more experienced workers

(Group B

). Moreover, eye-movement patterns of experienced

workers (Group B

) showed that they tended to spend signiﬁcantly

more time examining hazardous areas related to fall-protection sys-

tems in images (Welch’st=−1.740; p= 0.099 <0.1).

Training and Worker Attentiveness

Sixty percent of the participants claimed they had received the

OSHA 10-h training, 16% of the participants had informal or on-

site safety training, and the remaining 24% had no safety training at

all. To test whether having safety training could affect a worker’s

attentional allocation and hazard-detection skills, the participants

were divided into two groups: people who had received safety train-

ing, including the OSHA 10-hour training; informal or on-site

safety training (Group A

); and people who had not received any

safety training at all (Group B

). The descriptive statistics for these

groups’eye movement metrics across AOIs are summarized in

Table 4.

The descriptive analysis revealed that trained workers (Group

) were slower to ﬁxate on hazardous areas compared with work-

ers who had not received any safety training (Group B

). The results

of the run-count descriptive analysis showed that these trained

workers (Group A

) also tended to return their attention to hazard-

ous areas (e.g., ladder, fall-protection systems, struck-by hazards,

housekeeping hazards) more often, excluding falls-to-lower-level

hazards. Moreover, trained workers (Group A

) generally dwelt

less on hazards comparedwith Group B

. However, trained workers

spent more time examining ladder, struck-by, and housekeeping

Table 2. Groupings of Workers Based on Background: Group A versus Group B

Null

hypothesis Group A Group B

) Workers who had fewer than 5 years of work experience (B

) Workers who had more than 10 years of work experience

) Workers who had received safety training (formal or informal) (B

) Workers who had not received any safety training (formal or

informal)

) Workers who had been injured on the jobsite (B

) Workers who had not been injured on the jobsite

) Workers who had seen someone injured on the jobsite (B

) Workers who had not seen someone injured on the jobsite

) Workers who had been injured and/or had seen someone injured

on the jobsite

) Workers who had not been injured and/or had not seen someone

injured on the jobsite

) Workers who had trained and had more than 10 years of

experience

) Workers who had trained and had fewer than 5 years of experience

Table 3. Impact of Work Experience on Workers’Attentional Allocation across Different Types of Hazards

Hazards (AOIs) Group

First fixation time (ms) Dwell percentage Run count

NMean SD MD Mean SD MD Mean SD MD

General A

9 3,196.344 631.912 3,211.060 0.104 0.009 0.105 1.833

0.331 1.805

11 3,671.524 1,082.945 3,767.006 0.010 0.011 0.099 2.235

0.469 2.229

Ladder A

9 3,225.971 775.104 3,326.000 0.107 0.024 0.095 2.111

0.503 1.889

11 3,607.659 1,319.006 3,305.647 0.105 0.018 0.096 2.606

0.695 2.667

Fall to lower level A

9 1,181.896 306.796 1,101.116 0.266 0.047 0.276 3.788 0.886 3.535

11 1,412.959 586.633 1,125.949 0.240 0.050 0.250 4.402 1.006 4.326

Fall-protection system A

9 3,620.371 824.743 3,434.400 0.079

0.008 0.079 1.539

0.353 1.532

11 4,087.963 1,085.257 3,629.250 0.086

0.010 0.085 2.054

0.452 2.170

Struck by A

9 4,890.770 964.985 5,262.824 0.059 0.020 0.066 1.269

0.294 1.269

11 5,189.599 1,965.198 4,920.750 0.071 0.014 0.069 1.969

0.497 1.846

Housekeeping A

9 3,549.130 779.484 3,562.800 0.078 0.015 0.082 1.496

0.327 1.400

11 3,798.890 1,437.093 3,777.000 0.080 0.024 0.086 2.012

0.637 2.200

Note: N= number of subjects in each group; SD = standard deviation in each group; MD = median in each group.

p≤0.05.

p≤0.1.

J. Manage. Eng., 2017, 33(5): 04017024

hazardous areas. The reasons behind these ﬁndings, to be discussed

later, may relate to different affective factors, such as workers’

background and work experience.

To investigate the differences between these groups, 18 series

of permutation simulations were each run 1,000 times. The results

of these simulations meant that the authors could not reject the

null hypothesis H

: safety training (formal or informal) has no

impact on worker attentiveness to hazards on a construction site.

With the resulting p>0.05, it was found that training alone has a

minimal (nonsigniﬁcant) impact on workers’attentiveness to haz-

ards. The only signiﬁcant effect of informal or on-site safety

trainingwasonidentifyingstruck-byhazardsinimages(Welch’s

t= 2.693; p-value = 0.032).

Past Injury Exposure and Worker Attentiveness

Among the 27 participants who had reliable eye-tracking data,

some of them did not answer the survey questions about their past

injury exposure (i.e., four participants did not answer whether they

had been injured and three participants did not answer whether they

had seen someone injured on a jobsite). Of the remaining 23 work-

ers, 13 had been injured and the other 10 stated that they had never

been exposed to any kind of injuries at a jobsite. To better under-

stand the impact of injury exposure on attentional allocation of

workers, the research team divided the participants into the groups

identiﬁed in Table 2. Descriptive statistics of eye movement metrics

across AOIs are presented in Tables 5–7. For further investigation,

72 permutation simulations comparing the previously mentioned

groups were each run 1,000 times.

The descriptive statistics presented in Table 5demonstrate the

impact on workers’attentional allocation and hazard-detection

capabilities of having been or not having been injured (Group A

versus Group B

). Although workers in Group B

ﬁxated more

quickly on most hazardous areas, detailed analysis of the ﬁrst ﬁx-

ation time did not provide any meaningful discrimination

between groups (Group A

versus Group B

). Interestingly, while

run-count analyses showed that workers in Group A

returned

their attention to hazardous areas (all types of hazards) more fre-

quently, they tended to dwell longer than those in Group B

only

on ladder, fall-protection system, and struck-by hazards in the

images.

Table 6depicts differences in eye-movement metrics of workers

who had (Group A

) or had not (Group B

) seen someone else

injured on a jobsite. Workers in Group B

ﬁxated more quickly on

hazardous stimuli than workers in Group A

. However, further

investigation into their eye-movement patterns revealed that

although Group A

workers returned their attention to hazardous

areas more frequently, they did not dwell much on those areas;

Table 4. Impact of Safety Training on Workers’Attentional Allocation across Different Types of Hazards

Hazards (AOIs) Group

First fixation time (ms) Dwell percentage Run count

NMean SD MD Mean SD MD Mean SD MD

General A

19 3,707.500 896.162 3,636.453 0.101 0.009 0.101 2.201 0.459 2.217

5 3,233.559 621.273 2,997.412 0.106 0.010 0.106 2.149 0.401 2.117

Ladder A

19 3,498.433 1,044.360 3,326.000 0.108 0.022 0.096 2.608 0.673 2.667

5 3,305.928 675.298 3,666.250 0.106 0.017 0.106 2.400 0.666 2.722

Fall to lower level A

19 1,385.279 476.604 1,273.861 0.247 0.045 0.250 4.337 1.003 4.326

5 1,087.521 465.293 939.429 0.285 0.049 0.297 4.791 1.314 4.209

Fall-protection system A

19 4,170.635 1,254.556 3,629.250 0.082 0.009 0.082 1.933 0.470 2.064

5 4,065.010 939.264 3,874.703 0.082 0.010 0.079 1.838 0.351 1.936

Struck by A

19 5,606.011 1,829.265 5,328.000 0.068 0.015 0.068 1.767 0.497 1.731

5 4,686.430 805.299 5,149.000 0.056 0.018 0.064 1.315 0.275 1.269

Housekeeping A

19 4,049.956 1,197.663 4,144.000 0.078 0.019 0.079 1.912 0.529 2.000

5 3,669.257 902.123 4,104.800 0.071 0.022 0.059 1.707 0.393 1.733

Note: N= number of subjects in each group; SD = standard deviation in each group; MD = median in each group.

Table 5. Impact of Being Injured on Workers’Attentional Allocation across Different Types of Hazards

Hazards (AOIs) Group

First fixation time (ms) Dwell percentage Run count

NMean SD MD Mean SD MD Mean SD MD

General A

13 3,760.634 797.876 3,876.145 0.101 0.007 0.101 2.469 0.465 2.515

10 3,514.459 841.847 3,310.964 0.101 0.013 0.104 1.989 0.399 1.862

Ladder A

13 3,412.060 1,150.025 3,240 0.107 0.015 0.103 2.944 0.650 3.000

10 3,560.817 681.234 3,425.756 0.103 0.027 0.091 2.267 0.668 1.944

Fall to lower level A

13 1,346.532 446.439 1,289.581 0.245 0.045 0.244 4.911 1.040 4.488

10 1,271.499 520.490 1,107.387 0.254 0.060 0.279 3.928 0.872 3.977

Fall-protection system A

13 4,315.336 1,214.908 4,205.511 0.087 0.011 0.084 2.244 0.579 2.277

10 3,961.475 985.744 3,553.077 0.080 0.008 0.082 1.747 0.432 1.638

Struck by A

13 5,473.008 1,278.817 5,060.909 0.071 0.010 0.069 2.027 0.575 1.769

10 5,575.305 2,127.791 5,344.615 0.060 0.019 0.061 1.446 0.397 1.462

Housekeeping A

13 4,235.941 1,260.845 4,424 0.070 0.018 0.069 2.051 0.527 2.067

10 3,661.517 973.167 3,760.8 0.087 0.016 0.089 1.760 0.546 1.633

Note: N= number of subjects in each group; SD = standard deviation in each group; MD = median in each group.

J. Manage. Eng., 2017, 33(5): 04017024

instead, they distributed their attention across images, excluding

more complex scenes that included struck-by or fall-protection sys-

tem hazards. This result might imply that this group tended to main-

tain situational awareness of the environment by distributing their

attention across scene.

Notwithstanding, the results showed that Group B

(who had

neither been injured themselves nor had seen anyone else injured on

a jobsite) ﬁxated more quickly on hazardous areas than Group A

(who had both been injured and/or seen someone else injured at a

jobsite), but these groups (A

versus B

) did not differ signiﬁcantly

in ﬁrst ﬁxation time on hazards. Of note, Group A

spent less time

examining hazardous areas in images and tended to distribute their

attention across the scene; instead, Group B

tended to dwell on spe-

ciﬁc hazardous areas with the potential cost of failing to identify

other hazards (Table 7).

Next, to determine whether the effect of injury exposure on

eye movements was greater for certain hazard types, the authors

examined the extent to which the characteristics of each group

predicted ﬁrst ﬁxation time, dwell percentage, and run count on

each hazard type. The results of the permutation simulations are

shown in Table 8. Workers who had previous injury exposure dis-

tributed their attention signiﬁcantly differently in all hazardous

situations. Contrary to null hypothesis H

, injury exposure has a

signiﬁcant impact on workers’attentiveness to hazards on a con-

struction site. Workers who had injury exposure tended to return

their attention more frequently to hazardous areas (Welch’st=

−2.846; p=0.015<0.05). Also, workers who personally had

been injured at a jobsite distributed their attention signiﬁcantly

differently across scenes that included struck-by hazards. The

run-count and dwell percentage analyses veriﬁed that these work-

ers more frequently returned their attention to potential or active

struck-by hazards in images (Welch’st=−2.859; p=

0.013 <0.05) and gazed at these hazards for a longer duration

(Welch’st=−1.816; p=0.083<0.1). Comparing Groups A

and

revealed that workers in Group A

spent signiﬁcantly less time

on housekeeping hazards (Welch’st= 2.302; p=0.036<0.05).

Signiﬁcantly, workers in Group A

deployed frequent short ﬁxa-

tions on housekeeping hazardous situations to detect active or

potential hazards without dwelling on one source. The follow-up

investigation will be discussed later.

Work Experience and Worker Attentiveness while

Controlling for Training

Because the results demonstrate that the standard OSHA 10-h train-

ing has a minimal impact (not signiﬁcant) on search strategy, atten-

tional allocation, and the hazard-identiﬁcation skills of construction

workers, it can be concluded that standard OSHA 10-h training

alone cannot be considered a comprehensive tool for improving

workers’attention. Moreover, this study’s eye-movement data

Table 7. Impact of Being Injured and/or Seeing Someone Else Injured on Workers’Attentional Allocation across Different Types of Hazards

Hazards (AOIs) Group

First fixation time (ms) Dwell percentage Run count

NMean SD MD Mean SD MD Mean SD MD

General A

9 3,910.608 870.891 3,997.135 0.100 0.005 0.101 2.576 0.450 2.593

6 3,405.179 714.482 3,310.964 0.103 0.008 0.104 1.899 0.453 1.748

Ladder A

9 3,631.806 1,304.424 3,240.000 0.105 0.015 0.103 3.062 0.563 3.000

6 3,387.897 620.099 3,359.556 0.113 0.032 0.100 2.222 0.771 1.917

Fall to lower level A

9 1,316.858 487.587 1,243.721 0.244 0.045 0.244 5.207 1.046 5.628

6 1,103.080 184.127 1,081.177 0.263 0.044 0.280 3.791 1.033 3.523

Fall-protection system A

9 4,418.435 1,325.734 4,296.690 0.086 0.012 0.084 2.355 0.631 2.298

6 4,008.249 1,182.101 3,592.171 0.079 0.010 0.078 1.628 0.424 1.532

Struck by A

9 5,731.014 1,363.010 5,289.913 0.070 0.009 0.065 2.060 0.557 1.731

6 5,687.108 1,713.703 5,392.161 0.058 0.015 0.061 1.346 0.340 1.270

Housekeeping A

9 4,473.882 1,319.159 4,829.600 0.066 0.016 0.066 2.067 0.435 2.067

6 3,647.123 944.816 3,577.714 0.081 0.016 0.084 1.522 0.472 1.433

Note: N= number of subjects in each group; SD = standard deviation in each group; MD = median in each group.

Table 6. Impact of Seeing Someone Else Injured on Workers’Attentional Allocation across Different Types of Hazards

Hazards (AOIs) Group

First fixation time (ms) Dwell percentage Run count

NMean SD MD Mean SD MD Mean SD MD

General A

13 3,708.054 999.684 3,756.299 0.099 0.010 0.100 2.437 0.454 2.455

11 3,381.456 725.299 3,310.964 0.104 0.008 0.105 1.982 0.466 1.860

Ladder A

13 3,689.773 1,137.103 3,458.4 0.100 0.015 0.095 2.838 0.647 2.944

11 3,095.264 670.915 3,305.647 0.114 0.025 0.106 2.354 0.768 1.944

Fall to lower level A

13 1,380.634 567.458 1,187.905 0.242 0.056 0.244 4.877 1.048 4.488

11 1,199.450 308.056 1,101.116 0.259 0.044 0.275 3.922 0.881 3.767

Fall-protection system A

13 4,256.243 1,174.578 4,205.511 0.085 0.010 0.084 2.223 0.596 2.277

11 3,897.974 1,118.337 3,434.4 0.081 0.010 0.079 1.716 0.456 1.681

Struck by A

13 5,631.502 1,849.055 5,289.913 0.068 0.015 0.065 1.917 0.561 1.731

11 5,160.357 1,543.208 5,149 0.067 0.016 0.068 1.552 0.553 1.269

Housekeeping A

13 4,230.567 1,281.604 4,593.5 0.075 0.021 0.076 2.082 0.432 2.200

11 3,658.894 893.291 3,562.8 0.080 0.017 0.084 1.691 0.600 1.533

Note: N= number of subjects in each group; SD = standard deviation in each group; MD = median in each group.

J. Manage. Eng., 2017, 33(5): 04017024

results supported the idea that in addition to training, years of expe-

rience seem to be a prominent variable in shaping the safety percep-

tions of workers. To verify this ﬁnding, training was considered a

controlling factor, and workers were grouped based on their years

of experience (less than 5 years or more than 10 years). The descrip-

tive statistics relating eye-movement metrics to work experience

while controlling for training are provided in Table 9. Workers with

less experience (Group A

) generally had lower ﬁrst ﬁxation times

on hazardous areas; however, workers with more experience

(Group B

) had higher run counts, meaning they returned their

attention to hazardous areas in images more often.

To investigate the differences between these groups, 18 series of

permutation simulations were each run 1,000 times. Among trained

workers, workers who had more experience (Group B

) distributed

their attention signiﬁcantly differently across scenes and had signiﬁ-

cantly higher run counts across all types of hazards: ladder related

(Welch’st= 4.136; p= 0.003 <0.05), fall to lower level (Welch’st=

2.804; p= 0.023 <0.05), fall-protection system (Welch’st= 4.997;

p=0.002<0.05), struck by (Welch’st= 3.392; p= 0.004 <0.05),

and housekeeping (Welch’st=5.109;p= 0.001 <0.05).

Contrary to null hypothesis H

, the result of the permutation sim-

ulations conﬁrmed that training has minimal inﬂuence on the

attentional allocation and hazard detection of groups and what

differentiates trained workers in safety perception is their years of

experience in the ﬁeld.

Findings and Discussion

Work Experience

The ﬁrst null hypothesis, which aimed to test whether years of work

experience had an impact on workers’attentiveness to hazards on a

construction site, was rejected. The ﬁndings indicated that more

experienced workers (>10 years) returned their attention to hazard-

ous areas more often than less experienced workers. These hazards

may be comprehended by experienced workers as more dangerous

than others or perhaps extensive experience is required with certain

stimuli before one classiﬁes it as a potential hazard. Previous litera-

ture showed that hazard-detection skill can be impacted by safety

experience, i.e., one of the indicators of tacit knowledge (Zhang

et al. 2014;Hadikusumo and Rowlinson 2004), so the outcomes

conﬁrm existing theory.

The results presented in Table 3revealed that when the con-

struction environment includes potential or active sources of haz-

ards, more experienced workers tended to maintain a balance

between processing and searching the scene by spending less

time exploring hazardous areas and by more frequently tracking

back to those hazardous areas. The results of previous studies

(Choudhry and Fang 2008;Chietal.2005) indicated that both

young and less experienced workers are more prone to accidents,

and when these workers gain more experience, they become more

Table 8. Permutation Simulation Results for Workers Who Had Injury Exposure

Hazard Eye-tracking metrics

Group A

versus

Group B

Group A

versus

Group B

Group A

versus

Group B

Welch’stp-value Welch’stp-value Welch’stp-value

General Run count −2.656 0.019

−2.409 0.032

−2.846 0.015

Ladder Run count −2.441 0.029

−1.652 0.103 −2.292 0.047

Fall to lower level Run count −2.463 0.028

−2.425 0.026

−2.588 0.035

Fall-protection system Run count −2.357 0.025

−2.357 0.021

−2.667 0.014

Struck by Run count −2.859 0.013

−1.600 0.130 −3.080 0.009

Dwell percentage −1.816 0.083

−0.069 0.936 −1.752 0.096

Housekeeping Run count −1.288 0.210 −1.802 0.086

−2.258 0.053

Dwell percentage 2.302 0.036

0.702 0.482 1.769 0.113

p≤0.05.

p≤0.1.

Table 9. Impacts of Work Experience on Workers’Attentional Allocation across Different Types of Hazards, with Training as a Controlling Factor

Hazards (AOIs) Group

First fixation time (ms) Dwell percentage Run count

NMean SD MD Mean SD MD Mean SD MD

General A

9 3,931.732 992.628 3,997.135 0.097 0.010 0.098 2.379

0.359 2.320

6 3,044.915 603.285 3,161.181 0.105 0.009 0.106 1.701

0.205 1.748

Ladder A

9 3,933.498 1,219.092 3,442.889 0.100 0.016 0.095 2.852

0.479 2.944

6 3,132.274 866.041 3,057.133 0.108 0.028 0.094 1.917

0.392 1.833

Fall to lower level A

9 1,541.430 570.365 1,403.628 0.228 0.044 0.232 4.579

0.989 4.395

6 1,126.322 ,199.763 1,124.571 0.268 0.050 0.276 3.477

0.523 3.477

Fall-protection system A

9 4,250.217 1144.259 3,958.000 0.086 0.011 0.084 2.175

0.373 2.277

6 3,318.001 425.332 3,422.200 0.079 0.005 0.080 1.411

0.218 1.457

Struck by A

9 5,442.838 2,043.338 4,920.750 0.072 0.014 0.069 2.009

0.459 1.846

6 4,707.313 1,168.006 5,141.417 0.065 0.019 0.070 1.308

0.341 1.231

Housekeeping A

9 3,991.903 1,339.133 3,777.000 0.087 0.019 0.088 2.244

0.408 2.200

6 3,407.124 830.206 3,205.900 0.075 0.013 0.079 1.356

0.266 1.367

Note: N= number of subjects in each group; SD = standard deviation in each group; MD = median in each group.

p≤0.05.

p≤0.1.

J. Manage. Eng., 2017, 33(5): 04017024

aware of safety requirements. Moreover, a study conducted by

Sawacha et al. (1999) showed that there is a strong relationship

between the experience of operatives and their level of safety per-

formance; however, the Sawacha study was based on a question-

naire; thus, the present study takes this investigation a step further

by supporting this claim using empirical data.

First time of ﬁxation was not signiﬁcant for any of the AOIs,

which means that more experienced workers were not faster to ﬁx-

ate on potential hazards; however, these workers were more likely

to return to the AOIs multiple times. Generally, more experienced

workers spent less processing time than less experienced workers,

which is an outcome that agrees with the study by Dzeng et al.

(2016). Similarly, the duration of time that more and less experi-

enced workers spent dwelling on fall-protection system hazards

was signiﬁcantly (moderately) different (p-value = 0.099). The

results presented in Table 3revealed that when the construction

environment includes potential or active sources of fall-protection

system hazards, more experienced workers tended to maintain a

balance between processing and searching the scene by spending

more time exploring fall-protection system hazardous areas and fre-

quently tracked back tothose areas.

In a previous study that used eye tracking to compare the hazard

identiﬁcation of workers, Dzeng et al. (2016) noted that experienced

workers did not perform signiﬁcantly better than novices in identi-

fying hazards. Although their ﬁndings are in contrast with the

results of the current study, it is critical to note the points of depar-

ture between the two studies. First, the authors of the current study

have deﬁned experienced workers as those with more than 10 years

of experience and novices as those with less than 5 years of experi-

ence. Dzeng et al. (2016) qualiﬁed experienced workers as those

with a minimum of 5 years of experience and, more critically, their

novice sample consisted of students, not construction workers, who

had no work experience and never received safety training of any

substantive nature. In other words, Dzeng et al. (2016) compared

workers with experience with students without any experience,

which signiﬁcantly impacted the results of their study.

Comparatively, the central objective in the current study is to study

whether gaining more experience on a construction site improves

workers’hazard-identiﬁcation skills (as manifested by attention and

eye movements); thus, all participants in this study were required to

have experience working on a construction site. All of the partici-

pants in the current study are professional construction workers,

and the authors took rigorous steps to divide them into two distinct

groups.

Second, whereas the Dzeng et al. (2016) stimuli consisted of a

limited number (only four) of virtual images of hypothetical con-

struction scenarios, the current study’s 35 images were taken from

real construction sites and represented a wide range of common

hazards. In addition, there were at least 20 AOIs for each type of

hazard across the images, which helped to obtain more reliable out-

comes by measuring the average performance of workers in identi-

fying different types of hazards. As such, although the study Dzeng

et al. (2016) conducted is an important step in applying eye-tracking

technology to a construction setting, the present study considerably

expands the scope of what can be studied via eye tracking in con-

struction safety.

The ﬁndings of this study can further be compared with the

results of other previous eye-tracking studies. Underwood (2007)

suggested that the efﬁciency of visual search strategies relates to the

changes in driving skills that mark the transition from novice to

experienced drivers. Many other studies have provided support for

the notion that driving experience is a key predictor of crash rates,

with young novice drivers being at a greater risk (Chapman and

Underwood 1998;Konstantopoulos et al. 2010). Similarly, learning

by doing and gaining experience can be one of the best training

options for construction workers (Choudhry and Fang 2008). This

study shows that as construction workers gain experience, their

hazard-detection skills improve, enabling them to search and

examine scenes more intuitively. Thus, it can be concluded that

gaining experience in construction is strongly related to effective

and efﬁcient visual search and attentional distribution.

Safety Training

The second null hypothesis of the study stated that past safety train-

ing (formal or informal) would have minimal (nonsigniﬁcant)

impact on workers’attentiveness to hazards on a construction site.

The hypothesis could not be rejected for formal training. The results

showed no signiﬁcant differences in the search strategies or atten-

tional patterns of workers with or without the OSHA 10-h certiﬁcate

training when they were exposed to hazardous situations. The ﬁnd-

ings of this eye-tracking experiment can be alarming for construc-

tion safety professionals, because it implies that the most common

and basic training (i.e., OSHA 10-h certiﬁcate) may not consider-

ably improve hazard-detection skills. Although the results do not

state that the OSHA 10-h certiﬁcate is ineffective, they do suggest

that developing more innovative and interactive training techniques

can improve workers’hazard-detection skills [e.g., using OSHA

visual inspection training, which is interactive and game-based

training; using a three-sided virtual reality, such as cave automatic

virtual environment Perlman et al. (2014)].

Regarding informal training, however, the second null hypothe-

sis was rejected for run counton struck-by hazards. The results indi-

cated that the attentional pattern of workers who have had informal

training (i.e., through on-site training or participation in safety

meetings) was signiﬁcantly different for struck-by hazardous situa-

tions (Welch’st= 2.693; p-value = 0.032). To study this ﬁnding fur-

ther, attentional distributions (heat maps) of two workers, one who

had informal training and the other who had no safety training, were

created. As the heat map comparison shows (Fig. 3), the worker

with informal training appears more attentive to struck-by hazards

and ﬁxates on areas that show potential for struck-by accidents.

These workers distributed their attention across the scene and

scanned areas close to a potential hazard, checking the proximity of

equipment to workers and other equipment in the scene.

Past Injury Exposure

The third null hypothesis, which tested whether workers’past injury

exposure affected their attentional patterns, was also rejected. The

results of the analysis demonstrated that injury exposure signiﬁ-

cantly impacts the cognitive processes of those who had previously

been exposed to speciﬁc hazards (Table 8). The test subjects in this

group behaved more conservatively and tended to check the sur-

rounding environment frequently with short ﬁxations, seeking

potential and active sources of hazards. This ﬁnding conforms with

some previous studies (Törner and Pousette 2009;Shin et al. 2014),

which found that past injury exposure increases workers’risk

awareness. It was interesting that workers who had been injured

before had no signiﬁcant differences in attentional allocation to

housekeeping hazardscompared with those who had no injury expo-

sure. The reason behind this ﬁnding may correlate with the type(s)

of hazards they had been exposed to before. According to workers’

self-reports, none of the workers had experienced injury with house-

keeping or a cluttered environment as sources of hazard.

Workers who had past injury exposure became more conserva-

tive and careful regarding struck-by hazards in a dynamic

J. Manage. Eng., 2017, 33(5): 04017024

construction environment. Jacob and Karn (2003) showed that a

higher dwell percentage on particular elements is a sign of that ele-

ment’s importance for that group, leading to the conclusion that

workers with past injury exposure would dwell longer and return

their attention more often to comprehend the complexity of a

dynamic hazardous situation (struck-by hazard) than workers with

no injury exposure. The ﬁndings also indicated that workers who

had been injured or had witnessed another’s injury tended to not

dwell on speciﬁc hazards and distributed their attention over the

scene while frequently returning their attention to better compre-

hend all dimensions of a hazardous area. On the other hand, the

mean dwell time percentages for workers without any injury expo-

sure experience were longer, and they tended to focus their attention

on speciﬁc hazards, possibly misidentifying other active hazards or

failing to anticipate potential hazards on the construction site.

Furthermore, grouping workers based on their past experience

of seeing another worker injured provided additional insight into

the cognitive processes of construction workers. Having seen

another worker injured impacted the attentional allocation of the

current participants. These workers appeared to become more con-

cerned about their safety and health, and when they faced hazardous

situations, they tended to maintain situational awareness. Although

the workers who had witnessed someone else injured frequently

checked AOIs related to housekeeping (p-value = 0.086) and ladder

hazards (p-value = 0.103) (moderately signiﬁcant), their run count

to fall hazards was more signiﬁcant (p-value = 0.026). Self-reports

of workers revealed that approximately 70% of them had experi-

enced witnessing another worker injured due to fall incidents, such

as falling from scaffold, falling from an unprotected edge, or falling

from height due to broken cable. Scan paths for a representative

worker of each group veriﬁed the statistical ﬁndings that workers

who saw another worker falling from height exhibited differential

visual behaviors and followed a different search strategy (Fig. 4).

Moreover, the run count associated with struck-by–related hazards

was not signiﬁcantly different between these groups. Returning to

the workers’self-reports of their injury experience also conﬁrmed

that none of the workers acknowledged seeing another worker being

injured due to struck-by accidents.

Years of Experience while Controlling for

Training Impacts

The fourth null hypothesis, which aimed to test whether years of

work experience of trained workers had an impact on their atten-

tiveness to hazards on a construction site, was rejected. The results

suggested that as trained workers gain more experience in the con-

struction ﬁeld, they will be more aware of safety requirements on

jobsites and of the full extent of all potentially hazardous items/sit-

uations in the environment. This outcome is consistent with previ-

ous studies in which workers become better at detecting hazards

and behaving safely by gaining experience and being exposed to

hazardous situations (tacit knowledge) rather than by receiving for-

mal training with traditional educational methods (explicit knowl-

edge) (Podgórski 2010).

For example, as seen in Figs. 5(a–c), the less experienced worker

is more stimulus driven and concentrated on imminent hazards,

Fig. 3. Attentional distributions (heat maps) (images courtesy of David Ausmus, with permission): (a) original picture; (b) worker who had informal

safety training; (c) worker who had no safetytraining

Fig. 4. Difference in cognitive process (search strategy) (images courtesy of David Ausmus, with permission): (a) original picture; (b) worker who

had not seen someone injured due to fall hazards; (c) worker who hadseen someone injured due to fall hazards

J. Manage. Eng., 2017, 33(5): 04017024

such as the danger to the man standing on the scaffold, but the more

experienced workers showed a balance in focusing and dividing

their attention across the scene and are more goal driven. To better

understand the difference between the search strategies and cogni-

tive processes of these groups, scan paths were obtained from one

worker in each group. Scan paths suggest that experienced workers

optimized their attention by distributing it across the scene and scru-

tinized more areas to maintain their awareness about potential or

active sources of hazards in scene. The study of Aboagye-Nimo et

al. (2012) points out that “explicit knowledge […] will not be sufﬁ-

cient in practice in order to prevent accidents”(p. 419). They high-

lighted the important role of tacit knowledge, namely work experi-

ence, in learning safe work practices. In sum, these ﬁndings align

with the study of McBride and Cutting (2015), who veriﬁed that as

a result of years of experience and training, the controlled atten-

tional allocation for hazard detection in less experienced workers

turned into an automatic attentional allocation in more experienced

workers.

Conclusions

One of the root causes of accidents is human error (Abdelhamid and

Everett 2000), i.e., the lack of attention and the failure of workers to

identify hazards. If workers are distracted or lack attentiveness,

then they cannot identify and properly respond to a hazard. Because

there is a direct link between visual cues and attentional allocation,

one of the scientiﬁc methods of studying attention is tracking the

eye movements of people. Although eye tracking has been exten-

sively used to study attention in other domains of science

(Duchowski 2007), its applications remain relatively unexplored in

the ﬁeld of construction safety.

To address this knowledge gap, the research team used eye-track-

ing technology to measure the impact of safety knowledge (in terms

of training, work experience, and injury exposure) on construction

workers’attentional allocation. It was found that although work ex-

perience and injury exposure signiﬁcantly impact visual search strat-

egies and attentional allocation, the difference between workers with

and without the OSHA 10-h certiﬁcate is not signiﬁcant. One should

note that the results do not state that the OSHA 10-h certiﬁcate is

ineffective, but it calls for developing more innovative training tech-

niques, such as interactive and game-based training, which can

improve workers’hazard-detection skills. The results indicate that

by integrating both tacit knowledge (work experience and injury ex-

posure) and explicit knowledge (e.g., interactive training), future

endeavors to enhance worker safety knowledge will achieve supe-

rior outcomes in terms of worker safety awareness.

This study’s results increase the ﬁeld’s understanding of the

variables that impact attentional allocation and provide a novel

approach for improving construction site safety by using eye-

tracking technologies that have been widely accepted as the most

direct and continuous measure of attention. The results of the study

reported in this paper make a signiﬁcant contribution to the existing

literature. First, the eye-tracking metrics that characterize the varia-

tion in construction workers’attention while they search for hazards

were identiﬁed. Second, given the established link between eye

Fig. 5. Differences in cognitive process and attentional allocation (images courtesy of David Ausmus, with permission): (a and d) original picture;

(b) attentional distribution for the group of workers who had received safety training and fewer than 5 years of experience; (c) attentional distribution

for the group of workers who had received safety training and more than 10 years of experience; (e) visual search strategy for a worker who had

received safety training and fewer than 5 years of experience; (f) attentional distribution and visual search strategy for the group of workers who had

received safety training and more than 10 years of experience

J. Manage. Eng., 2017, 33(5): 04017024

movements and cognitive processes, the ﬁndings of this study lay

the foundation for using eye-tracking technologies to further study

the role of cognition inconstruction safety. The previously described

approach can be used by other researchers to advance knowledge

regarding the way that attentional allocation of construction workers

impacts the likelihood of accidents and can subsequently lead to the

development of new accident-causation theories.

The results also make a signiﬁcant contribution to practice.

Because of these results, one can conclude that traditional teaching

methods (e.g., lectures, media) for construction worker education

and training do not improve workers’visual search skills sufﬁ-

ciently to identify hazards. More effective interactive methods

should be applied instead of conventional teacher-student training.

For instance, safety knowledge could be more effectively acquired

using interactive and experiential learning methods. The need to

ﬁnd more effective training methods becomes more urgent in the

face of the fact that 185,000 new workers are predicted to join con-

struction in this decade (NCCER 2013). Eye-tracking technology

can provide a viable solution for this challenge by providing real-

time tracking of the eye movements of construction workers to mea-

sure the effectiveness of their training (via pretest and posttest mon-

itoring). Eye tracking can further be used for innumerable purposes

inclined toward achieving safer working conditions, such as identi-

fying hidden hazards, measuring the situation awareness of work-

ers, and improving the effectiveness of safety-training programs.

This study has some limitations and areas that future work can

address. First, the research team examined the gaze exhibited toward

static images of construction sites rather than dynamic real-world

construction sites. One might argue that since the images are merely

pictures ona screen,they do not translateto construction sites; work-

ers may act differently in interactions with real-world construction

sites. To completely address this issue, future research may examine

the workers’gazes and assess their hazard-detection skills on real-

world construction sites using a mobile eye tracker. Another limita-

tion of this study relates to sampling and the number of participants.

Because of practical limitations, the research team only recruited

workers located near Lincoln and Omaha. In addition, although the

sample of the present study was relatively large for an eye-tracking

study, the practical limitations of recruiting construction workers to

participate in the experiment in a laboratory setting restricted the

number of participants. Future studies might address this limitation

by replicating the study in other geographical regions and also

increasing the number of participants.

Acknowledgments

The research team thanks the University of Nebraska-Lincoln for

supporting the research reported in this paper through a Research

Council Faculty Seed Grant. The research team also thanks

workers who participated in this study. Appreciation is extended

to Mr. David W. Ausmus (image photographer), Mr. Ronald D.

Allen, and other professional safety managers for their

considerable and vital support in this study. Any opinions,

ﬁndings, and conclusions or recommendations expressed in this

material are those of the writers and do not necessarily reﬂect the

views of the University of Nebraska-Lincoln.

References

Abdelhamid, T. S., and Everett, J. G. (2000). “Identifying root causes

of construction accidents.”J. Constr. Eng. Manage.,10.1061

/(ASCE)0733-9364(2000)126:1(52),52–60.

Aboagye-Nimo, E., Raiden, A., Tietze, S., and King, A. (2012). “The use of

experience and situated knowledge in ensuring safety among workers of

small construction ﬁrms.”Proc., 28th Annual ARCOM Conf.,S.D.

Smith, ed., Association of Researchers in Construction Management,

Reading, U.K., 413–422.

Adams, D. C., and Anthony, C. D. (1996). “Using randomization techniques

to analyse behavioural data.”Animal Behav, 51(4), 733–738.

Anderson, J. R. (2005). Cognitive psychology and its implications,

Macmillan, New York.

Anderson, M. J. (2001). “Permutation tests for univariate or multivariate

analysis of variance and regression.”Can. J. Fish. Aquat. Sci., 58(3),

626–639.

Argote, L., and Miron-Spektor, E. (2011). “Organizational learning: From

experience to knowledge.”Org. Sci., 22(5), 1123–1137.

Ballard, J. C. (1996). “Computerized assessment of sustained attention: A

review of factors affecting vigilance performance.”J. Clin. Exp.

Neuropsychol., 18(6), 843–863.

Berger, V. W. (2000). “Pros and cons of permutation tests in clinical trials.”

Stat. Med., 19(10), 1319–1328.

Bhoir, S. A., Hasanzadeh, S., Esmaeili, B., Dodd, M. D., and Fardhosseini,

M. S. (2015). “Measuring construction workers’attention using eye-

tracking technology.”Proc., ICSC15: The Canadian Society for Civil

Engineering 5th Int./11th Construction Specialty Conf., Canadian

Society for Civil Engineering, Montreal, QC, Canada.

Bisley, J. W. (2011). “The neural basis of visual attention.”J. Physiol.,

589(1), 49–57.

Bojko, A. A. (2009). “Informative or misleading? Heat maps decon-

structed.”Human-computer interaction. New trends, J. A. Jacko, ed.,

Springer, Berlin, 30–39.

Borowsky, A., Oron-Gilad, T., Meir, A., and Parmet, Y. (2012). “Drivers’

perception of vulnerable road users: A hazard perception approach.”

Accid. Anal, Prev., 44(1), 160–166.

Chapman, P. R., and Underwood, G. (1998). “Visual search of driving situa-

tions: Danger and experience.”Perception, 27(8), 951–964.

Chi, C. F., Chang, T. C., and Ting, H. I. (2005). “Accident patterns and pre-

vention measures for fatal occupational falls in the construction indus-

try.”Appl. Ergon., 36(4), 391–400.

Chomsky, N. (1959). “Review: Verbal behavior by B. F. Skinner.”

Language, 35(1), 26–58.

Choudhry, R. M., and Fang, D. (2008). “Why operatives engage in unsafe

work behavior: Investigating factors on construction sites.”Saf. Sci.,

46(4), 566–584.

Chua, D., and Goh, Y. (2004). “Incident causation model for improving

feedback of safety knowledge.”J. Constr. Eng. Manage.,10.1061

/(ASCE)0733-9364(2004)130:4(542), 542–551.

Chun, M. M., and Wolfe, J. M. (2001). “Visual attention.”Blackwell hand-

book of sensation and perception, E. B. Goldstein, ed., Blackwell,

Malden, MA.

De Brigard, F. (2012). “The role of attention in conscious recollection.”

Front. Psychol.,3,1–10.

Drummond, G. B., and Vowler, S. L. (2012). “Different tests for a differ-

ence: How do we do research?”Adv. Physiol. Educ., 36(1), 3–5.

Duchowski, A. (2007). Eye tracking methodology: Theory and practice,

Springer Science and Business Media, Dordrecht, Netherlands.

Dzeng, R. J., Lin, C. T., and Fang, Y. C. (2016). “Using eye-tracker to com-

pare search patterns between experienced and novice workers for site

hazard identiﬁcation.”Saf. Sci., 82(Feb), 56–67.

Edgington, E. S. (1995). Randomization tests, 3rd Ed., Marcel-Dekker,

New York.

Ernst, M. D. (2004). “Permutation methods: A basis for exact inference.”

Stat. Sci., 19(4), 676–685.

Eudey, T., Kerr, J., and Trumbo, B. (2010). “Using R to simulate permuta-

tion distributions for some elementary experimental designs.”J. Stat.

Educ., 18(1), 1–30.

Field, A. (2013). Discovering statistics using IBM SPSS statistics, 4th Ed.,

Sage, London.

Fisher, D., Pradhan, A., Pollatsek, A., and Knodler Jr., M. (2007).

“Empirical evaluation of hazard anticipation behaviors in the ﬁeld and

on driving simulator using eye tracker.”Transportation Research

Record, 2018, 80–86.

J. Manage. Eng., 2017, 33(5): 04017024

Foerster, R. M., and Schneider, W. X. (2013). “Functionally sequenced

scanpath similarity method (FuncSim): Comparing and evaluating scan-

path similarity based on a task’s inherent sequence of functional (action)

units.”J. Eye Mov. Res., 6(5), 1–22.

Fuchs, A. F. (1971). “The saccadic system.”The control of eye movements,

Academic, New York, 343–362.

Garrett, J., and Teizer, J. (2009). “Human factors analysis classiﬁcation

system relating to human error awareness taxonomy in construction

safety.”J. Constr. Eng. Manage.,10.1061/(ASCE)CO.1943-7862

.0000034, 754–763.

Gleason, J. (2013). “Comparative power of the ANOVA, randomization

ANOVA, and Kruskal-Wallis test.”Ph.D. dissertation, Wayne State

Univ., Detroit.

Goldberg, J. H., Stimson, M. J., Lewenstein, M., Scott, N., and Wichansky,

A. M. (2002). “Eye tracking in web search tasks: Design implications.”

Proc., 2002 Symp. on Eye Tracking Research and Applications,

Association forComputing Machinery, New York, 51–58.

Ha, J. S., and Seong, P. H. (2009). “A human–machine interface evaluation

method: A difﬁculty evaluation method in information searching

(DEMIS).”Reliab. Eng. Syst. Saf., 94(10), 1557–1567.

Hadikusumo, B., and Rowlinson, S. (2004). “Capturing safety knowl-

edge using design-for-safety-process tool.”J. Constr. Eng. Manage.,

10.1061/(ASCE)0733-9364(2004)130:2(281), 281–289.

Hallett, P. (1986). “Eye movements (and human visual perception).”

Handbook of perception and human performance,Vol.1,Wiley,

New York.

Hallowell, M. R. (2012). “Safety-knowledge management in American

construction organizations.”J. Manage. Eng.,10.1061/(ASCE)ME

.1943-5479.0000067, 28(2), 203–211.

Hasanzadeh, S., Esmaeili, B., and Dodd, M. (2016). “Measuring construc-

tion workers’real-time situation awareness using mobile eye-tracking.”

Construction Research Congress 2016, ASCE, Reston, VA, 2894–

2904.

Hauland, G. (2003). “Measuring team situation awareness by means of eye

movement data.”Proc., HCI Int., J. Jacko and C. Stephanidis, eds.,

Lawrence Erlbaum Associates, Mahwah, NJ, 230–234.

Hoffman, J. E., and Subramaniam, B. (1995). “The role of visual attention

in saccadic eye movements.”Percept. Psychophys., 57(6), 787–795.

Ikuma, L. H., Harvey, C., Taylor, C. F.,and Handal, C. (2014). “A guide for

assessing control room operator performance using speed and accuracy,

perceived workload, situation awareness, and eye tracking.”J. Loss

Prev. Process Ind., 32(Nov), 454–465.

Jacob, R. J., and Karn, K. S. (2003). “Eye tracking in human-computer

interaction and usability research: Ready to deliver the promises.”Mind,

2(3), 4.

James, W. (1913). The principles of psychology, Vol. II, Dover, New York.

Java Graphical User Interface (GUI) for R 1.7-9 [Computer software].

Deducer package, RoSuDa, Univ. of Augsburg, Augsburg, Germany.

Jie, L., and Clark, J. J. (2008). “Video game design using an eye-movement-

dependent model of visual attention.”ACM Trans. Multimedia Comput.

Commun. Appl., 4(3), 22.

Just, M. A., and Carpenter, P. A. (1980). “A theory of reading: From eye ﬁx-

ations to comprehension.”Psychol. Rev., 87(4), 329.

Konstantopoulos, P., Chapman, P., and Crundall, D. (2010). “Driver’s vis-

ual attention as a function of driving experience and visibility. Using a

driving simulator to explore drivers’eye movements in day, night and

rain driving.”Accid. Anal. Prev., 42(3), 827–834.

Koskinen, K. U., Pihlanto, P., and Vanharanta, H. (2003). “Tacit knowledge

acquisition and sharing in a project work context.”Int. J. Project

Manage., 21(4), 281–290.

Laeng, B., and Teodorescu, D. S. (2002). “Eye scanpaths during visual im-

agery reenact those of perception of the same visual scene.”Cognit. Sci.,

26(2), 207–231.

LaFleur, B. J., and Greevy, R. A. (2009). “Introduction to permutation and

resampling-based hypothesis tests.”J. Clin. Child Adolesc. Psychol.,

38(2), 286–294.

Lavine, R. A., Sibert, J. L., Gokturk, M., and Dickens, B. (2002). “Eye-

tracking measures and human performance in a vigilance task.”Aviat.

Space Environ. Med., 73(4), 367–372.

Lopez, R., Love, P. E., Edwards, D. J., and Davis, P. R. (2010). “Design

error classiﬁcation, causation, and prevention in construction engi-

neering.”J. Perform. Constr. Facil.,10.1061/(ASCE)CF.1943-5509

.0000116, 399–408.

Ludbrook, J. (1994). “Advantages of permutation (randomization) tests in

clinical and experimental pharmacology and physiology.”Clin. Exp.

Pharmacol. Physiol., 21(9), 673–686.

Ludbrook, J., and Dudley, H. (1998). “Why permutation tests are superior

to t and F tests in biomedical research.”Am. Stat., 52(2), 127–132.

Manly, B. F. J. (1997). Randomization and Monte Carlo methods in biol-

ogy, Chapman &Hall, New York.

Mackenzie, A. K., and Harris, J. M. (2015). “Eye movements and hazard

perception in active andpassive driving.”Vis. Cognit., 23(6), 736–757.

McBride, D. M., and Cutting, J. C. (2015). Cognitive psychology: Theory,

process, and methodology, Sage, Thousand Oaks, CA.

Muttart, J., Fisher, D., Knodler, M., and Pollatsek, A. (2007). “Driving with-

out a clue: Evaluation of driver simulator performance during hands-

free cell phone operation in a work zone.”Transportation Research

Record, 2018, 9–14.

NCCER (National Center for Construction Education and Research).

(2013). “Craft workforce development 2013 and beyond.”hhttps://goo

.gl/qf09NUi(Apr. 25, 2016).

Palinko, O., Kun, A. L., Shyrokov, A., and Heeman, P. (2010). “Estimating

cognitive load using remote eye tracking in a driving simulator.”Proc.,

2010 Symp. on Eye-Tracking Research and Applications, Association

for Computing Machinery, New York, 141–144.

Pan, B., Hembrooke, H. A., Gay, G. K., Granka, L. A., Feusner, M. K., and

Newman, J. K. (2004). “The determinants of web page viewing behav-

ior: An eye-tracking study.”Proc., 2004 Symp. on Eye Tracking

Research and Applications, Association for Computing Machinery,

New York, 147–154.

Parasuraman, R., Warm, J. S., and See, J. E. (1998). “Brain systems of vigi-

lance.”The attentive brain, MIT Press, Cambridge, MA, 221–256.

Perlman, A., Sacks, R., and Barak, R. (2014). “Hazard recognition and risk

perception in construction.”Saf. Sci., 64(Apr), 22–31.

Pernice, K., and Nielsen, J. (2009). “How to conduct eye-tracking studies.”

Nielsen Norman Group, Fremont, CA.

Podgórski, D. (2010). “The use of tacit knowledge in occupational safety

and health management systems.”Int. J. Occup. Saf. Ergon., 16(3),

283–310.

Poole, A., and Ball, L. J. (2006). “Eye tracking in HCI and usability

research.”Encycl. Hum. Comput. Interact., 1, 211–219.

Pradhan, A., et al. (2009). “Attention maintenance in novice drivers:

Assessment and training.”Proc., Int. Driving Symp. on Human Factors

in Driver Assessment, Training, and Vehicle Design, Univ. of Iowa

Public Policy Center, Iowa City, IA.

R Development Core Team. (2012). “R: A language and environment for

statistical computing, version R2.15.0.”Report prepared for the R

Foundation for Statistical Computing, Vienna, Austria.

Raschke, M., Blascheck, T., Richter, M., Agapkin, T., and Ertl, T. (2014).

“Visual analysis of perceptual and cognitive processes.”2014 Int. Conf.

on Information Visualization Theory and Applications (IVAPP), IEEE,

New York, 284–291.

Rehder, B., and Hoffman, A. B. (2005). “Eye tracking and selective atten-

tion in category learning.”Cognit. Psychol., 51(1), 1–41.

Richardson, D. C., and Spivey, M. J. (2004). “Eye tracking: Characteristics

and methods.”Encycl. Biomater. Biomed. Eng., 3, 1028–1042.

Romoser, M. R., Fisher, D. L., Mourant, R., Wachtel, J., and Sizov, K.

(2005). “The use of a driving simulator to assess senior driver perform-

ance: Increasing situational awareness through post-drive one-on-one

advisement.”Proc., Third Int. Driving Symp. on Human Factors in

Driver Assessment, Training and Vehicle Design, Univ. of Iowa Public

Policy Center, Iowa City, IA, 456–463.

Rozenfeld, O., Sacks, R., Rosenfeld, Y., and Baum, H. (2010). “Construction

jobsafetyanalysis.”Saf. Sci., 48(4), 491–498.

Salvucci, D. D., and Goldberg, J. H. (2000). “Identifying ﬁxations and sac-

cades in eye-tracking protocols.”Proc., 2000 Symp. on Eye Tracking

Research and Applications, Association of Computing Machinery, New

York, 71–78.

J. Manage. Eng., 2017, 33(5): 04017024

Sanders, M. S., and McCormick, E. J. (1998). “Human factors in engineer-

ing and design.”Ind. Robot, 25(2), 153–153.

Sawacha, E., Naoum, S., and Fong, D. (1999). “Factors affecting safety per-

formance on construction sites.”Int. J. Project Manage., 17(5),

309–315.

Shepherd, M., Findlay, J. M., and Hockey, R. J. (1986). “The relationship

between eye movements and spatial attention.”Q. J. Exp. Psychol. Sect.

A, 38(3), 475–491.

Shin, M., Lee, H. S., Park, M., Moon, M., and Han, S. (2014). “A system dy-

namics approach for modeling construction workers’safety attitudes

and behaviors.”Accid. Anal. Prev., 68(Jul), 95–105.

Sohlberg, M. M., and Mateer, C. A. (1989). Introduction to cognitive reha-

bilitation: Theory and practice, Guilford Press, New York.

Suh, W., Park, P., Park, C., and Chon, K. (2006). “Relationship between

speed, lateral placement, and drivers’eye movement at two-lane rural

highways.”J. Transp. Eng.,10.1061/(ASCE)0733-947X(2006)132:

8(649), 649–653.

Sun, Q., Xia, J., Nadarajah, N., Falkmer, T., Foster, J., and Lee, H. (2016).

“Assessing drivers’visual-motor coordination using eye tracking,

GNSS and GIS: A spatial turn in driving psychology.”J. Spatial Sci.,

61(2), 1–18.

Sun, Y., Fisher, R., Wang, F., and Gomes, H. M. (2008). “A computer

vision model for visual-object-based attention and eye movements.”

Comput. Vision Image Understanding, 112(2), 126–142.

Theeuwes, J., Belopolsky, A., and Olivers, C. N. (2009). “Interactions

between working memory, attention and eye movements.”Acta Psychol.,

132(2), 106–114.

Todd,S.,andKramer,A.F.(1993).“Attentional guidance in visual

attention.”Proc., Human Factors and Ergonomics Society 37th

Annual Meeting, Human Factors &Ergonomics Society, Santa

Monica, CA.

Törner, M., and Pousette, A. (2009). “Safety in construction—A compre-

hensive descriptionof the characteristics of high safety standards in con-

struction work, from the combined perspective of supervisors and

experienced workers.”J. Saf. Res., 40(6), 399–409.

Torralba, A., Oliva, A., Castelhano, M. S., and Henderson, J. M. (2006).

“Contextual guidance of eye movements and attention in real-world

scenes: The role of global features in object search.”Psychol. Rev.,

113(4), 766.

Underwood, G. (2007). “Visual attention and the transition from novice to

advanced driver.”Ergonomics, 50(8), 1235–1249.

Underwood, G., Crundall, D., and Chapman, P. (2011). “Driving simulator

validation with hazard perception.”Transp. Res. F: Trafﬁc Psychol.

Behav., 14(6), 435–446.

Underwood, G., Phelps, N., Wright, C., Van Loon, E., and Galpin, A.

(2005). “Eye ﬁxation scanpaths of younger and older drivers in a hazard

perception task.”Ophthal. Physiol. Opt., 25(4), 346–356.

Vecera, S. P., Cosman, J. D., Vatterott, D. B., and Roper, Z. J. (2014). “The

control of visual attention: Toward a uniﬁed account.”Psychol. Learn.

Motivation, 60, 303–347.

Wedel, M., and Pieters, R. (2008). “A review of eye-tracking research in

marketing.”Review of marketing research, N. K. Malhotra (ed.), Vol. 4,

Emerald Group, Bingley, U.K., 123–147.

Weibel, N., Fouse, A., Emmenegger, C., Kimmich, S., and Hutchins, E.

(2012). “Let's look at the cockpit: Exploring mobile eye-tracking for

observational research on the ﬂight deck.”Proc., Symp. on Eye Tracking

Research and Applications, Association for Computing Machinery, New

York, 107–114.

Yantis, S. (1998). “Objects, attention, and perceptual experience.”Visual

attention, R. D. Wright (ed.), Oxford University Press, New York, 187–

214.

Yarbus, A. L. (1967). Eye movements and vision, Plenum, New York.

Zhang, L., and He, J. (2016). “Critical factors affecting tacit-knowledge

sharing within the integrated project team.”J. Manage. Eng.,10.1061

/(ASCE)ME.1943-5479.0000402, 04015045.

Zhang, S., Boukamp, F., and Teizer, J. (2014). “Ontology-based semantic

modeling of safety management knowledge.”Proc., 2014 Int. Conf. on

Computing in Civil and Building Engineering, R. Issa and I. Flood, eds.,

ASCE, Reston, VA, 2254–2262.

Zhao, F., Schnotz, W., Wagner, I., and Gaschler, R. (2014). “Eye tracking

indicators of reading approaches in text-picture comprehension.”

Frontline Learn. Res., 2(4), 46–66.

Zheng, B., et al. (2011). “Surgeon’s vigilance in the operating room.”Am. J.

Surg., 201(5), 673–677.

J. Manage. Eng., 2017, 33(5): 04017024

Using Eye-Tracking Technology to Assess the Effect of Daily Safety Training on Hazard Recognition Skills

Article

Jan 2024

Hazard recognition skills are a critical factor in construction safety. Eye-tracking technology is considered the direct and objective way to study hazard recognition issues. The paper aims to evaluate the effect of daily safety training on people's hazard recognition skills using eye-tracking technology. The study focuses on three aspects: (1) effect of safety backgrounds (safety certifications, construction-related education, and construction-related work experience) on people's hazard recognition performance; (2) impact of daily safety training on hazard recognition skills and on the scan path of workers with varying backgrounds; and (3) comparison of the impact of daily safety training on the scan path of people with different hazard recognition levels. An eye-tracking experiment was conducted to obtain participants' eye movement data and achieve the study's objectives. The results of the study indicate that people who are educated in a related field of study demonstrate an advanced level of hazard recognition abilities, while safety certificates and construction-related work experience do not appear to have a significant impact. Following daily safety training, the participants exhibited more careful examination of construction scenarios. Additionally, individuals with higher hazard recognition levels displayed increased confidence in their hazard recognition judgments. This study makes a significant contribution to the existing body of knowledge by demonstrating the potential of eye-tracking technology to investigate the effect of safety training on visual scan paths. The results of this study emphasize the importance of considering various safety training methods for workers with varying hazard recognition levels to enhance their performance in identifying hazards.

Comparing the safety awareness of workers on the virtual and real construction site using eye-tracking technology

Article

Dec 2023

A critical review of the feasibility of emerging technologies for improving safety behavior on construction sites

Article

Apr 2024
J SAFETY RES

Introduction: Advancements in the modern construction industry have contributed to the development of a range of technology-based interventions to improve the safety behavior of front-line construction workers. Notwithstanding the extensive research on safety behavior, there is still a paucity of research on assessing technology interventions for safety behavior to provide an overview of their strengths and limitations. The present study aims to bridge this gap in the literature and identify the main trends of research. Method: A systematic review and critical content analysis are adopted to capture an overview of the state of knowledge on safety behavior technologies. As a result of searching Scopus, Web of Science, and Google Scholar databases in the period from Jan 2010 to Feb 2023, a total of 359 potential studies went through the systematic screening process, and finally, 48 representative studies were selected, followed by an assessment of the feasibility and applicability of the safety behavior technologies. Results: It was found that safety behavior technology is characterized by seven technologies, including virtual-reality simulation (T1), eye-tracking technology (T2), prediction modeling of safety behavior (T3), computer-based training (T4), drone/sensor-based hazard monitoring (T5), vision-based behavior monitoring (T6), and real-time positioning (T7). Conclusions: This research improves understanding of the status of safety behavior technologies and provides a critical review of their feasibility from the perspective of four assessment criteria, including application, limitation, benefit, and feasibility. Practical applications: The categorization of technologies adds value to the body of knowledge in terms of generic requirements for their implementation and adaptation on construction sites.

Situation Awareness Study in the Construction Industry: A Systematic Review

Conference Paper

Mar 2024

Understanding Cognitive Anticipatory Process in Dynamic Hazard Anticipation Using Multimodal Psychophysiological Responses

Article

Full-text available

Jan 2024
J CONSTR ENG M ASCE

Given the dynamic and complex nature of a construction jobsite, workers' safety is significantly connected to both their ability to efficiently scan their surrounding environment and their ability to anticipate and identify potentially hazardous sources in a timely manner. However, existing studies have failed to consider the spatial-temporal characteristics of high-risk construction hazards as well as the limited training available for enhancing workers' hazard anticipation. Accordingly, this study (1) examines the impact of dynamic hazards' spatiotemporal characteristics on construction workers' hazard identification and anticipation performance, and (2) proposes and illustrates a new methodological approach that harnesses cue perception timing to assemble a coherent cognitive model of situational awareness. To achieve these objectives, this study combines subjective survey measurements with such technologies as 360°video panoramas and wearable psychophysiological sensors to examine 30 construction workers' hazard search strategies, timely hazard identification performance , and spatial attentional distribution. The results of this study demonstrate that workers' hazard anticipation abilities are differentially influenced by both spatial and temporal characteristics of dynamic hazards: (1) most workers failed to distribute their spatial attention across the 360°construction environment and missed hazards outside their field of view (behind or overhead), highlighting the importance of attentional orienting; (2) most workers failed to identify emerging hazard cues that would enable anticipating latent hazards; and (3) many workers failed to develop a correct situational awareness framework because they had improper cue-perception timing. The findings of this study highlight the importance of considering the unique spatiotemporal characteristics of dynamic hazards to properly predict workers' true hazard anticipation abilities. Furthermore, the findings of this study are expected to provide valuable insight into developing effective safety training strategies to enhance workers' situational awareness.

A Science Mapping Approach-Based Review of Construction Workers’ Safety-Related Behavior

Article

Full-text available

Apr 2024

Promoting safe behaviors among construction workers and mitigating unsafe behaviors is an effective approach to enhancing safety performance in the construction industry. Although progress has been made, the research themes concerning construction workers’ safety-related behaviors (CWSRB) and the detailed progress of each theme remain unclear due to differences in review perspectives and conceptual scopes. This study utilized CiteSpace software (V6.2R3 version) to conduct an analysis of co-authorship networks, co-word networks, and co-citations on 563 published articles in this field from 2013 to 2023. This study’s outcomes highlight several key insights: (1) journals such as Safety Science play a pivotal role in the domain; (2) institutions such as the City University of Hong Kong and Hong Kong Polytechnic University, along with prolific authors like Li, are major contributors to the field; (3) the focus of research has evolved from early organizational factors towards a more diverse range of topics, with deep learning emerging as a significant current research hotspot; (4) this study has identified high-cited literature and 11 primary clusters within the field. Current research focuses on five areas: safety-related behavior concepts, influencing factors and consequences, formation mechanisms, interventions, and applications of new technologies. Establishing clear classification criteria for unsafe behaviors, comprehensively understanding the formation mechanisms of safety-related behaviors, evaluating the effectiveness of intervention strategies, and exploring the practical applications of new technologies are future research directions. This study provides researchers with a holistic view of the present state of research and potential avenues for future exploration, thereby deepening the knowledge and comprehension of stakeholders within this domain.

Pioneering Research on a Neurodiverse ADHD Workforce in the Future Construction Industry

Conference Paper

Mar 2024

Safety Practice Assessment Plan Review in Railway Engineering Industrial Towards Sustainable Construction

Article

Full-text available

Feb 2024

The emerging growth of the universal railway development has inspired many nations, including Malaysia, as a crucial strategy for construction sustainable by improving the domestic transportation network and boost economic competitiveness. It serves as the foundation of a city’s public transportation system, to convey huge numbers of people effectively. An organization’s sustainable development strategy should be examined in conjunction with other factors such, occupational safety, and health. Hence, implementation safety practice assessment plan review in railway engineering industrial towards sustainable construction is important. This study was focused on the objectives of (i) to identify safety practice assessment for improvement strategy in railway industrial engineering, (ii) to investigate safety practice assessment challenges in railway engineering industry, and (iii) to analyses safety practice assessment elements in railway engineering industrial. There are 20 journals listed, with 100 articles selected articles were dissected and then the variables on challenges in safety practice has been identified. From the results of the framework, recommendation to improve safety practice assessment plan for future railway engineering industrial to sustain construction can establish. Thus, the research will be helpful for them to understand better upon the procedures and possible safety issues, and able to practice safety procedures to avoid undesired accidents to assist Malaysian construction sustainability. In addition, for the existing literatures, this research will be an additional input for the lacking details related to safety practices in railway project sites.

Impact of Immersive Virtual Reality Environment on Hazard Perception and Situation Awareness during Hazard Identification Tasks

Conference Paper

Nov 2022

An immersive virtual reality (IVR) environment has been proposed in various industries to improve workers' hazard recognition skills and safety awareness at the worksites. Even though the IVR environment would provide a rich visualization experience for hazard perception compared with traditional methods (e.g., videos, static images, verbal instructions, and paper-based texts), little is known about how the IVR environment would help workers and safety managers in hazard perception in the dynamic construction environment. To fill this gap, this research conducts an experimental study to investigate the cognitive process involved during construction hazard perception tasks in virtually simulated construction scenes, aiming to understand how different construction hazard scenarios would affect one's capability on safety awareness. To measure the cognitive process during hazard perception, a widely accepted and well-recognized situation awareness global assessment technique (SAGAT) was used by randomly freezing the scenario and asking participants to assess their situational awareness on the basis of predesigned queries. In accordance with the results, we discussed the feasibility of using virtually created construction hazards to understand the cognitive behavior associated with safety awareness. The findings of this research would help provide a theoretical foundation on how to better design the IVR module and contents to improve hazard perception skills.

Using Eye-Tracking to Examine the Cognitive Process of Hazard Recognition in Construction VR Scenarios

Conference Paper

Jan 2024

Eye Tracking Indicators of Reading Approaches in Text-Picture Comprehension

Article

Full-text available

Jan 2014

Despite numerous studies on reading and multimedia comprehension, the usage of text and picture with different reading strategies has rarely become a focus of research. The current study aims to explore whether the usage of text differs from the usage of picture when readers follow different strategies of knowledge acquisition. In a within-subjects design using eye tracking, seventeen secondary school students comprehended blended text and picture materials with three different strategies.(1) Initial coherence-formation strategy, which requires students to process text and picture unguidedly.(2) Consecutive task-oriented strategy, which requires them to gather information to answer the question (which explains the task) provided after prior experience with text and picture.(3) Initial task-oriented strategy, which requires them to comprehend text and picture to solve the task equipped with the prior information about the question from the beginning. Eye tracking data showed that text and picture play different roles in these processing conditions.(1) The results are in line with the assumption that text (rather than picture) is more likely used to construct mental models in initial coherence-formation processing of text and picture.(2) Students seem to primarily rely on the picture to answer the question after the prior experience with the material with consecutive task-oriented strategy.(3) Text and picture are both used heavily when the question is presented first, enabling students to selectively process question-relevant aspects of the material at first contact.

Measuring Construction Workers’ Real-Time Situation Awareness Using Mobile Eye-Tracking

Conference Paper

Full-text available

May 2016

Eye-tracking technology is an emerging methodology that has the potential to transform current practices in measuring workers’ situation awareness (SA). Since eye movements are assumed to indicate where a worker’s attention is directed, tracking such movements provides a practical way to measure workers’ attention and comprehension of construction hazards. Considering this technology’s potential applications, some researchers have started to use remote eye trackers to investigate workers’ abilities to identify hazards; however, these studies lack process capabilities to assess real-time situation awareness. In order to address this limitation, the main objective of this study is to use a mobile eye-tracker to measure workers’ SA in different scenarios within a real-world construction site. Using direct measures of SA (eye tracking) in parallel with subjective SA measures, this research resulted in some interesting associations between these two types of SA measurements. It also found that situation awareness and visual attention allocations of workers vary significantly according to the scenario’s workload, the state of the area of interest, and the workers’ level of experience. Results from this study may help to identify workers with lower SA and therefore, pinpoint opportunities to provide proactive training and develop guidelines for workers that will reduce human error and accidents in construction sites. Subsequently, this approach can measure the same workers’ SA level after training to determine whether their SA improved.

Using eye-tracker to compare search patterns between experienced and novice workers for site hazard identification

Article

Full-text available

Feb 2016
SAFETY SCI

The construction industry accounts for a high number of accidents. Although identifying hazards prior to commencing construction is widely employed to prevent accidents, it typically fails because of insufficient safety experience. The experience helps in training novice inspectors, although extracting and describing tacit knowledge explicitly is difficult. This study created a digital building construction site, and designed a hazard-identification experiment involving four workplaces featuring obvious and unobvious hazards (e.g., falls, collapses, and electric shocks), and an eye-tracker was used to compare the search patterns of the experienced and novice workers. The results indicated that experience assisted the experienced workers in assessing both obvious (p < 0.001) and unobvious hazards (p = 0.004) significantly faster than the novice workers could; however, it did not improve the accuracy with which they identified hazards, indicating that general work experience is not equivalent to safety-specific experience, and may not necessarily improve workers’ accuracy in identifying hazards. Nevertheless, the experienced workers were more confident in identifying hazards, they exhibited fewer fixations, and their scan paths for assessing hazards were more consistent. The experienced workers first assessed the high-risk targets—laborers working at heights—and subsequently assessed those working on the ground, followed by the equipment or environment. Furthermore, they typically inspected openings later than novice workers did. The search strategies identified may be incorporated into the training courses to improve the hazard awareness for novice workers.

Eye movements and hazard perception in active and passive driving

Article

Full-text available

Sep 2015
VIS COGN

Differences in eye movement patterns are often found when comparing passive viewing paradigms to actively engaging in everyday tasks. Arguably, investigations into visuomotor control should therefore be most useful when conducted in settings that incorporate the intrinsic link between vision and action. We present a study that compares oculomotor behaviour and hazard reaction times across a simulated driving task and a comparable, but passive, video-based hazard perception task. We found that participants scanned the road less during the active driving task and fixated closer to the front of the vehicle. Participants were also slower to detect the hazards in the driving task. Our results suggest that the interactivity of simulated driving places increased demand upon the visual and attention systems than simply viewing driving movies. We offer insights into why these differences occur and explore the possible implications of such findings within the wider context of driver training and assessment.

Assessing drivers’ visual-motor coordination using eye tracking, GNSS and GIS: a spatial turn in driving psychology

Article

May 2016

Vehicle-driving in real traffic can be considered as a human-machine system involving not only the attribute of the vehicle movement but also the human visual perception, cognition and motion of the driver. The study of driving behaviours, therefore, would integrate information related to driver psychology, vehicle dynamics and road information in order to tackle research questions concerning driving safety. This paper describes a conceptual framework and an integrated GIS data model of a visual-motor coordination model (VMCM) to investigate drivers’ driving behaviour via the combination of vision tracking and vehicle positioning. The eye tracker recorded eye fixations and duration on video images to exhibit the driver’s visual search pattern and the traffic scenes. Real-time kinematic (RTK) post-processing of multi-GNSS (global navigation satellite system) tracking generated the vehicle movement trajectory at centimeter-level accuracy, which encompasses precise lateral positioning and speed control parameters of driving behaviours. The eye fixation data were then geocoded and linked to the vehicle movement trajectory to represent the VMCM on the GIS platform. An implementation prototype of the framework and the VMCM for a study of older drivers is presented in this paper. The spatial-temporal visualisation and statistical analysis based on the VMCM data-set allow for a greater insight into the inherent variability of older drivers’ visual search and motor behaviours. The research framework has demonstrated a discriminant and ecologically valid approach in driving behaviour assessment, which can also be used in studies for other cohort populations with modified driving scenarios or experiment designs.

Eye tracking in HCI and usability research

Article

Jan 2005

Eye tracking is a technique whereby an individual’s eye movements are measured so that the researcher knows both where a person is looking at any given time and the sequence in which the person’s eyes are shifting from one location to another. Tracking people’s eye movements can help HCI researchers to understand visual and display-based information processing and the factors that may impact the usability of system interfaces. In this way, eye-movement recordings can provide an objective source of interface-evaluation data that can inform the design of improved interfaces. Eye movements also can be captured and used as control signals to enable people to interact with interfaces directly without the need for mouse or keyboard input, which can be a major advantage for certain populations of users, such as disabled individuals. We begin this article with an overview of eye-tracking technology and progress toward a detailed discussion of the use of eye tracking in HCI and usability research. A key element of this discussion is to provide a practical guide to inform researchers of the various eye-movement measures that can be taken and the way in which these metrics can address questions about system usability. We conclude by considering the future prospects for eye-tracking research in HCI and usability testing. Purchase this chapter to continue reading all 9 pages >

Ontology-Based Semantic Modeling of Safety Management Knowledge

Conference Paper

Jun 2014

Construction safety related knowledge and project specific information are scattered and fragmented. Despite technological advancements of information and knowledge management in the building and construction industry, a link between safety management and information models is still missing. The objective of this study is to investigate a new approach to organize, store and re-use construction safety knowledge. A construction safety ontology is proposed to formalize the safety management knowledge, which consists of three main domain ontology models, including Construction Product Model, Construction Process Model and Construction Safety Model. The interaction between safety ontology and Building Information Modeling (BIM) is also explored. A prototype application of ontology-based job hazard analysis and visualization is implemented to further illustrate the applicability and effectiveness of the developed ontology. The developed construction safety ontology is expected to enable more effective inquiry of safety knowledge, which is the first step towards automated safety planning using BIM.

Critical Factors Affecting Tacit-Knowledge Sharing within the Integrated Project Team

Article

Sep 2015

Sharing tacit knowledge is of great importance for an integrated project team to achieve success. However, research on the factors affecting tacit-knowledge sharing within an integrated project team has been insufficient in the construction management field. This paper aims to identify the critical factors affecting tacit-knowledge sharing and explore the interrelationships between these factors. Based on a literature review, the authors first identify 31 factors of tacit-knowledge sharing within integrated project teams. Through a questionnaire and statistical test, this research next analyzes the factors that are significant to tacit-knowledge sharing within an integrated project team. Then, factors analysis was used to study the critical factors of tacit-knowledge sharing. Finally, the authors analyze the interrelations among the critical factors and explore how they influence tacit-knowledge sharing within the integrated project team. The findings in this research demonstrate the impacts of trust on tacit-knowledge sharing with the evolution of trust taken into account. The authors also find that two personal factors are related to trust and tacit-knowledge sharing. This paper can provide useful guidance for an integrated project team to promote tacit-knowledge sharing and ultimately improve team performance.

Human Factors in Engineering and Design982 Human Factors in Engineering and Design . McGraw Hill International Editions, 1992. 790 pp, ISBN: 0‐07‐112826‐3 £22.99 Seventh edition

Article

Apr 1998
IND ROBOT

Human factors and systems human factors research methodologies information input and processing text, graphics, symbols and codes visual displays of dynamic information auditory, tactual and olfactory displays speech communications physical work and manual materials handling motor skills human control systems controls and data entry devices hand tools and devices applied anthropometry, work space design and seating arrangement of components within a physical space interpersonal aspects at work place design illumination climate noise motion human error, accidents and safety human factors and automobile human factors in systems design.

Randomization Tests

Article

Jun 1981
J AM STAT ASSOC

This article has no abstract.

Measuring the Impacts of Safety Knowledge on Construction Workers’ Attentional Allocation and Hazard Detection Using Remote Eye-Tracking Technology

Abstract and Figures

Recommended publications

Measuring the Impacts of Safety Knowledge on Construction Workers’ Attentional Allocation and Hazard...

Impact of Construction Workers’ Hazard Identification Skills on Their Visual Attention

MEASURING CONSTRUCTION WORKERS' ATTENTION USING EYE- TRACKING TECHNOLOGY

Measuring the Impact of Working Memory Load on the Safety Performance of Construction Workers