Professional Documents
Culture Documents
BY
M A Y 2003
The focus of this thesis is the impact of user knowledge on web usage. A framework is
proposed that brings together research from the fields of consumer research, cognitive
science and information systems. This framework sees knowledge of the web as an
influence on perceptions of the web, which in turn influences current web session usage.
These perceptions relate to perceived usefulness of the web and perceived ease of use.
What is the relationship between a user’s perceptions of the web and a person’s
What is the relationship between a user’s knowledge content of the web and a
What is the relationship between a user’s knowledge content of the web and a
was developed. In analysing the relationships between the constructs, respondents were
grouped into users with and without web-site design and maintenance experience.
Results show that what users think they know about how to use the web is a strong
predictor of both how easy and how useful they think the web is. This highlights the
importance of user perceptions, especially when considering how users use the web.
What users actually know about the web, especially what certain features and attributes
are, also has an influence on how easy and how useful they think the web is.
Significantly, these results apply for users with and without web-site design and
maintenance experience.
These findings help us to understand the relationship between a user’s confidence with
technology and how easy and useful that person finds the technology. This is
technologies (e.g., PCs, notebooks, mobile phones, touch-screen e-kiosks, ATMs, email
and the web). Potentially, this understanding can be a source of new product ideas, of
innovative designs, and of new uses. It also might suggest communication themes for
ii
A CKNOWLEDGEMENTS
-Robert Frost -
As taking the road less travelled makes all the difference, so too does the support,
guidance and dedication of so many individuals that provided the necessary foundation
for this thesis. I take great pleasure in acknowledging those who knowingly and
unwittingly contributed to this task; a task that I set before myself in January 1998.
wisdom and encouragement, and whose dedication and support provided the pillars
upon which this document is based. Mark, your knowledge, your thoughts and your
presence enabled me not only to start my PhD but to also bring it all the way to
completion. I thank you for your patience, your friendship and the insights and direction
you have afforded me, I will remain always your loyal colleague. I would also like thank
Dr Cynthia Webster, a wonderful colleague and a cherished friend, Cynthia thank you
for your support, guidance and the mentor you became. I would also like to thank
Jack Cadeaux, Ms Debra Caldow and Dr Timothy Bock for their helpful comments and
A special thanks must also go to the heart, the soul and the essence of the School of
‘Girls’, thank you for your friendship, your good humour and your cherished ear – you
iii
Special thanks must also go to the commercial supporters of this study, The Campaign
Palace, DoubleClick Australia, ZDNet, Australian NetGuide and The Australian IT, for
On a more personal note, I would like to thank my parents, Doug and Donata Page, my
sisters Tara and Nicci Page, and my friends Tracey Kluck and Shelly Coughran, for their
continued support and encouragement. Last but not least, to David, for his forbearance
while I worked long hours and in isolation and for his continued love and support
throughout everything.
iv
D EDICATION
- My Parents -
Doug and Donata Page, whose continual encouragement and belief in me throughout
my life afforded me the motivation to commence and follow this life path. Mum and
Dad, you stayed with me throughout it all and, as when I crossed the road so many
times as a child, along this adult path you also held my hand.
- My Fiancé -
David Stanley Thomas, the Welshman whose integrity, spirit and soul provided the
inspiration, the love and the support that enabled me to achieve this milestone in my life.
David, you afforded me not only your love and rational presence throughout this
v
TABLE OF C ONTENTS
Abstract .............................................................................................................................................i
Acknowledgements....................................................................................................................iii
Dedication.......................................................................................................................................v
Table of Contents........................................................................................................................ vi
List of Figures............................................................................................................................... ix
List of Tables ..................................................................................................................................x
vi
5.7 User Knowledge Content and web Perceptions .....................................................77
5.8 User Web Knowledge Content: Summary ...............................................................83
vii
12.2 Research Contributions.............................................................................................196
12.3 Research Limitations .................................................................................................202
12.4 Extensions and Future research..............................................................................208
Reference List.............................................................................................................................212
Appendices .................................................................................................................................224
Appendix A: Previous TAM Research ..........................................................................226
Appendix B: Primary Exploratory Research ................................................................228
Appendix C: Variable Conceptualisation & Operationalisation (Pre/Post Test)..243
Appendix D: Scale Development (Student Sample One: n=128) .............................245
Appendix E: Scale Development (Student Sample Two: n=153) .............................246
Appendix F: Web Site & Web Survey Design..............................................................247
Appendix G: Web Survey Advertising & Publicity....................................................260
Appendix H: Web Site Performance Statisticzs...........................................................263
Appendix I: DoubleClick™ Banner Ad Campaign Report/s....................................264
Appendix J: Scale Validation (Web Sample: n=2077) .................................................265
Appendix K: Scale Performance Comparison (Student & Web Sample)...............267
Appendix L: Variable Distribution.................................................................................268
Appendix M: Multiple Regression – Residual Plots...................................................271
Appendix N: Normality P-P Plots ..................................................................................274
Appendix O: Multiple Regression Assumption Check .............................................277
Appendix P: Sample & Variable Description...............................................................278
Appendix Q: Bivariate Analysis – Convergent Validation .......................................283
Appendix R: Nonparametric Correlation Coefficients: Spearman Rho.................335
Appendix S: ANOVA Reported Mean Scores .............................................................337
viii
L IST OF F IGURES
ix
L IST OF TABLES
x
Table 46: Web User Group B - WSD/M Experience: RQ1 .......................................................................... 170
Table 47: Web User Group A – No WSD/M Experience: RQ2 .................................................................. 170
Table 48: Web User Group B – With WSD/M Experience: RQ2 ............................................................... 171
Table 49: ANOVA: Effect of Actual and Perceived Knowledge on PWU.............................................. 172
Table 50: Multiple Regression Results MRA7: PWU = F (Actual & Perceived Knowledge)............. 173
Table 51: Web User Group A - No WSD/M Experience: RQ2................................................................... 175
Table 52: Web User Group B – With WSD/M Experience: RQ2 ............................................................... 175
Table 53: Web User Group A – No WSD/M Experience: RQ3 .................................................................. 176
Table 54: Web User Group B – With WSD/M Experience: RQ3 ............................................................... 176
Table 55: ANOVA: Effect of Actual and Perceived Knowledge on PEWU ........................................... 177
Table 56: Multiple Regression Results MRA8: PEWU = F (Actual & Perceived Knowledge) .......... 179
Table 57: Web User Group A – No WSD/M Experience: RQ3 .................................................................. 180
Table 58: Web User Group B – With WSD/M Experience: RQ3 ............................................................... 180
xi
C HAPTER 1: I NTRODUCTION
- Plato -
(427 BC - 347 BC)
The Republic
1.1 INTRODUCTION
It was argued by Norman (1990) that computers of the future will be invisible in the
sense that users will be unaware that they are even using a computer. This argument is
naked to the human eye within products such as the automobile, the telephone handset,
the microwave oven, the cassette and CD player, electronic calculators, vending
machines and even household whitegoods (i.e., refrigerator, dryer and washing
machine).
directed system interface is highly apparent to those currently using the system and
those who will use it in the future (Barwise 2001). This is evident from the personal
computer (PC) and notebook, to the mobile phone and personal digital assistant (PDA)
and from touch-screen e-kiosks to ATM’s to the World Wide Web, electronic mail and
other internet-based resources. Thus the use and perception of these complex and user-
knowledge of the system and its uses. For example, to use a PC, one needs to know how
to install programs, open and save documents; to use a mobile phone one needs to know
how to create an address book entry and retrieve messages; to use the World Wide Web
1
on the internet, one needs to know how to use search engines and search directories,
technologies (ICT), as evident above. Consequently the effective use and understanding
of electronic technologies, and user interaction with theses systems, has become an
essential user requirement – but little research has been undertaken to investigate
unprecedented control over the use and management of the system and the contents of
the system with which they interact (Rust and Oliver 1994). Drawing on studies in
perceptions, this study investigates user knowledge and perceptions of the World Wide
Web (simply described as the web)1. It also proposes a framework for investigating the
effect user knowledge and perceptions of this highly complex and technologically driven
1. Its aim is to better understand and determine the influence certain user characteristics
(i.e., knowledge content and perception) have on current web session usage.
1 Although the core focus of this research is the web, the context and variables under investigation have a
much wider application to current and future forms of user-directed electronic technologies and thus
should not be limited by the technology discussed.
2
Figure 1: Graphical Representation of the Dissertation (RQ1-RQ3)
The relationships proposed in this dissertation are further compared across two web user
groups; namely, those users with, and those users without, web site design and
experience of users currently adopting the web and that are predicted to use the web in
the future. The early penetration of the web within society was dominated by a
population of users with direct work-related experience and knowledge of the system
that they were using (e.g., web site developers and information technology architects).
However as the population of web users grows, the profile is changing and we are
beginning to see more users not defined by the same work-related and system
development parameters as early web adopters. Therefore, users with and without web
site design and maintenance experience are compared for each relationship proposed in
proven to be one of the most challenging issues in the study of human behaviour. This is
3
hypermedia computer-based systems, like the web, research into the usage of these
systems is required.
Current research investigating an individual’s usage of the web specifically goes very
little beyond descriptive profiles (e.g., demographics) and thus offers limited accounts of
explanatory variables of usage behaviour. However, Korgaonkar and Wolin (1999) state
that understanding why and how consumers use the web may be the key to unlocking
the web’s capacity. Furthermore, many researchers do not take into account the temporal
aspect of ‘usage’ when measuring web usage. Ram and Jung (1990) contend that in
contrast to the discrete event of purchase, usage is a continuous event which may change
over the length of time of exposure or ownership to the stimuli in question. Hence,
researchers should measure both past and current (present) usage experience. In this
research current usage experience is of central interest and is defined as the act of using
the web for some purpose at the time the measurements were made (i.e., recalled over
the last week or month). Past usage experience refers to the act of using the web for some
purpose prior to the time current measurements were made (Delbridge and Bernard
hypotheses about current web session usage behaviour. These are tested in the empirical
factors such as demographics (Communications and Interactive 1999b; Gefen and Straub
1997), perceptions (Agarwal and Prasad 1998; Colley, Henry, Holmes, and James 1996;
Morris and Dillion 1997), computer experience (Handzic and Low 1999; Knapp, Miller,
and Levine 1987; Novak and Hoffman 1997; Schumacher and Morahan-Martin 2001;
Taylor and Todd 1995), and attitudes (Communications and Interactive 1999b; Diaz,
Hammond, and McWilliam 1997; Hubona and Geitz 1997; Kay 1993) have been
4
(Rumpradit and Donnell 1999; Swanson 1988) and organisational constraints have also
been taken into account in past studies. Of specific interest to this dissertation is the body
Davis (1986) in his Technology Acceptance Model (hereafter TAM), and how these
The main focus for a large percentage of TAM studies has been the investigation of the
relationship between user perceptions of the system and system use in an organisational
the changing profile of users (i.e., with both advanced and limited computing
experience), motivations for usage maybe be changing. So too might the environment
In this dissertation it is argued that a distinct relationship will exist between users’
perceptions of the web and current web session usage because of:
the changing roles and profiles of users and their motivations for system usage;
the complexity of the user interface of the web compared to other electronic devices
It is also examined if this relationship differs across users who have experience with web
site design and maintenance (e.g., webmaster/IT specialist) and users without this
and their relationship with system adoption and use. However only a few studies have
extended beyond looking at characteristics of the system and basic usage experience (i.e.,
5
it is argued that characteristics of the user play a large role in forming a user’s perception
influencing the role performed by users and the level of interdependence between a user
and the web interface this study particularly examines the relationship between user
knowledge content of the web and the users’ perceived ease of web use and perceived
web usefulness.
From an examination of the literature it is apparent that consumers can have different
task), and declarative (i.e., knowledge of what features and terms are). Furthermore, it is
shown in this dissertation that the scope of knowledge content among users may differ:
from common (i.e., knowledge of basic features and terms) to specialised (i.e.,
knowledge of more advanced features and terms). The very nature and complexity of
user-directed systems like the web provide a basis for investigation of both the scope and
type of user knowledge content and how these may influence user perceptions of the
One of the most common methods for measuring consumer knowledge of electronic
technologies has been the use of proxies to infer knowledge (i.e., usage experience and
purchase behaviour). But, the use of proxies to infer knowledge stored in memory
assumes that people learn from experience at the same rate when presented with
different products. By contrast, Hoch and Deighton (1989) and Brucks (1985) contend
that the more complex the product, the wider the gap between experience and true
expertise is likely to become. In other words less may be learnt from experience with a
complex product than from experience of a simple product. Thus, to aid the examination
of the relationship between user knowledge content of the web and user web
perceptions, measures of user knowledge content (i.e., type and scope) are developed
further.
6
Here it is argued that a relationship will exist between user knowledge content of the
web and user web perceptions for the same reasons as given before, namely:
the changing roles and profiles of users and their motivations for system usage;
the complexity of the user interface of the web compared to other electronic devices
As before, it is also examined if this relationship differs across users who have experience
with web site design and maintenance (e.g., webmaster/IT specialist) and users without
between the constructs depicted in Figure 1, with a view to answering the question:
What is the relationship between a user’s perceptions and knowledge of the web, and
a person’s current web session usage?
This general question is divided into three research questions, each with specific
underlying hypotheses:
What is the relationship between a user’s perceptions of the web (i.e., both ease of
use and usefulness) and a person’s current web session usage? (RQ1).
What is the relationship between a user’s knowledge content of the web and a
What is the relationship between a user’s knowledge content of the web and a
Chapters 3-5 and summarised in Chapter 6. The methods used to examine the specific
constructs and test the detailed hypotheses are described in Chapters 7 and 8.
7
1.4 SUMMARY FINDINGS
Descriptive and empirical findings from this study are presented in full in Chapters 9 to
10. The purpose here is simply to highlight a few of the main results to be kept in mind
when reading earlier sections of the dissertation. The summary results of each research
Perceived ease of web use had a positive effect on web session usage frequency and
Perceived web usefulness had a positive effect on web session situational variety;
Perceived web usefulness firstly, and perceived ease of web use secondly, had a
Perceived ease of web use and perceived web usefulness had a positive effect on
Perceived web usefulness had a positive effect on web session usage frequency; web
Perceived ease of web use had a positive effect on duration of web session use;
Perceived ease of web use firstly, and perceived web usefulness secondly, had a
Perceived procedural and perceived declarative web knowledge had a positive effect
8
Actual common procedural and common declarative web knowledge had a positive
Perceived procedural web knowledge content had the strongest positive impact on
Perceived procedural and perceived overall web knowledge content had a positive
Perceived procedural web knowledge content had the strongest positive impact on
Perceived procedural and perceived declarative web knowledge had a positive effect
Actual common procedural web knowledge had a positive effect on perceived ease
of web use;
Perceived procedural web knowledge content had the strongest positive impact on
Perceived procedural and perceived declarative web knowledge had a positive effect
Actual common declarative web knowledge had a positive effect on perceived ease
of web use;
Perceived procedural web knowledge content had the strongest positive impact on
9
1.5.1 MANAGERIAL MOTIVATIONS
The number of consumers using the web across differing consumer segments is
changing. Users with more advanced knowledge and experience with computing
technology are becoming more stable and users with less computing experience are
activity on-line, and the poor performance of the dot coms provide evidence that a better
understanding of the user and the user-web interaction is required. However little is
known about how and why users are using this electronic technology and what are the
profile developing web user populations and distinguish different groups of web
to increase user knowledge of the web and thus achieve wider penetration,
describing the evolution of the number of hosts, number of users, and general
characteristics of key players with this electronic technology. As stated by Berthon, Pitt,
and Watson (1996b), back in 1996, ‘most of the work done so far has been of a descriptive
nature – what the medium is’. The lack of understanding of the consumer and of
10
consumer behaviour on the internet is motivation enough to further investigate users of
the web.
empirical research investigating the responses of users to interactive media and user-
directed (externally paced) technologies (e.g., Web), in contrast to the bias of most
elaboration of these presented in Chapter 12. In brief, these contributions include (but are
The development of more refined, tested and validated self-report measures of ‘post-
Empirical support for the argument put forward but not tested by Moore (1991) and
Adams et al. (1992) that usage context influences the effect of user perceptions of a
system on usage,
Further refinement and validation of measures for the study of perceived ease of web
use and perceived web usefulness. Measures are developed to assess a user’s
perceived ease of use and perceived usefulness of certain web functions (e.g.,
11
This study further extends work conducted on the Technology Acceptance Model
(TAM) toward predicting system perceptions and usage. Limited research had been
conducted that explored how and why system perceptions were formed. This
factor, knowledge content, on a user’s perception of the stimuli of interest, i.e., the
perceptions,
Within the knowledge literature, terms have been used and misused, giving rise to
One of the most common methods for measuring consumer knowledge, especially in
the technology area, has been the use of proxies. This study goes beyond the use of
validating and testing objective and subjective measures of user knowledge content
of the web across two very different user groups – users with and without web site
Given the rapidity with which technologies and the profile of those using the
technologies change, the relevance of the results of this study might vary over time.
Specific limitations are identified with respect to how some of the measures were
Limitations further exist as a result of the recruitment method used (i.e., banner ad
12
an inability to screen completely for multiple-survey responses; and also the impact
of survey length.
limited to:
Further development and refinement of the actual knowledge scales for validation of
the difference between common and specialised knowledge (scope), as well as the
the research findings. For example, a comparative analysis between users and non-
How, and from where, do users acquire their knowledge about how to use the web;
is it from experience and use of the technology, from media sources, from personal
explore the relationships between user knowledge content, perceived web usefulness,
perceived ease of web use and current web session usage. The relationships proposed
Acceptance Model (TAM). In the next chapter, these themes are discussed in a general
way. The attributes, features and uses of electronic technologies are considered. This
provides background information and offers a rationale for why further research and
investigation is important.
13
Following on from Chapter 2, Chapter 3 presents past research on consumer usage
technology, and Chapter 5 examines user knowledge content of the web. In sum, the
discussion presented from Chapter 2 through to Chapter 5 sets the foundation for the
14
C HAPTER 2: R ESEARCH B ACKGROUND –
’The same technology that simplifies life by providing more functions in each device
also complicates life by making the device harder to learn and harder to use’
- Norman -
(1990)
2.1 INTRODUCTION
In this chapter, a review of the management and development of user-driven technology
is presented. Conceptual ideas and empirical findings are considered, drawing on the
computer-based technology of the World Wide Web (hereafter web) on the internet. This
study examines the web as a developing electronic technology, drawing from established
services to personal computing, and networks to broadcast devices. Due to the range of
15
devices available, and their system complexity, it is difficult to categorise them to form
and communication devices exist, one technology has especially captivated the attention
of the public and business. This is the internet and the web. Without question, interest in
this has increased more rapidly and widely than for any other electronic device in recent
decades. Its penetration has not surpassed that of broadcast devices like TV and Radio or
the telephone, however the impact of the internet and the web on not just business but
of as referring to the same entity. However, these terms actually make reference to two
distinct electronic technologies, although they are very closely inter-linked. The
Protocol (hereafter IP) (Ainscough and Luckett 1996; Hoffman and Novak 1996; Pallab
1996). The development of IP and software that understands this protocol has become
the means of transmission for the internet. In scope, this network of connected
Since the early 1970’s, the connections and networks have been run by universities,
school systems, libraries, federal and state governments and the military. Broader
business use of the internet only became possible in the early 1990’s when it was
16
identified that the commercialisation was essential for further investment and
search environment, called the web that added a hypermedia5 capability to the internet.
As the development of the web was the key to the commercialisation and rapid growth
of the internet, the web is the communication technology further examined in this study.
The commercial potential of the Internet was realized with the development of the web,
a distributed hypermedia system. Prior to this development, the Internet was text-based,
command driven and user unfriendly (Lawrence, Corbitt, Fisher, Lawrence, and Tidwell
2000). However, following its inception, the web - supported at the client level by a
graphical user interface (GUI) – had a profound influence on the usability and
Technically, the web rests on three enabling protocols: hypertext mark-up language
(HTML), hypertext transfer protocol (HTTP), and Uniform Resource Locators (URL’s).
HTML specifies a simple mark-up language for describing and displaying information;
HTTP specifies the form and nature of the request/retrieval process that occurs between
a web client (e.g., a web browser housed on a PC) and a web server connected to the
internet. URL’s are used to specify the location of documents housed on web servers on
structure and process of the web. Hypertext and hypermedia development and research
extend back to the early 1960’s. They basically enable the structuring of information as an
associative network of nodes and links, free from the linear sequential structure that
Oinas-Kukkonen 1997).
5 Hypermedia is the combination of the node and link access of hypertext and the synchronisation facility of
multimedia, as discussed in section 2.2.1.2 (Bornman and Von Solms 1993).
17
Hypertext systems are databases in which chunks of information are linked together in a
non-sequential way. This provides a vehicle for intuitive non-linear access to information
(Bornman and Von Solms 1993; Smith and Wilson 1993). In a typical hypertext system, a
window on the computer screen would correspond to a node in the database. Users
navigate through the system by selecting buttons or hotspots within the window that
activates links to other nodes. Smith and Wilson (1993) specify three methods of
graphics, and dynamic audio, video and animation content (Bornman and Von Solms
1993). These are time based. Multimedia, like video and audio, typically have
temporal order (Rada 1995). For example, time dependencies exist in the sequence of
Hypermedia is the combination of the node and link access of hypertext and the
synchronisation facility of multimedia (Bornman and Von Solms 1993). It is from the
communications structure and process of the web that provided the impetus for
Hoffman and Novakʹs (1996) reference to the web, and supported interfaces, as the most
referred to as HCME).
HCME’s relate to a range of electronic technologies (i.e., web, personal digital assistant
(PDA), touch-screen e-Kiosks, etc.) and these electronic technologies differ to non-HCME
based electronic technologies (i.e., TV, radio, etc.) in terms of vividness, interactivity,
media pacing (i.e., external/internal) and the flow of information and communication
18
transfer. Given the amount of research already conducted on non-HCME based
aforementioned characteristics.
which enables users to provide and interactively access hypermedia content (i.e.,
interactivity) (Hoffman and Novak 1996; p53). The key distinguishing features of the
web are:
It uses the network structure of the internet (Ainscough and Luckett 1996; Hofacker
Based on hypermedia (Rada 1995; Bieber et al. 1997; Lawrence et al. 2000);
Information transfer is externally and internally paced6 (van Raaij 1998; Stangelove
1996; McWilliam, Hammond, and Diaz 1997; Blattberg and Deighton 1991; Berthon,
Both person and machine interactivity are facilitated (Steuer 1992; Hoffman and
Novak 1996);
(Rumpradit and Donnell 1999; Encarnacaeo, Loseries, and Sifaqui 1999; Church
1999);
6 Pacing refers to who controls the speed and sequence of information transfer. With some electronic
technologies the speed and sequence of information transfer is controlled by the sender (i.e., it is externally
paced) or the receiver of the information (i.e., it is internally paced). For example, broadcast television and
radio are two externally paced electronic technologies, where as the web has the capability to enable both
internal and external pacing.
19
It is an electronic technology that tends to be media rich, information rich and fairly
presented with more and more electronic information and communication resources.
This turns the spot light on users, the focus of this dissertation. The background to this
parties that have quite different sending and receiving capabilities. The user interface is
which the two make contact, interact and communicate (Church 1999).
Moran (1981) defines the ‘interface’ as consisting of everything the user comes in contact
with while using the system – physically, perceptually, and conceptually. Marchionini
(1995) builds on this and his discussion of an interface is most useful to illustrate the
between the user and the database [of knowledge being searched]’ and that this database
can reside in ‘people, books, libraries, and maps as well as in a variety of electronic
20
form of exchange with both the interface and user encoding and decoding a stream of
symbols flowing to the users, from the user and to-and-from the user to accomplish
The core motivation for the investigation of HCI is the end goal of system usability -
system ‘ease of use’ and ‘usefulness’. Hartson (1998) specifies that usability is seated
within the user’s perspective of the interaction with a computer system and the interface,
and not just about the components of the interface itself. The underlying principle of the
dialogue that involves few spoken words, but the exchange of meaningful symbols. This
facilitate the acquisition of meaning from the symbols by all participants (Church 1999;
meaning can thus be seen as the key mediating factor in the success of communication
An example where communication fails, and meaning is not acquired during a HCI, is
the interaction between a user and the electronic interface of a home video recorder
(VCR). Many users of a VCR do not understand how to use the pre-record and date/time
settings of the device. This interaction is confounded by the design of the VCR and its
functions and also the design of the user manual used to aid VCR users. In summary,
knowledge of the VCR and its functions is the primary ingredient that would facilitate
increased ease of use and usefulness of this device, thus enabling system usability.
The specific theme of this study examines the user interface component of the web on the
internet. The most common user interface that individuals use to access the web on the
internet is a ‘web browser’ - more technically known as a ‘web client’. A web browser is
a graphical user interface (GUI) that allows the user to access information in the form of
sound, text, graphics, and video clips on the internet providing visual representations of
21
interface to the web on the Internet via HTTP. Figure 2 provides a screen shot of two
Figure 2: The Web Browser: A GUI for the World Wide Web
Novak 1996), involves the use of a web browser Navigation of this system is determined
by the characteristics of the system being used. The nature and structure of the
configuration of the web, supports two navigational activities: ‘system browsing’ and
browsing (Smith and Wilson 1993), whereas retrieval is the direct search for queried
associations (Rada 1995). Brown (1988) differentiates between browsing and searching
22
Rada (1995) further states that a characteristic of browsing is going from node to node in
the course of discovering what is in the information space and whether a vaguely
conjunction with Boolean operators (e.g., AND, OR, etc.), statistical word weighting and
document relevance rankings for query modification (Brown 1988). In searching, people
express a query and the information system returns a set of node contents which match
or are related to the query. Knuth and Brush (1990) suggest that the ability to browse the
Both browsing and information retrieval navigational activities on the web are non-
linear in nature and provide essentially unlimited freedom of choice and greater control
for the user about information alternatives. This is in contrast to the restrictive
technologies (i.e., television and offline newspapers). Hoffman and Novak (1996)
comment that network navigation in HCME’s permit greater freedom of choice than the
shopping networks.
interactivity, both person and machine interactivity, which comprise web interactivity.
The hypermedia computer based communication technology of the web, and its
graphical user interface, provides users with unprecedented control over the
management and use of the delivery system with which they interact. However, despite
the fact that the interactive delivery system offers the user a wider range of choices and
23
simultaneously greater individualization than non-interactive delivery systems (Rust
and Oliver 1994), many aspects of graphical user interface use are not well understood.
In particular, the influence of the physical constraints of the user (e.g., memory
capabilities, information transfer limits, and computational ability) on the success of the
1999). Smith and Wilson (1993) further state that from a behavioural viewpoint, even
though hypertext systems have much to contribute in regards to the effectiveness and
usability of network systems, they may also create a new set of usability problems,
particularly in terms of network navigation and the use of graphical user interfaces.
complex and information-rich nature of the communication technology of the web, the
changing role of the user in the communication transmission, and the heightened role of
interdependence that has evolved between the communication delivery system and the
user. Specific problems in these areas may inhibit system and behaviour adoption and
continued use.
For example, significant problems exist in using graphical browsers. Conklin (1987)
indicates that graphical browsers rely on the highly developed visual spatial processing
of the human visual system. As nodes and links are placed in two or three dimensional
space users have to orientate themselves by visual cues just as when they are walking
through a familiar city. Conklin (1987) also points out that there is no natural typography
for an information space like the web. So, until a person is familiar with any given layout
disorientated by a large number of nodes and links, frequent changes in the network,
slow or awkward response to inputs, and limited visual orientation (Conklin 1987).
environments:
24
Lack of Closure: the problems that arise from unfamiliarity with the structure or
conceptual organization of the network. For example, not knowing the extent of a
Cognitive Overhead: the problems that stem from the cognitive demands placed on
the user of a hypertext system. The user must decide which path to take through the
network but may find interesting sidetracks that distract attention from the main
task.
and/or understanding the semantic content of nodes. This results in a lack of detailed
memory of any particular item and an inability to summarise what has been learned.
These problems of using hypermedia computer-mediated systems give rise to the need
for an understanding as to what inhibits user adoption and use of the web. Current
research examines certain psychological characteristics of web users and user segments.
For example research investigating the determinants of web use have looked at certain
user navigational behaviours (Hoffman and Novak 1996), the principle of flow
(Hoffman, Novak, and Yung 1998; Novak and Hoffman 1997), and individual
and McWilliam (1997) further conceptually reinforced Petty, Cacioppo, and Schuman
(1983) and Petty and Priester (1994) supposition that an elective, interactive and novel
medium naturally brings greater consumer involvement. Research also has examined
different user segments, such as novice and more experienced web users (Diaz et al.
1997), teenagers, young adults (Napoli and Ewing 1998), people in different geographic
markets (Teo, Lim, and Lai 1997), and different user perceptions (Briggs and Hollis 1997;
Ducoffe 1996; Eighmey 1997; Maddox and Mehta 1997; Teo et al. 1997).
In 1984 Rice (1984) flagged that little effort had been allocated to the analysis of the
interaction between the characteristics of media technology and the characteristics of the
users themselves, this position still stands today with hypermedia computer-mediated
25
valid, a theoretical framework of media use and the implications of the interplay of
media-user characteristics and their effect on system adoption and usability is altogether
interface and design issues (i.e., button style or window placement) are often the focus of
research papers in this area (Rada 1995). Most attention in the study of communication
delivery systems has concentrated on the components of the message (i.e., source,
content, etc.) and the effects of such message characteristics on the audience (i.e., recall,
intention, etc.) (Weaver 1988). With respect to the web, research has examined user
responses to web sites (Eighmey 1997; Eighmey and McCord 1998), web site features and
design (Napoli and Ewing 1998), and specific usage trends such as newsgroup usage
Hodkinson and Kiel (1997) also identify a lack of scholarly investigations of user-directed
technologies, although there has been some. For instance, researchers have examined the
web in the context of general marketing communications tools (Hoffman and Novak
1996), for offline and online web advertising (Briggs and Hollis, 1997; Ducoffe 1996;
Maddox and Mehta 1997) and for examining the impact of new digital media on the use
of other communication technologies (Coffey and Stipp 1997; Napoli and Ewing 1998). It
is evident that progress has been made by some researchers, but much remains to be
studied.
understand how the content of the technology influences behaviour. Due to the increase
kiosks, electronic organizers such as PDA’s and wireless system technologies such as
WAP and iMode), understanding and determining consumer usage of the technology
26
As outlined in this chapter, users of user-directed hypermedia computer-based
communication technologies can exercise unprecedented control over the use and
management of the system and the contents of the system with which they interact (Rust
and Oliver 1994). Drawing from the cognitive science literature, and studies in consumer
also proposes a framework for investigating the effect user knowledge and perceptions
of a highly complex and technologically driven system may have on system usage. This
What is the relationship between a user’s perceptions and knowledge of the web, and
a person’s current web session usage?
two core components theorized in this dissertation as the determinants of current web
27
C HAPTER 3: P REDICTING
‘Riches do not consist in the possession of treasures, but in the use made of them.’
- Napoleon Bonaparte -
(1769-1821)
3.1 INTRODUCTION
Understanding why people use certain products and engage in certain behaviours is a
computer-based systems, like the web, research into the usage of these systems is
required. Current research is limited and in many instances it goes little beyond profiling
web users. However, Korgaonkar and Wolin (1999) state that understanding why and
how consumers use the web may be the key to unlocking the web’s capacity. Therefore,
this chapter presents an overview of research into current usage experience of the web
and provides a foundation for a model of ‘current web session usage’. As presented in
Figure 3, the construct ‘current web session usage’ is one of the key dependent variables
Draws a distinction between current usage experience and past usage experience and
28
Lays the foundation for discussion in subsequent chapters of user perceptions of the
Before, addressing these themes in some detail, three leading perspectives adopted in the
literature for the investigation of consumer usage behaviour are discussed (section 3.2).
investigated from three research perspectives - the social interaction perspective, the
social interaction perspective examines the symbolic aspects of usage. It examines social
socially conspicuous products such as a car or house (Belk, Bahn and Mayer 1982;
usage, especially consumer experiences such as ‘fantasies, feelings and fun’ - the hedonic
utilisation perspective examines the functional use of products and their attributes in
29
different situations (McAlister and Pessemier 1982; Srivastava, Shocker and Day 1978).
The functional utilization perspective is particularly relevant here because of the way the
web is used. The web is an electronic technology that offers multiple features and
different situations for each application. Furthermore, Ram and Jung (1990) note that the
usage of durables, such as personal computers and VCR’s, are the focal point of the
and use. This approach arose out of the functionalist perspective of mass media first
articulated during the 1940s in research concerning the effects of radio programs on
members of the listening audience. A psychologist and mass media researcher, Herzog
people say they derive from using a particular mass medium (Herzog 1944). These self-
reported perceptions and motivations gave researchers insight into the factors that
particularly television (Rubin 1994). For example, certain types of television programmes
This perspective has been applied to understanding user motivations and behaviour in
the context of cable television (Donohew, Palmgreen and Rayburn 1987); message
content (Swanson 1987); TV remote control devices (Walker and Bellamy 1991);
30
computer aided instructional settings (Kuehn 1994); television commercials (Schlinger
1979) and the opportunities for relaxation offered by use of digital media (Barwise and
Hammond 1998). Furthermore, McGuire (1974) points out that this perspective appears
to be particularly useful in explaining the continuing use of a medium. While initial use
of a medium may result from accidental exposure, curiosity about new things or
Furthermore, the uses and gratifications perspective rests on the basic assumption that
the audience is actively involved in media usage as opposed to being a passive recipient
audience members can actively search for information, can view information in a wide
range of content formats, and can interact with numerous information sources. The uses
and gratifications perspective is thus very much in line with web use, and this
perspective is adopted here. However, before discussing the antecedents of web usage
experience, an overview of the literature on product and media usage is presented. What
hard to find because so many different interpretations and measures have been
proposed. Within the literature, current usage experience and past usage experience are
often treated as one and the same (e.g., Bettman and Park 1980; Dishaw and Strong
1999). However, this might hide important effects, particularly for internet applications
like the web (Kraut, Mukhopadhyay, Szczypula, Kiesler, and Scherlis 1998). This may be
due to the web’s more adaptable and more complex characteristics in comparison to
many household devices currently in use. For example, many consumer products and
media alternatives, such as the VCR, the television, the radio, have a narrow range of
uses. By contrast, the web offers a variety of applications. Hence a user’s early
31
experiences are likely to have an impact on later usage, but the two should not be treated
as one and the same – the behaviours of people who experiment with the web in the first
few months are not going to be the same as those with five years experience.
Ram and Jung (1990) contend that in contrast to the discrete event of purchase, usage is a
continuous event which may change over the length of time of exposure or ownership to
the stimuli in question. Hence, researchers are advised to measure both past and present
usage experience. In this research current usage experience is defined as the act of using
the web for some purpose at the time the measurement were made. Past usage
experience refers to the act of using the web for some purpose prior to the time the
measurements were made (Delbridge and Bernard 1998). The former construct – current
ideas and can lead to ideas for new product uses or product design and development.
Furthermore, new markets for existing products can be indicated, as well as appropriate
economic success and also the high rate of new product failure of new or existing
adoption and use. Understanding how products, or in this case, how electronic
technologies are used, and what determines their usage, is thus an important part of
Usage also has important implications for the communication of product information to
the consumer. Ram and Jung (1990), for example, showed that only a small number of
respondents reported the use of certain features of durable goods, with some
respondents not even aware of these features. This result is extremely apparent in
studies of technology-related products (Higgins and Shanklin 1992); for example, among
American VCR owners one-third do not record programs while absent from home –
32
despite this being a key feature of a VCR (Rosen and Weil 1995). This suggests that
Current usage could also be used as the basis for segmenting product markets. For
example, Potter et al. (1988) attempted to identify the profiles of five usage segments for
VCRs. Studies of computer usage in the workplace have had a wide range of uses too.
They have been used to determine training needs, to determine the effectiveness of
system implementation, to establish time costs associated with certain work tasks, and/or
has provided ample illustration of how usage estimates facilitate the evaluation of
system success. For example, user receptivity towards computers (Sarris, Sawyer and
Quigley 1993; Saltz, Saltz and Rabkin 1985) and the effect of computer implementation
and use (Knapp, Miller and Levine 1987). Having shown the importance of studying
frequency and usage variety. Zaichkowsky (1985b) also discusses two categories of
product usage - depth of consumption and breadth of consumption - and relates these to
equivalent to Ram and Jungʹs (1990) category of usage frequency. Seeley and Targett
(1997) further showed that computer use comprised three categories - frequency, depth
and breadth of use. Therefore, it is possible to derive the following four categories of
product usage:
Usage Frequency refers to how often a product is used within a certain time frame.
Usage Variety refers to the different use motivations and different situations in
which a product is used. For example, using the web for information search and
33
shopping (motivational variety) and accessing the web at home and at work
(situational variety).
Breadth of Use refers to the number of different types or brands of a product owned
or used in a category within a given current time frame. For example, the number of
different types of web sites (e.g., search engine, shopping, email, etc.) used within the
last week.
Depth of Use refers to the overall number of items within a category used within a
certain current time frame. For example, the total number of web sites accessed in the
last week.
Ram and Jung (1990) differentiate frequency and variety of consumption on several
characteristics. The authors propose that usage frequency can be driven mainly by task
requirements of the consumer, while usage variety depends both on the variety of
features offered by the product and the variety of usage situations (Ram and Jung 1990).
Furthermore, there are likely to be temporal variations in the two categories. Usage
frequency may be high immediately after purchase, whereas usage variety may be
dependent on a user’s skill and knowledge and thus high after continued use (Ram and
Jung 1990). Both also could be considered manifestations of different consumer needs:
frequency - routine needs, and usage variety - variety seeking needs. In addition, an
increase in usage variety is likely to have a positive impact on the market development
of the product. For instance, the higher the technical sophistication, the greater number
of potential uses, and thus the greater the number of potential users.
Much of the research on usage experience has been product-specific. Such as studies of
VCR’s (Harvey and Rothe 1986; Levy 1980; Levy 1981; Potter et al. 1988) and personal
computers (Dutton, Kovaric and Steinfield 1985; Mentzer, Schuster and Roberts 1987).
Typically, product usage has been studied in the context of pre-purchase decision-
making (Belk 1979; Bettman and Park 1980; Johnson and Russo 1984; McAlister and
Pessemier 1982; Srivastava et al. 1978) and seldom post-purchase consumption. The
focus in this dissertation is somewhat different – the focus is on current usage of media
34
technology, specifically the web on the internet, during post-purchase consumption. This
demands by media owners for high quality usage data. For example, from the 1950’s TV
viewing information was collected through organizations such as Arbitron in the US and
AGB Ltd in the UK. Similar data have been collected for newspapers and magazine
readership, radio listening, cable viewing, etc. Systematic studies of media usage
included Goodhardt, Ehrenberg and Collins (1975) and Barwise and Ehrenberg (1988) on
web usage data has been collected by firms such as Forrester Research in the US and UK
When looking at the relationship of ʹproduct usageʹ to ‘media usageʹ one can begin to see
possible example. Bawa and Shoemaker (1987) discussed an extensive body of literature
on segmenting consumers based on their use of broadcast and direct response media.
of media vehicles can be found in much of the media research. For example, Urban
(1976) suggested that heavy and light magazine readers might respond differently to ads
As further discussed by Chatterjee, Hoffman and Novak (1998), segmenting users on the
basis of their media usage frequency yields insights on whether the medium attracts and
35
Frequency of media usage has been the most popular measure of media usage
experience. However, Olney, Holbrook and Batra (1991) also identified viewing time as
Holbrook and Gardnerʹs (1993) argument that duration time is a critical outcome
depth of web-site usage as important, defined as the number of pages accessed. The
following categories of web use are further discussed and investigated: frequency,
variety, breadth, depth and duration of use. In this study the latter three are proposed to
comprise three dimensions of ‘usage extent’. Therefore, ‘usage frequency’, ‘usage variety’
and ‘usage extent’ are further conceptualised and discussed in relation to current web
usage experience. These categories are presented in Table 1, together with examples of
broad media types (e.g., web versus magazines) and specific media vehicles (e.g., web
36
Number of
Number of Number of
Number of new/different
Breadth new/different new/different
new/different site magazines
(Familiarity) sections read in web sites used in
features used in a visit read/bought in a
an issue a session
month
Overall number
Overall number of Overall number Overall number
Depth of magazines
pages visited at the of pages read in of sites visited in
(Amount) read/bought in a
site in a visit an issue a session
month
3.4.2 CURRENT WEB USAGE: WEB SESSION VERSUS WEB SITE VISIT
Use of a hypermedia system like the web is termed network navigation. Hoffman and
through an HCME. This non-linear search and retrieval process provides essentially
unlimited freedom of choice and greater control for the user about information
alternatives. This may be contrasted with the restrictive navigation options available in
traditional media such as television and print media or even the centrally controlled
During a web session the user might visit a series of web sites (site visitation) to acquire
to the macro perspective of ʹweb usageʹ, and ʹvisitationʹ the micro perspective of ʹweb site
usageʹ. This study discusses the implications of current (present) usage experience of the
web from the macro perspective of web usage (i.e., column three of the examples in
Usage frequency is defined as how often a product is used within a certain time frame,
either within a product context (Zaichkowsky 1985b; Ram and Jung 1990; Seeley and
Targett 1997) or with respect to media (Urban 1976; Bawa and Shoemaker 1987;
Chatterjee et al. 1998). In this dissertation usage frequency is defined with respect to
current web session use: ‘how often a session on the web is undertaken within a certain
37
time frame, such as a day, a week, or a month. This definition is consistent with existing
descriptive measures of web usage by academics. For example, Novak et al. (1998)
examined web usage frequency and Sivadas et al. (1998) asked respondents about their
over a day, a week and a monthʹ, and Jupiter (1999) also use similar definitions of current
A unique position taken by Novak et al. (1998) was that in addition to measuring present
frequency of use of the web, they should also measure ‘anticipated or future usage
frequency of the medium over the next year’. However, Nunes (2000) found that
individuals are unable to predict their own future usage or, at the very least, they find
this difficult to do. For the purpose of this study, current session usage frequency of the
web is only examined in terms of present usage frequency, not future usage.
Potential segmentation of consumers on the basis of their frequency of visits yields many
insights. Chatterjee et al. (1998) segmented their sample into four groups according to
site visit frequency and they noted that as frequency of usage increases, consumers tend
to stay longer at sites and are exposed to more passive ads, but they click on few passive
ads or browse through active ads during visits. Napoli and Ewing (1998) also segmented
users based on their frequency of use, classifying them into heavy, moderate or light
The influence of user perceptions of the web on the frequency of current web usage
The term variety refers to the extent to which items in a set are different or distinct (Desai
and Hoyer 2000). Thus, this second category refers to the different motivations for which,
and situations in which, a product is used (Ram and Jung 1990; Zaichkowsky 1985b). In
38
the context of this dissertation, the focus is on current session usage variety of the web.
This comprises the variety of situations and variety of motivations in which the web is
Usage consumption can occur in a variety of different situations and thus the number
and type of situations in which usage takes place are worth investigating (Desai and
Hoyer 2000). With respect to media, we are exposed to, and may use, various media in
the home, at work, at school, while in transit, and in a number of other situations.
Certain external factors will influence the situations available for media use, for example
proposed that internal user characteristics also play a role in influencing the type and
Due to the user-driven and complex nature of the web, individual characteristics are
believed to heavily influence a user’s choice of usage situation and thus the variety of
situations in which the web is accessed. Usage variety is also determined by the number
and type of motivations for web use. Motives are regarded as general predispositions
that influence an individual’s action to fulfil a need or want – such as the need for
information or the need to communicate (Schiffman, Bednall, Watson, and Kanuk 1997).
Rubin (1993) contends that motives are key components of audience activity. In addition,
different categories of motivations may appeal to differing types of people, may engage
The uses and gratifications perspective has been applied successfully to a range of new
media and related technologies to explain user motivations for media use. As previously
discussed (section 3.2), the web is more adaptable and has more complex characteristics
than many household devices currently in use. Typically, television has a fairly narrow
range of uses and an individual’s usage motivation is heavily influenced by the capacity
of this medium to entertain. By contrast, due to the complexity of features and attributes
39
For example, Rafaeli (1986) examined audience member reports regarding the use of
electronic bulletin boards. These users report recreation, entertainment and diversion as
the primary motivations for use, followed by learning what others think and
controversial content and communication. This study revealed a wide range of uses and
communication.
Korgaonkar and Wolin (1999) further identified seven motivations and concerns (Table
2). These suggest that consumers use the web for many more reasons than to retrieve
five primary motivations for using the internet: interpersonal utility, passing time,
Motivation Description
From a review of the literature it is evident that a number of descriptive profiles of web
usage motivation are evident. Users are driven to satisfy needs of hedonism,
40
in the literature are the predictors of web usage motivations. For example, an individual
might be driven to satisfy a need for information and thus searches the web; but, what
factors directly influence this individual’s use of the web to satisfy the felt need?
Characteristics of the medium certainly influence its ability to fulfil this need, but
characteristics of the user may also influence the choice to use the web above other
alternatives.
The influence of user perceptions of the web on the variety of current web usage
The final category to be discussed is the current extent of web session use. This is defined
as the degree of session use of the web, as opposed to variety and frequency of current
use. This construct comprises the dimensions of: duration of use (time), breadth of use
(range) and depth of use (amount) of the web (See Table 1).
Visit duration has been defined as the time between consumer entry to and exit from a
web site (Chatterjee et al. 1998; Dreze and Zufryden 1997d). From the macro perspective
adopted in this study, duration of web use in a session is thus defined as the time
between logging-on and logging-off the web. Segmenting consumers on the basis of their
session duration yields many insights. Holbrook and Gardner (1993) argued that
Holbrook and Batra (1991) further identified viewing time as a dependent variable in a
Breadth of use, when applied to product usage, refers to the range of different and/or
new types/brands of products owned or used in a product class within a given time
frame (e.g., number of different and/or new brands – Sony™, TEAC™, etc.). Thus,
breadth of web session use is defined as the range of different sites or tools that are used.
41
For example, this would correlate with the number of new, unfamiliar or different web
Session depth is defined as the total number of web sites or search tools used within a
given time frame. For example the total number of VCR’s owned, irrespective of brand.
Depth of web usage has been measured from a micro perspective, thus depth of site use
during a site visitation involves the measurement of the number of pages accessed. From
a macro perspective this would correspond to the total number of web sites accessed
during a web session (regardless of whether they are the same type of web site such as
Dreze and Zufryden (1997a) developed two effectiveness measures that were
particularly relevant for the study of music web sites. The model of web site effectiveness
is based on the number of pages accessed (visit depth), and time spent during site
visitation (visit duration), with these measures explained by site attributes (e.g., pp 53).
Dreze and Zufrydenʹs (1997d) model provides a potentially useful approach for
evaluating and designing web site contents and configurations (i.e., background, image
size, sound file display, and celebrity endorsement, use of java and frames, as well as
operating system).
Descriptive profiles of the extent of web usage are reviewed in the literature (e.g., Diaz et
al. 1997; Eighmey 1997; Novak and Hoffman 1997; Seeley and Targett 1997; Spink,
Bateman, and Jansen 1999). However, as with other categories of web usage, there has
been little investigation of the predictors of the extent of web usage, at either the macro
and micro levels. It is evident that system characteristics such as access, modem, and
computer configuration would influence the extent of usage of the web. However, due to
the user-directed nature of the medium, it is proposed here that user characteristics may
also be an influence. Therefore, the influence of user perceptions of the web on current
web session usage experience will be discussed in section 3.5 and chapter 4.
42
In summary, web usage behaviour is defined in terms of current web session usage
frequency, usage variety, and extent of use. In the next section the proposed antecedents
of web usage are introduced, leading to the main theme of this dissertation, i.e., the
influence of consumer perceptions of the web on current web session usage (i.e.,
patterns, but also gives insight into the dynamics of usage shifts over time – including
changing patterns of consumer needs, variations in consumer skill levels, and likely
product market developments. Determining usage may also influence forecasts of user
investigating the determinants of web usage experience. So, what influences the
Many studies profile and provide a descriptive account of usage, however very few
actually propose explanations for web usage. Nevertheless, a number of theories can be
drawn upon from the areas of consumer behaviour, communications and management
information systems. These theories stem from characteristics of the medium itself (i.e.,
from the graphical user interface and system design) (Conklin 1987; Foss 1989; Davis
1986); characteristics of the individual (i.e., from web experience) (Diaz, Hammond and
McWilliam 1997), and web involvement (McWilliam, Hammond and Diaz 1997); and
from features of the usage situation (i.e., from accessibility) (Cheung, Chang, and Lai
system use. In contrast, in this dissertation, the influence of individual characteristics are
believed to determine the use of media in general, the motives for using media, and the
conditions for contact with media. For example, gender, age, income and occupation are
43
three heavily used variables to explain web use by industry and academic researchers’
alike. In addition, Swoboda (1998) concluded that the conditions under which consumers
use interactive media are primarily determined by their involvement, and their
selective media consumption and direct media contact are primarily determined by
However, in an area such as web usage and network navigation of hypertext systems,
user perceptions of the medium may play a significant role in influencing current web
session usage frequency, usage variety and the usage extent. Drawing from the uses and
perceptions from the micro perspective of usage – i.e., web site use. Eighmey (1997)
investigated the impact of perceptions of site design and site satisfaction on web use and
found that users are assisted by information being placed in an enjoyable context and the
site being ‘easy to use’. In a further study, Eighmey and McCord (1998) identified that
site factors associated with entertainment value, personal relevance and information
involvement accounted for the largest proportion of total variance in web site
satisfaction. The responses of the research participants in this study confirmed the earlier
study, that an enjoyable context is important. Napoli and Ewing (1998) further identified
a number of attributes of web sites that are deemed most important by users and found
These studies show the value of investigating user perceptions of the web. However, the
focus has been on the micro perspective investigating user perceptions of the structure
and content of specific web sites. As outlined in section 3.4, this study investigates the
web itself, not specific web sites. Therefore, the first of three main research questions for
44
RQ1: What is the relationship between a user’s perception of the web and a person’s
current web session usage?
This study draws from a model developed in information technology and management
challenging issue. This is particularly so where the environment changes, as is the case
with the increasing use of electronic technologies. Therefore, to aid successful design and
usage of these systems is required. This was discussed in this chapter. Specifically:
The distinction between current usage and past usage was drawn.
The foundations were laid for a discussion of user perceptions of the web as an
The next chapter discusses user perceptions of the web, and specifically reviews the
45
C HAPTER 4: PERCEIVED EASE OF WEB USE AND WEB
USEFULNESS
‘The world changes according to the way people see it, and if you alter even by a
millimetre the way people look at reality, than you can change it’
4.1 INTRODUCTION
Computer adoption, acceptance and usage by consumers depends on internal factors
context). The acceptability, adoption and use of the web – as against computer adoption
in general – presents its own challenges. For instance, in Chapter 2, problems using
graphical browsers were discussed (Conklin 1987; Foss 1989). Such problems may
influence a user’s perception of the web and inhibit use. Therefore, in this chapter, the
impact on usage of the web of two specific perceptions is examined - perceived ease of
46
The process of perception is investigated, drawing on principles from consumer
psychology. These principles have been applied in the fields of Information Technology
(IT) and Management Information Systems (MIS) to gauge the impact of consumer
perceptions on information system adoption and use. Particularly relevant in this context
is the Technology Acceptance Model (TAM). This discussion provides a platform for
addressing RQ1 – the relationship between user perceptions of the web and current web
session usage.
4.2 PERCEPTION
Perception is formally defined as the ‘process by which an individual selects, organises
and interprets stimuli into a meaningful and coherent picture of the world’ (Schiffman,
The study of perception, according to Schiffman et al., (1997) is largely the study of what
we subconsciously add to, or subtract from, raw sensory inputs to produce our own
private picture of the world. Our perception of the world is thus formed by our
individual needs, drives, past experiences, motives, personality and learning. Therefore,
different individuals derive different meanings from the same sensory information, i.e.
47
In the context of purchase and information search behaviour, the buying decision is
individual’s perception that the item is capable of fulfilling particular needs (Kassarjian
In the more specific context of a user’s perception and adoption of technology, the
Technology Acceptance Model (hereafter TAM) has been the basis of most research. The
TAM model is drawn from the management information systems discipline. It was
technology. The approach is consistent with the diffusion theory proposed by Rogers
observability. Moore and Benbasat (1991), however, found that only three innovation
information technology (see Chapter 2), TAM is examined in this dissertation to assess
the influence of perceptions of the web on current web session usage (RQ1).
the most challenging issues in information systems research (Swanson 1988). A long-
standing objective of MIS research has been to improve our understanding of the factors
48
in organizations (Keen 1980). Studies from these areas have investigated the impact of
behaviour (Srinivasan 1985; Swanson 1987). Furthermore, they have examined how these
internal beliefs and attitudes are influenced by various external factors (e.g., system
technical design) (Benbasat and Dexter 1986) and user characteristics (e.g., cognitive
style) (Huber 1983). However, research findings have been mixed and inconclusive
about perception as a determinant of user adoption, acceptance and use of the system.
From this premise, Davis (1986) developed and tested an adapted form of Fishbein and
Ajzenʹs (1975) theory of reasoned action (TRA), called the technology acceptance model
(TAM) (Figure 6). This was meant to explain computer usage behaviour and specifically
looked at the development and testing of the effect of system characteristics on user
major determinant of whether or not he/she actually uses it. This attitude toward using is
in turn a function of two major beliefs: perceived usefulness and perceived ease of use.
Perceived usefulness is defined as ‘the degree to which an individual believes that using
a particular system would enhance his or her job performance’ and perceived ease of use
is defined as ‘the degree to which an individual believes that using a particular system
would be free of physical and mental effort’ (Davis 1986). Perceived ease of use further
49
has a causal effect on perceived usefulness and design features directly influence
Tests of the model confirmed several of the relationships hypothesized and refuted
others. Both a survey and an experiment were conducted. In the survey, Davis (1986)
found that:
Perceived ease of use had a significant effect on both usefulness and attitude;
Therefore, the TAM motivational variables – attitude toward using, perceived usefulness
and perceived ease of use – taken together, fully mediate between system design features
and self-reported usage behaviour. The survey results are shown in Figure 7. Behaviour
in this case relates to use of PROFs e-mail and XEDIT™ file editor.
50
Experimental data further supported the theoretical causal structure of TAM. Perceived
usefulness had a significant effect on attitude toward using the system. It also had a
contrast, had a limited effect on attitude toward using and no direct effect on behaviour.
Perceived usefulness was thus found to be more important than perceived ease of use at
determining self-predicted system usage. The experimental results are shown in Figure
WriteOne™, Davis, Bagozzi and Warshaw 1989a; Bagozzi, Davis and Warshaw
1992).
Perceived usefulness and ease of use are significantly correlated with self-reported
indicators of system use (for the PROFS e-mail system and the XEDIT™ file editor,
Davis 1989b; and for micro-computers, Igbaria, Guimaraes and Davis 1995).
Perceived usefulness has a significantly stronger relationship with system usage than
perceived ease of use (for the PROFS™ e-mail system and the XEDIT™ file editor,
Davis 1989b; for e-mail and voice mail, Adams, Nelson and Todd 1992; for word
processing, Bronson 1999; and for CONFIG™, Gefen and Keil 1998).
mail system and the XEDIT™ file editor, Davis 1989b; for word processing, Bronson
51
1999; for Microsoft™ Word and Excel, Chau 1996; for micro-computers Igbaria et al.
Perceived ease of use has a significant positive effect on the attitude toward using an
information system (for debugger DBG™ program, Bajaj and Nidumolu 1998).
toward using the system (for the PROFS™ e-mail system and the XEDIT™ file editor,
Perceived developer response has a significant effect on both perceived ease of use
However, in contrast to the above findings, further research has also reported that:
Perceived ease of use does not have a significant relationship with perceived
Perceived ease of use does not have a significant relationship with behavioural
The importance and explanatory power of perceived ease of use and usefulness on
and channel disposition have also supported TAM. As further outlined above,
52
relationships because of inconsistent findings. One of the major modifications considered
has been the removal of the constructs of intention to use and/or attitude and instead the
investigation of the direct effect of perceived ease of use and perceived usefulness have
directly on usage.
setting, for the use of information processing and communication systems, and with
reference to improving performance and outcomes. Adams et al. (1992) suggests that
problems might arise in testing TAM in these contexts as system usage may be
considered mandatory or, if not mandated, the system has become a defacto standard.
For example, the low explanatory power of TAM for Wordperfect™, found by Adams et
al. (1992), was possibly influenced by the view of what others thought users should be
doing – thus, the subjective norm is having an influence. Moore and Benbasat (1991)
report that mandatory use of information technology has a positive impact on usage and
that, in situations of mandated use, other factors tend to have less ability to explain
TAM has also been applied to a number of technological system developments such as
the web. With specific reference to the web it has been found that (Appendix A):
Perceived usefulness and perceived ease of use predict usage but that usefulness has
a stronger effect (Fenech 1997; Teo, Lim and Lai 1999; Lederer, Maupin, Sena and
Zhaung 2000).
53
Perceived ease of use predicts usage (for Netscape™, Morris and Dillion 1997);
perceived usefulness and attitudes (Moon and Kim 2001) (for Netscape™, Morris
al. 1992).
Perceived ease of use has a significant effect on perceived enjoyment (Teo et al. 1999).
and Benbasat 1991); and on attitudes (Moon and Kim 2001) (for Netscape™, Morris
Perceived playfulness has a significant effect on attitudes toward using the Web and
Attitudes have a significant effect on behavioural intention and this has had a
significant impact on usage (Moon and Kim 2001) (for Netscape ™, Morris and
Dillion 1997).
However, it should be noted that these applications of TAM to the web and browser
commented by Fenech (1997), ‘the application of TAM, as proposed by Davis (1986) and
further modified by Davis et al. (1989a) and Davis (1989b), to different technological
testing TAM on an information system that is not characterised by mandatory use, the
factors highlighted by Moore and Benbasat (1991) and the problems suggested by
Dishaw and Strong (1999) further contend that one of the major weaknesses of TAM for
understanding IT usage is its lack of task focus. They indicate that IT is a tool for which
users accomplish certain organisational goals and thus inclusion of task characteristics
ought to provide a better model of IT usage. In this dissertation, therefore, TAM is tested
54
4.4 WEB PERCEPTIONS AND CURRENT WEB SESSION USAGE
The first research question of this dissertation was stated in Chapter 3:
RQ1: What is the relationship between a user’s perceptions of the web and a person’s
current web session usage?
Current web session usage is defined as the act or fact of using the web for some purpose
within the current time period (i.e., within the period in which the measurement
occurred) (Delbridge and Bernard 1998) (Chapter 3). It was further noted that current
web session usage comprises three categories: usage frequency, usage variety and extent
of use. Prior research has focused on the impact of user perceptions on usage frequency
and, to a lesser extent, usage volume and duration. Here all three categories are
Figure 9: Perception of the Web and Current Web Session Usage (RQ1)
frequency of the information system is the most heavily used measure of system usage.
For non-web based systems, overall usage frequency (Davis 1986; Bajaj and Nidumolu
1998; Bagozzi et al. 1992; Adams et al. 1992; Davis 1989b) and frequency during a certain
time-frame (e.g., a week) (Bajaj and Nidumolu 1998), have been consistently used to
measure system usage. In addition, duration of usage (Davis 1986; Adams et al. 1992),
55
amount of use (Adams et al. 1992), and potential usage (Bajaj and Nidumolu 1998) have
been used.
For web-based systems, overall usage frequency (Moon and Kim 2001; Teo et al. 1999;
Lederer et al. 2000; Fenech 1997), frequency during a certain time-frame (e.g., daily,
weekly) (Teo et al. 1999) and volume of usage (Moon and Kim 2001; Fenech 1997), have
been used as indicators of web-based system usage. Teo et al. (1999) also defined internet
usage in terms of the diversity of usage – this is consistent with the fact that there are
many tasks/motivations for using the web (i.e., getting information, getting product
These indicators of frequency, duration, volume and diversity of use are consistent with
the three categories of usage proposed in Chapter 3. Each is examined in order to derive
session on the web is undertaken within a certain time frame’ (i.e., how often the web is
accessed). Adams et al. (1992) in study 1, Davis (1989b) and Igbaria et al. (1995) found a
positive relationship between perceived ease of use and usage frequency for non-web
based systems. With respect to the internet and web-based systems, there is some
support for a relationship between perceived ease of web use and current web session
usage frequency (Teo et al. 1999; Karahanna and Straub 1999; Lederer et al. 2000; Fenech
1997).
However, inconsistency in findings is evident in the literature with studies also reporting
a minimal or no relationship between perceived ease of use and usage frequency (Davis
1986; Adams et al. 1992 in study 2, Bagozzi et al. 1992; Taylor and Todd 1995). One
explanation for these mixed results is the nature of the relationship. A curvilinear
56
relationship may exist which, if assessed by linear analytical techniques, might be classed
as ‘no relationship’.
For example, frequency of use may be very low among those with few skills or those
Woodruffe-Burton 1999). But frequency of use might rise among those who are
developing skills – they respond to the novelty and challenge of the system. Frequency
of use may fall again among those who see the system as easy to use – this might be
It is hypothesized that:
H1A: Perceived ease of use of the web will have a curvilinear relationship with current
It is further evident in the literature that perceived usefulness has a stronger relationship
than perceived ease of use with usage frequency for non-web based (Davis 1986; Adams
et al. 1992; Davis 1989b), and web-based systems (Teo et al. 1999; Karahanna and Straub
1999, Gefen and Straub 1997; Lederer et al. 2000; Fenech 1997). Thus the more useful the
H2A: Perceived usefulness of the web will have a positive relationship with current web
situations and different motivations for web usage, independent of how frequently it is
used’. Teo et al. (1999) measured diversity of internet usage by looking at tasks and
motivations. They found that perceived ease of use had a positive impact on diversity of
use and further argued that users are driven to adopt and use a system primarily
useful, the user is more than willing to cope with an element of usage difficulty. Igbaria
57
et al., (1995) further found support for the relationship between perceived ease of use
H3A: Perceived ease of use of the web will have a curvilinear relationship with current
H3B: Perceived ease of use of the web will have a curvilinear relationship with current
H4A: Perceived usefulness of the web will have a positive relationship with current web
H4B: Perceived usefulness of the web will have a positive relationship with current web
usage and comprises the degree of duration (time), breadth (range) and depth (amount)
of session use’.
When perceived ease of use is low, use of the web will also be low. Users will not find
the web easy to use at first. They will not bookmark a lot of sites, they are unlikely to
visit a lot of sites and/or different new sites, and they will spend relatively less time
surfing the web. However, as perceived ease of use rises to a medium level, current
usage will increase – users will explore the web and thus the time spent and the number
and breadth of applications accessed will rise. As ease of use becomes high, efficiency in
use will occur. There might be a fall in the number of web sites that are accessed (e.g.,
because of bookmarking) (depth of web use), web sessions might take less time (duration
of use), and fewer new sites might be accessed (breadth). Firstly, then, it is hypothesized
that:
58
H5A: Perceived ease of use of the web will have a curvilinear relationship with current
H5B: Perceived ease of use of the web will have a curvilinear relationship with current
H5C: Perceived ease of use of the web will have a curvilinear relationship with current
Overall the central proposition is that the less useful you perceive the web to be to use,
the fewer sites will be accessed and bookmarks recorded. At a medium level of
usefulness, the number of sites accessed and bookmarked will increase, as will the time
spent online. Number of sites accessed and so forth will further increase as you come to
H6A: Perceived usefulness of the web will have a positive relationship with current web
H6B: Perceived usefulness of the web will have a positive relationship with current web
H6C: Perceived usefulness of the web will have a positive relationship with current web
It is acknowledged that external factors will have an impact too, such as where the
session usage and perceived ease of use will have a curvilinear relationship with current
session usage. Overall this indicates that perceived usefulness is the primary influence,
and perceived ease of use the secondary influence, on web usage behaviour.
59
4.5 DETERMINANTS OF WEB PERCEPTION
Due to the hypothesised influence of user perceptions on behaviour, it is worth
examining what may predict and/or determines a user’s perception of the web.
aid marketers and decision makers. They might, for instance, influence users to engage
in increased usage behaviour, explore the web for new and different sites and tools, etc.
Two determinants are noted: characteristics of the system, and characteristics of users.
This builds on the ideas of Krech et al. (1962) who suggested human perception is
influenced by two distinct factors: stimulus factors (e.g., browser) and personal factors
(e.g., experience), and further specified that perception is a result of both. As outlined in
Kassarjian and Robertson (1968), the perceptual organization of stimuli in the nervous
system is related directly to the nature of the physical object and, furthermore, is in part
determined by the motivations and need value-systems of the observer (Kassarjian and
Robertson 1968). Hence an interaction occurs between the stimuli and the observer in the
Much of the TAM literature focuses on characteristics of the system (Davis, 1986). By
factor, knowledge content, on a user’s perception of the web. This relates to issues
concerning the impact of experience and learning, training and education, exposure to
the technology itself and the level of use, comfort, skill and/or expertise of the individual
using the technology (Davis and Bostrum 1993; Nelson 1990). This gives rise to the
RQ2: What is the relationship between a user’s knowledge content of the web and a
person’s perceived usefulness of the web?
60
RQ3: What is the relationship between a user’s knowledge content of the web and a
person’s perceived ease of web use?
using graphical browsers and these problems might inhibit web use – or influence a
user’s perception of the system. Here, the influence of two specific perceptions are
relationships are hypothesized. In doing so, the work draws upon the Technology
Acceptance Model (TAM). In addition, this chapter lays the foundation for the
introduces the influence that user knowledge content of the web may have on a user’s
perceived ease of use of the web (RQ2) and perceived usefulness (RQ3) of the web.
61
C HAPTER 5: U SER K NOWLEDGE C ONTENT
OF THE W EB
- Socrates -
(470-399 BC)
5.1 INTRODUCTION
Drawing from the cognitive science literature, and studies in consumer research on
specifically the web. In the previous chapter, user perceptions were reported to influence
the adoption and acceptance of web-based technology. This chapter examines human
User knowledge content of the web and perceived usefulness of the web (RQ2);
User knowledge content of the web and perceived ease of web use (RQ3);
62
A discussion follows on the conceptualisation of consumer knowledge content.
Nicosia (1966), Howard and Sheth (1969), Engel, Kollat and Blackwell (1968). In a
revision to their models of the buyer’s decision-making process, Engel et al. (1990)
memory, external search and internal search. Furthermore, Bettman (1979) included
attention, information acquisition and evaluation, memory search, external search and
decision processes. Both of these models address the concept of gathering and using
ʹknowledge.ʹ These models, however, have come under criticism and debate regarding
their testability and plausibility as realistic views of consumer behaviour (Fletcher 1988;
Notwithstanding the above criticisms, support for the stages of the decision making
literature and as such the decision making process has been reported in a number of
generic marketing texts (i.e., Assael 1995; Barry 1986; Cravens and Woodruff 1986;
Hawkins et al. 1997; Kindra, Laroche and Muller 1994; Kotler and Armstrong 1994;
Schiffman et al. 1997; Solomon 1994). The model in Figure 11 shows a five step sequence
that a consumer goes through during the decision making process. These steps include:
or post-consumption.
63
Figure 11: A Simple Model of the Consumer Decision Making Process
(Schiffman et al 1997; Hawkins et al 1997)
behaviour, such as information search and decision making, for example Feick et al.
(1992). However, the role of knowledge is not explicit in the simple model presented in
Figure 11; rather, it is implied that the acquisition of knowledge takes place within the
second stage (i.e. information search) and is used in the third stage (i.e., evaluation of
alternatives). Researchers, however, have recognised both the importance of and need
for research on consumer knowledge stating that ‘despite the recognised importance of
independent area of research and theorising’ (Alba and Hutchinson 1987). Alba and
Marmorstein (1987) further stressed the need for research of consumer knowledge
(stored in memory) about a domain (Delbridge and Bernard, 1998). This information is
its type of contents (knowledge content) and may be measured in different ways
knowledge - knowledge scope. Following Brucks and Mitchell (1981) and Kanwar et al.
64
(1981), consumer knowledge is characterized by the content, structure and measurement
Presented in this study are two well-accepted types of consumer knowledge content
(declarative and procedural knowledge) and two additional types (common and
knowledge content has been drawn from recent empirical studies and discussions of
familiarity and expertise. This results in the conceptualisation in this chapter, and later
memory is important, however the concept of schemata, and similar concepts, such as
frames and scripts, provide the background for the current area of investigation. This
knowledge content that is structured in a consumer’s memory, but not the nature or
diagrammatic profile of its organization and thus knowledge structure is not directly
proposed that the type and measurement of knowledge content stored in a consumers
memory guides, controls and influences human behaviour. In addition, structure is not
examined, in keeping with the view that appropriate measures of content need to be
devised before the structure or organization of the content can be assessed (Mitchell
or facts required for ‘driving a car’. The process of how humans learn information can
assist our understanding of the different types of knowledge content stored in memory.
For example, as will be discussed later in this chapter, learning is viewed as a three stage
process (Anderson 1990; Fitt 1964) involving firstly a cognitive stage in which factual
65
knowledge of the domain is learnt (i.e., learning what a car is and what gears are), then
an associative stage, in which a method for forming productions and procedural skills is
developed (i.e., learning how to drive a car and how to change gears) and a third stage,
an autonomous stage, in which use of the productions becomes faster and more
automatic (i.e., habitual). From this process of learning, two distinctive types of
knowledge content consistently emerge, namely declarative knowledge (i.e., what) and
procedural knowledge (i.e., how) (Anderson 1976; Anderson 1983; Best 1989; Dodd and
White 1980).
Baddeley (1991) further states that learning is used for the acquisition of domain specific
knowledge and the mastering of new skills. The discussion by Baddeley (1991) lends
itself to the proposal by Alba and Hutchinson (1987) of two components of knowledge,
familiarity and expertise. The treatment of familiarity and expertise as synonymous with
knowledge (Dacin and Mitchell 1984; Johnson and Russo 1984; Spence and Brucks 1997;
Chi, Glaser and Rees 1982; Hastie 1982) provides impetus in this study for the
knowledge content. This classification is guided by the assumption that consumers not
only have differing knowledge of ‘what’ gears are or ‘how’ to use them, but can also
have stored in their memory, more common knowledge about how to change gears
These components of the 2x2 typology are discussed further in the following sub-
sections.
Anderson 1983; Best 1989; Dodd and White 1980). These two types of knowledge content
are critical concepts for this research, and differ both in their nature and in the functions
66
which they allow humans to perform. Declarative and Procedural knowledge content are
Referred to as knowledge about concepts, objects or events (Brucks and Mitchell 1981;
static in nature which is usually describableʹ (Best 1989, p7). This factual information is
and is organised as schema to aid comprehension. Best (1989) further added that
ʹdeclarative knowledge is flexible and can often be organised to suit our purposesʹ (p7).
Anderson (1983) indicated that declarative memory has strength, and that this strength
reflects the frequency of use and importance to the person of the item of information,
partly explaining why some facts are remembered and others forgotten. Brucks, (1986)
facts and usage situations of the domain of interest. For example, ‘this car is fast’ is a
factual statement about the car, whereas ‘colour’ and ‘model’ may be considered
ʹattributesʹ of a car.
underlying skilful actions (Best 1989, p7); thus, the knowledge of rules for taking action
that is believed to be stored and organized into production systems (Brucks 1986).
Brucks and Mitchell (1981) specified that the basic elements of production systems are
knowledge comprised information about the process and procedures for domain usage
and decision-making. For example, the statement, ‘to stop a fast car, the brake pedal is
condition (stop car) - action (push brake pedal to floor). Best (1989) added that the
67
organization of procedural knowledge content is not well understood, and that
The basic distinction, then, is between ʹknowing what ʹ (declarative) and ʹknowing howʹ
(procedural). While both declarative and procedural knowledge can guide behaviour,
the latter is considered of greater influence on actual behaviour. For example, in the
either not available or not so easy to interpret and thus require procedural knowledge
(Anderson 1983). Therefore, to drive a car, knowing ʹhow to drive a carʹ (procedural) will
have more influence on the end behaviour, (i.e., driving) than knowing ʹwhat driving a
While these two terms appear to be concise and definitive, it has been noted that the
difference between declarative and procedural knowledge may not be as distinct as the
definitions imply (Anderson 1983; Best 1989; Baddeley 1991). There is evidence to
indicate that some knowledge is initially encoded in a declarative format, and if used
either in whole or in part (Best 1989). Anderson (1983) referred to the acquisition of
ʹproductionsʹ as indicating that they are not ʹdirectly acquiredʹ but evolve within a
68
procedurally encoded format which allows people to do things, for example, to learn a
commences with an interpretive stage using declarative representations and, with time,
skill specific productions are compiled (Anderson 1983). However, during this
transformation process there is a phase when the difference between declarative and
procedural encoding formats is unclear and difficult to differentiate. As the true nature
of this interaction is not well understood (Anderson 1983), no clear determination of the
A number of researchers have studied buyer ʹproductʹ knowledge content with mixed
findings and mixed support for the measurable difference between declarative and
procedural knowledge content. For example, Brucks and Mitchell (1981) identified the
difference between declarative and procedural knowledge, but did not test them
separately; Mitchell (1981) reviewed a number of methods for measuring declarative and
procedural knowledge; and Dacin and Mitchell (1984) measured existing declarative
knowledge and identified that some aspects of declarative product knowledge could be
measured.
To summarise, two types of knowledge content – declarative and procedural – have been
different roles in memory. It is also apparent from the above discussion that in some
circumstances the clarity between these two types of knowledge can become muddied
and unclear. However, the ability of procedural knowledge content to control behaviour
further research into these types of knowledge content. In addition, from the discussions
69
5.4.2 COMMON AND SPECIALISED KNOWLEDGE CONTENT
The terms ‘expertise’, ‘novice’, and ‘familiarity’ give rise to discussion about the
classification of these elements within the knowledge framework. Do they pertain to the
consumer memory, how information is used by the consumer, or the type of knowledge
content stored in memory? For example, expertise and familiarity has been defined in the
literature with reference to the source (Alba and Hutchinson 1987; Spence and Brucks
1997), the outcome (Alba and Hutchinson 1987; Enis 1995), and the amount (Chi et al.
Alba and Hutchinson (1987) use the term expertise ‘very broadly that includes both the
cognitive structures and cognitive processes for taking action.’ Therefore, discussing
addition, their definition of product familiarity (i.e., the number of product related
experiences that have been accumulated by the consumer) represents neither the quality
nor kind of experience accumulated, just the quantity. In many of these studies,
familiarity and expertise are treated as synonymous with knowledge (Johnson and
Russo 1984; Bettman and Park 1980). These considerations give rise to the view that the
different types of knowledge content. From this, specialized and common knowledge
Conceptual and empirical studies of expertise have made reference to the quantity and
type of knowledge content that a consumer has acquired as key characteristics that
differentiate experts from novices. Defined by Spence and Brucks (1997), an expert or
someone with expertise has acquired domain specific knowledge through experience
and training. Chi et al. (1982) further identify that expertise is the possession of a large
body of experience, knowledge and procedural skill. In the marketing literature expertise
is further regarded as the high level of relevant skill and knowledge an individual has to
70
perform product related tasks successfully (Alba and Hutchinson 1987; Homer and
The Macquarie Concise English Dictionary (1998) defines an expert as a person who has
special skill or knowledge in some particular field, a specialist (Delbridge and Bernard
1998). For example a ‘mechanic’ has a level of specialised skill with the mechanics of a
car engine. It is further outlined that having expertise is the possession of special skill or
these definitions stems from the terms special and specialized, thus distinguished or
different from what is ordinary or usual. From the above discussion, this study defines
interest required to perform skilled domain related tasks successfully.ʹ An expert would
be classified as having a high level of specialized knowledge and a novice a low level of
specialised knowledge.
de Bont and Schoormans (1995) find evidence to support the view that the problems
inherent to the early stages of the product-development process poses less of a concern
for consumers with much product expertise (i.e., those having specialized knowledge),
than those with little product expertise. This is because of an expert’s detailed cognitive
structure with respect to products in the category, the availability of more product
related information in their cognitive structure, and their ability to discriminate between
product attributes.
class (Johnson and Russo 1984) and the number of ‘product’ related experiences
involve advertising exposure, product purchase, or product usage. This term, familiarity,
has been used in association with expertise (Johnson and Russo 1984), however Hastieʹs
71
(1982) commentary on generic knowledge actually provides more depth as to the type of
Hastie (1982) specifies that generic knowledge includes general information about
classes of products, and instances exemplifying those products. The Macquarie Concise
(Delbridge and Bernard, 1998)). For example, a familiar and common element of a car
would be the gears and a familiar or common procedure would be using them. The
foundation of this definition stems from the term ‘common’ – i.e., widespread,
ordinarily, generally or publicly known (Delbridge and Bernard 1998). Based on these
information of the domain of interest required to perform general and common domain
Common Declarative
‘General and/or publicly known static information of facts, terms, attributes (what) of
Specialised Declarative
Common Procedural
(how) of using X, required to perform general and common domain related tasks
successfully’
72
Specialised Procedural
stored in memory. For example, Johnson and Russo (1984) classify respondents into
three categories of familiarity, low, moderate and high and refer to them as ‘snapshots’
As previously discussed, expertise has also been defined as the possession of a large
body of experience, knowledge and procedural skill (Chi et al. 1982). The information-
However, Shanteau (1992) contends that the amount of information used or stored in
memory does not necessarily reflect a degree of expertise, but that the type of
information used or stored does. Shanteau (1992) classifies the type of information as
either relevant/irrelevant. This study develops his argument, but adopts a different
approach to classifying expertise with reference to the type of knowledge content stored
the type of knowledge content stored in a consumer memory. This then can be used as a
73
basis for classifying consumers: the scope of specialized knowledge content that a
consumer acquires will indicate their classification as an expert or novice, and the scope
domain of interest. This enables consumers to be classified as both having common and
For example, a car mechanic with 15 years vocational experience might be categorized as
highly familiar and as having specialised knowledge of cars. However, a school teacher
with 15 years vocational experience might be categorized as highly familiar but with low
specialized knowledge of cars. See Table 4 for a depiction. Although in this example,
vocation could be used as a proxy for the type of knowledge content, depending on the
domain of interest, this might not always be the case. For example, what about the school
teacher whose hobby is car remodelling or the mechanic who specialises in motorbike or
powerboat engines?
particularly the use of proxies and the use of objective and subjective methods for
One of the most common methods for measuring consumer knowledge has been the use
of proxies to infer consumer knowledge. For example, heavy use has been made of
domain usage and purchase experience to measure consumer knowledge (Bettman and
74
Park 1980; Johnson and Russo 1984; Cole, Gaeth and Singh 1986; Woodside, Trappey and
MacDonald 1997; Park, Mothersbaugh and Feick 1994). This might be a convenient
In addition, objective and subjective methods for measuring consumer knowledge have
been well documented (Brucks 1985; Dacin and Mitchell 1984; Rao and Olson 1990).
prejudice, used to assess the information about a domain (i.e., that belong to a domain)
assess ‘actual’ knowledge of the domain of interest through the undertaking of tasks,
about the information of a domain (i.e., that belong to the individual) that is believed by
In practice, these methods are sometimes confused with ‘types of knowledge content’. A
like ‘types of knowledge content’ and in addition provide definitions with respect to the
domain of interest and the type of knowledge content investigated. For example, Alba
and Hutchinson (1987) and Homer and Kahle (1990) define objective knowledge as
expert a user perceives that they are about a productʹ (Alba and Marmorstein 1987;
Brucks 1985). Thus, methods of knowledge measurement are defined with reference to
The definitions adopted in this study do not see the treatment of the method as a ‘type of
thinks they knowʹ (perception) and ʹwhat an individual actually knowsʹ (stored in
75
5.6.2 COMPARATIVE DISCUSSION
Proxies are heavily used as a way to infer a consumer’s actual knowledge. These
methods assume that people learn from experience and at the same rate. In addition, the
argument has been presented by Kanwar et al. (1981), Marks and Olson (1981), and Alba
and Hutchinson (1987), that objective measures of knowledge were operationally and
However, there is some doubt about the correlation between what people think they
know, what their experience is, and what is actually stored in a consumers memory
(DeNisi and Shaw 1977; Fischoff, Slovic and Lichtenstein 1977; Lichtenstein and Fischoff
1977; Nelson, Leonesio, Shimamura and Landwehr 1982; Park 1982; Schacter 1983). For
example, Brucks (1985) concluded that experience measures and subjective measures,
while useful in other contexts, do not represent measures of the actual knowledge stored
in a consumer’s memory. Feick et al., (1992a) further found only modest correlations (0.4
to 0.6) between subjective and objective measures of knowledge. Park et al. (1994)
cues that underlie information retrieved by self-assessed and objective measures. Both
these results support the findings of Brucks (1985). However, Selnes and Gronhaug
(1986) reported a significant relationship between subjective and objective measures and
Cole et al. (1986) found significant correlations between objective and subjective
measures in the case of a single product (convergent validity), and low correlations in the
Selnes and Gronhaug (1986) and Mitchell (1981) contend that selection of objective or
measures are preferable when the research is related to the consumer’s ability to encode
consumer’s memory is identified by Brucks and Mitchell (1981) and Engel, Blackwell and
Miniard (1990), concluding that objective measures of knowledge were better than
76
experience and self-report measures. Engel et al. (1990), however, also cautioned that
objective measurement ʹis by no means an easy task, given the vast array of relevant
knowledge content of the web. This is important for a variety of reasons. As presented
in Chapter 2, arguments presented by Mandelli (1997) and Hoffman and Novak (1996)
support the view that developments in electronic technology are changing the roles
technologies and the user is also changing. Because of the active and changing role
played by the user, he or she makes a contribution towards the success of the
interact with the web rely on the highly developed visual spatial processing of the
human visual system and as there is no natural typography for the web, until one is
ʹcognitive overhead’ and ʹlearning by browsingʹ (Foss 1989) – were also discussed in
the hypermedia system and the user-interface. Reed and Oughton (1997), for example,
found that hypermedia knowledge in general had a large impact on user productivity. It
was found that learning style and hypermedia knowledge play an important role in how
knowledge content presented empirical evidence that perceived skill and challenge
Diaz, Hammond and McWilliam (1997), identified that experience with the web (i.e., as a
77
proxy for knowledge) might be an important moderator of: attitude toward the medium,
the reported success of the activity undertaken, and the placement of value on web-
based information.
Their results further suggested that experience may not influence behaviour and
attitudes in a linear fashion, but that the most experienced users might be enthusiasts for
the medium, while moderate users may be technically competent, but enjoy the web less
than the enthusiasts (and less than novices). Experienced (i.e., heavy) users, according to
Diaz et al., (1997), further found the web more legible, more stimulating and possessed
higher company-brand recall than novices. They found no specific difference between
novices and those who are moderate users of the web for any of the question asking
about information on the web. However, major differences existed between the most
experienced users and the moderate and novice users, with the former placing a higher
value on web-based information. Heavy users were more than likely to feel that the web
experience matched their expectations and they were more likely to agree that the
These applications and findings are very suggestive, however there are significant
problems with these studies. For this reason, further investigation and measurement of
consumer knowledge of the web is needed. For example, the study by Diaz et al. (1997)
differentiated novice and expert users of the web is a very loose way, based on hours
spent on-line. By contrast, using the definition discussed in section 5.4, a web expert
would be defined as having a high level of specialized web knowledge, thus ‘skilled
and/or extraordinary information (procedural and declarative) about the web required to
perform skilled web related tasks successfully.’ Thus, Diaz et al. (1997) only measures a
In fact this use of proxy measures of a consumer’s knowledge of the web is widespread
in the academic literature and commercially, with most investigations of web users
moving very little beyond measures of direct experience (i.e., hours, months and years
medium used) and ʹperceived knowledgeʹ of the web (i.e., ʹI am extremely skilled at
78
using the web, compared to most usersʹ (Novak, Hoffman and Yung 1998). However, as
previously discussed, a poor correlation exists between what people think they know,
This argues for further research to examine consumer knowledge content of hypermedia
computer-based technologies like the web. It also argues for the development of reliable
measures of consumer knowledge of the web. By extension, this means looking at the
role consumer knowledge of the medium plays in influencing medium use and
knowledge about both the products they are interested in and about the environments in
which they access these products. This dissertation’s main objective is thus to investigate
the relationship between a user’s knowledge content of the web and a user’s perception
of the web.
RQ2: What is the relationship between a user’s knowledge content of the web and a
person’s perceived usefulness of the web?
In this study is hypothesised that a strong positive relationship exists between common
and specialised declarative knowledge content and perceived usefulness, and that an
and perceived usefulness, and a weak relationship will occur with specialised procedural
knowledge content.
With increasing declarative knowledge, users should acquire a perception of how the
web and its features ʹshouldʹ work. Thus a strong positive relationship is initially
content about the web, it is theorised they will start to realise the negative aspects of
79
usage, and thus become aware of system based problems that inhibit use. Thus high
common and specialised declarative knowledge makes them more aware of the
Handzic and Low (1999) found that more experienced users of processing programs
(Microsoft Word, etc.) had more favourable perceptions of the usefulness of the
technology. They felt that as users become more experienced with using processing
programs, they become more aware of certain program features and also more efficient
in the use of its attributes. However, the present study is concentrating on the web, not
H7A: Actual common procedural knowledge of the web will have a curvilinear
H8A: Actual common declarative knowledge of the web will have a positive relationship
H9A: Actual specialized procedural knowledge of the web will have a positive
H10A: Actual specialised declarative knowledge of the web will have a positive
H11A: Perceived procedural knowledge of the web will have a positive relationship with
H12A: Perceived declarative knowledge of the web will have a positive relationship with
H13A: Perceived general knowledge of the web will have a positive relationship with
80
5.7.2 WEB KNOWLEDGE CONTENT AND PERCEIVED EASE OF WEB USE
It is asked:
RQ3: What is the relationship between a user’s knowledge content of the web and a
person’s perceived ease of web use?
Individuals are constantly making decisions about accepting, adopting and using
information gathering; i.e., information gathering about the existence of the innovation
as well as its characteristics and features through a social system within which adopters
are situated. Nelson (1990) states that the acceptance of computer technology depends on
the technology itself and the level of use, comfort, skill or expertise of the individual
using the technology. Of relevance to this dissertation, research has reported perceived
and use. An understanding of this means practitioners and researchers might be better
However, little attention has been paid to what leads to the development of certain
perceptions about an innovation (Agarwal and Prasad 1998; Venkatesh and Davis 1996).
Handzic and Low (2000) found that subjects with moderate-high experience tended to
perceive information technology as substantially more useful than those with low
experience. They also found that a relatively modest level of experience may be sufficient
for individuals to gain most of the relevant knowledge about various aspects of a given
technology. With respect to new media, King and Xia (1997) note that an individual’s
perceptions of media will vary widely according to that person’s skill, comfort and use of
the media.
Karahanna and Straub (1999) examined the psychological origins of perceived ease of
use and perceived usefulness. They found that usefulness of email technology was
81
determined by social influences, perceived ease of use, and social presence. A
considerable body of evidence discusses the impact experience and learning has on user
perceptions of information systems. Hubona and Geitz (1997) note that standardised
interfaces promote ease of use, but training and education are also important.
One of the most crucial aspects of these findings is that with experience and exposure to
Prasad (1998) who state that the relationship between how information is obtained and
the development of perceptions about the innovation has not been extensively studied.
The propositions presented here are consistent with existing theory in this area – as
discussed above. However, this study goes one step further and hypothesises that ease of
web use is actually affected by web knowledge content. Specifically, it is proposed that
due to the fact that the web is experience driven, a strong relationship will exist between
perceived ease of web use and the acquisition of common and specialised procedural
knowledge, and a weak relationship will exist with common declarative knowledge, and
experienced, they gain more procedural knowledge, which makes the web seem easier to
use.
knowledge content and a user’s perception of ease of web use than with actual
knowledge content of the web stored in memory. This proposition is motivated by the
H14A: Actual common procedural knowledge of the web will have a positive relationship
82
H15A: Actual common declarative knowledge of the web will have a positive relationship
H16A: Actual specialized procedural knowledge of the web will have a positive
H17A: Actual specialised declarative knowledge of the web will have a positive
H18A: Perceived procedural knowledge of the web will have a positive relationship with
H19A: Perceived declarative knowledge of the web will have a positive relationship with
H20A: Perceived general knowledge of the web will have a positive relationship with
dissertation is the investigation of actual and perceived knowledge content of the web.
a 2x2 typology of knowledge content. Further, this chapter theorised the influence of
actual and perceived knowledge content of the web on a users perceived ease of use and
RQ2: What is the relationship between a user’s knowledge content of the web and a
person’s perceived usefulness of the web?’
RQ3: What is the relationship between a user’s knowledge content of the web and a
person’s perceived ease of web use?
83
C HAPTER 6: R ESEARCH Q UESTIONS AND
H YPOTHESES
6.1 INTRODUCTION
Outlined in this chapter are the main research questions and hypotheses that were
discussed in the first section of the dissertation and that will be tested in the final sections
of this dissertation. Overall, this dissertation investigates the relationships between the
“What is the relationship between a user’s knowledge and perception of the web and
a person’s current web session usage?”
84
Actual common/specialised declarative knowledge content is a primary determinant
(RQ3).
H1A: Perceived ease of use of the web will have a curvilinear relationship with current
H2A: Perceived usefulness of the web will have a strong positive relationship with current
H3A-B: Perceived ease of use of the web will have a curvilinear relationship with current
H4A-B: Perceived usefulness of the web will have a strong positive relationship with
H5A-C: Perceived ease of use of the web will have a curvilinear relationship with current
H6A-C: Perceived usefulness of the web will have a strong positive relationship with
85
H7A: Actual common procedural knowledge of the web will have a curvilinear
H8A: Actual common declarative knowledge of the web will have a positive relationship
H9A: Actual specialized procedural knowledge of the web will have a positive
H10A: Actual specialised declarative knowledge of the web will have a positive with
H11A: Perceived procedural knowledge of the web will have a positive relationship with
H12A: Perceived declarative knowledge of the web will have a positive relationship with
H13A: Perceived general knowledge of the web will have a positive relationship with
H14A: Actual common procedural knowledge of the web will have a positive relationship
H15A: Actual common declarative knowledge of the web will have a positive relationship
H16A: Actual specialized procedural knowledge of the web will have a positive
H17A: Actual specialised declarative knowledge of the web will have a positive
H18A: Perceived procedural knowledge of the web will have a strong relationship with
86
H19A: Perceived declarative knowledge of the web will have a strong positive relationship
H20A: Perceived general knowledge of the web will have a strong positive relationship
The next section of this dissertation discusses the methodological approach that was
87
C HAPTER 7: C ONSTRUCT O PERATIONALISATION
(Jacoby, 1978)
7.1 INTRODUCTION
The above statement supports the argument presented by Churchill (1979) that the
measures of the variables investigated. For this reason it is important to develop reliable
and valid instruments to measure the constructs that are of interest in this dissertation.
This is especially relevant given the lack of standardised measures in some of the more
and further discussion of item/scale development by Gerbing and Anderson (1988) and
Rossiter and Kayande (1999), assisted in the development, adjustment and purification of
of web use, perceived web usefulness and actual and perceived web knowledge content.
scale/item testing stage of item generation. First, in order to establish initial content
validity, the item generation process involved a number of primary exploratory studies –
88
an expert survey, a novice observational study, web site and help file content analyses
and a small number of in-depth interviews. A brief overview of the findings of these
analysis of existing scales and items (i.e., academic and industry measures) was
samples was implemented as the method for data collection for scale/item testing and
89
validation. One questionnaire was administered to two student samples across two
disciplines. This tested the properties of the scale items measuring perceived ease of web
use, perceived web usefulness and current web session usage experience. The second
questionnaire was administered to three student samples across three disciplines. This
tested the properties of the scales measuring actual and perceived web knowledge
content. Samples were selected based on web-usage experience and web-course content
integration.
The total sample size for each study consisted of at least 100 respondents in accordance
with the sample size requirements for conducting exploratory factor analysis (Hair,
Anderson, Tatham, and Black, 1995). The samples for the two studies were convenience
can be alarmingly low (Yu and Cooper, 1983), the questionnaire was administered in a
controlled environment to increase response rates and also to allow for differences in
Prior to scale testing and analysis an assessment was made of the appropriateness of
aggregating the samples (that is, the combination of two samples for one study, and the
combination of three samples for the other study). Then a principal components
exploratory factor analysis was conducted on the each scale developed to assess scale
analysis was conducted on all multi-item measures. This was done to determine the
stability and dimensional structure of the scales and to examine the reported consistency
90
7.2.3.1 Internal Consistency Reliability Analysis
Reliability refers to the extent to which a scale produces consistent results if repeated
measurements are made (Perreault and Leigh, 1989). Malhotra et al. (1996) suggests that
reliability can be further defined as the extent to which measures are free from random
the scale. If the association is high, the scale yields consistent results and is therefore
reliable. Hair, Anderson, Tatham and Black (1995) state that reliability is a measure of the
internal consistency of the construct indicators, depicting the degree to which they
indicate the common latent construct (Hair et al., 1995: p641). More reliable measures
provide the researcher with greater confidence that the individual indicators are all
consistent in their measurements. Approaches for assessing reliability include the test-
retest, alternative forms, and internal consistency methods (Malhotra et al., 1996).
Internal consistency reliability is used to assess the reliability of a scale where several
items are summed to form a total score (Malhotra et al., 1996). The items should be
consistent in what they indicate about the characteristic. As the item generation process
developed scales that would require the summation of respondents’ results, internal
consistency methods were used to assess the overall reliability of these scales.
The coefficient alpha is the average of all possible split-half coefficients resulting from
different ways of splitting the scale items (Cronbach, 1951 in Malhotra et al., 1996). This
coefficient varies from 0 to 1. Malhotra et al. (1996) and Tull and Hawkins (1993) indicate
reliability. Hair et al. (1995) and Nunnally (1978) stipulate a commonly used threshold
value for acceptable reliability is a Cronbach Alpha of 0.7 or higher. However, they
suggest that values below 0.7 have been deemed acceptable if the research is exploratory
91
From this review, measures with coefficient alpha of 0.6 or higher were regarded as
reliable in both the item testing phase of this study and also during the process of scale
Factor analysis is a multivariate statistical technique that can be ‘utilised to examine the
whether or not the information can be condensed or summarised into a smaller set of
technique’ in which factors are formed to maximise their explanation of the entire set of
variables, the two primary reasons for conducting a factor analysis are summarisation
and data reduction (Hair et al., 1995). These objectives can be achieved from either an
summarise the variables and explore the structure of the measurement instruments.
A principal component factor model was deemed appropriate, as the primary concern is
prediction of the minimum number of factors needed to account for the maximum
A measure of sampling adequacy (MSA) was applied. This provides a measure of the
extent to which variables belong together and are thus appropriate for factor analysis.
Kaiser and Rice (1974) as stated in Hair et al (1995) suggest excluding individual
The next step involves the selection of the number of components to be retained for
further analysis. Principal components analysis was used for this step. The following
92
Percent of Variance Criterion - Using the cumulative percentages of the variance
Scree Test Criterion - Visually examining the ʹplot of latent rootʹ (eigenvalues) against
the number of factors in order of extraction and assessing the optimum number of
factors that can be extracted before the amount of unique variance begins to
A Priori Criterion - Extract according to scale development and design and prior
studies.
The question then arises of whether the constructs are uni- or multi-dimensional. As
with any factor analysis, there is no exact quantitative basis for deciding the number of
dimensions to extract. A standard latent root criterion is used, where only those
dimensions with initial eigenvalues of at least 1 are retained. In a number of cases here
this results in uni-dimensional constructs. Given prior theory, this is not unexpected.
Moreover, it has the advantage of greatly simplifying subsequent analyses. However, the
latent root criterion tends to be quite conservative and therefore, for completeness,
results are also presented based on the percentage of variance criterion (Hair et al. 1995,
work with the greater number of dimensions that this criterion generates.
Important to this statistical application are factor loadings. Factor loadings are the
correlations of each variable within the factor under analysis and they indicate the
degree of correspondence between the variable and the factor, with higher loadings
indicating that the variable is representative of the factor (Hair et al., 1995). The criteria
for determining the significance of factor loadings within this research were based on the
guidelines reported by Hair et al. (1995) in relation to sample size. These authors suggest
that in a sample of 120 respondents, factor loadings of 0.50 or higher are significant and
for a sample of 150 respondents factor loadings of 0.45 are considered significant.
Therefore, scale items that report a factor loading lower than 0.50 for study one (i.e., web
93
usage and user perceptions) and scale items that report a factor loading lower than 0.45
interest. In the next two sections specific details are given of the item generation and item
testing and purification process for the scales measuring web usage and user web
perceptions (i.e., study one) and actual and perceived knowledge content of the web (i.e.,
study two).
measures of the three areas of current web usage experience and user web perceptions.
The results identified items that reliably measure current web usage frequency, usage
variety (situational), usage variety (motivational), usage extent (breadth), usage extent
(depth) and usage extent (duration). Existing scales from research in Management
Information Systems (MIS) were also used to develop measures of perceived ease of web
use and perceived usefulness of the web. The results identified scale items that reliably
measured perceived ease of the web use and perceived usefulness of the web.
Firstly the research and sampling design for this study will be discussed followed by the
item generation process for each scale and the presentation of the scale testing results for
the items measuring current web session usage and user web perceptions.
To assess the properties of the scale items developed to test perception and current usage
experience of the web, the sample was drawn from postgraduate (61%) and
94
undergraduate (39%) university courses (refer to Table 6). Convenience and domain
The postgraduate marketing class was sampled because these were expected to hold
variable perceptions. As the course being undertaken was not administered on the web,
it is expected that a variance in perceptions of the web may exist. Some would be aware
of the web as a marketing tool, others not. This group also matched Australian web-users
in general, these being described as predominantly ʹhigher educatedʹ and ʹmore affluentʹ
A lower level of web use and experience was expected among the first-year
undergraduates. These would have been exposed to only a limited amount of online
course material. Exposure to the web would thus be either in association with personal
usage or usage for other educational programs (this data was not acquired from
responses).
The survey instrument to measure user web perceptions (i.e., PEWU and PWU) and
current web usage (frequency, variety and extent) was administered in paper-format.
This was intended to increase the variance in domain experience. If conducted online,
course convenors emphasized the importance of the students participation. Despite high
95
expected response rates due to the method of administration, it was expected that the
quality of the responses would be variable because of the limited intrinsic motivation to
independent-sample means comparison analysis was conducted. It was found that the
means for the 20 scale items measuring perceived ease of web use, and 23 scale items
measuring perceived web usefulness, were comparable. The samples were also
It was found that the samples performed as expected on the descriptive items measured.
For example, both samples had a high incidence of full-time enrolments (P=87%, U=98%)
and a comparable gender distribution (Female: P=63%, U=78%). The age distribution for
the samples was reasonably comparable, although the undergraduate sample was
slightly younger (P= 73% - 27 and younger, U=80% - 19 and younger). In addition, the
samples were not comparable in terms of past web usage experience, with 75% of the
postgraduate sample having 3-6 years web experience and 68% of the undergraduate
sample having less than 2 years web experience. This result, however was expected and
the primary motivation for sample selection. Based on the above analyses, aggregation of
the independent samples was conducted to facilitate scale analysis and development.
For the scales developed measuring current web session usage and user web
perceptions, a principal components factor model with a varimax rotation was used. The
number of dimensions to be extracted for further analysis was selected based on the
following four criteria: the latent root criterion, cumulative variance percentages, the
scree plot, and prior knowledge. In addition, factor loadings of +/- 0.50 and above were
considered significant and thus for interpretation purposes, only items with factor
loadings above this threshold were considered. Reliability analyses were also performed
96
on all four scales to check for internal consistency, with items deleted if the corrected
item-total correlation fell below 0.6. In the following sections, each scale is discussed in
turn.
(69%) and 40 male respondents (31%). The age distribution was also skewed, with 78% of
the sample aged less than 25 years of age. Given the sampling design described above, a
significant bias towards the young age group is not surprising. However, even though
this sample is not representative of the demographic profile of the Australian web
population, this should not affect the results as the purpose here is scale development.
With regard to respondents direct web usage experience, the following results were
identified: 84% of the sample had 3 or more year’s computer experience, 40% had less
than 2 years and 43% had 3 to 4 years web experience; 56% accessed the web on a daily
basis and 42% on a weekly basis; 43% use the web on average for less than 1 hour per
occasion and 46% for 1-3 hours. Furthermore, 42% of the sample access the web from 2 to
3 different locations and 56% of the sample had between 2-3 email accounts. From this
description it is inferred that the sample has a medium to high level of direct web usage
experience.
categories of constructs of current usage experience, frequency of use, variety of use and
extent of use were conceptualised. Based on this conceptual understanding and the
endeavour to develop reliable and valid measures of current web session usage, an item
pool was generated. The items generated were designed to build on existing measures,
97
and extend them into the context of web usage. Therefore existing measures of usage
Existing research on consumer usage experience was reviewed in the item generation
process to assess existing item structure and design. These items were modified for this
dissertation to correspond with the study context, current web session use. Examined
from a purchase context, self-reports were found to suffer from biases such as forgetting,
(Wind and Lerner, 1979). These biases are likely to exist in the usage context as well. In
addition, unlike purchase which is a discrete event, usage is a continuous event which
may change over the period of ownership or access. Despite these problems with self-
reports, Ram and Jung (1990) suggest that systematically designed self-reports can
provide reliable and valid measures of usage, and also can save a considerable amount of
time and effort compared to the use of purchase diaries. Therefore, self-report measures
Item Generation
Ram and Jung (1990) indicated the use of a number of items, with multiple-choice
response formats, for the self-report measurement of product usage frequency. This
approach is consistent with research studies that have measured the frequency of use of
personal computers and computer-based software (Swoboda, 1998; Bagozzi et al., 1992;
Davis, 1989b; Davis et al., 1989a); the internet and internet related services (Bronson,
1999; GVU, 1998; Napoli and Ewing, 1998; Sivadas et al., 1998; Teo et al., 1999); and other
technologically based systems such as ATM’s, VCR’s, and camera’s (Ram and Jung, 1990;
Sinkovics et al., 1999; Zaichkowsky, 1985b). In addition, this item structure is consistent
with not only academic research but also industry and government research on internet
and web usage frequency (ABS, 1998; GVU, 1998; www.consult.com, 1999). Thus, one
98
item with an 8-category multiple-choice question will be used to measure current web
Item Testing
From the 8-categories ranging from 1=once a month to 8=5 or more times a day, the
7.3.3.4 Scale/s: Current Web Session Usage Variety - Situational and Motivational
Item Generation
Motivational Variety
In industry and government research the reason for internet and web usage has been
measured using one check-list question, where respondents are asked to check the
reasons for product usage (ABS, 1998; eMarketer, 1999; GVU, 1998; Jupiter, 1999a;
measured using a single-item 12-category checklist. This item will be coded according to
the number of items checked, indicating the number of motivations for web session use,
and thus will comprise a final measure of the number of motivations for which the web
is used.
Situational Variety
The locations from which the internet, the web, electronic kiosks and/or computers have
been used was measured using a single item, with either a check-list question or a
multiple choice question (ABS, 1998; www.consult.com, 1998; GVU, 1998; Napoli and
Ewing, 1998). In this study, situation variety was measured using one item: a 6-category
99
measure, with a multiple choice question, with the respondent indicating the number of
Item Testing
Motivational Variety
A 12-category check-list is used to measure the motivations for current web session use.
This is coded according to the number of items checked and comprises a 1-item scale.
Situational Variety
measure the number of locations from which the web is accessed. From the 6-categories
7.3.3.5 Scale/s: Current Web Session Usage Extent – Breadth, Depth and Duration
Item Generation
extent comprises breadth, depth and duration of web session use. Dreze and Zufryden
(1997d) developed models for two effectiveness measures that they found most relevant
to their study of web sites: number of pages accessed (depth) and time spent during a
site visit (duration). These authors used mechanical observation of log-file data for their
study. However, as this study is concerned with the macro perspective of web use, as
opposed the micro perspective of web site visitation and use, mechanical observation is
of item formats that would be appropriate for the measurement of web session usage
100
Session Breadth
With respect to breadth of web session use (i.e., the number of new and/or different web
sites and/or search tools accessed), a review of the literature revealed only a few suitable
measures. For example, Teo et al., (1999) uses a 7-item Likert scale to measure diversity
of internet usage. Hoffman et al., (1998) used an 8-item Likert scale to measure if
respondents visit the same or different web sites (α = 0.79) Therefore to measure user
perceptions of the number of new and/or different web sites they currently access, four
Likert scales (with 7-point strongly agree-strongly disagree responses) were developed.
Session Depth
With respect to the depth of web session use (i.e., the total number of web sites and/or
search tools accessed), a review of the literature revealed a variety of possible measures.
Depth of product usage has been measured with, for example, open end-ended
questions asking the respondent to specify how many messages have been sent and
received (Davis, 1989b; Karahanna and Straub, 1999), and multiple choice responses
have been used to measure the number of ATM cards owned (Sinkovics et al., 1999) and
the number of purchases made (ABS, 1998; Midgley, 1983). A check-list response format
has also been used to measure the number of brands trialed (Schaefer, 1997). With
respect to the web, it was felt a better measure would be to ask respondents to indicate
their level of agreement or disagreement with certain statements with respect to the
number of web sites they might visit, search tools they might use in a given session on
the web, and the number of bookmarks in their favourites folder. Thus four Likert scales
Session Duration
Duration of web session use also seemed an interesting scale item to review. In the
literature, one scaled multiple choice question is used to measure the time spent with
media technologies such as the television (Napoli and Ewing, 1998), computers and
software (Davis, 1986; Davis, 1989b; Dishaw and Strong, 1999; GVU, 1998; Hubona and
Geitz, 1997), and the internet and internet-related services such as the web (GVU, 1998;
101
Jupiter, 1999a; Moon and Kim, 2001; Napoli and Ewing, 1998; Novak et al., 1998; Teo et
al., 1999; www.consult.com, 1999). Duration of web session use is measured here with
Item Testing
Session Breadth
Initial data screening and analysis of correlation patterns identified that individually and
collectively all 4-items meet the necessary threshold of sampling adequacy (KMO
rotation was conducted. This analysis extracted 2 dimensions that explained 70% of the
variance of current web session usage extent - breadth. Further examination resulted in
1-item being deleted as the corrected item-total correlation fell below the threshold value
of 0.6. The final scale thus comprises 3-items with a total scale reliability of 0.7. These 3-
items measure 2 dimensions that explain a corrected 82% of the variance of the construct.
See Table 7 for scale and dimension variance and Appendix D for final item factor
loadings.
Initial data screening and analysis of the correlation patterns identified that individually
and collectively all 4-items meet the necessary threshold for sampling adequacy (KMO
102
= 193.032, df = 6, Sig. = 0.000). Thus a principal components analysis with a varimax
rotation was conducted. This analysis extracted 3 dimensions that explained 94% of the
variance of current web session usage extent - depth. Results showed dimension 1
accounting for 43% of the variance, dimension 2 for 26% of variance and dimension 3
25% of the variance. The final scale thus comprises 4-items with a total scale reliability of
0.81. See Table 8 for scale and dimension variance and Appendix D for final item factor
loadings.
From the 8-categories ranging from 1 = Less than 15 minutes to 8 = 13 or more hours, the
7.3.3.6 Item Generation and Testing Summary: Current Web Session Usage
In summary, the structure of the items developed to measure the categories of current
web session usage frequency, usage variety and usage extent were developed in
accordance with existing academic, industry and government based scales. The specific
content of the scale items also was improved by conducting a number of preliminary
From the item generation process this study operationalised constructs measuring
frequency of session use (1-item), variety of session use - situational (1-item), variety of
session use – motivational, extent of session use - breadth (4-items), extent of session use
– depth (4-items), and extent of session use – duration (1-item). The reliability and
dimensionality of all multi-item scales was tested using exploratory factor analysis and
103
reliability analysis. A summary is presented in Appendix C and the final item factor
applicable scale for measuring usage perceptions of web users, an item pool was
generated. The items generated were designed to build on existing measures, and extend
them into the context of user perceptions of the web. Therefore, both content analysis of
existing scales and a number of primary exploratory studies (as presented in Appendix
B) were conducted. Earlier research on the perceived ease of use and usefulness (Davis,
1986; Davis, 1989a; Adams, 1992; Segars, 1993; Davis, 1989b) was reviewed to assess
existing item structure and design. The results of this review are detailed below.
Item Generation
It is evident from a review of the literature that the main way to measure perceived ease
of use is with multi-item scales, using Likert questions. This variable is usually seen as a
and more recently the internet and internet-based systems such as the web, email and
individual web sites. However, a task focus needs to be introduced here too. From
content analysis of the existing scales, and from qualitative studies outlined later in this
chapter, a number of scale items were developed to assess the predicted dimensions of
perceived ease of web use. Specifically, 20-scaled items with a Likert question response
format were developed to measure a user’s perceived ease of web use for certain tasks.
104
Item Testing
Initial data screening and analysis of correlation patterns identified that individually and
collectively all 20-items meet the necessary threshold of sampling adequacy (KMO
analysis was conducted. This analysis extracted 16-items with factor loadings equal to or
above 0.6 that explained 70% of the variance of perceived ease of web use. Following
further analysis to check the internal consistency, a further 2-items were removed – these
fell below the corrected item-total correlation threshold of 0.6. After the removal of items
due to low factor loadings and low item-total reliability, a final factor analysis was
conducted.
The final perceived ease of web use scale thus comprises 14-items with a total scale
reliability of 0.94. These 14-items measure 3 dimensions that explain a corrected 72% of
the variance of the construct. See Table 9 for scale and dimension variance and Appendix
Item Generation
A user’s perceived usefulness of the web for certain tasks is measured using the same
principles and procedures as for perceived ease of web use. This resulted in the
105
Item Testing
Initial data screening and analysis of correlation patterns identified that individually and
collectively all 23-items meet the necessary requirements for exploratory factor analysis
(KMO Measure of Sampling Adequacy = 0.874, Bartlettʹs Test of Sphericity: Approx. Chi-
Square = 1717.465, df = 253, Sig. = 0.000). Thus a principal components analysis with a
varimax rotation was conducted. This analysis extracted 20-items that explained 6
dimensions and 70% of the variance of perceived usefulness of the web. Following
further analysis to check for the internal consistency of the scale, a further 6-items were
removed – these fell below the corrected item-total correlation threshold of 0.6. After the
removal of items due to low factor loadings and low item-total reliability a final factor
The final perceived web usefulness scale comprised 14-items with a total scale reliability
of 0.9. These 14-items measure 4 dimensions that explain a corrected 75% of the variance
of the construct. See Table 10 for scale and dimension variance and Appendix D for final
7.3.4.3 Item Generation and Testing Summary: User Perceptions of the Web
In summary, from the item generation process 20-scaled items were developed to
measure a user’s perceived ease of web use and 23-scaled items were developed to
106
Following item generation, all items were tested using an exploratory factor analysis and
reliability analysis. The results identify 14-items that reliably measure 3 dimensions of
perceived ease of web use and 14-items that reliably measure 4 dimensions of perceived
web usefulness. A summary is presented in Appendix C and the final factor loadings are
shown in Appendix D.
KNOWLEDGE CONTENT
To assist in the measurement of independent variables in this dissertation, objective and
assessment of the type and scope of knowledge content a consumer actually has of the
system they are using. Results identify scale items that reliably measure actual ‘Common
Procedural’ knowledge content of the web and measures that reliably measure perceived
The research and sampling design for this study is presented first. This is followed by a
discussion of the item generation process for each scale and the scale testing results.
To assess the properties of the scales, four independent samples were recruited. The
samples were drawn from one postgraduate and three undergraduate university student
groups (see Table 11). Convenience and domain experience were the primary
motivations for selection. Across the samples respondents were expected to have varying
107
Table 11: Sample Description - Independent Student Groups
No. Program Discipline (Subject) EKL Sample Response Useable Admin
1 PG Marketing
L/M/H 70 24 24 TH
(Masters) (Elements of Marketing)
2 UG Science (Conservation,
L/M 63 57 48 IC
(First Yr) Biology and Biodensity)
3 UG Communications 31
M/H 100 33 TH
(Third Yr) (New Technology A)
4 UG Education
L/M 64 55 50 IC
(First Yr) (Edu. Psychology)
PG = Postgraduate; UG = Undergraduate; EKL = Expected Knowledge Level (L=Low, M=Moderate, H=High)
TH = Take Home; IC = In-class
The survey instrument was administered in paper-format. This was intended to increase
the variance in domain experience and reduce the chances of perceptual bias (a potential
problem if the study had been conducted on the web). Some 297 surveys were
To assist in the process of data aggregation and scale item analysis, the original 5-
This followed Park et al. (1994) and the recommendations of Malhotra, Hall, Shaw and
Crisp (1996). Data were cross-tabulated to see how the four samples compared. For both
the multiple-choice and true/false items measuring actual web knowledge there was
variance in the data, with all four samples behaving as expected (i.e. the
Education and Science were low-med). In addition the three samples performed as
expected across the three scales measuring perceived knowledge content of the web. The
individual samples were also comparable on the descriptive items measured (based on a
108
7.4.1.4 Scale Development and Analysis
For the four scales developed measuring actual knowledge content of the web and the
three scales measuring perceived knowledge content of the web, a principal components
factor model with a varimax rotation was used. The number of dimensions to be
extracted for further analysis was selected based on the following four criteria: the latent
root criterion, cumulative variance percentages, the scree plot, and prior knowledge. In
addition, factor loadings of +/- 0.45 and above were considered significant and thus for
interpretation purposes, only items with factor loadings above this threshold were
considered. Reliability analyses were also performed on all four scales to check for
internal consistency, with items deleted if the corrected item-total correlation fell below
(69%) and 47 male respondents (31%). The age distribution was also skewed with 75% of
the sample less than 24 years of age. With regard to respondents’ direct web usage
experience, the following results were found: 13% of the sample had less than 2 years
computer experience, 18% had 3-4 years experience, 29% had 5-6 years experience and
40% had 7 or more years experience. With respect to web experience, 22% had been
using the web for between 6-11 months, 30% between 1-2 years and 39% between 3-4
years.
CONTENT
To assist in the successful design and adoption of hypermedia computer-based systems,
like the web, objective measures are developed to assess the type of knowledge content a
consumer actually has of the system. Results identify reliable scale items that objectively
109
The focus of the study is actual consumer knowledge content of the web. However, as
easy task, given the vast array of relevant knowledge that a consumer may possess’.
Thus, before developing measures of the variables defined in chapter five of this
Previous research shows that actual consumer knowledge has been assessed using
brand/attribute recall and elicitation methods (Brucks 1985; Selnes and Gronhaug 1986;
Brucks 1986; Mitchell 1981; Park, Feick and Mothersbaugh 1992), in-depth interviews
(Dacin and Mitchell 1984), task-allocation methods (Mitchell 1981; Russo and Johnson
1980), responses to multiple choice questions (Zaichkowsky 1985b; Park et al. 1994),
open-ended lists (Brucks 1985; Selnes and Gronhaug 1986), and true/false questions (Rao
Multiple-choice and true/false item structures were adopted here. A response option of
‘don’t know’ was added to both item structures to increase the chances of measuring
actual knowledge as opposed to capturing the respondent’s ability to guess correctly. All
items were recoded as correct/incorrect prior to analysis with ‘don’t know’ recoded as
Brucks and Mitchell (1981) proposed a typology for classifying the knowledge that a
consumer has about a particular product category or purchase decision. ‘Product class
knowledge’ was seen as comprising: terminology, specific facts, relationships, criteria for
knowledge, while relationships and criteria for evaluation relate to more abstract ideas.
110
Brucks (1986) further developed this typology, arguing that “consumer knowledge can
be classified and measured by its content and that the typology proposed was
categories” (p58). The Brucks (1986) typology comprises eight items giving rise to
Brucksʹ (1986) research has two important implications for research into consumer
product knowledge. Firstly, measures of consumer knowledge should cover the full
range of product knowledge, and secondly that the different types of knowledge content
(i.e., procedural and/or declarative) will have varying effects on the decision making
process. The first of these conclusions has direct implications for the development of
involved a number of qualitative steps. Firstly, a set of free response questions were
constructed on the basis of the typology developed by Brucks and Mitchell (1981) and
later defined by Brucks (1985) and Brucks (1986). These free response question were used
as the foundation of an expert survey conducted among a panel of web designers and
web-marketing experts – they were able to comment on those aspects of the web that are
required or used for web navigation from a design perspective (i.e., terminology,
attributes, facts, evaluative criteria, usage situations). Further information was obtained
from an observational study of novice users, from web help files and web site content
analyses, and from in-depth interviews with web users. The results of these studies are
detailed in Appendix B.
With this procedure, a pool of 110 items was developed to measure consumer
benefits, and condition-action statements for web navigation. The panel of experts rated
all 110 items - with first round ratings grouping items as measuring either declarative or
procedural knowledge and second round ratings grouping items as measuring either
111
specialised or common knowledge of the web. Nine items were deleted as categorisation
was not consistent across the panel. A final set of items was derived, consisting of 13-
items measuring specialised declarative knowledge of the web. These measures were
measures, and thus are considered to be objective measures. The conceptualisation and
Initial data screening and analysis of correlation patterns identified that individually and
collectively all 13 items met the necessary threshold of sampling adequacy (KMO
= 410.173, df = 78, Sig. = 0.000). Thus a principal components analysis with a varimax
rotation was conducted. Only those items with factor loadings above +/-0.45 were
considered. This reduced the scale from 13 items to 9 items, with a further 3 items
removed due to low item-total correlations. The final factor analysis identified 6-items
that extracted 3 dimensions and explained 75% of the variance. Total scale reliability was
0.8. See Table 12 for scale and dimension variance explained and Appendix E for the
112
Actual Common Declarative Knowledge Content
Initial data screening and analysis of the correlation pattern showed that 28 of the 31
items, individually and collectively, met the necessary threshold of sampling adequacy
Bartlettʹs Test of Sphericity: Approx. Chi-Square = 1381.676, df = 378, Sig. = 0.000). Thus a
principal components analysis with a varimax rotation was conducted. Only those items
with factor loadings above +/-0.45 and item-total correlations at 0.6 or above were
considered. This reduced the items from 28 items to 10 items. The final factor analysis
identified 10-items that formed 2-dimensions and explained 57% of the variance, with a
total scale reliability of 0.9. See Table 13 for scale and dimension variance and Appendix
Initial data screening and analysis of correlation patterns identified 1 of the 19 items
measuring specialized procedural knowledge had a MSA level lower than .50. This item
was removed resulting in the remaining set of 18 variables, collectively and individually,
Sig. = 0.000). The principal components analysis with a varimax rotation was conducted.
Only those items with factor loadings above +/-0.45 and item-total correlations at 0.6 or
above were considered. This reduced the scale from 19 items to 11 items. These 11-items
extracted 3 dimensions and explained 59% of the variance, with a total scale reliability of
0.8. See Table 14 for scale and dimension variance and Appendix E for final item factor
loadings.
113
Table 14: Variance Explained of Specialised Procedural Web Knowledge
Initial Extraction Rotation
Eigenvalues Sums of Sums of
Squared Squared
Loadings Loadings
Component Total % of Cumulative Total % of Cumulative Total % of Cumulative
Variance % Variance % Variance %
1 4.2 38 38 4.1 38 38 2.3 21 21
2 1.3 12 50 1.3 12 50 2.2 20 40
3 .9 9 59 .9 9 59 2.0 18 59
Extraction Method: Principal Components Analysis with a Varimax Rotation
For the items measuring specialised declarative knowledge of the web, 4 of the 29 items
had MSA levels of less than .50. These items were omitted resulting in the reduced set of
exploratory factor analysis with a varimax rotation was conducted. Only those items
with factor loadings above +/-0.45 and item-total correlations at 0.6 or above were
considered. This reduced the scale from 25 items to 11 items that extracted 3 dimensions
and explained 59% of the variance. Total scale reliability of 0.9. See Table 15 for scale and
114
7.4.3.3 Item Generation and Testing Summary: Actual Web Knowledge Content
In summary, the above analysis identified scale items that reliably measured actual
the web.
KNOWLEDGE CONTENT
Subjective measurement is defined as a method used to assess individual perceptions
measures. For the purpose of this dissertation, subjective measures of a user’s web
knowledge content are developed to assess the knowledge content a consumer thinks
they have of the web. Results identify scale items that reliably measure perceived overall,
A number of existing scales were analysed to determine the measurement level and
question response format used to measure perceived knowledge content of the web. Of
specific interest to this study has been the prior use of subjective measures to measure
with multiple-choice questions, have been used to measure the perceived knowledge
content of personal computers (Selnes and Gronhaug, 1986). Measures, with semantic
machines (Brucks, 1985). In a number of studies Likert scales have been used to measure
perceived knowledge of CD players (Park et al., 1992; Park et al., 1994) and cars (Johnson
and Russo, 1984), and the perceived level of skill in relation to the web (Novak and
115
other products and/or areas identified that a majority measure perceived knowledge
content with Likert questions (Cole et al., 1986; Park et al., 1992; Johnson and Russo,
1984; Rao and Sieben, 1992; Yale and Gilly, 1995; Hulland and Kleinmuntz, 1994).
The exploratory studies were also used to help establish the content of the scale items to
From this process, a pool of 20 items was developed that subjectively measured
usage procedures, benefits, and condition-action statements for web navigation. A panel
of experts rated all 20 items - with first round ratings grouping items as measuring either
generic (i.e., about attributes in general and not specifically questioning a user’s
knowledge of ‘cookies’ for example), it was difficult to differentiate the scope and thus
the ‘common’ and ‘specialised’ nature of the items. Therefore, the scope of knowledge
content was not measured subjectively. A final set of items was derived, consisting of 3-
content.
Initial data screening and item-total reliability analysis on the 3-items measuring
reported low item-total reliability, thus resulting in 2-items measuring perceived overall
116
knowledge content of the web. These two items were further investigated using
between the two items (rs = .757, p<.01). Thus the items are deemed appropriate for
Initial data screening and analysis of correlation patterns identified that individually and
collectively all 8 items meet the necessary threshold of sampling adequacy (KMO
= 632.463, df = 21, Sig. = 0.000). Thus a principal components analysis was conducted.
Further assessment of item-total correlations found that 4 of the 8-items have low scores,
i.e., below the 0.6 threshold. These items were omitted. Final results showed 1 dimension
accounting for 77% of the variance with a reliability of 0.9. See Table 17 for scale variance
For the items measuring perceived declarative knowledge content of the web, 2 of the 9
items had MSA levels of less than .50. These items were omitted. The reduced set of 7-
117
items collectively and individually met the necessary threshold of sampling adequacy
(KMO Measure of Sampling Adequacy = 0.947, Bartlettʹs Test of Sphericity: Approx. Chi-
Square = 1327.274, df = 36, Sig. = 0.000). The principal components analysis extracted 1
dimension that explained 75% of the variance. Total scale reliability was 0.9. See Table 18
for dimension variance and Appendix E for final item factor loadings.
7.4.4.3 Item Testing and Generation Summary: Perceived Web Knowledge Content
In summary, the results identify 2-items that reliably and subjectively measure perceived
overall web knowledge content; 7-items that reliably and subjectively measure perceived
declarative knowledge content and; 4-items that reliably and subjectively measure
user behaviour on the web. The procedure results in the development of reliable
measures of current web session usage frequency (1-item); current web session usage
variety – situational (1-item), current web session usage variety – motivational (1-item),
current web session usage extent – breadth (3-items), current web session usage extent –
depth (4-items), perceived ease of web use (14-items), perceived web usefulness (14-
items), actual common procedural web knowledge content (6-items), actual common
knowledge content (11-items), actual specialised declarative web knowledge content (11-
118
web knowledge content (7-items), and perceived overall web knowledge content (2-
Following scale and instrument development, the relationships discussed in this study
are tested using a web-based survey. The research design for hypothesis testing is
119
C HAPTER 8: R ESEARCH M ETHODOLOGY
8.1 INTRODUCTION
The methodology is described. A cross-sectional web-based survey instrument was used
to test the questions posed in this dissertation. This consisted of six sections with single
and multi-item measurement scales. The entire survey consisted of 180 formally
structured pre-coded questions, the response to which was entered into Filemaker Pro™
selected sampling design was used with the sample recruited using monetary incentives,
help generate measurement items. Once developed, these items then were used
objectives. There are three basic research designs: exploratory, conclusive descriptive and
120
conclusive causal. For the purpose of testing the hypotheses proposed in this
dissertation, a descriptive conclusive research design was adopted. This design is used to
Rao (1980) reports that theoretical relationships in consumer behaviour are tested and
validated with the assistance of data collected through descriptive research, either cross-
longitudinal research involves the repeated measurement of the variable of interest from
a fixed sample over time (Malhotra et al. 1996). As it was desirable to have a snapshot of
web users, this dissertation uses a single cross-sectional research design and measures
The two main methods of data collection for descriptive research are either survey or
methods to explore the constructs and generate scale items. However, the main data
through a structured formalised technique (Malhotra et al. 1996). With the advantage of
minimising interviewer bias, which may exist in focus groups or interviews, survey
knowledge and behaviours (Tull and Hawkins (1993). Questionnaires also have the
advantage of ease of coding, time efficiency and cost effectiveness (Tull and Hawkins
1993). A survey design is therefore easy to administer, the data obtained are reliable, and
121
8.2.2.3 Computer Assisted Web-based Mode of Survey Administration
instrument was used to collect information from a sample of experts (see Appendix B)
and a self-administered paper-based survey instrument was used to collect the data to
test the scale items generated from the preliminary exploratory studies (see Chapter 7).
The effectiveness and feasibility of survey instruments have been profoundly influenced
more recently Internet-related technologies such as email and the web. Electronic
technologies have improved the efficiency of personal interviews (face-to-face) and mail
surveys and also provided completely new modes of survey administration such as disk-
personal interviewing (CAPI), and Internet-based electronic survey modes such as email
and web-based survey instruments. For the purpose of this study, a web-based Internet
These surveys have many advantages over traditional surveying methods. Table 19
compares the attributes of Internet (web), personal, telephone, and mail surveys and
shows that web-based Internet surveys have relative low cost, fast turnaround and
survey completion, high geographic coverage and high response rates. However, while
a large proportion of the general population still does not have access to the Internet
online, researchers continue to face the problem of biased samples. Apart from surveys
specifically for the Internet population, such as in this study, Internet-based surveys
true for countries outside the United States where there are significantly smaller
122
Table 19: Comparative Attributes of Differing Modes of Survey Administration
Attributes Internet (web) Personal Telephone Mail
Costs Very Low Very high Medium Low
Speed of turnaround Fast Instant Instant Slow
Response rate High Very high Medium Low
Population segments Possible access Less than mail but more
Few Many
accessible to all segments than Internet
Feasible geographic
Very High Very Low Medium High
reach of survey
Accessibility of
Low Varies Medium Very high
medium to respondents
Time taken Fast Long Medium Long
Source: Pope, Tam, Forrest and Henderson (1997), pg 22.
Despite some research studies reporting virtually identical internal reliability across
and also limited influence of survey mode on the data quality received (Yun and
Trumbo, 2000), concern has been expressed as to the quality of Internet survey
responses. Nevertheless, Joinson (1999) found that respondents actually reported lower
social anxiety and less social desirability bias when using the Internet-based survey
There are several factors that should be considered when employing the Internet as a
surface mail surveys, with web-based surveys showing the lowest cost (as stated in Yun
and Trumbo, 2000). However, Watt (1999) indicated that for some web-based surveys the
cost might out weigh that of mail-based surveys if the population size is held constant
(this is because of the costs incurred in survey programming and network monitoring
and administration – however labour costs are hard to calculate). An overview of some
123
Table 20: Advantages and Disadvantages of Internet Surveys
Internet Surveys
Advantages Disadvantages
High response rates Sample self-selection
Greater response accuracy Un-representativeness of the general
Have the potential to be more enjoyable population
and aesthetically pleasing Multicultural considerations
Less expensive Shorter attention span of respondents
Faster turnaround Lack of interpersonal nuances
Customised surveys Untruthfulness of respondents
Easier data transfer Multiple entries
User convenience User experience and system knowledge
Geographic coverage effects on survey completion
No interviewer bias Privacy and security concerns
Flexible graphical representation Respondent time and financial access costs
Browser incompatibility
Sources: Forrest (1999), Oppermann (1995), Smith (1997)
In summary, a web-based survey instrument was used as the primary form of data
Filemaker Pro™ database configured to record and store the completed survey entries.
The survey consisted of 180 formally structured pre-coded questions with response
categories provided for 176 questions. Only 85 of these questions are used in this
dissertation to assess the hypotheses. The generation of these items was discussed in
Chapter 7 and factor analysis results are presented in Appendices D and E. The web site
Prior to pilot study administration, testing of both the survey instrument and web site
sample of end-users who accessed the survey from a number of differing computing
platforms and geographic locations. Feedback on the survey and processing accuracy
system was conducted, giving rise to a number of changes to the format of the survey
(i.e., colours and layout) and the removal of interaction errors between the database and
124
8.2.3 RESEARCH DESIGN: SUMMARY
In summary, a single cross-sectional descriptive research design, using a computer-
assisted web-based mode of survey administration to quantitatively collect and treat the
usage behaviour have used a number of parameters for sample selection: product usage
experience (Kouchakadjian and Fietkiewicz 2000; Nunes 2000; Ram and Jung 1990);
education (Sinkovics et al. 1999); computer and media use (Coffey and Stipp 1997; Lin
1992); geographic region (Van den Bulck 1999; Jeffres and Atkin 1996), and employment
classification (Seeley and Targett 1999; van Braak 2001). Research into usage behaviour
on the web has predominately used samples defined by web access and use (Korgaonkar
and Wolin, 1999; Kraut et al., 1998; Chatterjee et al., 1998); senior education (Eighmey
and McCord, 1998; Diaz et al., 1997; Schumacher and Morahan-Martin, 2001;
Papacharissi and Rubin, 2000); newsgroup use (Sivadas et al., 1998); and employment
classification (Stevens, Williams, and Smith, 2000; Henry and Stone, 1999).
As the core of this dissertation is the profiling of current web usage behaviour, the main
parameter used for sample delineation is web usage experience. This is consistent with
sampling (Huck, Cormier and Bounds 1974). Random or probabilistic sampling creates a
sample based on probability theory in which each individual from the population has an
equal chance of being selected (Tull and Hawkins 1993). Such sampling methods include
125
simple random sampling, cluster sampling, stratified random sampling and multi-stage
individuals have a greater chance of being selected from the sample population (Huck et
al. 1974). Neuman (1997) describes convenience, quota, snowball and purposive
relies on the judgement of the researcher and is limited in its ability to provide an
(Weiers 1988; Neuman 1997). However, the advantages of non-probability sampling are:
1) the speed with which respondents are identified, 2) the ease of access to a sample
group, especially with respect to the internet population, and 3) the low costs involved in
The internet, by its very nature, poses a number of unique problems in guaranteeing a
random sample. Unlike telephone and mail surveys in which samples can be produced
through census lists and electoral rolls, the internet has no central registry of users (Kaye
and Johnson 1999; Taylor 2000). A large number of people also still do not access and use
the internet and thus attempting to reach users of internet-related technologies like the
web through traditional methods such as the telephone or mail may result in high levels
there would be a large number of nonusers, who are not the focus of this study.
Therefore, to reach an audience of web users, a user survey was posted on the web (as
previously discussed, see also Appendix F), and various mechanisms were used to drive
126
8.3.3 SAMPLE RECRUITMENT
In an attempt to reduce possible non-sampling and sampling error, participation
To encourage participation of web users, a neutral incentive was employed. This offered
participants the chance to win one of four cash rewards ranging from between $A49.30
and $A642.15. Monetary incentives were selected over the offering of tangible goods
industry funded surveys (e.g., www.consult.com 1999). This was done to keep down the
costs of prizes and to ensure the incentives would appeal to all types of respondents.
a target web page – are commonly used to attract users to web sites and web-based
surveys. User responses to this form of advertising are measured by counting ‘click-
throughs’. ‘Click-throughs’ are necessary to move respondents from the site and into the
survey itself – they can be considered similar to a cover letter in a mail survey (Tuten,
Thus, web banner advertisements were used in this study to increase awareness of the
survey and to entice users to visit the web site and complete the survey. Although a
quite inexpensive means to attract survey respondents, the risks with this method are
two fold. Firstly, the probability of seeing any banner advertisement or banner campaign
is directly proportional to the user’s total amount of internet usage. For example, if the
frequency of web use is 20 times a month the user is 20 times more likely to see the
banner campaign than a user who accesses the web just once a month. Thus, using
banner recruitment may bias the sample toward heavy internet users. Secondly, the
127
saturation of banner ads on one type of web site will result in a heavy sample bias
toward site type and topic area, and, if this is not the intention, the problem needs to be
addressed. Therefore, for this dissertation, banner ads were spread across a large number
of differing sites.
Three banner ads were developed, corresponding to the first three prizes of the
researcher (see Appendix G for the banner advertisements). The Australian DoubleClick
banner advertising network (http://www.doubleclick.net) was the main provider used for
amount of inventory was also supplied by individual site and portal vendors to further
(http://www.ruralpress.com), which is targeted at the four out of every 10 people who live
Additional online advertising took the form of web site URL links from various vendors
registration and email recruitment was also undertaken to drive online traffic.
Offline Publicity
To minimise sampling error a number of methods of offline publicity were also used to
drive traffic to the web survey. These included a media release targeting local, regional
and state newspapers throughout Australia; a newspaper article in The Australian (IT
128
a review of the study in the January 2001 issue of the Australian NetGuide, targeted at
average web users. The Australian NetGuide has an audit of 40,814 copies per issue,
making it the number 1 selling internet magazine in the country (this ranks it in the top
100 magazines in any category in Australia). The magazine has national distribution
incentives, and both offline publicity and online advertising, were used to drive
measured and to test the hypotheses proposed. These techniques are discussed in the
determine the response rate. The exact number of web users exposed to the online
banner ad campaign and the offline publicity is not known. However, three sources of
‘audience behaviour data’ from the web site, the banner ads, and the database, were
129
Specifically, the web site log-file was used to assess the total number of unique visitors to
the web site compared with the total number of completed surveys received.
Furthermore, the banner ad campaign reports presenting data of the total number of
unique users exposed to the banner ads was compared with the total number of
completed surveys recruited from the banner ad campaign. Finally, as the study was
conducted over a period of three and a half months, it was possible to examine the ‘date
of creation’ in the database to assess any patterns in the timing of usable responses. In
addition, this latter information was also used to test for any significant differences
between those who were recruited early on and those who were recruited at a late stage.
graphical analysis of the sample characteristics were used to describe the sample.
was conducted on all multi-item measures. For single-item measures, means and
medians were assessed. This was conducted to see whether the scales remained
consistent from the development stage (described in Chapter 7) to the final stage
(presented in Chapter 9). See chapter 7 for a discussion of the techniques and their
In this study comparison is made of the observed differences between two user groups:
those users with and without web site design and maintenance experience. To compare
these two groups, the sample was split based on the variable ’Web Site Design and
130
1=Experience. In subsequent chapters, these two user groups are referred to as ‘with no
variables examined in this study, scatterplots were produced. These help to show
visually: 1) the nature of the hypothesised relationships (i.e., linear or non-linear), 2) the
3) the strength of the relationships (none, weak, moderate or strong). Simple scatterplots
were prepared with the dependent variable plotted on the y-axis and the independent
variable plotted on the x-axis. As this study also examined the observed differences
between users with and without web site design and maintenance experience, two
scatterplots were assessed for each hypothesis. The results of these visual comparisons
Of the 26 hypotheses proposed in this dissertation, eighteen are ‘linear’, seven are
Linear Relationships
For metric data Pearson’s correlation coefficient can be used to assess linear
relationships. However, the eighteen linear relationships that are hypothesised in this
dissertation make use of a mix of data types. Certain data violate the assumptions of
scores it was found that seven out of the eight violated the assumption of normality,
which is required for the conduct of parametric statistics (see Appendix L). Therefore,
the rank-order correlation coefficient (Spearman’s rho) was used to test eleven of the
relationships.
131
Furthermore, from a visual and statistical assessment of the distribution of sample scores
for the measures it was found that all seven violated the assumption of normality (See
Appendix L). Therefore, the rank-order correlation coefficient (Spearman’s rho) was also
used to test the eight relationships hypothesised to exist between the independent and
dependent variables.
Spearman rho correlation coefficient, although suitable for measures and when
parametric assumptions have been violated, is not the most suitable coefficient for
testing for a curvilinear relationship. De Vaus (2002) indicates that coefficients designed
for nominal variables (e.g., Cramerʹs V and Goodman and Kruskal’s tau) will detect non-
linear relationships.
Therefore, the variables between which curvilinear relationships are hypothesised will
be reduced to 3-category level measures and the association tested using Cramer’s V and
Goodman and Kruskal’s tau. These chi-squared coefficients are selected, as against Phi
and Yule’s Q, as the variables examined have more than two categories each (i.e., are
larger than a 2x2). It is expected that some information may be lost in this process,
however this is the price to pay for testing whether a curvilinear relationship exists.
association (e.g., Gamma) and a non-linear measure (e.g., Goodman and Kruskalʹs Tau).
As a general rule, if Goodman and Kruskalʹs tau is higher than Gamma, it indicates a
non-linear relationship may exist between the two variables examined. At which point
De Vaus, (2002) indicates that a closer examination of the 3x3 cross-tabulation should be
132
Significance Testing
For large samples, statistical testing should be based on the stringent significance level of
0.01. As this dissertation has an entire sample of n=2077, which has been split into two
smaller groups (with experience = 1177 and no experience = 900), 0.01 is the primary and
hypotheses were reduced to 8 specific relationships (MRA1-8) and were tested using a
split-wise multiple regression method. The level of significance for these tests was held
To conduct the step-wise multiple regressions, all 8 relationships were assessed against
the assumptions set down for the conduct of regression analysis (summarised in
Linearity. The bivariate analysis conducted earlier examined the presence or absence
assumption of linearity was further assessed by visually examining the residual plots
Normality. Univariate analysis of linearity was assessed using graphical analysis (i.e.,
histograms) for non-metric data and statistical analysis (i.e., normality test) for metric
data. In addition, further assessment of the normal probability plot for users with
of the residuals was conducted for users with and without WSD/M experience.
Independent Errors. According to Field (2000) and Hair et al. (1995), an assessment of
133
independent errors. According to Field (2000), this test statistic can vary from 0 to 4
with a value of 2 meaning that the residuals are uncorrelated. However, the Durban-
model and the number of observations and thus should be assessed accordingly
(Field, 2000). The number of independent variables used in MRA1-6 is two and for
MRA7 and MRA8 the number is seven. With 100 observations or above, and two
1.63 (p<0.05) residual correlation is present. With 100 observations or above, and five
Multicollinearity. As cited in Field (2000) and reiterated in Hair et al. (1995) if the
largest variance inflation factor (VIF) is greater than 10 or the average VIF is
with multicollinearity may exist, and if below 0.2 a potential problem may exist.
From an assessment of the above, limitations were noted in conducting the step-wise
shows great area of concern. These are noted in Chapter 10 and discussed further in
Chapter 12. Data were not transformed to ‘remedy’ these violations, partly because
transformations may change relationships between variables in ways that are hard to
databases, to check the structure of the web-site, to check for clarity and errors, and to
see the sample response rate generated from the banner advertising. In contrast to earlier
samples used, the pilot study was drawn from a sample of web users similar to those
expected for the final sample. It was scheduled to run until a quota of over 150
134
respondents were recruited and/or the aforementioned issues were ironed out. A
web site during the pilot study (unique visitors: n=1232) a total of 20% completed the
survey. 140 responses were recruited in the first few days with a total of 253 responses
The timing of responses showed that 41% of the usable responses were received on the
first day of the pilot, 54% on the second day and 5% on the third day. In addition, 80% of
the usable respondents became aware of the study from the banner ad campaign, 13%
from a web site hyperlink, 2.4% from a search engine query and 1.6% from word-of-
mouth.
uneven gender distribution with 36% female and 64 % male respondents. From a
comparison of the gender distribution of the pilot sample with industry and government
statistics, this result indicated that the pilot study represented Australian web users in
1999, not in 2000 (Table 21). This was important as the sample to be recruited for the final
135
Some 57% of the pilot sample were aged between 25-44 years of age. This age
larger than the 46% reported by NetRatings (2000) (for the age range 25-49 years). 47% of
the pilot sample were employed full-time, and 47% and 22% resided in NSW and
Victoria respectively.
In summary, it was evident from these results that additional advertising and/or
promotion of the web survey offline would be required to ensure the recruitment of a
more representative sample of Australian web users and to avoid some non-response
biases.
for a period of 3 ½ months until January 31st 2001, with offline publicity and online
banner advertising scheduled throughout this period. The pre-Christmas period was
selected for a number of reasons. Firstly, this period followed the summer Olympics in
Sydney 2000, thus excluding any sampling bias and response error that might have
occurred if administered during the Olympic period. Secondly, it was an opportune time
to maximise responses because of the reported increase in web usage due to pre-
Christmas shopping and browsing online (Jupiter, 1999a; Jupiter and Interactive, 1999b).
was used. Non-probability sampling was employed, with sample recruitment aided by
136
C HAPTER 9: D ESCRIPTIVE R ESEARCH R ESULTS
‘The voice of the people hath some divineness in it, else how should so many men
agree to be of one mind’
- Francis Bacon -
(1561-1626)
9.1 INTRODUCTION
This chapter provides a descriptive account of the web sample, and an assessment of the
reliability and dimensional structure of the scale items used to measure the variables in
this study. An analysis of responses is presented first (section 9.2). This is followed by a
review of the sources used to recruit the web sample, and a descriptive profile of those
who responded (section 9.3). The final section of this chapter (section 9.4) then compares
and contrasts the reliability and dimensionality of the scale items used in this web
52=non-residents). Hair et al., (1995) recommend that for multivariate and bivariate
analysis the sample size should be at least 5 times the number of variables (i.e.,
parameters) in the model. Since the proposed model has 15 variables, the minimum
response necessary would be 75 observations. The sample size of 2077 is thus far in
Comley (2000) summarises the response rates of a number of virtual surveys in 1999 (i.e.,
email and web-based), and most were in the range 15% to 29%. Ray, Griggs, & Tabor
(2001) summarised the response rates in their survey as follows: 41% (academic sector),
137
31% (general web sector), and 19% (business sector). From the web site log file analysis
presented in Appendix H, of those who visited the web site during the main study
(unique visitors: n=5104) a total of 41% submitted a usable survey. This is comparable to
response rates identified by Ray et al. (2001). However, this response rate has been hard
to determine given the methods used to recruit respondents (i.e., banner advertising),
and therefore the rate of 41% might be regarded as quite generous. The fact that well
over half those who visited the site did not submit a usable survey indicates that perhaps
As indicated in the next section of this chapter and reported in Appendix I, according to
the banner advertising reports a total of 867, 617 unique users were exposed to the
banner ads placed in October and December. Of this number, a total of 893 usable
responses were obtained. This provides a response of 0.1% of web users exposed to the
When assessing the pattern of responses over the study period, the last two weeks in
October 2000 and the first two weeks in December 2000 yielded the greatest portion of
Table 22: Timing of Survey Receipt (Oct 2000 to Jan 2001) - Main Web Sample
Date of Survey Receipt (Usable Surveys Only)
Month October November Decembera January
Date 18-31st 1-14th 15-30th 1-14th 15-27th 6-14th 15-31st
% 41 10 10 26 4 3 43
n 851 219 216 540 90 65 96
a Due to downtime of the hosting server on the 27th of December and a resulting conflict between the server
the database, no entries were received between the 27th December to the 6th of January.
138
Figure 13: Survey Creation Date - Main Web Study (Oct 2000 to Jan 2001)
With¾ indicating the main phases of banner advertising, from Table 22 and Figure 13, it
is evident that survey responses are coincident with the banner advertising campaigns.
A test was conducted to compare the means of the early versus late respondents. The
sample (n=2077) was split into equal groups (n=207) to compare means for each variable
measured. The variable means for the first 207 (Group 1) and the last 206 (Group 10)
respondents were compared. The tests revealed no significant differences in the means
reported for WSUEB (p=0.036), WSUED (p=0.000), WSUEDUR (p=0.000), WSUF (p=0.042)
and Age (p=0.026). In general, there appear to be some differences between early and late
were recruited from the banner ad campaigns, a further 27% (n=551) from a link on a
web site, and 13% (n=270) from an Email or Email list. Promotion off the web (2%),
advertising off the web (2%), and WOM (2%) only accounted for a very limited number
of usable responses.
139
Figure 14: Source of Respondent Study Awareness
A Banner Ad 43.04%
Study Awareness
An Email or Email List Advertising off the Web (e.g., TV, Radio)
A Banner Ad Promotion off the Web (e.g., Poster, Flyer)
Pies show counts
A Link on a Web Site From the Actual Researcher
A Search Engine Query Other
Word of Mouth (eg., friend, co-worker)
the survey – i.e., “Are you an Australian resident?” – if not, an open ended form field
was incorporated that allowed respondents to indicate their place of residence. It was
found that 98% of completed survey responses were from Australian residents. The
additional 2% were residents in Canada, China, East Timor, Hong Kong, Indonesia,
Netherlands, NZ, Norway, Samoa, Switzerland, Taiwan, USA, and the UK. Prior to data
collection those survey responses derived from non-Australian residents were going to
be removed from the final sample of respondents. However one discrepancy was
overlooked in that although resident in another country, respondents might have been
located in Australia for a number of years on temporary work or student visas and thus
were included in the final sample. However for future reference - in addition to
140
information should be obtained as to their current place of residence and period at this
location.
9.3.2 DEMOGRAPHICS
The sample recruited has a fairly even gender distribution with 56% of respondents male
comparable to the gender distribution reported in industry and government reports for
2000 and 2001. Furthermore, as opposed to the sample description obtained from the
pilot (see section 8.5.2 and Table 21), the main study sample is a far better representation
With respect to age, 52% of respondents are aged 30 years and younger and 48% of
46 - 54 years 11.17%
18 - 24 years 22.92%
40 - 45 years 10.30%
Age Category
Younger than 18 years
18 - 24 years
25 - 30 years
31 - 39 years
40 - 45 years
46 - 54 years
31 - 39 years 21.67% 55 years and over
25 - 30 years 20.75%
Pies show counts
141
In addition:
45% of the sample were full-time wage earners. 5.5% did not indicate their labour
force status and 5.5% indicated “other”. This might comprise a number of categories
19% of the sample worked in Profession-IT; 8% Professional - Business; 18% did not
indicate their current occupation and the remainder of the sample was distributed
and clerical/admin.
25% did not indicate the industry worked in, and the remainder of the sample were
(6%); Retail Trade (6%) and Banking, Finance and Insurance Sector (5%).
highly educated sample. This is consistent with the Australian web user being highly
educated (www.consult.com 1999; ABS 2000 and NOIE 2001). In summary, 5% left
school before 15; 7% achieved their junior school certificate; 17% their senior
North Territory.
also examined. Within the sample recruited for this study (n=2077), 57% (n=1177)
indicated they had web-site design and maintenance (WSD/M) experience and 43%
(n=900) indicated they did not have this experience. For all analyses presented and
discussed in Chapters 10, 11 and 12 the sample is split into these two user groups.
142
9.4 MEASUREMENT ASSESSMENT AND TREATMENT
To assess the reliability and dimensionality of the scales, reliability and factor analyses
were conducted on all multi-item measures. Descriptive statistics (median’s) were also
development. This was conducted in order to assess the stability and dimensional
structure of the scales and the reported consistency with scale properties reported during
scale development. The method for scale development was based on analysis of two
students samples (n = 128 & n =152) (documented in Chapter 7), whereas results here for
scale validation are for the main web sample (n = 2077). The method for scale validation
is consistent with the method conducted scale development. See Chapter 7, section 7.2
for a discussion as to the method that was also used for scale validation. In all cases items
‘Current web session usage frequency’ was operationalised using a single-item scale (see
Chapter 7). The item responses were coded from lowest frequency (1 = “once a month”)
to highest frequency (8 = “5 or more times a day”), with a median for the web sample of
7 = “2-4 times a day”. By comparison, the median usage frequency category for the
student sample was 6 = “once a day”. Thus, although comparable in many respects, the
web sample used the web more frequently than the student sample.
‘Current web session usage variety – situational’ was measured using 2 measures. The
first item measuring the number of situations from which the web is accessed was a scale
item asking respondents to indicate the number of locations from which they accessed
the web in a typical week. The second item measured the types of locations from which
respondents accessed the web with a 10-category check-list. This second item was
143
recoded 0 = “not used” and 1 = “used”, with responses summated to form a 1-item scale
that reported the number of types of locations from which the web is accessed with a
range 1 to 10 locations.
To measure current web session usage variety – situational, the two items were further
positive relationship between the two items (rs = .598, p<.01). Thus the items will be
Table 24: Situational Variety: Location No. & Location Type No.
Spearman’s Rho Correlation Coefficient
Situational Variety
Location Type No.
Spearman’s Correlation Coefficient .598 **
Rho Situational Variety
Sig. (2-tailed) .000
Location No.
N 2077
** Correlation significant at the .01 level (1-tailed)
‘Current web session usage variety – motivational’ was operationalised using a single-
item with a 12-category checklist (Chapter 7). Following data collection this item was
coded 0 (“no motive”) and 1 (“motive”) for each motive listed. These codes were then
summated to form a 1-item measure of the number of motives for current web session
usage – motivation, with a range of 0 = “no motives” to 12 = “12 motives”. From the
resultant 13-category response format, the median for the web sample was 6 (6 motives
for usage). In comparison, the median for the student sample was 4 (4 motivations for
usage). Thus the web sample, on average, used the web for a larger number of reasons.
Exploratory factor analysis was performed on the initial 3-item Likert scales measuring
‘Current Web Session Usage Extent – Breadth’ to summarise the data in terms of a set of
underlying dimensions that make up the multi-item scale. Initial data screening and
144
analysis of correlation patterns identified that individually and collectively all 3 items
= 0.000). Thus a principal components exploratory factor analysis was conducted. This
analysis extracted 2 dimensions which explain 83% of the variance of ‘Current Web
Session Usage Extent – Breadth’ with a total scale reliability of 0.7. See Table 25 for scale
Table 25: Variance Explained of Current Web Session Usage Extent - Breadth
Initial Extraction Sums of Squared
Eigenvalues Loadings
Component Total % of Cumulative Total % of Cumulative
Variance % Variance %
1 1.7 59 59 1.5 49 49
2 .7 24 83 1.0 34 83
Extraction Method: Principal Components Analysis with a Varimax Rotation
These results are consistent with those reported for the student sample (Chapter 7 and
Appendix K). In fact, the 3-items developed and first tested on the small student sample
explained 82% of the variance, which compares with 83% for the web sample.
Exploratory factor analysis was performed on the initial 4-item scale with a Likert
question response format measuring ‘Current Web Session Usage Extent – Depth’. Initial
data screening and analysis of correlation patterns identified that individually and
collectively all 4 items meet the necessary threshold of sampling adequacy (KMO
was conducted. This analysis extracted 3 dimensions that explained 89% of the variance
of current web session usage extent – depth, with a total scale reliability of 0.6. See Table
26 for scale and dimension variance and Appendix J for item factor loadings.
145
Table 26: Variance Explained of Current Web Session Usage Extent - Depth
Initial Extraction Sums of Squared
Eigenvalues Loadings
Component Total % of Cumulative Total % of Cumulative
Variance % Variance %
1 1.9 47 47 1.5 38 38
2 1.0 24 71 1.0 26 64
3 .7 18 89 1.0 25 89
Extraction Method: Principal Components Analysis with a Varimax Rotation
These results are consistent with those reported for the student sample (see Chapter 7
and Appendix K). The 4-items developed and first tested on the small student sample
explained 94% of the variance. In the web sample, these 4-items explained slightly less
variance at 89% of the variance, with consistent results in scale dimensionality. However,
the total scale reliability is less in the web sample (alpha = 0.6) as compared to a higher
‘Current web session usage extent – duration’ was operationalised using a single-item
measure with an 8-category multiple-choice question. Following data collection this item
duration”). The median for the web sample was category 4 = “1-3 hours”. This result is
consistent with the median for the student sample, which is also category was also
Exploratory factor analysis was performed on the 14-item scale with Likert questions
measuring ‘Perceived Ease of Web Use’. Initial data screening and analysis of correlation
patterns identified that individually and collectively all 14-items meet the necessary
Test of Sphericity: Approx. Chi-Square = 11372.124, df = 91, Sig. = 0.000). Thus a principal
146
dimensions that explained 64% of the variance. Three items were removed from the
analysis as they had factor loadings below +/- 0.6, resulting in a total of 11-items with a
total scale reliability of 0.9. See Table 27 for scale and dimension variance and Appendix J
These results are slightly inconsistent with those reported from the student sample
(Chapter 7 and Appendix K). The 14-items developed and first tested on the small
student sample explained 72% of the variance. However, In the web sample, these 14-
items were reduced to 11-items and explained slightly less variance (64%). In addition,
the dimensionality of the scale changed from 3 dimensions during scale testing to 4
dimensions. Despite these changes in the dimensionality of the scale, and also a
reduction in items, the reduced 11-item scale reported consistent total scale reliability of
0.9.
Exploratory factor analysis was performed on the initial 14-item scale with Likert
questions measuring ‘Perceived Web Usefulness’. Initial data screening and analysis of
correlation patterns identified that individually and collectively all 14-items meet the
0.881, Bartlettʹs Test of Sphericity: Approx. Chi-Square = 12614.496, df = 91, Sig. = 0.000).
Thus a principal components exploratory factor analysis was conducted. This analysis
extracted 4 dimensions that explained 70% of the variance, with a total scale reliability of
147
0.9. See Table 28 for scale and dimension variance and Appendix J for item factor
loadings.
These results are extremely consistent with those reported from the student sample (see
Chapter 7 and Appendix K). In fact, the 14-items developed and first tested on the small
student sample explained the same amount of variance (70%), extracted the same
number of dimensions (4) and reported the same degree of total scale reliability at 0.9.
exploratory factor analysis outlined below, these items were then summated to from an
overall measure of the four scales measuring actual common procedural, actual common
Exploratory factor analysis was performed on the initial 6-item scale measuring ‘Actual
Common Procedural Knowledge Content of the Web’. Initial data screening and analysis
of correlation patterns identified that individually and collectively all 6-items meet the
0.792, Bartlettʹs Test of Sphericity: Approx. Chi-Square = 2781.046, df = 15, Sig. = 0.000).
148
Thus a principal components exploratory factor analysis was conducted. This analysis
Procedural Web Knowledge Content’ with a total scale reliability of 0.7. See Table 29 for
scale and dimension variance and Appendix J for item factor loadings.
Table 29: Variance Explained of Actual Common Procedural Web Knowledge Content
Initial Extraction Sums of Squared
Eigenvalues Loadings
Component Total % of Cumulative Total % of Cumulative
Variance % Variance %
1 2.7 45 45 2.1 36 36
2 0.9 15 60 1.2 21 57
3 0.8 13 73 1.0 16 73
Extraction Method: Principal Components Analysis with a Varimax Rotation
These results are consistent with those from the student sample (see Chapter 7 and
Appendix K). The 6-items developed and first tested on the small student sample
explained 75% of the variance. In the web sample, these 6-items explained a slightly
lower, yet comparable 73% of the variance, with consistent results in the dimensionality
Exploratory factor analysis was performed on the initial 11-item scale measuring ‘Actual
Specialised Procedural Web Knowledge Content’. Initial data screening and analysis of
correlation patterns identified that individually and collectively all 11-items meet the
0.870, Bartlettʹs Test of Sphericity: Approx. Chi-Square = 4486.578, df = 55, Sig. = 0.000).
Thus a principal components exploratory factor analysis was conducted. This analysis
extracted 4 dimensions that explained 61% of the variance, with a total scale reliability of
0.8. See Table 30 for dimension and scale variance and Appendix J for item factor
loadings.
149
Table 30: Variance Explained of Actual Specialised Procedural Web Knowledge
Content
Initial Extraction Sums of Squared
Eigenvalues Loadings
Component Total % of Cumulative Total % of Cumulative
Variance % Variance %
1 3.7 33 33 2.2 20 20
2 1.1 10 43 1.8 17 37
3 1.0 9 52 1.6 14 51
4 0.9 9 61 1.1 10 61
Extraction Method: Principal Components Analysis with a Varimax Rotation
These results are consistent with those from the student sample (see Chapter 7 and
Appendix K). The 11-items developed and first tested on the small student sample
explained 58% of the variance. In the web sample, these 11-items explained a slightly
higher 61% of the variance, extracted an additional dimension and reported consistent
Exploratory factor analysis was performed on the initial 10-item scale measuring ‘Actual
Common Declarative Web Knowledge Content’. Initial data screening and analysis of
correlation patterns identified that individually and collectively all 10-items meet the
0.876, Bartlettʹs Test of Sphericity: Approx. Chi-Square = 5297.685, df = 45, Sig. = 0.000).
Thus a principal components exploratory factor analysis was conducted. This analysis
extracted 3 dimensions that explain 59% of the variance, with a total scale reliability of
0.8. See Table 31 for dimension and scale variance and Appendix J for item factor
loadings.
Table 31: Variance Explained of Actual Common Declarative Web Knowledge Content
Initial Extraction Sums of Squared
Eigenvalues Loadings
Component Total % of Cumulative Total % of Cumulative
Variance % Variance %
1 3.9 39 39 2.4 24 24
2 1.1 11 50 2.0 20 44
3 0.8 8 58 1.4 15 59
Extraction Method: Principal Components Analysis with a Varimax Rotation
150
These results are consistent with those from the student sample (see Chapter 7 and
Appendix K). The 10-items developed and first tested on the small student sample
explained 57% of the variance. In the web sample, these 10-items explained a slightly
higher 59% of the variance, extracted an additional dimension and reported consistent
Exploratory factor analysis was performed on the initial 10-item scale measuring ‘Actual
Specialised Declarative Web Knowledge Content’. Initial data screening and analysis of
correlation patterns identified that individually and collectively all 10-items meet the
0.879, Bartlettʹs Test of Sphericity: Approx. Chi-Square = 6080.714, df = 45, Sig. = 0.000).
Thus a principal components exploratory factor analysis was conducted. This analysis
extracted 5 dimensions that explain 76% of the variance, with a total scale reliability of
0.8. See Table 32 for scale and dimension variance and Appendix J for item factor
loadings.
These results are inconsistent with those from the student sample (see Chapter 7 and
Appendix K). The 10-items developed and first tested on the small student sample
explained 59% of the variance, whereas in the web sample these 10-items explained a
higher 75% of the variance, extracted two additional dimensions and reported consistent
151
9.4.4 SCALE VALIDATION: PERCEIVED WEB KNOWLEDGE CONTENT
To measure current overall web knowledge content, two items were used. These two
items were further investigated using Spearman’s Rho correlation coefficient. Table 33
shows a significant positive relationship between the two items (rs = .750, p<.01). Thus
the items will be summated to measure perceived overall web knowledge content.
These results are consistent with those reported from the student sample (see Chapter 7).
In fact, the 2-items developed and first tested on the small student sample reported a
correlation coefficient of r=.757, (p<0.01), very similar to that reported by the web sample.
Exploratory factor analysis was performed on the initial 7-item scale measuring
‘Perceived Declarative Web Knowledge Content’. Initial data screening and analysis of
correlation patterns identified that individually and collectively all 7-items meet the
0.939, Bartlettʹs Test of Sphericity: Approx. Chi-Square = 11559.443, df = 21, Sig. = 0.000).
Thus a principal components exploratory factor analysis was conducted. This analysis
extracted 1 dimension that explained 73% of the variance, with a total scale reliability of
0.9. See Table 34 for scale and dimension variance and Appendix J for item factor
loadings.
152
Table 34: Variance Explained of Perceived Declarative Web Knowledge Content
Initial Extraction Sums of Squared
Eigenvalues Loadings
Component Total % of Cumulative Total % of Cumulative
Variance % Variance %
1 5.1 73 73 5.1 73 73
Extraction Method: Principal Components Analysis with a Varimax Rotation
These results are consistent with those from the student sample (see Chapter 7 and
Appendix K). The 7-items developed and first tested on the small student sample
explained 75% of the variance, and in the web sample these 7-items explained 73% of the
Exploratory factor analysis was performed on the initial 4-item scale measuring
‘Perceived Procedural Web Knowledge Content’. Initial data screening and analysis of
correlation patterns identified that individually and collectively all 4-items meet the
Thus a principal components exploratory factor analysis was conducted. This analysis
extracted 1 dimension that explained 79% of the variance, with a total scale reliability of
0.9. See Table 35 for sale and dimension variance and Appendix J for item factor
loadings.
These results are consistent with those from the student sample (see Chapter 7 and
Appendix K). The 4-items developed and first tested on the small student sample
explained 77% of the variance, and in the web sample these 4-items explained 79% of the
153
9.4.5 SCALE ASSESSMENT: SUMMARY
In summary, all the scales were broadly consistent across the web sample and the
that they explained, and total scale reliability. Where inconsistency was identified, in the
case of the scale items measuring perceived ease of web use and actual specialised
declarative web knowledge, the resulting scale structure and reliability was an improved
result on the scale development studies. Overall, the scales developed here appear to be
valid.
For the remainder of this dissertation, analyses will only refer to the ‘overall construct’ or
‘variable’ and not its underlying dimensions (factors) that were extracted. This is to
firstly explore the relationships between the constructs with further analysis
knowledge, perceived web knowledge, web perceptions, and current web session usage,
are reported in Appendix P (Table P1 and Figures P1-P15). The frequency percentages
reported are partitioned into ‘low’, ‘medium’ and ‘high’ sample scores for the entire
sample (n=2077), and the two sample groups, users with WSD/M experience (n=1177)
Users with WSD/M experience have consistently high actual web knowledge scores,
whereas users with no WSD/M experience report medium to high scores, and low-
medium scores for actual specialised declarative knowledge content of the web.
154
Users with WSD/M experience have consistently high perceptions of how much they
know about the web, whereas users with no WSD/M experience report medium to
Users with WSD/M experience and with no WSD/M experience report consistently
medium to high levels of perceived ease of web use and perceived web usefulness.
Users with WSD/M experience report consistently higher level of usage frequency,
situational variety, motivational variety, usage depth and usage duration than users
with no WSD/M experience. However both user groups report similar levels of usage
breadth.
Thus, overall, the web sample recruited performs consistently med-to-high across all
the variables surveyed, and both users with and without WSD/M experience differ
significant difference between the two users groups for their performance across the
variables investigated in this dissertation. The mean scores for both web user groups
(Table P2) and the t-test results (Table P3) are presented in Appendix P. The results are
summarized here.
frequencies, the results identify that users with no WSD/M experience on average have
lower mean scores for actual web knowledge, perceived web knowledge, perceived ease
of web use and perceived web usefulness, and current web session usage – except for
breadth of usage extent. On this item the mean scores between the two groups are
have significantly less actual knowledge of the web, less perceived knowledge of the
155
web, perceive the web as less easier and less useful to use, and use the web less
frequently, with less number of motivations, at a lower number and type of locations, for
a lower duration of time, accessing fewer web sites and search engines than users with
and without WSD/M experience was found, for the number of different and/or similar
web sites or search engines each group accessed (i.e., usage breadth).
design and/or maintenance (WSD/M) experience and those users without this experience
on their level of actual knowledge, perceived knowledge, how easy and useful they think
the web is, whereas their usage of the web, and usage breadth of the web is similar
usefulness, a relationship proposed by Davis (1986) and further validated by other TAM
researchers was replicated here. Davis (1986) hypothesised that perceived ease of use
will have a significant direct effect on perceived usefulness stating that, all else being
equal, a system which is easier to use will result in greater usefulness for the user. He
reported a relatively strong relationship between perceived ease of use and perceived
usefulness (r=.64).
For non-web-based electronic systems, this hypothesis has been further supported
within the literature (Davis et al. 1989a; Davis 1989b; Adams et al. 1992; Taylor and Todd
1995; Igbaria et al. 1995; Chau 1996; Gefen and Keil 1998; Bronson 1999; Karahanna and
Straub 1999). In addition, with respect to web-based systems, and as tested from the
supported (Morris and Dillion 1997; Teo et al. 1999; Moon and Kim 2001). Given that the
156
variables were not normally distributed in the present sample, a non-parametric
correlation was conducted (Spearman’s Rho) (see Chapter 8 for a discussion). Results are
Table 36: Perceived Ease of Web Use & Perceived Web Usefulness – Spearman’s Rho
Correlation Coefficient
Web Site Design and
Maintenance Perceived Web
Experience (0/1) Usefulness (Sum)
No Experience Spearman’s Perceived Ease of Correlation Coefficient .755 **
Rho Web Use (Sum)
Sig. (1-tailed) .000
N 900
Experience Spearman’s Perceived Ease of Correlation Coefficient .731 **
Rho Web Use (Sum)
Sig. (1-tailed) .000
N 1177
**. Correlation significant at the .01 level (1-tailed)
Table 36 shows a significant positive relationship between PEWU and PWU for both web
users with no web site design and maintenance experience (rs = .755, p<.01) and those
with this experience (rs = .731, p<.01). Therefore, the easier the web is to use, the more
useful users find the web. The statistically significant positive relationship found in this
study, is consistent with the findings of Davis (1986) and other researchers.
an impressive 41% response rate from those who visited the web site, but a response of
only 0.1% of those who were exposed to the banner ad campaign. The scales used in the
web survey were broadly consistent in their dimensionality and reliability with those
developed from the student samples and the measures of perceived ease of web use and
mean scores across the constructs measured were identified to exist between users with
and without WSD/M experience. This descriptive finding highlights the importance of
segmenting the sample according to their level of web site design and/or maintenance
157
The next chapter, Chapter 10, presents the results of the relationships hypothesized in
Chapters 3-to-5 and summarized in Chapter 6. In brief, Chapter 10 presents the results of
Appendix Q.
158
C HAPTER 10: E MPIRICAL R ESULTS
M ULTIVARIATE A NALYSES
‘Truth in science can be defined as the working hypothesis best suited to open the
way to the next better one’
10.1 INTRODUCTION
The hypotheses that were introduced in this dissertation are examined here using
analysis of variance (ANOVA) and stepwise multiple regression analysis (MRA). The
motivation for using these techniques, and the assumptions that were assessed to
confirm the suitability of this procedure, were described in Chapter 8. The results of
at least one of the assumptions has been violated (e.g., it is hard to avoid the problem of
multicollinearity when so many of the measures refer to closely related constructs). This
is not uncommon and, as is often the case with multivariate techniques, results need to
analyses have also been used to further validate these findings and are presented in
Appendix Q.
explore the variables and the relationships between them, a correlation analysis was
159
conducted, the results of which are discussed here with the correlation matrix presented
in Appendix R. The total sample is presented in Table R1 and the split sample results for
users with and without WSD/M experience are presented in Table R2 in Appendix R. In
totality, it was identified that 80% of the correlations for the total sample (n=2077), 70%
for the users with no WSD/M experience, and 60% for users with WSD/M experience,
were statistically significant. More specifically, some observed trends identified in these
tables are:
the web;
Perceived knowledge of the web has a significant positive correlation with current
web session usage (except breadth of web use), perceived ease of web use and
Perceived ease of web use and perceived web usefulness has significant positive
correlation with the number of motives for web use and the depth of web use;
The number of motives for web session use has a significant positive correlation with
the number and types of situations and frequency of web session use.
When comparing the two web users groups (see Table R2 in Appendix R), it is evident
that the significant correlations reported for the users with no WSD/M experience are on
average stronger than those reported for users with WSD/M experience. Furthermore,
slight variation in some of the correlations is also reported identifying key differences
The specific research questions and hypotheses put forward in this dissertation will each
160
From the bivariate analyses presented in Appendix Q, 6 out of 12 hypotheses tested
were supported for users with no web site design and maintenance experience (n = 900).
Some 5 out of 12 hypotheses were supported for users with experience (n = 1177). These
hypotheses form the basis of the first 6 multiple regression analyses conducted in this
that the independent variables perceived web usefulness (PWU) and perceived ease of
161
web use (PEWU) have on the dependent variable, current web session usage frequency
(WSUF).
Table 39 contains the results of the stepwise multiple regression. Of the 6 assumptions, 3
For users with no WSD/M experience, PEWU has a significant positive effect on WSUF (t
= 3.92, p<.01), however it only explains 1.6% of the variance in WSUF (Adj. R2 = .016).
Under the stepwise method, PWU was excluded from this model.
By contrast, for users with WSD//M experience, PWU has a significant positive effect (t =
4.64, p<.01), however it only explains 1.7% of the variance in WSUF (Adj. R2 = .017).
10.3.1.2 Summary
Perceived ease of web use (PEWU) has a weak positive relationship with current web
session usage frequency for users with no WSD/M experience, and perceived web
usefulness (PWU) has a positive relationship for users with WSD/M experience.
162
10.3.2 MRA2: WSUVS = F (PWU & PEWU)
To test H3a and H4a, multiple regression analysis two (MRA2) examines the influence
that the independent variables perceived web usefulness (PWU) and perceived ease of
web use (PEWU) have on the dependent variable, current web session usage variety –
situational (WSUVS).
Table 40 contains the results of the stepwise multiple regression. Of the 6 assumptions,
For users with no WSD/M experience, PWU was found to be the best predictor of
WSUVS having a positive effect (t = 6.91, p <.01), however it only explained 0.06% of the
variance in WSUVS (Adj. R2 = .006). PEWU was removed from the model.
For users with WSD/M experience, both PEWU and PWU were removed from the model
10.3.2.2 Summary
Perceived web usefulness (PWU) was the best determinant of the variety of situations
from which the web is accessed for users with no WSD/M experience. However for users
with this experience, neither perceived ease of use, nor perceived web usefulness, was a
significant predictor.
163
10.3.3 MRA3: WSUVMNO1 = F (PWU & PEWU)
To test H3b and H4b, multiple regression analysis three (MRA3) examines the influence
that the independent variables perceived web usefulness (PWU) and perceived ease of
web use (PEWU) have on the dependent variable, current web session usage variety –
Table 41 contains the results of the stepwise multiple regression. Of the 6 assumptions, 4
For users with no WSD/M experience, both PEWU and PWU had a significant positive
effect on WSUVMNO1 (t = 5.50, p <.01, and t = 2.30, p <.01, respectively), with this model
explaining 13.6% of the variance in WSUVMNO1. PWU had a greater impact than
PEWU.
For users with WSD/M experience, PEWU was removed from the model, with only PWU
having a significant positive effect on WSUVMNO1 (t = 7.47, p <.01). This explained only
164
10.3.3.2 Summary
Perceived ease of use and perceived web usefulness both had a significant positive effect
on the number of motivations for which the web is used by users with no WSD/M
experience. Perceived web usefulness was the stronger predictor. In comparison, for
users with WSD/M experience, only perceived usefulness of the web had a significant
that the independent variables perceived web usefulness (PWU) and perceived ease of
web use (PEWU) have on the dependent variable, current web session usage extent -
breadth (WSUEB).
Table 42 contains the results of the stepwise multiple regression. Of the 6 assumptions, 4
For users with no WSD/M experience, both PEWU and PWU were excluded from the
model. For users with WSD/M experience, PWU had a significant negative effect on
WSUEB (t = -2.05, p <.05), but it only explained 0.3% of the variance in WSUEB (Adj. R2 =
165
10.3.4.2 Summary
Neither PEWU nor PWU is a good predictor of WSUEB for users with no WSD/M
experience. Only perceived web usefulness is a predictor of WSUEB for users with
WSD/M experience, and the magnitude of this negative relationship is quite small.
that the independent variables perceived web usefulness (PWU) and perceived ease of
web use (PEWU) have on the dependent variable, current web session usage extent -
depth (WSUED).
Table 43 contains the results of the stepwise multiple regression. Of the 6 assumptions, 5
For users with no WSD/M experience, both PEWU and PWU had a significant positive
effect on WSUED (t = 5.38, p <.01, and t = 3.73, p < .01, respectively), and they explained a
total 17.5% of the variance in WSUED (Adj. R2 = .175). It is evident from these results that
PEWU was the primary determinant of WSUED, with PWU in a secondary role.
For users with WSD/M experience, both PEWU and PWU had a significant positive
effect on WSUED (t = 3.18, p <.01, and t = 2.36, p < .01, respectively), and they explained a
total 4.8% of the variance in WSUED (Adj. R2 = .048). It is evident that PEWU was the
166
Table 43: Multiple Regression Results
MRA5: Web Session Usage Extent - Depth = F (Perceived Web Usefulness & Perceived
Ease of Web Use)
WSD/M Adj. Independent S.E. Sig.
Dep. Var. R2 b Β t. stat. F
Exp. R2 Variable (b) Level
No Exp Web Session .177 .175 Constant 7.157 .791 9.052 96.170 .000
Usage Extent – Perceived Ease of .111 .021 .262 5.377 .000 +
Depth Web Use
Perceived Web 0.066 .018 .182 3.731 .000 +
Usefulness
With Web Session .049 .048 Constant 13.712 .770 17.819 30.478 .000
Exp Usage Extent - Perceived Ease of 0.058 .018 .136 3.180 .000 +
Depth Web Use
Perceived Web 0.035 .015 .101 2.364 .000 +
Usefulness
10.3.5.2 Summary
Perceived ease of web use was the primary determinant, and perceived web usefulness
the secondary determinant, of the depth of web use for both web users with and without
WSD/M experience. These positive relationships were somewhat stronger for users with
that the independent variables perceived web usefulness (PWU) and perceived ease of
web use (PEWU) have on the dependent variable, current web session usage extent -
duration (WSUEDUR).
Table 44 contains the results of the stepwise multiple regression. Of the 6 assumptions, 3
For users with no WSD/M experience, PWU was excluded from the model and PEWU
was the primary determinant of WSUEDUR (t = 7.06, p <.01), but this positive
167
For users with WSD/M experience, PWU was again excluded, and PEWU had a
significant positive effect on WSUEDUR (t = 3.68, p <.01), although it explained only 1.1%
10.3.6.2 Summary
Perceived ease of web use was the primary determinant of how long the web is used, for
both web user groups. However, this positive relationship was stronger for users with no
WSD/M experience.
PEWU was the primary determinant of how frequently the web is used for users
primary determinant for users with WSD/M experience, also having a positive effect.
(MRA1)
PWU was the best predictor of the variety of situations from where the web is
accessed for users with no WSD/M experience, having a positive effect. Neither ease
of use, nor usefulness, were determinants for users with WSD/M experience (MRA2)
For web users with no WSD/M experience, PWU was the primary and PEWU the
secondary determinant of the number of motivations for web use, both having a
positive effect. By comparison, only PWU had a positive effect on the number of
motivations for web use for users with WSD/M experience (MRA3).
168
Neither, PEWU or PWU were determinants for breadth of web session use for user
with no WSD/M experience. However PWU was the main determinant for users with
WSD/M experience, having a negative effect. PEWU was removed for this user
group (MRA4).
PEWU was the primary, and PWU the secondary, determinant of the depth of web
session use for both web users groups, having a positive effect. However, the model
explained more variance for users with no WSD/M experience, than those with this
experience (MRA5).
Perceived ease of use was the primary determinant of how long the web is used, for
both web user groups, and was stronger for users with no WSD/M experience. PWU
was removed from the model. All relationships found were positive (MRA6).
These results show that a significant difference is present between users with and
without WSD/M experience. This use of the web is influenced by ease of web use and
web usefulness, but the influence is somewhat different across the two user groups. It is
also evident that a difference exists between the type of usage behaviour being
determined (i.e., frequency, duration, variety etc) for each user group. For example, both
perceived ease of web use and perceived web use are important for predicting usage
behaviour for users with no WSD/M experience, however for users with this experience,
perceived web usefulness is of core importance. See summary Table 45 and Table 46
below for comparative results from the multivariate analyses conducted in this chapter
169
H6B PWU Positive WSUED Accept (Reject Null)
H5C PEWU Curvilinear WSUEDUR Reject (Accept Null)
PEWU (+)
H6C PWU Positive WSUEDUR Accept (Reject Null)
the web and a person’s perceived usefulness of the web? As summarized in Table 47
below, for web users with no web site design and maintenance experience (n = 900), 6
below, for web users with experience, 5 out of 7 hypotheses were supported. For more
170
Table 48: Web User Group B – With WSD/M Experience: RQ2
Bivariate Bivariate Result Multivariate
Independent Relationship Dependent
Label (Appendix Q) Label
H7A ACPWK Curvilinear PWU Reject (Accept Null)
H8A ACDWK Positive PWU Accept (Reject Null)
H9A ASPWK Positive PWU Accept (Reject Null)
H10A ASDWK Positive PWU Reject (Accept Null) MRA7
H11A SWPK Positive PWU Accept (Reject Null)
H12A SWDK Positive PWU Accept (Reject Null)
H13A SWOK Positive PWU Accept (Reject Null)
Two multivariate techniques will now be used to further test these hypotheses. ANOVA
will firstly be used to explore these 7 hypotheses and determine the mean effect that
actual and perceived knowledge content of the web has on a user’s level of perceived
web usefulness. Following this analysis, stepwise multiple regression analysis will be
used to more specifically identify which types of actual and perceived web knowledge
content are statistically significant determinants of perceived usefulness of the web for
perceived web knowledge content a user has influences his/her perceptions of how
useful the web is to use. There are four measures of actual knowledge content and three
As presented in Table 49, the result of the ANOVAs reveals that the level of actual and
perceived web knowledge content (i.e., low, medium and high) that a web user has, has
a significant effect on their level of perceived usefulness of the web. A closer look at the
mean scores (see Appendix S) shows that the level of actual common procedural, actual
positive impact on the level of perceived web usefulness. Thus, the more users know and
171
think they know about the web, the more useful they will find the web to use. This result
is consistent for both web users with and without WSD/M experience. See Table 49.
Table 49: ANOVA: Effect of Level of Actual [4] and Perceived [3] Web Knowledge
Content (L/M/H) on Perceived Web Usefulness
No WSD/M Experience With WSD/M Experience
Knowledge Content Groups
df F. Stat. Sig. Level df F. Stat. Sig. Level
Between 2 2
Actual Common Procedural Within 897 71.207 .000* 1174 13.872 .000*
Total 899 1176
Between 2 2
Actual Common Declarative Within 897 75.007 .000* 1174 9.841 .000*
Total 899 1176
Between 2 2
Actual Specialised Procedural Within 897 25.927 .000* 1174 11.044 .000*
Total 899 1176
Between 2 2
Actual Specialised Declarative Within 897 11.214 .000* 1174 4.923 .000*
Total 899 1176
Between 2 2
Perceived Overall Within 897 96.596 .000* 1174 65.219 .000*
Total 899 1176
Between 2 2
Perceived Procedural Within 897 135.57 .000* 1174 67.264 .000*
Total 899 1176
Between 2 2
Perceived Declarative Within 897 115.109 .000* 1174 66.462 .000*
Total 899 1176
* p<0.01
10.4.1.2 Summary
In summary, from the above ANOVA, it is evident that for both web users with and
without WD/M experience that actual and perceived web knowledge content has a
actual common procedural web knowledge (ACPWK), actual common declarative web
172
(SWPK), perceived declarative web knowledge (SWDK) and perceived overall web
knowledge (SWOK).
Table 50 contains the stepwise multiple regression results. Of the 6 assumptions, 2 are
For users with no WSD/M experience, SWPK (t = 6.97, p <.01), ACPWK (t = 3.73, p <.01),
ACDWK (t = 3.76, p <.01) and SWDK (t = 3.24, p <.01) had a significant positive effect,
and ASDWK (t = -4.79, p <.01) and SWOK (t = -2.86, p <.01) had a significant negative
effect, on PWU. Collectively, these variables explained 31.7% of the variance in perceived
173
web usefulness. ASPWK was excluded from the model. SWPK was the strongest
determinant of PWU.
For users with WSD/M experience, SWPK (t = 7.69, p <.01) and SWOK (t = 2.53, p <.01)
had a positive effect, and ASDWK (t = -4.73, p <.01) a negative effect, on PWU, with the 4
other variables excluded. The 3-variable model explained 17.4% of the variance in
10.4.2.2. Summary
Perceived procedural web knowledge content was the strongest positive predictor of
perceived web usefulness for users with no web site design and maintenance experience.
This was followed closely by actual common declarative and actual common procedural
web knowledge content. Actual specialised declarative web knowledge and perceived
In comparison, for users with WSD/M experience, perceived procedural web knowledge
and perceived overall web knowledge are the key positive determinants. Actual
usefulness.
the web and what certain web features are, and what they actually know about common
procedures and attributes of the web, increases how useful they perceive the web to be.
Whereas the more they actually know about specialised features and attributes of the
web and what they think they know overall, decreases how useful they think the web is.
In comparison, for users with WSD/M experience, what they think they know about how
to use the web and what they think they know overall, increases how useful they think
174
the web is. Whereas what they actually know about specialised attributes and features of
the web decreases how useful they perceive the web to be. See summary Table 51 and
Table 52 below for comparative results from the bivariate (Appendix Q) and multivariate
the web and a person’s perceived ease of web use? As summarized in Table 53 and Table
54 below, for web users with no web site design and maintenance experience, 7 out of 7
hypotheses were supported in the bivariate analyses and for web users with experience,
175
Table 53: Web User Group A – No WSD/M Experience: RQ3
Bivariate Bivariate Result Multivariate
Independent Relationship Dependent
Label (Appendix Q) Label
H14A ACPWK Positive PEWU Accept (Reject Null)
H15A ACDWK Positive PEWU Accept (Reject Null)
H16A ASPWK Positive PEWU Accept (Reject Null)
H17A ASDWK Positive PEWU Accept (Reject Null) MRA8
H18A SWPK Positive PEWU Accept (Reject Null)
H19A SWDK Positive PEWU Accept (Reject Null)
H20A SWOK Positive PEWU Accept (Reject Null)
Two multivariate techniques will now be used to further test these hypotheses. ANOVA
will firstly be used to explore these 7 hypotheses and determine the mean effect that
actual and perceived knowledge content of the web has on a user’s level of perceived
ease of web use. Following this analysis, stepwise multiple regression analysis will be
used to more specifically identify which types of actual and perceived knowledge
content are statistically significant determinants of perceived ease of web use for each
perceived web knowledge content a users has effects their perceptions of how easy the
web is to use. There are four measures of actual knowledge and three measures of
176
10.5.1.1 Statistical Assessment
As presented in Table Table 55, the results of the ANOVAs reveal that the level of
perceived and actual web knowledge content (i.e., low, medium and high) has a
significant effect on a user’s perceived ease of use of the web. A closer look at the mean
scores (see Appendix S) shows that the level of actual common procedural, actual
positive impact on the level of perceived ease of web use. Thus, the more users know
and think they know about the web, the easier users will find the web to use. This result
is consistent for web users with and without WSD/M experience. See Table 55.
Table 55: ANOVA: Effect of Level of Actual [4] and Perceived [3] Web Knowledge
Content (L/M/H) on Perceived Ease of Web Use
No WSD/M Experience With WSD/M Experience
Knowledge Content Groups
df F. Stat. Sig. Level df F. Stat. Sig. Level
Between 2 2
Actual Common Procedural Within 897 57.215 .000* 1174 11.518 .000*
Total 899 1176
Between 2 2
Actual Common Declarative Within 897 49.728 .000* 1174 8.981 .000*
Total 899 1176
Between 2 2
Actual Specialised Procedural Within 897 29.077 .000* 1174 14.269 .000*
Total 899 1176
Between 2 2
Actual Specialised Declarative Within 897 13.139 .000* 1174 7.257 .001*
Total 899 1176
Between 2 2
Perceived Overall Within 897 198.167 .000* 1174 83.681 .000*
Total 899 1176
Between 2 2
Perceived Procedural Within 897 328.571 .000* 1174 115.911 .000*
Total 899 1176
Between 2 2
Perceived Declarative Within 897 263.789 .000* 1174 110.421 .000*
Total 899 1176
* p<0.01
10.5.1.2 Summary
In summary, from the above ANOVAs, it is evident that for both web users with and
without WD/M experience that actual [4] and perceived [3] web knowledge content has a
177
10.5.2 MRA8: PEWU = F (ACTUAL & PERCEIVED WEB KNOWLEDGE)
To further test H14a to H20a, multiple regression analysis eight (MRA8) examines the
influence on perceived ease of web use (PEWU) of the following independent variables:
actual common procedural web knowledge (ACPWK), actual common declarative web
(SWPK), perceived declarative web knowledge (SWDK) and perceived overall web
knowledge (SWOK).
For users with no WSD/M experience, SWPK (t = 12.93, p <.01), ACPWK (t = 6.01, p <.01)
and SWDK (t = 6.57, p <.01) had a significant positive effect, and ASDWK (t = -5.90, p
<.01) and SWOK (t = -5.06, p <.01) had a significant negative effect, on PEWU. These 5
variables explained a total 54.4% of the variance in perceived ease of web use, with
perceived procedural web knowledge being the strongest determinant. Actual common
knowledge (ASPWK) were removed from the model for this user group.
For users with WSD/M experience, SWPK (t = 9.34, p <.05), SWDK (t = 2.72, p <.01) and
ACDWK (t = 2.45, p <.01) all had a significant positive effect, and ASDWK (t = -7.11, p
<.01) had a significant negative effect, on perceived ease of web use. These 4 variables
explained a total 27.9% of the variance in perceived ease of web use. Actual common
(ASPWK), and perceived overall web knowledge (SWOK) were all removed from the
model.
178
Table 56: Multiple Regression Results
MRA8: Perceived Ease of Web Use = F (Actual [4] & Perceived [3] Knowledge Content)
WSD/M Dep. Adj. Independent S.E. F Sig.
R2 b β t. stat.
Exp. Var. R2 Variable (b) Level
No Expa Perceived .546 .544 Constant 22.964 1.058 21.700 215.441 .000
Ease of Perceived
Web Use Procedural Web 1.308 .101 .662 12.934 .000 +
Knowledge
Actual Specialised
Declarative Web -.584 .099 -.166 -5.899 .000 -
Knowledge
Actual Common
Prodcedural Web 1.184 .197 .163 6.005 .000 +
Knowledge
Perceived
Declarative Web .405 .062 .333 6.566 .000 +
Knowledge
Perceived Overall
-1.046 .207 -.272 -5.063 .000 -
Web Knowledge
With Perceived .282 .279 Constant 24.374 1.813 13.441 114.890 .000
Expb Ease of Perceived
Web Use Procedural Web 1.219 .130 .457 9.340 .000 +
Knowledge
Actual Specialised
Declarative Web -.970 .136 -.239 -7.114 .000 -
Knowledge
Perceived
Declarative Web .191 .070 .135 2.721 .001 +
Knowledge
Actual Common
Declarative Web .464 .189 .078 2.452 .003 +
Knowledge
a No WSD/M experience – excluded variables: Actual Common Declarative Web Knowledge (ACDWK);
Actual Specialised Procedural Web Knowledge (ASPWK)
b With WSD/M experience – excluded variables: Actual Common Procedural Web Knowledge (ACPWK);
Actual Specialised Procedural Web Knowledge (ASPWK); Perceived Overall Web Knowledge (SWOK)
10.5.2.2 Summary
perceived ease of web use for users with no WSD/M experience, with perceived
declarative web knowledge and actual common procedural web knowledge being
For users with no WSD/M experience, perceived procedural web knowledge was the
179
10.5.3 RESEARCH QUESTION THREE: SUMMARY
For web users with no WSD/M experience, what they think they know about how to use
the web, their actual common knowledge about web procedures, and what they think
they know about common attributes and features of the web, increases how easy they
think the web is. However, what they actually know about specialised features and
attributes of the web, and what they think they know overall, has a negative effect on
For users with WSD/M experience, what they think they know about how to use the
web, and what certain attributes and features are, and what they actually know about
common attributes and features, increases how easy they think the web is to use. What
they actually know about specialised features and attributes of the web decreases how
easy they think the web is to use. These results are broadly consistent across both user
groups. See Table 57 and Table 58 below for comparative results from the bivariate and
multivariate analyses.
180
10.6 MULTIVARIATE ANALYSIS: SUMMARY
In summary, for users with no WSD/M experience, how easy the web is to use is a
determinant. Furthermore, what this user group thinks they know about the web, and
what they actually know, are key determinants of how easy and useful they see the web.
For users with WSD/M experience, how useful the web is to use is a primary
Furthermore, what this web user group thinks they know about how to use the web is a
primary determinant of how easy and how useful they perceive the web to be.
The results reported in this chapter are further validated with bivariate analyses that are
presented in Appendix Q. Next, in Chapter 11, the results reported here are discussed,
181
C HAPTER 11: E MPIRICAL D ISCUSSION
‘Free and fair discussion will ever be found the firmest friend to truth’
- G. Campbell -
11.1 INTRODUCTION
This chapter discusses the main results of the study, and compares and contrasts them
with findings from previous studies of web system use and usability. Results for
research question 1 are discussed first. This question investigated the relationship
between web user perceptions and usage. Secondly, the implications drawn from the
relationship between a user’s knowledge of the web and their perceived usefulness of
the web are discussed. Thirdly, the implications of the findings derived from the
investigation of user knowledge of the web and their perceived ease of web use are
discussed. All results are compared across two web user groups, those with and without
web and that person’s current web session usage? This was investigated by looking at
the relationship between perceived ease of web use and perceived web usefulness with
of use; 5) depth of use and 6) duration of use. These are discussed in turn.
session usage frequency for both web user groups. No curvilinear relationship existed
between perceived ease of use and web session usage frequency for either user group.
182
Further investigation using multiple regression revealed that how easy a user with no
WSD/M experience perceives the web is in fact the primary positive determinant of how
frequently they will use the web. Perceived usefulness was not found to be a good
predictor of web session usage frequency for this user group. Thus, the easier this group
perceive the web to be, the more frequently they will actually use it.
In contrast, how useful users with WSD/M experience perceived the web was a primary
positive determinant of how frequently they use it. Thus, for this group, the more useful
they perceive the web to be the more frequently they will use it.
11.2.1.1 Discussion
The findings reported for users with WSD/M experience are consistent with the results
obtained in the original TAM study on PROFs™ email, XEDIT™, Chartmaster™ and
Pendraw™ by Davis (1986) where perceived usefulness of the system had a profound
effect on self-predicted usage behaviour and no statistically significant effect was found
for perceived ease of use. This is also consistent with subsequent studies of other
Bagozzi et al., 1992); PROFs™ email and EXIDT™ (Davis, 1989b); of micro-computers
(Igbaria et al., 1995); of electronic and voice mail (Adams et al., 1992); of word processing
Also, the relationship found for users with no WSD/M experience between perceived
ease of use and usage frequency is consistent with research conducted by Morris and
Dillion (1997) for the browser program Netscape™ and for web studies by Fenech (1997),
Teo et al. (1999) and Lederer et al. (2000). These researchers found that perceived ease of
These findings provide further support for the suggestion put forward by Moore and
Benbasat (1991) and Adams et al. (1992) that the mandatory use of the system in an
organisational setting might have an impact on perceptions and usage of a system. These
183
authors also emphasise the influence of subjective norms – i.e., perceptions influenced by
the view of what others think users should be doing. In the current study, use of the
system was segmented according to those that ‘had mandatory use experience’ and
those that ‘did not’, as defined by WSD/M experience. Differences were found between
these two groups. If your use of the web was not defined by a work-related role (i.e., no
experience designing and maintaining web sites) how easy you perceive the web is the
primary determinant of how frequently you use it. In contrast, if your use is defined by a
work-related role (i.e., experience with designing and maintaining web sites) then how
useful you perceive the web is the primary determinant of your usage frequency.
This finding has implications not only for web site design, but also for marketing
communications directed at certain user groups. In these cases, the aim might be to
promote an online presence or use of the web through advertising, publicity, PR, or even
word-of-mouth. For example, for communications aimed at those users with less
experience, emphasis of the ease of use of the system is paramount. In contrast, for users
with more specialised experience, communication of the more useful benefits of the
(WSUVS)
Perceived web usefulness was found to have a positive relationship for both users with
perceived ease of use and web session usage variety–situational for either group.
Investigation using multiple regression revealed that how useful a user with no WSD/M
experience perceives the web was the primary positive determinant of the number of
situations from which they will access the web, thus increasing situational variety.
Perceived ease of use was not a good predictor for this user group.
184
In contrast, for users with WSD/M experience, neither how easy, nor how useful the web
was perceived to be, had any impact on the number of situations from which the web
was accessed.
11.2.2.1 Discussion
These findings show that no direct significant relationship exists between perceived ease
of web use and the number of situations from which the web is accessed for either web
users with or without WSD/M experience. Limited research is available with which to
compare these results; however, if compared to the investigation by Igbaria et al. (1995)
that identified that perceived ease of use of a system had a positive impact on usage
variety and the Teo et al. (1999) study of the diversity of system use, the current results
are inconsistent.
By contrast, it was found that perceived web usefulness was a good determinant of the
number of situations/locations from which the web would be accessed for users with no
WSD/M experience. Thus, the more useful the web is seen to be, so it will be accessed
from more situations and/or locations. This result is consistent with the findings reported
by Igbaria et al. (1995) and Teo et al. (1999) who argued that the more useful a system the
more variety and diversity of use will occur. The implication is that to increase the
number of situations and/or locations from which those with no WSD/M experience
access the web, their perception of the web’s usefulness needs to be increased. This, in
turn, has implications for the design of web sites and the marketing communications
used to communicate to this user group about web sites and services – e.g., advertising
to direct users to retail and brand sites for online shopping or promotions to encourage
Perceived usefulness was not a good predictor for users with WSD/M experience. Again,
this result appears to be at odds with earlier studies. However, it must be kept in mind
that other external factors (e.g., physical access, cost, location, etc.) might be very
185
important determinants of the number of situations/locations from where the web is
accessed.
(WSUVMNO1)
Perceived web usefulness was found to positively relate to the number of motivations for
web use. No curvilinear relationship was found between perceived ease of web use and
the number of motivations for web use for either user group.
Investigation using multiple regression revealed that for users with no WSD/M
experience how useful they perceived the web to be, and then how easy they perceived it
to be, had a positive effect on the number of motivations they had for using the web. In
contrast, for users with WSD/M experience how useful they thought the web was, was
the primary positive determinant of the number of motivations they had for using it.
How easy they thought the web was had no additional effect.
11.2.3.1 Discussion
How useful the web is perceived to be was the primary determinant of the number of
motivations for its use by both user groups. Thus, the more useful the web is perceived,
the greater the number of motivations users have for using it. This result is consistent
with that reported by Teo et al. (1999) in that users come to adopt and use a system
the argument put forward by Ram and Jung (1990) that product usage variety is
and 3, the web is a very complex system with numerous functions and uses – users who
In this sample the relationship is more moderate in strength for users with no WSD/M
experience than it is for users with WSD/M experience. Perhaps for the latter group other
186
elements such as work related factors (i.e., job description) also influence motivations for
web use and thus both perceived usefulness and work task expectations (i.e., subjective
norms) will influence the number of motivations for use. This latter distinction is
consistent with the argument put forward by Moore and Benbasat (1991) and Adams et
al. (1992) that the mandatory use of the system in an organisational setting, and the
influence of the subjective norm, impact end usage – in this case they influence the
The results for users with no WSD/M experience are consistent with the findings
reported by Igbaria et al. (1995) and Teo et al. (1999) that perceived ease of system use
will have a positive impact on usage variety and diversity. This begs the question as to
the importance of usefulness over ease of use with respect to providing increased
motivation for use. It seems that no matter how easy a system is to use, how useful it is
for various tasks provides the main influence on the number of motivations for system
usage. The implication is that web sites must be useful, then users will have more
reasons to use and return to the site than if it was merely easy to use. If the site is not that
easy to use, users will still have a number of reasons to use the site if it is deemed useful.
By contrast, for those users who do have experience with WSD/M, perceptions of ease of
use do not influence the number of motivations for usage. Usefulness alone is the
use and breadth of web session use for users with no WSD/M experience. This was not
found for users with this experience. Furthermore, no relationship was found between
perceived web usefulness and breadth of web session use for either user group.
Investigation using multiple regression revealed that neither how useful nor how easy
the web was thought to be were determinants of breadth of web session use for users
with no WSD/M experience. In comparison, how useful the web was perceived to be was
187
a primary negative determinant of web session usage breadth for users with WSD/M
experience. These experienced users access fewer different and/or new web sites, and
11.2.4.1 Discussion
As specified above for users with no WSD/M experience, perceived usefulness in the
bivariate and multivariate analysis was not found to influence breadth of web session
use. Thus, no matter how useful or not they perceive the web, this will not influence the
breadth of different types of web sites or search tools used. A u-shaped relationship was
identified between perceived ease of use and breadth of session use in the bivariate
analysis - possibly explaining why no linear relationship was found for this user group
in the multivariate analyses (i.e., it was curvilinear not linear). The relationship implies
that at first, when the web is seen as not very easy to use, the number of new or different
sites/search engines used is quite low (i.e., exploration may be low). As the web becomes
moderately easier to use, and as users explore a little, the number of new/different sites
or search engines accessed grows. Then, as the web is seen as very easy to use, the
number of new/different web sites or search engines decreases – perhaps as use becomes
more routine and as favourite sites and/or search tools are bookmarked and repeatedly
used.
For users with WSD/M experience, as the perceived usefulness of the web increases, the
number of new and/or differing sites visited or search tools used slightly decreases. The
implication is that the more useful the web becomes, so the less exploration across new
and/or differing sites occurs, perhaps as users develop a ‘choice set’ (i.e., a list of regular
bookmarked sites). This suggests experience gives rise to online loyalty toward web sites
and/or search tools. And, perhaps, implies a need to see how quickly users develop their
188
11.2.5 PREDICTING WEB SESSION USAGE EXTENT: DEPTH (WSUED)
Perceived web usefulness was found to have a positive relationship with depth of web
session use for both user groups. No curvilinear relationship was found between
Further investigation using multiple regression revealed that for both user groups, how
easy they thought the web was, followed by how useful they thought it was, will have a
positive effect on the total number of web sites and search tools used.
11.2.5.1 Discussion
The main difference between the two groups is that perceptions of ease of use and
usefulness have a larger impact on the depth of current web session usage for users
without WSD/M experience than for users with this experience. Dreze and Zufryden
(1997a) found that site visit depth was explained by web site attributes; that is, the more
site attributes the greater the site depth, the more effective the site attributes the more site
depth, and so forth. Therefore, perhaps perceived ease of use mediates the relationship
between system features and usage depth. For example, the better designed the
attributes the easier the web is to use, and the more useful it is seen, and thus the more
This finding also might relate to the ‘use effectiveness’ variable explored by Segars and
Grover (1993) as an additional factor of TAM that influences usage. This states that usage
increases as the perceived effectiveness of a system increases. Thus, the more effective a
system, the greater the total number of sites and/or search tools accessed.
In sum, this finding opens up the question of the role of user perceptions of a system in
relation to system attributes and drivers of systems use. This has implications for the
189
11.2.6 PREDICTING WEB SESSION USAGE EXTENT: DURATION
(WSUEDUR)
Perceived web usefulness was found to have a positive relationship with duration of
web session use for both user groups. No curvilinear relationship was found between
Further investigation using multiple regression revealed that the easier both user groups
found the web, the longer they used it, irrespective of its perceived usefulness (i.e.,
11.2.6.1 Discussion
Dreze and Zufryden (1997a) report that the duration of visits was explained by web site
attributes. Therefore, attributes of the web site and the web system have a direct impact
on session usage duration, not an indirect effect through perceived ease of use. The
findings in this dissertation are in contrast, with perceived ease of use possibly being a
moderator, not a mediator, between system features and session duration. Further
support for this comes from a study conducted by Hoffman and Novak (1996) who
argue that perceived ease of use of a system promotes seamless navigation and when
consumers are ‘seamlessly’ navigating though a web site, they are in a state of ‘flow’,
Holbrook and Gardner (1993) also argue that duration time is a critical outcome measure
versus goal-directed orientations. Perhaps, then, the easier a system is to use, the longer
it will be used for, and the more experiential or exploratory information seeking
These findings further open up the discussion as to what other factors may influence
how long the web is used, in addition to perceived ease of web use. For example,
intrinsic factors such as task/motivation for which the web is used (i.e., checking email,
190
navigating, shopping, etc.) might play a role. So too might extrinsic factors, such as cost,
system capability (i.e., download speed), and web site/system features (i.e., browser
easy to use and they will use it more often and for longer;
firstly useful, and then easy to use, and they will have more motivations for using it;
firstly easy to use, followed by useful, and they will access and/or use a larger
useful and they will use it more frequently, have more motivations for using it, and
will access an increased number of different and/or new sites and search tools;
firstly easy to use, and then useful, and they will access overall a larger number of
perceptual beliefs about the systems in question. Past research has concentrated on
explaining how the beliefs in the model lead to system use, not what leads to these
perceptions. In contrast, this study investigated user knowledge content of the system as
questions 2 and 3.
Difficulty arises when attempting to discuss and compare the results with earlier work
because that work has been so limited. Nevertheless, a number of studies have used
191
domain-related experience as a proxy measure for actual knowledge content and
subjective measures to infer ‘actual knowledge’, and this provides a basis for
comparison.
they know how to use the web, when they think they know about certain common
features and attributes of the web, when they actually know about common procedures
for using the web, and when they actually know about certain common features and
attributes. In contrast, what they actually know about specialised features and attributes
and what they think they know overall, will decrease how useful they regard the web.
For users with WSD/M experience, perceived usefulness increases when they think they
know how to use the web and when they think they know about the web overall.
Consistent with users with no WSD/M experience, what they actually know about
specialised features and attributes will decrease how useful they regard the web.
11.3.1.1 Discussion
From the above it is evident that to make users with WSD/M experience see the web as
more useful, they need to think they know a lot overall about the web and more
specifically how to use it. An interesting thing to note here is that these users rely more
on what they think they know about the web as an indicator of how useful it is, than
what they actually know – despite having more technical experience. This result is
consistent with results found by Handzic and Low (1999) if ‘experience’ is used as a
proxy for knowledge. They found that more experienced users of processing programs
had more favourable perceptions of the usefulness of the technology. Perhaps, as users
become more experienced with using processing programs, they become more aware of
certain program features and also more efficient in the use of its attributes. This notion is
supported by Reed and Ouchton (1997) who found that hypermedia knowledge in
general had a large impact on user productivity and Thompson et al. (1994) who found
192
For users with no WSD/M experience, to see the web as more useful they need to not
only think they know a lot about how to use the web, and what common features and
attributes are, but also have a good actual common knowledge of how to use the web
and what certain web features are. An additional factor here may be the effect of overall
educational level has a significant negative effect on computer anxiety and a significant
supported by Brancheau and Wetherbe (1990) who found that early adopters of
spreadsheet software were likely to be more highly educated than late adopters. In some
cases, for users with no WSD/M experience, knowledge acts as a means by which to
increase confidence with the system (i.e., reducing anxiety). It may also increase the
perceived and actual knowledge of the electronic system. This is true for both user
groups, although there are some differences in the type of perceived and actual
knowledge and the degree of effect. Diaz, et al (1997) further identified that experience
with the web, as a proxy for knowledge, was an important moderator of attitude toward
the medium and that experienced users found the web more legible and more
stimulating.
to use the web, what they think they know about web features and attributes, and their
common knowledge about how to use the web, all have a positive influence on how easy
they think the web is to use. In contrast, what these groups actually know about
specialised attributes and features has a negative effect on how easy they think the web
is to use.
193
11.3.2.1 Discussion
It was theorised in this study that a stronger relationship would exist between perceived
knowledge content and a user’s perception of ease of use than with actual knowledge
content of the web. This proposition was motivated by the finding that experience is
more accessible in memory than stored knowledge (Park et al. 1994). Perceptions – to the
extent that they are based on experience – might also be more accessible in memory. This
Handzic and Low (1999) also reported that ease of use is related to information
technology experience; that is, the more experience you have, the easier you will find a
system to use. In this dissertation it was found that for both user groups, what they
thought they knew about how to use a system had the strongest positive influence on
how easy they thought the system was to use. This relationship was somewhat stronger
for those with no WSD/M experience, than those with this experience. In these
build confidence. This, in turn, is likely to result in them regarding the system as easier to
use.
Venkatesh and Davis (1996), in an expansion of TAM that focused on the antecedent
variables of perceived ease of use, theorized that direct experience with software
moderates the relationship between objective usability and perceived ease of use.
comparing what it would take for an expert to complete a task using the system to what
it would take for a novice to complete the same task using the same system. Venkatesh
and Davis (1996) predicted that objective usability would be a predictor of perceived
ease of use only after an individual had direct experience with the software. They found
support for their predictions and the results reported here appear to be consistent with
these conclusions.
194
Make them think they know a lot about how to use the web, and also about what
certain features and attributes are, and they will think the web is more useful and
easier to use.
Make sure they actually have a good common knowledge of how to use the web, and
what certain features and attributes are, and they will see the web as more useful.
Make sure they actually have a good common knowledge of how to use the web and
Make them think they know a lot about how to use the web, and that overall they
know a lot about the web, and they will think the web is more useful.
Make them think they know a lot about how to use the web, and what certain
features and attributes are, and they will think the web is easy to use.
Make sure they actually have a good common knowledge about what certain
features and attributes are and they will think the web is easy to use.
user’s confidence with technology and how easy and useful they find the technology.
Although actual common knowledge of the system has a positive effect on user
perceptions, what a user thinks they know about how to use a system (i.e., perceived
procedural knowledge) is evidently the strongest predictor of both how easy and how
useful they think the system is. This is the case for users with and without WSD/M
experience, although there are some differential effects across these two groups.
Discussed in the next chapter, Chapter 12, are the academic and managerial
contributions of this dissertation, the areas for future research, and also a number of
limitations.
195
C HAPTER 12: I MPLICATIONS , C ONTRIBUTIONS ,
- Bertrand Russell -
(1872 - 1970)
12.1 INTRODUCTION
This study drew on studies from consumer research and information technology to
investigate user knowledge and user perceptions of the web. A framework was
developed to depict the effect user knowledge and perceptions of this highly complex
and technologically driven system might have on system usage. A number of hypotheses
were proposed, tested and analysed. In this final chapter, Chapter 12, we discuss the
academic and managerial contributions of this study, the areas for future research, and
products can be an important source of new product ideas and can lead to ideas for new
product uses or product design and development. Furthermore, new markets for
product promotion. Considering the economic importance of new products and their
high rate of failure, it becomes crucial to identify factors fostering and inhibiting
consumer adoption and use. Understanding how products and electronic technologies
196
are used, and what determines their usage, is thus an important part of researching and
Usage has important implications for the communication of product information to the
consumer. Ram and Jung (1990), for example, showed that only a small number of
respondents reported the use of certain features of durable goods, with some
respondents not even aware of these features. This result is extremely apparent in the
could also be used as the basis for segmenting product markets. For example Potter et al.
(1988) attempted to identify the profiles of five usage segments for VCRs. Studies of
computer usage in the workplace have had a wide range of uses too. They have been
implementation, to establish time costs associated with certain work tasks, and to
has provided ample examples of how usage estimates facilitate the evaluation of system
success. For example, user receptivity towards computers (Sarris, Sawyer and Quigley
1993; Saltz, Saltz and Rabkin 1985) and the effect of computer implementation and use
predictors of use. This study, in particular, contributes to our understanding in three core
measurement of system usage; 2) the role of usage context on system use; and 3)
197
12.2.1.1 Measurement of System Usage
In this study, the development of more refined, tested and validated self-report measures
three specific areas – the measurement of frequency, variety and extent of system use.
Furthermore, the study differentiates measures of ‘current web session usage’ from
measures of ‘past web session usage’ – something that is not done in much of the
Empirical support was presented in this dissertation for the argument put forward but
not tested by Moore (1991) and Adams et al (1992) that the usage context influences the
effect of user perceptions of a system on usage. It was found that for users with no
WSD/M experience their usage was primarily influenced by how easy to use they
thought the web was, and then secondly by how useful. Whereas for users with WSD/M
experience, their usage was primarily influenced by how useful they regarded it. This
result has important implications not only for the Technology Acceptance Model (TAM)
This study further extends work conducted on TAM toward predicting system usage.
Firstly, most of the earlier TAM studies only test the effect that perceived ease of use and
measures of usage was included to obtain a better overall picture of the impact of user
perceptions on system usage. Thus, the measures included not only usage frequency, but
situational and motivational variety, and breadth, depth and duration of system use.
These are important measures of usage and thus should be included in TAM to obtain a
Secondly, this dissertation identified that PEWU and PWU had differential effects on
usage frequency, depending on which user group a person belongs to. In essence, if you
198
had experience with designing and maintaining web sites, than how useful you found
the web was the primary predictor of how frequently you used it. In comparison, for
users with no experience designing and maintaining web sites, then how easy you
thought the web was to use became the main predictor of how frequently you used it.
This result helps to explain conflicting results in the literature, relative to TAM, of the
Thirdly, usefulness was found to be the primary predictor of the number of motivations
for web session use. Thus, the more useful the system is, the increased number of
motivations users will have for using it. This result is interesting as both commercially
and academically there is a strong emphasis on first improving how easy a system is to
use by looking at how it is designed before addressing the functions for which it is used
and thus why a user might use it. In this study, how easy the system was to use had only
a secondary influence on the number of motives for use. Whereas perceived usefulness
Fourthly, how easy the web is to use has a positive effect on the total number of web sites
and/or search tools a user will use. Thus, the easier the system is to use, the increased
number of functions and/or features will be used. This might be related to another core
finding – that the easier the system is to use, the longer it will be used – usage duration.
Thus, in this case, because the system is perceived as easy to use, more features and
understand the core difference between the effect of how useful they perceive the system
to be and how easy they perceive it to be. This has a large impact on usage frequency of
the system, the number of locations from which it is accessed, the number of motives
they have for using it, the different functions/attributes of the system they use, the total
number of functions they use and how long they will use it.
199
12.2.2 USER PERCEPTIONS
This study contributes two core elements to system perceptions, namely 1) the
The study further refines and validates measures of perceived ease of web use (PEWU)
and perceived web usefulness (PWU). Most measures of PEU and PU are global
measures of user perceptions (i.e., overall evaluation) and they do not take into
consideration the functions for which the system might be used. This has mainly been as
a result of their testing on systems dictated by only one or two core functions -
Therefore, an extra effort was made in this study to develop measures that actually
measure perceived ease of use and usefulness for certain system functions (e.g.,
evaluation. This increases not only the usefulness of the scales developed, but also the
This study further extends work conducted on TAM toward predicting system
relationships among perceptual beliefs, attitudes and the adoption and acceptance of
system technologies. This aspect of the model has received great attention. However,
very few studies explore the determinants of a user’s perceptual beliefs about the
systems in question. Although highly valuable, in the development and further testing of
TAM, past research has concentrated on explaining how the beliefs in the model lead to
system use. By contrast, little has been done to explore how and why these beliefs were
formed. As stated by Karahanna and Straub (1999), what explains how a user comes to
200
believe that a system is useful in his or her job? What would be the antecedents for the
The original model posited by Davis (1986), and later modified by a number of
researchers, put considerable emphasis on the characteristics of the system (i.e., design,
attributes etc) as antecedents of perceived ease of use and perceived usefulness of the
system in question. But authors such as Krech et al. (1962) suggest this is only half the
story. They categorise the human perception process as influenced by two distinct
factors: stimulus factors (e.g., browser) and personal factors (e.g., experience), further
specifying that perception is a result of both. Due to the heavy focus in the literature on
Within the knowledge literature, use and misuse of knowledge terminology, has led to a
declarative and subjective knowledge have all been defined as types of knowledge. Thus,
it was necessary to draw from the cognitive sciences and marketing to establish a
certain formats (knowledge structures), differs in its type of content (procedural and
201
declarative) and it scope (novice and expert) and may be measured in different ways
Heavy emphasis was also placed on both the objective and subjective measurement of
consumer knowledge content. From the existing literature a number of issues arise such
as the use of proxies to measure knowledge content and the treatment of knowledge
measurement as types of knowledge content. One of the most common methods for
measuring consumer knowledge, especially in the technology area, has been the use of
proxies to infer consumer knowledge. For example, domain usage and purchase
experience have been heavily used. This study went beyond the use of proxies.
In addition, objective and subjective methods for measuring consumer knowledge have
been well documented (Brucks 1985; Dacin and Mitchell 1984; Rao and Olson 1990).
Nevertheless, heavy use of subjective measures in the technology sector has necessitated
Brucks and Mitchell (1981) and Engel, Blackwell and Miniard (1990), concluding that
objective measures of knowledge were better than experience and self-report measures.
Thus, in this study, objective and subjective measures of user knowledge content of the
web were developed, validated and tested across two very differing user groups – user
limitations have been identified and are outlined in the following sub-sections.
202
12.3.1 GENERALISABILITY LIMITATIONS
Even though the internet and web are global electronic technologies, this study was
more homogeneous sample with respect to basic cultural differences in the assessment of
user perceptions, knowledge and use of the web. Care must be taken when generalising
the results of this study across geographical boundaries to web users in other countries,
using the technologies, the results are likely to be somewhat time dependent. For
example, the browser software and web sites upon which the analyses were based will
change and thus the measures developed and tested in this study will need to be
continually updated. The measurement scales developed for this study will require
final study (Chapter 9). This gives confidence in the procedure and the results. But, a few
specific limitations were identified with respect to how some of the measures were
Further refinement of the items developed to measure actual knowledge content of the
web would be beneficial. This is because of the possibility of response error due to
203
In addition, the items used to measure actual common declarative, actual specialised
declarative, and actual specialised procedural knowledge, only explain 57%, 59% and
50% of the variance respectively. Further scale development and refinement might
increase the percentage of the construct explained by including ‘other’ features, terms
Furthermore, actual procedural web knowledge content (common and specialised) was
measured here using objective tests (i.e., true/false, etc.) which are in fact more suitable
be carried out. This might further increase the differentiation between procedural and
As evident during the scale development stage, a limitation exists with the difficulty of
measuring the perceived scope of knowledge content that a user has. Further scale
development and testing is required to ascertain subjective measures of the scope of web
knowledge content. A further review of existing scales revels that in some cases
respondents are asked to rate their perceived knowledge content as compared to ‘an
expert’ or ‘a novice’. This might be one means by which to measure perceived common
It was felt that two items would be better to measure current web usage variety –
situational. This is because situations for web use can differ in variety by the type and
number of locations from which the web is access. This was a post hoc adjustment, made
204
12.3.4 SAMPLE RECRUITMENT & INSTRUMENT LIMITATIONS
A contribution is made in terms of creating, developing and maintaining the web-based
survey. This resulted in over 2,000 responses, a figure that compares extremely
favourably with most off-line surveys (see Appendix F for details). Nevertheless, some
The banner advertising campaign was not run for the entire duration of the study due to
both cost considerations and the availability of banner ad inventory. Thus, although the
level of ad exposure was highly effective for the study at hand (with n = 2077), this
approach might have had an impact on the type and number of respondents that were
The survey method allowed participants to select themselves for inclusion after initial
awareness of the study from both online and offline advertising and publicity. Thus,
there is likely to be some self-selection bias associated with this. Although a test for
response bias was performed between early, mid and late respondents, it is still entirely
possible that those web users with more experience and increased knowledge content of
the web formed the majority of those responding – this has been documented in other
web-based studies. To try to minimise this problem efforts were made to increase
awareness of the study among those with less experience and/or knowledge content of
the web; for example, through the choice of media vehicles that novices might use.
In addition, usage and demographic statistics were self-reported, rather than observed,
and thus care should be taken in interpreting and generalising from these results.
205
12.3.4.3 Multi-Response
Every attempt was made to track users by collecting personal details. This was to ensure
respondents only answered the survey once. However, this process was ad-hoc and thus
it cannot be said for sure that there were not any case of multiple-response. The use of
voluntary ‘opt-in’ mail lists, a web-based panel, or the use of advanced technical features
to lock-out multiple responses from the one client location (i.e., to prevent multiple
responses from one computer) could have been used to reduce the occurrence of
The design and length of the survey might have influenced the quality of the information
that some respondents provided. While pilot study results showed that people were
willing to complete the survey, it undeniably was seen as lengthy. This could have had
an impact on who chose to complete the whole survey. Also, it is likely that time
pressure and fatigue would have had an impact on the accuracy of the responses from
some of the respondents, especially as many uses pay for their internet access and thus
completing the survey would mean not only a time cost, but a financial cost too.
both qualitative pre-analysis and quantitative testing (see Appendix B). Again, however,
reduce the measures with more than 3 categories to 3-categories (i.e., low, medium and
high); this facilitates chi-square analyses (see Appendix Q), but it also results in the loss
of information. Such data reduction may impact the results of the bivariate analyses. An
206
alternative approach is to transform the data. However, the number of possible
transformations is large and this then runs the danger of introducing new difficulties
each contruct might prove for more detailed results. As noted in this dissertation, only
the relationships between the overriding constructs were examined and not the
informative to examine the greater number of dimensions that some of the exploratory
10 were violated (see Appendix O). One particular concern is the possibility of
multiple regression analysis was deemed to be adequate and appropriate. It is noted that
many studies in this field of research are likely to suffer similar limitations, although
surprisingly few of the published studies report whether the assumptions were violated
or not. In general, the results of these analyses should be looked at with an element of
al., 1995). Robust regression techniques are needed to deal with this. However, with a
There are alternatives ways to carry out the regression analysis. The dependent variables
could have been converted into dichotomous variables, or a small number of categorical
variables, and then logistic regressions might have been run. There is also scope for
structural equation modelling, to identify the set of causal relationships between the
207
12.4 EXTENSIONS AND FUTURE RESEARCH
At least four areas for further research are noted: replication of this study with a similar
sample and across geographic samples, a comparison of both users and non-users of the
web, extensions to the model, and the testing of the framework across differing electronic
exact replication of the study on an Australian web sample under similar conditions and
with similar methods of recruitment. Secondly, differentiated replication over space and
time. For example, the study could be replicated using web samples from the United
Kingdom, Europe and, most importantly given the longevity of use and the origins of
the web, the United States of America. This form of replication would aid both the
validation of framework proposed and the measures used, and would also assist in
increasing the generalisability of the results. Given changes in the development and use
Another way to confirm the results is to use multi-mode sampling (i.e., online and
an increased number of novice users and secure responses from some non-users, thereby
increasing the variance in the sample. This would further aid the comparison of different
types of user, and allow for some comparisons to be drawn with non-users (‘usage
This is important because by no means all demographic groups have participated in the
information revolution – those who are poorer, less educated, from rural areas, and
208
females consistently have been slower to use both computers and the internet (Bikson
and Panis, 1997; Tapscott, 1998). Although the gender differential is changing, the gap
between those with and without computer and/or internet access has serious
consequences. In some cases this usage differential is not just based on access to
stated by Tapscott (1998) ‘the issue is not just access to … new (technology), but rather
opportunities to learn may lead to a two tiered world of knowers and know-nots, doers
Martin (2001) as having important educational and economic benefits. The stratification
between those with and those without internet access is creating a digital divide. The
rich are going to be getting even richer in terms of information. The information poor
and corporations shift resources from their ordinary channels of communication onto the
differences between users and non-users of the web. The results could help those who
and education programs. The results also highlight to both commercial and government
age, by an urban-rural split, etc.), and different user samples (e.g., by profession, by
online shopping experience, by level of involvement in the web, etc.). Given the capacity
209
of on-line surveys to result in large sample sizes there would seem to be greater scope
here for dividing the data into sub-groups and drawing comparisons between these.
been investigated here. As a consequence, many potentially important factors have been
set to one side – for example, usage intentions and situational factors. The importance of
these has been emphasised in other consumer behaviour studies (Foxall 1980). Clearly,
the model needs to be extended to make proper allowance for these. Extensions of this
nature might also entail the use of slightly different methods, such as the creation of an
on-line user panel to monitor people’s attitudes, intentions and usage over time.
Other dependent variables can be envisaged. Usage, for example, is measurable in terms
of hits, page views, visits, visitors, etc. A certain amount of this information was collected
during the course of the on-line survey (see Appendix H), and this holds out the promise
of further analysis.
to existing, developing and new electronic tools, technologies and media. The digital
revolution has the potential to bring about many changes (Barwise and Hammond 1998).
Within the context of HCME-based electronic systems, like the web, the framework
developed here could be applied to a range of electronic technologies (e.g., the web itself,
technologies differ from non-HCME based electronic technologies (i.e., TV, radio, etc.) in
terms of vividness, interactivity, media pacing (i.e., external/internal) and the flow of
information and communication transfer. The way consumers see and use all these
210
technologies is of significance for media owners, systems operators, government users,
211
R EFERENCE L IST
Aaker, D. A., Kumar, V., & Day, G. S. (1995). Marketing Research. Toronto: John Wiley & Sons.
ABS. (1998). Use of the Internet by Households . Canberra: Australian Bureau of Statistics (Govt).
ABS. (2000). Use of the Internet by Households (8147.0). Canberra: Australian Bureau of Statistics
(Govt).
Adams, D. A., Nelson, R. R., & Todd, P. A. (1992). Perceived Usefulness, Ease of Use, and Usage of
Technology: A Replication. MIS Quarterly (June), 227-247.
Agarwal, R., & Prasad, J. (1998). The Antecedents and Consequents of User Perceptions in
Information Technology Adoption. Decision Support Systems, 22, 15-29.
Ainscough, T. L., & Luckett, M. G. (1996). The Internet for the Rest of Us: Marketing on the World
Wide Web. Journal of Consumer Marketing, 13(2), 36-47.
Alba, J. W., & Hutchinson, J. W. (1987a). Dimensions of Consumer Expertise. Journal of Consumer
Research, 13, 411-454.
Alba, J. W., & Marmorstein, H. (1987a). The Effects of Frequency Knowledge on Consumer
Decision Making. Journal f Consumer Decision Making, 14(June), 14-25.
Anderson, J. R. (1976). Language, Memory and Thought. New York: Lawrence Erlbaum.
Anderson, J. R. (1983). The Architecture of Cognition. Massachusetts: Harvard University Press.
Anderson, J. R. (1990). Cognitive Psychology and Itʹs Implications. (3rd ed.). New York: Freeman.
Ariely, D. (2000). Controlling the Information Flow: Effects on Consumersʹ Decision Making and
Preferences. Journal of Consumer Research, 27(September), 233-248.
Assael, H. (1995). Consumer Behaviour and Marketing Action. (5th ed.). Ohio: South-Western College
Publishing.
Baddeley, A. D. (1991). Human Memory: Theory and Practice. London: Lawrence Erlbaum.
Bagozzi, R. P., Davis, F. D., & Warshaw, P. R. (1992). Development and Test of a Theory of
Technological Learning and Usage. Human Relations, 45(7 (July)), 659-686.
Bajaj, A., & Nidumolu, S. R. (1998). A Feedback Model to Understand Information System Usage.
Information and Management, 33, 213-224.
Barry, T. E. (1986). Marketing: An Integrated Approach. New York: Dryden.
Barwise, P., & Hammond, K. (1998). Media. London: Phoenix.
Barwise, P. (2001) TV, PC, or mobile? Future media for consumer e-commerce. Business Strategy
Review, Spring, Vol. 12:1, 35-42
Bawa, K., & Shoemaker, R. (1987). The Effect of Direct Mail Coupon on Brand Choice Behaviour.
Journal of Marketing Research, 24(November), 370-376.
Beatty, S. E., & Smith, S. M. (1987). External Search Effort: An Investigation Across Several
Product Categories. Journal of Consumer Research,, 14(June), 83-95.
Belch, G. E., & Belch, M. A. (1995). Introduction to Advertising and Promotion: An Integrated
Marketing Communications Perspective. U.S.A: Irwin.
Belk, R. W. (1979). A Free Response Approach to Developing Product Specific Consumption
Situation taxonomies. In A. D. Shocker (Ed.), Analytical Approaches to Product and Marketing
Planning (pp. 177-196). Cambridge, MA: Marketing Science Institute.
Belk, R. W., Bahn, K. D., & Mayer, R. N. (1982). Developmental Recognition of Consumption
Symbolism. Journal of Consumer Research, 9(June), 4-17.
Benbasat, I., & Dexter, A. S. (1986). An Investigation of the Effectiveness of Colour and Graphical
Presentation under Varying Time Constraints. MIS Quarterly(March), 59-84.
212
Berthon, P., Pitt, L., & Watson, R. T. (1996a). Re-surfing W3: Research Perspectives on Marketing
Communication and Buyer Behaviour on the World Wide Web. International Journal of
Advertising, 15, 287-301.
Berthon, P., Pitt, L. F., & Watson, R. T. (1996b). The World Wide Web as an Advertising Medium:
Toward an Understanding of Conversion Efficiency. Journal of Advertising Research, 36(1),
43-54.
Best, J. B. (1989). Cognitive Psychology. (2nd ed.). New York: West Publishing.
Bettman, J. R. (1979). An Information Processing Theory of Consumer Choice. Massachusetts: Addison-
Wesley.
Bettman, J. R., & Park, C. W. (1980). Effects of Prior Knowledge and Experience and Phase of the
Choice Process on Consumer Decision Processes: A Protocol Analysis. Journal of Consumer
Research, 7(December), 234-248.
Bieber, M., Vitali, F., Ashman, H., Balasubramanian, V., & Oinas-Kukkonen, H. (1997). Fourth
Generation Hypermedia: Some Missing Links for the World Wide Web. International
Journal of Human-Computer Studies, 47, 31-65.
Bikson, T., & Panis, C. (1997). Computers and Connectivity: Current Trends. In S. Kiesler (Ed.),
Culture of the Internet (pp. 407-430). Mahwaw, NJ: Lawrence:Erlbaum.
Blattberg, R. C., & Deighton, J. (1991). Interactive Marketing: Exploring the Age of Addressability.
Sloan Management Review, 33(1), 5-14.
Boote, J., & Mathews, A. (1999). ʺSaying is one thing; doing is anotherʺ:the role of observation in
marketing research. Qualitative Market Research: An International Journal, 2(1), 15-21.
Booth, P. A. (1989). An Introduction to Human-Computer Interaction. Hillsdale, N. J.: Lawrence
Erlbaum Associates.
Booth-Kewly, S., Edwards, J. E., & Rosenfeld, P. (1992). Impression management, social
desirability, and computer administration of attitude questionnaires: Does the computer
make a difference? Journal of Applied Psychology, 77, 562-566.
Bornman, H., & Von Solms, S. H. (1993). Hypermedia, Multimedia and Hypertext: Definitions and
Overview. Electronic Library, 11(4/5), 259-268.
Briggs, R., & Hollis, N. (1997). Advertising on the Web: Is There Response Before Click-through.
Journal of Advertising Research, 37(2), 33-45.
Bronson, M. J. (1999). Modelling Technophobia: A Case for Word Processing. Computers in Human
Behaviour, 15, 105-121.
Brown, P. (1988). Linking and Searching within Hypertext. Electronic Publishing, 1(1), 45-53.
Brucks, M. (1985). The Effects of Product Class Knowledge on Information Search Behaviour.
Journal of Consumer Research, 12(June), 1-16.
Brucks, M. (1986). A Typology of Consumer Knowledge Content. Paper presented at the Advances in
Consumer Research, Provo, UT.
Brucks, M., & Mitchell, A. (1981). Knowledge Structures, Production Systems and Decision
Strategies. Advances in Consumer Research, 8, 750-7.
Chatterjee, P., Hoffman, D. L., & Novak, T. P. (1998). Modelling the Clickstream: Implications for Web-
based Advertising Efforts, [Web Site]. Project 2000 - Manuscripts. Available:
http://www2000.ogsm.vanderbilt.edu/papers/clickstream/clickstream.html [1998, 3rd
December].
Chau, P. Y. K. (1996). An Empirical Assessment of a Modified Technology Acceptance Model.
Journal of Management Information Systems, 13(2), 185-204.
Cheung, W., Chang, M. K., & Lai, V. S. (2000). Prediction of Internet and World Wide Web Usage
at Work: A Test of an Extended Triandis Model. Decision Support Systems, 30(December),
83-100.
213
Chi, M. T. H., Glaser, R., & Rees, E. (1982). Expertise in Problem Solving. In R. Sternberg (Ed.),
Advances in Psychology and Human Intelligence (Vol. 1, pp. Chapt 1, 7-73). Hillsdale:
Lawrence Erlbaum Associates.
Choo, F. (1989). Expert-Novice Differences in Judgement/Decision Making Research. Journal of
Accounting Literature, 8, 106-136.
Church, G. M. (1999). The Human-Computer Interface and Information Literacy: Some Basics and
Beyond. Information Technology and Libraries(March), 3-21.
Churchill, G. (1979). A Paradigm for Developing Better Measures of Marketing Constructs. Journal
of Marketing Research, 16(Feb), 64-73.
Coffey, S., & Stipp, H. (1997). The Interactions between Computer and Television Usage. Journal of
Advertising Research, March/April, 61-67.
Cole, C. A., Gaeth, G., & Singh, S. N. (1986). Measuring Prior Knowledge. Paper presented at the
Advances in Consumer Research.
Colley, A., Henry, O., Holmes, S., & James, L. (1996). Perceptions of Ability to Program or to Use a
Word Processor. Computers in Human Behaviour, 12, 329-337.
Comley, P. (2000). Pop-up surveys: what works, what doesnʹt work and what will work in the future.
Paper presented at the ESOMAR Net Effects Internet Conference, Dublin.
Communications, J., & Interactive, N. (1999b). Attitudes, Behaviours and Demographics of the Online
User (Industry Vol. 2.). NY: Jupiter Communications & NFO Interactive.
Conklin, J. (1987). Hypertext: An Introduction and Survey. IEEE Computer, 20(9), 17-41.
Cravens, D. W., & Woodruff, R. B. (1986). Marketing. Massachusetts: Addison-Wesley.
DʹAngelo, J., & Little, S. K. (1998). Successful Web Pages: What are They and Do They Exist?
Information Technology and Libraries (June), 71-81.
Dacin, P. A., & Mitchell, A. A. (1984). The Measurement of Declarative Knowledge. Advances in
Consumer Research, 11, 454-459.
Davis, F. D. (1986). A Technology Acceptance Model For Empirically Testing New End-User
Information Systems: Theory and Results. Unpublished Ph.D., Massachusetts Institute of
Technology (MIT), Massachusetts.
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989a). User Acceptance of Computer Technology:
A Comparison of Two Theoretical Models. Management Science, 35(8), 982-1003.
Davis, F. D. (1989b). Perceived Usefulness, Perceived Ease of Use, and User Acceptance of
Information Technology. MIS Quarterly(September), 319-340.
Davis, S., & Bostrum, R. P. (1993). Training End Users: An Experimental Investigation of the Roles
of Computer Interface and Training Methods. MIS Quarterly, 17(1), 61-85.
de Bont, C. J. P. M., & Schoormans, J. P. L. (1995). The Effects of Product Expertise on Consumer
Evaluations of New-Product Concepts. Journal of Economical Psychology, 16, 599-615.
De Vaus, D. A. (2002). Survey in Social Research. Crows Nest: Allen & Unwin.
Delbridge, A., & Bernard, J. R. L. (Eds.). (1998). The Macquarie Concise English Dictionary (3rd ed.).
Macquarie: Macquarie University.
DeNisi, A. S., & Shaw, J. B. (1977). Investigation of the Uses of Self-reports of Abilities. Journal of
Applied Psychology, 62(October), 641-4.
Desai, K. K., & Hoyer, W. D. (2000). Descriptive Characteristics of Memory-based Consideration
Sets: Influence of Usage Occasion Frequency and Usage Location Familiarity. Journal of
Consumer Research, 27(December), 309-323.
Diaz, A. N., Hammond, K., & McWilliam, G. (1997). A Study of Web Use and Attitudes Amongst
Novices, Moderate Users and Heavy Users. Paper presented at the 25th EMAC Conference
Proceedings.
Dishaw, M. T., & Strong, D. M. (1999). Extending the Technology Acceptance Model with Task-
Technology Fit Constructs. Information and Management, 36, 9-21.
214
Dodd, D. H., & White, R. M. J. (1980). Cognition: Mental Structures and Processes. Boston: Allyn and
Bacon.
Donohew, L., Palmgreen, P., & Rayburn, J. D. (1987). Social and Psychological Origins of Media
Use: A Lifestyle Analysis. Journal of Broadcasting and Electronic Media, 31(3 (Summer)), 255-
278.
Dreze, X., & Zufryden, F. (1997a). Testing Web Site Design and Promotional Content. Journal of
Advertising Research, 37(2), 77-91.
Dreze, X., & Zufryden, F. (1997d). A Web Based Methodology for Product Design Evaluation and
Optimisation (Working Paper Series ): University of Southern California, Graduate School
of Business.
Ducoffe, R. H. (1996). Advertising Value and Advertising on the Web. Journal of Advertising
Research, 36(5), 21-34.
Durbin, J., & Watson, G. S. (1951). Testing for Serial Correlation in Least Squares Regression II.
Biometrika, 30, 159-178.
Dutton, W., Kovaric, P., & Steinfield, C. (1985). Computing in the Home: A Research Paradigm.
Computers and the Social Sciences, 1, 5-8.
Easterby-Smith, M., Thorpe, R., & Lowe, A. (1991). Management Research: An Introduction,. London:
Sage.
Eighmey, J. (1997). Profiling User Responses to Commercial Web Sites. Journal of Advertising
Research(May/June), 59-66.
Eighmey, J., & McCord, L. (1998). Adding Value in the Information Age: Uses and Gratifications
of Sites on the World Wide Web. Journal of Business Research, 41, 187-194.
eMarketer. (1999). The eGlobal Report (Industry Vol. 1). NY: eMarketer.
Encarnacaeo, J. L., Loseries, F., & Sifaqui, C. (1999). Human Media Technology - The Human-Centered,
Sustainable Technology Development. Paper presented at the Hawaii International
Conference on System Sciences, Hawaii.
Engel, J. F., Blackwell, R. D., & Miniard, P. W. (1990). Consumer Behaviour. (6th ed.). Fort Worth:
Dryden Press.
Engel, J. F., Kollat, D. T., & Blackwell, R. D. (1968). Consumer behaviour: Holt, Rinehart and
Winston.
Enis, C. R. (1995). Expert-Novice Judgements and New Cue Sets: Process versus Outcome. Journal
of Economic Psychology, 16, 641-662.
Esrock, S. L., & Leichty, G. B. (1999). Corporate World Wide Web Pages: Serving the News Media
and Other Publics. Journal of Mass Communication Quarterly, 76(3), 456-467.
Feick, L., Park, C. W., & Mothersbaugh, D. L. (1992a). Knowledge and Knowledge of Knowledge:
What We Know, What We Think We Know and Why the Difference Makes a Difference.
Advances in Consumer Research, 19, 190-192.
Fenech, T. (1997, 1-3 Dec). Using Perceived Ease of Use and Perceived Usefulness to Predict
Acceptance of the World Wide Web. Paper presented at the ANZMAC 1997, Melbourne.
Field, A. (2000). Discovering Statistics Using SPSS for Windows. London: Sage Publications.
Fischoff, B., Slovic, P., & Lichtenstein, S. (1977). Knowing with Certainty: The Appropriateness of
Extreme Confidence. Journal of Experimental Psychology: Human Perception and Performance,
3(November), 552-64.
Fishbein, M., & Ajzen, I. (1975). Belief, Attitude, Intention and Behaviour: An Introduction to
Theory and Research. MA: Addison-Wesley.
Fitt, P. M. (1964). Perceptual Motor Skill Learning. In Melton (Ed.), Categories of Human Learning
(pp. 243-85). New York: Academic Press.
Fletcher, K. (1988). An Investigation into the Nature of Problem Recognition and Deliberation in
Buyer Behaviour. European Journal of Marketing, 22(5), 58-66.
Forrest, E. (1999). Internet Marketing Research. Sydney: McGraw-Hill.
215
Foss, C. L. (1989). Tools for Reading and Browsing Hypertext. Information Processing Management,
25(4), 407-418.
Foxall, G. R. (1980). Marketing Models of Buying Behaviour. European Research, 8, 195-206.
Fred van Raaij, W. (1998). Interactivity and Channel Paper.
Gefen, D., & Keil, M. (1998). The Impact of Developer Responsiveness on Perceptions of
Usefulness and Ease of Use: An Extension of the Technology Acceptance Model. The
Database for Advances in Information Systems Research, 29(2 (Spring)), 35-49.
Gefen, D., & Straub, D. W. (1997). Gender Differences in the Perception and Use of E-mail: An
Extension to the Technology Acceptance Model. MIS Quarterly, 21(4 (December)), 389-400.
Gerbing, D. W., & Anderson, J. C. (1988). An Updated Paradigm for Scale Development
Incorporating Uni-dimensionality and Its Assessment. Journal of Marketing Research,
25(May), 186-192.
Glaser, S. (1997). The real world and virtual world. Internet Research, 7(4), 246-251.
GVU. (1998). GVUʹs WWW 10th User Survey, [Web Site]. Graphic, Visualisation & Usability Centre
(GVU). Available: www.gvu.gatech.edu/uer_surveys/ [1998, October].
Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1995). Multivariate Data Analysis. (4th
ed.). New Jersey: Prentice Hall.
Handzic, M., & Low, G. (1999). The Role of Experience in User Perception of Information
Technology: An Empirical Examination . Sydney: UNSW.
Handzic, M., & Low, G. (2000). The Role of Experience in User Perceptions of Information
Technology: An Empirical Examination. Paper Presented in School Seminar Series: School of
Information Technology, UNSW, February, 1-8.
Hartson, H. R. (1998). Human-Computer Interaction: Interdisciplinary Roots and Trends. The
Journal of Systems and Software, 43, 103-118.
Harvey, M. G., & Rothe, J. T. (1986). Video cassette Recorders: Their Impact on Viewers and
Advertisers. Journal of Advertising Research, 11(March), 19-27.
Hastie, R. (1982). Comment: Consumers Memory For Product Knowledge. Paper presented at the
Advances in Consumer Research.
Hawkins, D., Neal, C., Quester, P., & Best, R. (1997). Consumer Behaviour: Implications for Marketing
Strategy. Sydney: McGraw-Hill Australia.
Heinich, R., Molenda, M., & Russell, J. D. (Eds.). (1989). Instructional media and the new technologies
of instruction (3rd ed.). New York: Macmillan.
Henry, J. W., & Stone, R. W. (1999). The impact of end-user gender, education performance, and
system use on computer self-efficacy and outcome expectancy. Southern Business Review,
Fall, 10-16.
Herzog, H. (1944). What Do We Really Know About Day-Time Serial Listeners. In P. F. Lazarsfeld
& F. N. Stanton (Eds.), Radio Research 1942-1943 (pp. 3-33). New York: Duel, Stone and
Pierce.
Higgins, S., & Shanklin, W. L. (1992). Seeking Mass Market Acceptance for High Technology
Consumer Products. Journal of Consumer Marketing, 9(1), 5-14.
Hoch, S. J., & Deighton, J. (1989). Managing What Consumers Learn From Experience. Journal of
Marketing, 53(April), 1-20.
Hodkinson, C., & Kiel, G. (1997). Electronic Interactivity in the Australian Marketplace: Some
Observations, Issues and Predictions. Asia Pacific Advances in Consumer Research, 2, 116-
127.
Hofacker, C. F. (1999). Internet Marketing. USA: Digital Springs.
Hoffman, D. L. (2000). The Revolution Will Not Be Televised: Introduction to the Special Issue on
Marketing Science and the Internet. Marketing Science, 19(1), 1-3.
Hoffman, D. L., & Novak, T. P. (1996). Marketing in Hypermedia Computer Mediated
Environments: Conceptual Foundations. Journal of Marketing, 60(July), 50-68.
216
Hoffman, T. P., Novak, D. L., & Yung, Y.-F. (1998). Measuring the Flow Construct in On-line
Environments: A Structural Modelling Approach. Project 2000(Date Accessed: 9/08/98).
Holbrook, M. B., & Gardner, M. P. (1993). An Approach to investigating the Emotional
Determinants of Consumption Durations: Why Do People Consume What They
Consumer for as Long as They Consumer It? Journal of Consumer Psychology, 2(2), 123-42.
Holbrook, M. B., & Hirschman, E. C. (1982). The Experiential Aspects of Consumer Behaviour:
Consumer Fantasies, Feelings and Fun. Journal of Consumer Research, 9(September), 132-
140.
Homer, P. M., & Kahle, L. R. (1990). Source Expertise, Time of Source Identification and
Involvement in Persuasion: An Elaborate Processing Perspective. Journal of Advertising,
19(1), 30-39.
Howard, J. A., & Sheth, J. N. (1969). The Theory of Buyer Behaviour. New York: John Wiley.
Huber, G. P. (1983). Cognitive Style as a Basis for MIS and DSS Design: Much Ado about Nothing.
Management Science, 29, 567-582.
Hubona, G. S., & Geitz, S. (1997). External Variables, Beliefs, Attitudes, and Information
Technology Usage Behaviour. IEEE, 25, 21-28.
Huck, S. W., Cormier, W. H., & Bounds, W. G. J. (1974). Reading Statistics and Research. New York:
Harper and Row.
Hulland, J., & Kleinmuntz, P. (1994). Factors Influencing the Use of Internal Summary Evaluations
versus External Information in Choice. Journal of Behavioural Decision Making, 7, 79-102.
Igbaria, M., Guimaraes, T., & Davis, G. B. (1995). Testing the Determinants of Microcomputer
Usage via a Structural Equation Model. Journal of Management Information Systems, 11(4
(Spring)), 87-114.
Jacoby, J. (1978). Consumer Research: A State of the Art Review. Journal of Marketing, 42(April), 87-
96.
Jeffres, L., & Atkin, D. (1996). Predicting use of technologies for communication and consumer
needs. Journal of Broadcasting and Electronic Media, 40, 318-330.
Johnson, E. J., & Russo, J. E. (1984). Product Familiarity and Learning New Information. Journal of
Consumer Research, 11(June), 542-550.
Joinson, A. (1999). Social desirability, anonymity and Internet-based questionnaires. Behaviour
Research Methods, Instruments and Computers, 31(3), 433-438.
Jupiter. (1999a). Online Customer Service Strategies: Strategies For Improving Satisfaction and
Retention (Industry ): Jupiter Communications.
Jupiter, & Interactive, N. (1999b). Attitudes, Behaviours and Demographics of the Online User
(Industry Vol. 2.). NY: Jupiter Communications and NFO Interactive.
Kanwar, R., Olson, J. C., & Sims, L. S. (1981). Toward Conceptualising and Measuring Cognitive
Structures. Paper presented at the Advances in Consumer Research, MI.
Kaplan, A. (1964). Chapter 5: Measurement, The Conduct of Inquiry: Methodology for Behavioural
Science. (pp. 171-214). San Francisco: Chandler Pub. Co.
Karahanna, E., & Straub, D. W. (1999). The Psychological Origins of Perceived Usefulness and
Ease of Use. Information and Management, 35, 237-250.
Kassarjian, H. H., & Robertson, T. S. (Eds.). (1968). Perspectives in Consumer Behaviour. Illinois:
Scott, Foresman and Company.
Kay, R. H. (1993). An Exploration of Theoretical and Practical Foundations for Assessing Attitudes
towards Computers: The Computer Attitude Measure (CAM). Computers in Human
Behaviour, 9, 371-386.
Kaye, B. K., & Johnson, T. J. (1999). Research Methodology: Taming the Cyber Frontier. Social
Science Computer Review, 17(3), 323-337.
Kindra, G. S., Laroche, M., & Muller, T. E. (1994). The Canadian Perspective: Consumer Behaviour.
(2nd ed.). Ontario, Canada: I.T.P - Nelson.
217
King, R. C., & Xia, W. (1997). Media Appropriateness: Effects of Experience on Communication
Media Choice. Decision Sciences, 28(4), 877-910.
Knapp, R. G., Miller, M. C., & Levine, J. (1987). Experience and Attitude towards Computers in
Medicine Students and Clinical Faculty Members at One School. Journal of Medical
Education, 62, 344-346.
Knuth, R. A., & Brush, T. A. (1990). Results of the Hypertext 89 Design Survey. Hypermedia, 2(2),
91-107.
Kolbe, R. H., & Burnett, M. S. (1991). Content Analysis Research: An Examiniation of Applications
with Directives for Improving Research Reliability and Objectivity. Journal of Consumer
Research, 18(September), 243-250.
Korgaonkar, P. K., & Wolin, L. D. (1999). A Multivariate Analysis of Web Usage. Journal of
Advertising Research, March/April, 53-68.
Kotler, P., & Armstrong, G. (1994). Principles of Marketing. (6th ed.). Englewood Cliffs: Prentice
Hall.
Kouchakadjian, A., & Fietkiewicz, R. (2000). Improving a Product with Usage-based Testing.
Information and Software Technology, 42, 809-814.
Kraut, R., Mukhopadhyay, T., Szczypula, J., Kiesler, S., & Scherlis, W. (1998). Communication and
Information: Alternative Uses of the Internet in Households. CHI - Papers, April, 18-23.
Krech, D., Cruchfiled, R. S., B, & Ballachey, E. L. (1962). Individual in Society. (Vol. Part I and Part
II). New York: McGraw-Hill.
Kuehn, S. A. (1994). Computer Mediated Communication in Instructional Settings: A Research
Agenda. Communication Education, 43(April), 171-183.
Laitenberger, O., & Dreyer, H. M. (1998). Evaluating the Usefulness and the Ease of Use of a Web-
based Inspection Data Collection Tool. IEEE, 122-132.
Lawrence, E., Corbitt, B., Fisher, J., Lawrence, J., & Tidwell, A. (2000). Internet Commerce. (2nd ed.).
Brisbane: John Wiley & Sons.
Lederer, A. L., Maupin, D. J., Sena, M. P., & Zhaung, Y. (2000). The Technology Acceptance Model
and the World Wide Web. Decision Support Systems, 29, 269-282.
Levy, M. R. (1980). Home Video Recorders: A User Study. Journal of Broadcasting, 24(3), 327-336.
Levy, M. R. (1981). Home Video Recorders and Time Shifting. Journalism Quarterly, 58(3), 401-405.
Lichtenstein, S., & Fischoff, B. (1977). Do Those Who Know More Also Know More About How
Much They Know? The Calibration of Probability Judgements. Organisational Behaviour
and Human Performance, 20(December), 159-83.
Li, X. (1998). Web Page Design and Graphic Use of Three US Newspapers. Journal of Mass
Communication Quarterly, 75(2), 353-365.
Lin, C. A. (1992). The Functions of the VCR in the Home Leisure Environment. Journal of Broadcast
and Electronic Media, Summer, 345-351.
Lutz, R. J. (1975). Changing Brand Attitudes Through Modification of Cognitive Structure. Journal
of Consumer Research, 1(March), 49-59.
Maddox, L. M., & Mehta, D. (1997). The Role and Effect of Web Addresses in Advertising. Journal
of Advertising Research, 37(2), 47-59.
Malhotra, N. K., Hall, J., Shaw, M., & Crisp, M. (1996). Marketing Research: An Applied Orientation.
Sydney: Prentice Hall.
Mandelli, A. (1997, 20th-23rd May). The Internet and the New Media: Mass Communication for
Relationship Marketing. Paper presented at the 26th EMAC Conference.
Marchionini, G. (1995). Information Seeking in Electronic Environments. Cambridge ; New York:
Cambridge University Press.
Marks, L. J., & Olson, J. C. (1981). Toward a Cognitive Structure Conceptualisation of Product
Familiarity. Paper presented at the Advances in Consumer Research, MI.
218
McAlister, L., & Pessemier, E. (1982). Variety Seeking Behaviour: An Interdisciplinary Review.
Journal of Consumer Research, 9(December), 311-322.
McGuire, W. J. (1974). Psychological Motives and Communication Gratification. In E. Katz (Ed.),
The Uses of Mass Communication . Beverly Hills, CA: Sage Publications.
McKeithan. (1981). Knowledge Organization and Skilled Differences in Computer Programmers.
Cognitive Psychology, 307-25.
McWilliam, G., Hammond, K., & Diaz, A. (1997). Going Places in Webtown: A New Way of
Thinking About Advertising and the Web. The Journal of Brand Management, 4(4), 261-270.
Mentzer, J. T., Schuster, C. P., & Roberts, D. J. (1987). Microcomputer versus Mainframe Usage by
Marketing Professionals. Journal of Academy of Marketing Science, 15(Summer), 1-9.
Midgley, D. F. (1983). Patterns of Interpersonal Information Seeking for the Purchase of a
Symbolic Product. Journal of Marketing Research, 20(February), 74-83.
Mitchell, A. A. (1981). Models of Memory: Implications for Measuring Knowledge Structures. Paper
presented at the Advances in Consumer Research, MI.
Moon, J., & Kim, Y. (2001). Extending the TAM for the World Wide Web. Information and
Management, 38(217-230).
Moore, G. C., & Benbasat, I. (1991). Development of an Instrument to Measure the Perceptions of
Adopting an Information Technology Innovation. Information Systems Research, 2(3), 192-
222.
Moran, T. (1981). An Applied Psychology of the User. ACM Computing Surveys, 13(1 (March)), 1-
12.
Morris, M. G., & Dillion, A. (1997). How User Perceptions Influence Software Use, Decision
Support Systems. IEE Software, July-August, 58-65.
Napoli, J., & Ewing, M. (1998). The Media Habits and Internet Practices of the Net Generation (Working
Paper Series 9811). Perth: Curtin University of Technology.
Nelson, R. R. (1990). Individual Adjustment to Information-driven Technologies: A Critical
Review. MIS Quarterly, 14(1), 87-98.
Nelson, T. O., Leonesio, R. J., Shimamura, A. P., & Landwehr, R. F. (1982). Over learning and the
Feeling of Knowing. Journal of Experimental Psychology: Learning, Memory and Cognition,
8(July), 279-88.
NetRatings, A. (2000). News Release: Male Surfers Dominate Surfing Activity . Stamford:
ACNeilsen.
Neuman, W. L. (1997). Social Research Methods: Qualitative and Quantitative Approaches. (3rd
ed.). Boston: Allyn and Bacon.
Nicosia, F. M. (1966). Consumer Decision Processes. Englewood Cliffs: Prentice Hall.
Nielsen, J. (1999). Designing Web Usability. Idianapolis: New Riders Publishing.
NOIE. (2000). The Current State of Play (Government ). Canberra: National Office of Information
Economy (NOIE).
NOIE. (2001). The Current State of Play - Australiaʹs Participation in the Information Economy
(Government ). Canberra: National Office of Information Economy (NOIE).
Norman, D. (1990). The Design of Everyday Things. New York: Doubleday.
Novak, T. P., & Hoffman, D. L. (1997). Measuring the Flow Experience Among Web Users.
Working Paper - Vanderbilt University.
Novak, T. P., Hoffman, D. L., & Yung, Y. (1998). Measuring the Flow Construct in Online
Environments: A Structural Modelling Approach. Project 2000 - Manuscripts, 1-26.
Nunes, J. C. (2000). A Cognitive Model of Peopleʹs Usage Estimations. Journal of Marketing
Research, 37(November), 397-409.
Nunnally, J. C. (1978). Psychometric Theory. (2nd ed.). New York: McGraw-Hill.
219
Olney, T. J., Holbrook, M. B., & Batra, R. (1991). Consumer Response to Advertising: The Effects of
Ad Content. Emotions, and Attitude toward the Ad on Viewing Time. Journal of Consumer
Research, 17(March), 440-53.
Oppermann, M. (1995). E-mail surveys - potential and pitfalls. Marketing Research, 7(3), 29-33.
Pallab, P. (1996). Marketing on the Internet. Journal of Consumer Marketing, 13(4), 27-39.
Papacharissi, Z., & Rubin, A. M. (2000). Predictors of Internet Use. Journal of Broadcasting and
Electronic Media, Spring, 175-196.
Park, C. W. (1982). Joint Decisions in Home Purchasing: A Muddling-through Process. Journal of
Consumer Research, 9(September), 151-162.
Park, C. W., Feick, L., & Mothersbaugh, D. L. (1992). Consumer Knowledge Assessment: How
Product Experience and Knowledge of Brands, Attributes, and Features Affects What We
Think We Know. Advances in Consumer Research, 19, 193-198.
Park, C. W., Mothersbaugh, D. L., & Feick, L. (1994). Consumer Knowledge Assessment. Journal of
Consumer Research, 21(1 (June)), 71-82.
Perreault, W. D. J., & Leigh, L. E. (1989). Reliability of Nominal Data Based on Qualitative
Judgements. Journal of Marketing Research, 25(May), 135-148.
Petty, R. E., Cacioppo, J. T., & Schuman, D. (1983). Central and Peripheral Routes to Advertising
Effectiveness: The Moderating Role of Involvement. Journal of Consumer Research,
10(September), 135-146.
Petty, R. E., & Priester, J. R. (1994). Mass Media Attitude Change: Implications of the Elaboration
Likelihood Model of Persuasion. In J. Bryant & D. Zilman (Eds.), Media Effects, Advances in
Theory and Research . Hillsdale, NJ: Lawrence Erlbaum Associates.
Pope, N. K., Tam, T., Forrest, E. J., & Henderson, K. (Aug. 10-13, 1997). Survey research on the web.
Paper presented at the 14th International Conference on Technology and Education, Oslo,
Norway.
Potter, S. J., Forrest, E., Sapolsky, B. S., & Ware, W. (1988). Segmenting VCR Owners. Journal of
Advertising Research, 28(2), 29-39.
Rada, R. (1995). Hypertext, Multimedia and Hypermedia. The New Review of Hypermedia and
Multimedia, 1, 1-19.
Rafaeli, S. (1986). The Electronic Bulletin Board: A Computer Driven Mass Medium, Computers and
the Social Sciences 2 (pp. 123-131). Osprey, FL: Paradigm Press.
Ram, S., & Jung, H. J. (1990). The Conceptualisation and Measurement of Product Usage. Journal of
the Academy of Marketing Science, Winter, 67-76.
Rao, A. R., & Olson, E. M. (1990). Information Examination as a Function of Information Type and
Dimension of Consumer Expertise: Some Exploratory Findings. Paper presented at the
Advances in Consumer Research.
Rao, A. R., & Sieben, W. A. (1992). The Effect of Prior Knowledge on Price Acceptability and the
Type of Information Examined. Journal of Consumer Research, 19(September), 256-270.
Rao, V. R. (1980). Books on Quantitative Methods for Consumer Research. Journal of Consumer
Research, 7(September), 198-210.
Ray, N., Griggs, K., & Tabor, S. (2001). Web Based Survey Research Workshop : WDSI.
Reed, M. W., & Oughton, J. M. (1997). The Effects of Hypermedia Knowledge and Learning Style
on the Construction of Group Concept Maps. Computers in Human Behaviour, 14(1), 1-22.
Rice, R. E. (1984). The New Media: Communication, Research and Technology. Beverly Hills:
Sage.
Robertson, T. S. (1974). A Critical Examination of the Adoption Process Models of Consumer
Behaviour. In J. N. Sheth (Ed.), Models of Buying Behaviour (pp. 271-95): Harper and Row.
Rogers, E. M. (1995). Diffusion of Innovations. (4th ed.). New York: Free Press.
220
Rosen, L. D., & Weil, M. (1995). Adult and Teenage Use of Consumer, Business and Entertainment
Technology: Potholes on the Information Superhighway. Journal of Consumer Affairs, 29(1),
55-84.
Rosenfeld, P., Booth-Kewlwy, S., Edwards, J., & Thomas, M. D. (1996). Responses on computer
surveys: Impression management, social desirability, and the big brother syndrome.
Computers in Human Behaviour, 12(2), 263-274.
Rossiter, J. R., & Kayande, U. (1999). Construct Measurement in Marketing. Unpublished
Manuscript(June), 1-47.
Rubin, A. M. (1993). Audience Activity and Media Use. Communication Monographs, 60, 98-103.
Rubin, A. M. (1994). Media Uses and Effects: A Uses and Gratifications Perspective. In J. Bryant &
D. Zillmann (Eds.), Media Effects: Advances in Theory and Research (pp. 281). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Rumpradit, C., & Donnell, M. L. (1999). Navigation Cues on User Interface Design to Produce Better
Information Seeking on the World Wide Web. Paper presented at the Hawaii International
Conference on System Sciences, Hawaii.
Russo, J. E., & Johnson, E. J. (1980). What Do Consumers Know About Familiar Products. Paper
presented at the Advances in Consumer Research.
Rust, R. T., & Oliver, R. W. (1994). Notes and Comments: The Death of Advertising. Journal of
Advertising, 23(4 (December)), 71-77.
Saltz, C. C., Saltz, J., & Rabkin, M. (1985). Perceptions and Knowledge of Medical Students
Regarding Computer Applications in Medicine. Journal of Medical Education, 60, 726-728.
Sarris, A., Sawyer, M. g., & Quigley, R. (1993). Attitude of Clinicians toward Use of Computers in
Hospital Settings. Professional Psychology: Research and Practice, 24, 220-223.
Schacter, D. L. (1983). Feeling of Knowing in Episodic Memory. Journal of Experimental Psychology:
Learning, Memory and Cognition, 9(January), 39-54.
Schaefer, A. (1997). Consumer Knowledge and Country Of Origin Effects. European Journal of
Marketing, 31(1), 56-72.
Schiffman, L., Bednall, D., Watson, J., & Kanuk. (1997). Consumer Behaviour. Sydney: Prentice Hall.
Schlinger, M. J. (1979). A Profile of Response to Commercials. Journal of Advertising Research, 19(2
(April)), 37-46.
Schumacher, P., & Morahan-Martin, J. (2001). Gender, Internet and Computer Attitudes and
Experiences. Computers in Human Behaviour, 17, 95-110.
Seeley, M., & Targett, D. (1999). Patterns of Senior Executives Personal Use of Computers.
Information and Management, 35, 315-330.
Seeley, M. E., & Targett, D. (1997). A Senior Executive End-User Framework. Information Systems
Journal, 7(4), 289-308.
Segars, A. H., & Grover, V. (1993). Research Note: Re-examining Perceived Ease of Use and
Usefulness: A Confirmatory Factor Analysis. MIS Quarterly(December), 517-525.
Selnes, F., & Gronhaug, K. (1986). Subjective and Objective Measures of Product Knowledge. Paper
presented at the Advances in Consumer Research.
Shanteau, J. (1992). How Much Information Does an Expert Use? Is it Relevant? Acta Psychological,
81, 75-86.
Sinkovics, R., Stottinger, B., Schlegelmilch, B. B., & Woodruffe-Burton, H. (1999). Diffidence about
the Use of Technology-Related Products: Development of a Technophobia Scale. Paper presented
at the 28th EMAC Conference, Berlin.
Sivadas, E., Grewal, R., & Kellaris, J. (1998). The Internet as a Micro Marketing Tool: Targeting
Consumer Through Preferences Revealed in Music Newsgroup Usage. Journal of Business
Research, 41, 179-186.
Smith, C. (1997). Casting the Net: Surveying an Internet population. Journal of Computer Mediated
Communication, 3(1).
221
Smith, P. A., & Wilson, J. R. (1993). Navigation in Hypertext Through Virtual Environments.
Applied Ergonomics, 24(4 (August)), 271-278.
Solomon, M. R. (1983). The Role of Products as Social Stimuli: A Symbolic Interactionism
Perspective. Journal of Consumer Research, 10(December), 319-329.
Solomon, M. R. (1994). Consumer Behaviour: Buying, Having and Being. Needham Heights, MA:
Allyn and Bacon.
Spence, M. T., & Brucks, M. (1997). The Moderating Effects of Problem Characteristics on Expertsʹ
and Novices Judgements. Journal of Marketing Research, 34(May), 233-247.
Spilich. (1979). Text Processing of Domain Related Information for Individual with high or Low
Domain Knowledge. Journal of Verbal Learning and Verbal Behaviour, 275-290.
Spink, A., Bateman, J., & Jansen, B. J. (1999). Searching the Web: A Survey of Excite Users. Internet
Research: Electronic Networking, Applications and Policy, 9(2), 117-128.
Sproull, N. L. (1995). Handbook of Research Methods. (2nd ed.). New Jersey: Scarecrow Press.
Srinivasan, A. (1985). Alternative Measures of System Effectiveness: Associations and
Implications. MIS Quarterly(September), 243-253.
Srivastava, R. K., Shocker, A. D., & Day, G. S. (1978). An Exploratory Study of the Influences of Usage
Situations on Perceptions of Product Markets. Paper presented at the Advances in Consumer
Research, Chicago.
Stangelove, M. (1996). Internet Advertising Review: The Theory and Practice of Internet
Advertising. Internet Business Journal, 2(8), 14-19.
Steuer, J. (1992). Defining Virtual Reality: Dimensions of Determining Telepresence. Journal of
Communication, 42(4), 73-93.
Stevens, P. M., Williams, K. P., & Smith, M. C. (2000). Organisational Communication and
Information Processes in an Internet-Enabled Environment. Psychology and Marketing,
17(7), 607-632.
Sudman, S. (1976). Applied Sampling. New York: Academic Press.
Swanson, D. L. (1987a). Gratification Seeking, Media Exposure and Audience Interpretations:
Some Directions for Research. Journal of Broadcasting and Electronic Media, 31(3 (Summer)),
237-254.
Swanson, E. B. (1987b). Information Channel Disposition and Use. Decision Science, 18, 131-145.
Swanson, E. B. (1988). Information System Implementation: Bridging the Gap Between Design
and Utilization. Homewood IL: Irwin.
Swoboda, B. S. (1998). Conditions of Consumer Information Seeking: Theoretical Foundations and
Empirical Results of Using Interactive Multimedia Systems. The International Review of
Retail, Distribution, and Consumer Research, 8(4), 361-381.
Tapscott, D. (1998). Growing Up Digital: The Rise of the Net Generation. New York: McGraw-Hill.
Taylor, H. (2000). Does Internet Research Work? International Journal of Market Research, 42(1), 51-
63.
Taylor, S., & Todd, P. (1995). Assessing IT Usage: The Role of Prior Experience. MIS Quarterly,
19(4 (December)), 561-570.
Teo, T. S. H., Lim, V. K. G., & Lai, R. Y. C. (1997). Users and Uses of the Internet: The Case of
Singapore. International Journal of Information Management, 17(5), 325-336.
Teo, T. S. H., Lim, V. K. G., & Lai, R. Y. C. (1999). Intrinsic and Extrinsic Motivation in Internet
Usage. Omega, 27, 25-37.
Tull, D. S., & Hawkins, D. I. (1993). Marketing Research: Measurement and Method. (6th ed.). New
York: Macmillian.
Tuten, T. L., Bosnjak, M., & Bandilla, W. (2000). Banner-advertised Web Surveys. Marketing
Research, Spring, 17-21.
Urban, C. (1976). Correlates of Magazine Readership. Journal of Advertising Research, 19(3 (June)), 7-
12.
222
Valacich, J. S., Paranka, D., George, J. F., & Nunamaker, J. F. (1993). Communication Concurrency
and the New Media. Communication Research, 20(2 (April)), 249-276.
van Braak, J. (2001). Factors Affecting the Use of Computer-mediated Communication by
Teachers in Secondary Schools. Computers and Education, 36, 41-57.
Van den Bulck, J. (1999). VCR-Use and Patterns of Time Shifting and Selectivity. Journal of
Broadcasting and Electronic Media, Summer, 316-326.
Venkatesh, V., & Davis, F. D. (1996). A Model of the Antecedents of Perceived Ease of Use:
Development and Test. Decision Sciences, 27(3), 451-481.
Walker, J. R., & Bellamy, R. V. (1991). Gratifications of Grazing: An Exploratory Study of Remote
Control Use. Journalism Quarterly, 68(3 (Fall)), 422-431.
Watt, J. H. (1999). Internet Systems for Evaluation Research. Paper presented at the Information
technologies in evaluation: Social, moral, epistemological and practical implications, San
Francisco.
Weare, C., & Lin, W. (2000). Content Analysis of the World Wide Web: Opportunities and
Challenges. Social Science Computer Review, 18(3), 272-292.
Weaver, D. (1988). Mass Communication Research. In W. S. N. (Ed.), Communication Research .
Syracuse: Syracuse University Press.
Weiers, R. M. (1988). Marketing Research. (2nd ed.). New Jersey: Prentice-Hall.
Wind, Y., & Lerner, D. (1979). On the Measurement of Purchase Data: Surveys versus Purchase
Diaries. Journal of Marketing Research, 16(February), 39-47.
Woodside, A. G., Trappey, R. J., & MacDonald, R. (1997). Measuring Linkage-Advertising Effects
on Customer Behaviour and Net Revenue: Using Quasi-Experiments of Advertising
Treatments with Novice and Experienced Product-Service Users. Canadian Journal of
Administrative Sciences, 14(2), 214-228.
www.consult.com. (1998). Australian Internet Users: Attitudes Towards Online Travel Booking
(Industry ). Sydney.
www.consult.com. (1999). Australian Internet User Report (Industry ). Sydney: www.consult.com.
Yale, L. J., & Gilly, M. C. (1995). Dyadic Perceptions in Personal Source Information Search. Journal
of Business Research, 32, 225-237.
Yun, G. W., & Trumbo, G. W. (2000). Comparative response to a survey executed by post, email
and web form. Journal of Computer-Mediated Communication, 6(1), 1-18.
Zaichkowsky, J. L. (1985b). Familiarity: Product Use, Involvement or Expertise. Paper presented at the
Advances in Consumer Research, Provo, UT.
223
A PPENDICES
In the pages that follow are a series of technical appendices that provide background
support and/or further technical information of the steps undertaken for the conduct of
Appendices F, G, H, and I, present the survey instrument, details of the web site and
advertising and publicity tactics used to create survey awareness, drive traffic and
relationship between each construct (i.e., correlation matrix) and the performance of
bivariate and multivariate analyses used to test the hypotheses set down in this
study.
224
A PPENDICES
- TABLE OF C ONTENTS -
Appendix K: Scale Performance Comparison (Student Samples and Web Sample)....................... 267
225
APPENDIX A: PREVIOUS TAM RESEARCH
Authors Constructs System Context Methodology Context Findings (Indep. > Dep.)
Survey (Organisational) EV(S)>EOU, EV(S)>A, EOU>A, U>A, A>Usage,
Davis (1986) EV(S), U, EOU, A, BI, Usage PROFs™, XEDIT™,
Experiment (Academic) U>Usage
Chartmaster™, Pendraw™
Davis, Bagozzi, &
U, EOU, A, BI, Usage WriteOne™ Experiment (Academic) EOU>U, U>A, EOU>A, A>BI, U>BI, BI>Usage
Warshaw (1989a)
PROFs™, XEDIT™, Survey (Organisational) U>Usage
Davis (1989b) U, EOU, Usage
Chartmaster™, Pendraw™ Experiment (Organisational) EOU>Usage
Bagozzi, Davis, & EOU, U, BI (Two Time Intervals),
WriteOne™ Experiment (Academic) U>BI, EOU>BI, BI>Usage
Warshaw (1992) Usage
E-mail, V-mail, Wordperfect™,
Adams, Nelson, & Todd, EOU>Usage
U, EOU, Usage Lotus Notes™ 123, Harvard Survey (Organisational)
(1992) U>Usage, EOU↔U
Graphics™
Re-examination of Adams et al. (1992) Three Factor Model:
Segars & Grover (1993) EOU, U Email
(Organisational) U, E, EOU
U, EOU, A, Subjective Norm,
EOU>U, U>A, EOU>A, A>BI, SN>BI, PBC>BI, BI>B,
Taylor & Todd (1995) Perceived Behavioural Control, BI, Computing Resource Centre Survey (Academic)
PBC>B
Behaviour
Igbaria, Guimaraes, &
EV, EOU, U, Usage Micro-computer Survey (Organisational/Academic) EV>EOU, EV>U, EOU>U, EOU>Usage, U>Usage
Davis (1995)
EOU>Near-term U, EOU>BI, Near-term U>Long-
Chau (1996) EOU, Near-term U, Long-term U, BI Microsoft™ Word and Excel Survey (Organisational)
term U, Near-term U>BI, Long-term U>BI
Morris & Dillion (1997) EOU, U, A, BI, Usage Netscape™ Survey EOU>U, U>A, EOU>A, U>BI, A>BI, BI>Usage
Gender>SPIR, Gender>U, Gender>EOU, SPIR>U,
Gefen & Straub (1997) Gender, U, EOU, Usage, SPIR E-Mail Survey (Organisational)
U>Usage
Fenech (1997) CS, U, EOU, A, UV, UF Word Wide Web Survey (Academic - For Work) U>Usage, EOU>Usage,
Past Usage>EOU
Bajaj & Nidumolu (1998) EV, U, EOU, A, Usage Debugger™ (DBG) Survey (Organisational) A>Usage
EOU>A
Gefen & Keil (1998) PDR; U, EOU, U CONFIG™ Survey (Organisational) EOU>U, U>Usage, PDR>EOU, PDR>U
Laitenberger & Dreyer Web based Inspection system
U, EOU, SP Usage Experiment (Work) EOU > U, U > SPUsage, EOU > SP Usage
(1998) (WIPS)
226
EOU, U, A, BI, Usage, TE, TTF, TF,
Dishaw & Strong (1999) COBOL™ Survey (Organisational) TF > EOU, TE > EOU, TE > U, TTF > EOU,
TC
Bronson (1999) U, CA, S, EOU, Usage Word Processing Survey (Academic – Job Performance) CA>U, CA>EOU, S>U, EOU>U, PU>Usage,
Teo, Lim, & Lai (1999) U, EOU, Usage, PE Internet Web Survey – EOU>U, EOU>Usage, EOU>PE, U>Usage, PE>Usage
Karahanna & Straub SI>U, EOU>U, SP>U, ACC>EOU
U, EOU, SP, SI, ACC, SUPP, Usage E-mail Survey (Organisational)
(1999) EOU>Usage, U>Usage
Dishaw and Strong (1999) EOU, TE, TTF, U, A, BI, TU, TC, TF MVS COBOL.CICS Survey (Organisational) TF>EOU, TE>EOU, TE>U, TTF>EOU
Handzic (2000) EV(PE), U, EOU MetaEdit™, Microsoft™ Word Survey (Academic) EV(PE)>EOU, EV(PE)>U
EOU>Usage, U>Usage
Lederer, Maupin, Sena, & Web Site (Respondent EOUA-U>-EOU
UA, EOUA, Usage, U, EOU E-mail Survey (Organisational)
Zhaung (2000) Specified) EOUA-F>EOU
UA-IQ>U
EOU>PP, EOU>U, PP>A, EOU>A, U>A, A>BI, U>BI,
Moon & Kim (2001) PP, EOU, U, A, BI, Usage World Wide Web Survey (Organisational)
PP>BI, BI-Usage
a Legend
227
APPENDIX B: PRIMARY EXPLORATORY RESEARCH
B.1 INTRODUCTION
As outlined in the dissertation itself, to develop a pool of items to measure each of the
conducted to ascertain the structure and content of the scale items. In addition, in order
to establish the content validity of the scale items, the item generation process also
Due to the exploratory nature of these studies, and the core focus of this dissertation
being hypothesis testing not scale development, only a brief review of these studies will
be outlined in the following sections of this appendix. The overall results of all studies
et al. 1996). It is used to gain insights into the general nature of the problem, the possible
decision alternatives, and/or the relevant variables that may need to be considered.
Exploratory research may consist of secondary data analysis (e.g., academic literature)
and primary data analysis (e.g., qualitative studies). Aaker et al., (1995) contends that the
research methods adopted are highly flexible, unstructured and typically qualitative.
The preliminary exploratory studies conducted in this dissertation to aid item generation
228
B.3 EXPERT SURVEY
To generate a concise and representative list of items a panel of web and web site design
experts were approached to discuss the concept of web knowledge. The purpose was to
available attributes, and evaluative criteria of attributes and usage situations. From this
discussion and further exploratory research, a list of terms, attributes and usage
situations was derived to help further develop items measuring actual web knowledge.
This approach is adapted from Brucks (1985) and his attempt to measure sewing
machine knowledge.
To assess the appropriateness of the language and terminology used and the readability
and wording of the items generated to measure the construct of web knowledge, a
convenience sample was selected as derived from a ‘media’ publications list. The total
list was screened with respect to occupations and a final list of 122 senior web site
designers and consultants from leading Australian firms were included in the final
sample. The 122 respondents were contacted by email inviting them to participate in the
expert open-ended survey. 25 emails were returned with permanent fatal errors
a total of 97 successful emails were sent out. Of the sample contacted, only 17 survey
entries were received, however only 12 of these entries were usable as 5 of the surveys
The survey contents were adapted from Brucks (1985) measurement of sewing machine
terminology, web attributes and features, knowledge of evaluative criteria used with
respect to the web and web usage situations. See Table B1 for a description.
229
Table B1: Expert Survey Free Response Question Description
Knowledge Context Question Description
A free response question ask subjects to list and define terms (i.e., those commonly found in web
Web Terminology guides and book glossaries) used to describe features of the web and web sites.
A free response question asks subjects to list all possible features of the web, including the items
not perceived as being important, that a user might encounter when navigating a the web. To
reduce measurement error caused by the variance in subjects listing of standard and obvious
Web Attributes attributes, Brucks (1985) introduced another question asking the subjects to list features that are
common to sewing machines. In this study, subjects are asked to tick the most commonly found
features of the web.
Criteria for The free response question directs respondents to write down everything (i.e., attributes and
features) that users deem important with the web and the criteria by which they are important
Evaluating the Web (i.e., why).
& Web Sites
The free response question asks subjects to List the situational characteristics that might influence
Web Usage
web site navigation, to explain how each situation might influence navigation and to rank how
Situations each situational characteristic according to its level of influence on navigation
Table B3, B4, and B5 present the final summation scores in descending order of most
frequent to least frequent frequency of term, feature and procedure use in the experts
sampled. These results present the frequency of occurrence in the surveys returned.
novice web users learning what the web is and how to use it. The study was intended to
be exploratory (Bootee and Mathews, 1999). This approach to data collection was
undertaken to acquire first-hand knowledge and observation of novice web users, their
queries and the type of information they might first learn about with respect to what the
This study is based on observation of two 2-hour short courses about navigating the web:
ʹSurf the net - Level 1ʹ and ʹSurf the net - Level 2ʹ. These courses were offered by a
community college and aimed to provide participants with base level information abut
‘how to navigate the World Wide Web and find information’. In both classes it could be
said that participants were members of the general community with no/limited
knowledge of the web and how to use it. Observation occurred in a natural setting and
was disguised to both the instructor and the class participants to minimize the risk of
230
interference. In this instance the observer acted as a full participant In the course. The
observer did not at any point ask questions or influence the class direction, but
undertook the instructors activities and took notes as if a novice. The course instructor
was debriefed at the completion of the two short cou5rses and permission was sought for
The unit of analysis consisted of both content material discussed and taught regarding
‘how to surf the web’ and content of student questions. The notes taken were later
imported into Nudist™ and the script coded according to the subject matter discussed.
The unit of analysis included web terminology, features of the web and web sites
discussed, the discussion of various web behaviours, and information about other
Internet-supported systems.
B.4.3 Results
Table B3, B4 and B5 present the final summation scores in descending order from most
frequent to least frequent frequency of term, feature and procedure used. These results
communication. Malhotra et al. (1996) and Neuman (1997) indicate that the unit of
(individuals and/or objects), themes (propositions), space and time measures (length or
duration of the message), or topics (subject of the message). Web site design and
promotional content has also been the focus of a number of research studies that
incorporate content analysis methodology. Dreze and Zufryden (1997a) for example
looked at the impact of background, image size, sound file display, celebrity
endorsement, use of java, frames and operating system on the number of pages accessed
and time spent at a web site. Li (1998) conducted a content analysis of three US
newspapers and found that Internet newspapers gave more priority to providing textual
231
information than graphical. Esrock and Leichty (1999) revealed through content analysis
that web pages are not used to their fullest potential by corporate entities to
an overview of different guidelines for effective web site design that have been
published, despite minimal research supporting the guidelines proposed. For example,
guidelines for the use of navigational tools (e.g., links to home page and help page),
practical considerations for images (e.g., how many per page and size), colour (e.g.,
number of colours per screen and type), audio and video (e.g., file size), content (e.g., text
only option, text position, search capability), and general visual characteristics such as
layout (e.g., page segmentation, use of white space). DʹAngelo and Little (1998), from this
review, compiled a list of ten characteristics specific to the visual and practical
considerations for web page design and assessed the use of these characteristics on
twenty web sites. However, they only looked at the visual and practical considerations
To aid the development of a number of scale items required for the conduct of this
dissertation, content analyses were conducted of three help files and over 80 web sites
Three help files were content analysed. The help files for both Netscape™ (204KB) and
Microsoft Explorer™ (237KB), and a leading ISP’s navigation help file for its members
(Ozemail™ - 63KB). These files were chosen as they act as a consistent user-based
resource to assist in web and web browser use. To identify consistent web sites that offer
help tools and/or information about the web proves very difficult due to the large
number of web sites that are available. Therefore, by sampling the help files associated
with the most ‘used’ and/or ‘preferred’ browser software, a degree of consistency is
upheld.
Only sections of the Help/FAQ files that relate to actual usage (i.e., navigation, shopping,
etc.) of the web/browser were analysed. Sections corresponding to web site design and
232
development were not analysed as the purpose of the analysis was to ascertain user
Australian web users were Microsoft Internet Explorer 4 (42%) and Netscape 4/Netscape
Communicator (32%). In addition, the help file of one of the largest ISPs in Australia,
The contextual unit of analysis is the ‘Help/FAQ’ files of both browsers and the ISP.
Manifest coding was used to count the frequency with which certain terms and features
were mentioned in the files, as identified from the literature analysis and expert survey.
The help files in text (.txt) format were imported into Microsoft™ Word and Nudist to
conduct the analysis. The frequency of features was assessed by counting the number of
times these features were mentioned in the help files. In addition, latent coding was used
to assign certain meanings to sections of the help files in accordance with past research
B.5.1.3 Results
Tables B3, B4 and B5 present the final summation scores in descending order from most-
to-least frequent use of terms, features and procedures in the help-files sampled. These
The contextual units of analysis are web pages however, it is first necessary to explain
233
B.5.2.1 Sampling Design
Data for this study comes from a number of individual web pages at a number of
sampled web sites. As the purpose of this study is the development of scale items that
validly measure a user’s knowledge, perception and use of the web, the most trafficked
Weare and Lin 2000). The top 20 web sites accessed by Australians in 1999 and 2000
(ranked monthly by traffic) were used to create a sampling frame. The source, Microplex
requests by Australians. These rankings were calculated using raw traffic, with some
sites aggregated to a common URL. This gave a minimum of 480 web sites (20 web sites
x 24 months = 480).
A web site, however, can range from 1 to thousands of web pages. Thus a certain
number of pages were selected to increase the manageability of the task at hand. Thus an
average of 5 pages per web site, 2450 web pages (html files), could be analysed. To
identify a workable sample size, a two-month window was selected for analysis (January
and February; 20 web sites x 2 months in each of 2 years = 80 web sites). Only 37 unique
web sites were sampled as analysis of the top 20 web sites took place on four different
occasions, and a number of sites were analysed more than once. For example,
http://www.ninemsn.com.au was in the top 20 web sites on all four occasions, whereas
http://www.careerone.com.au only made it into the top 20 on one out of the 4 occasions.
Due to the dynamic nature of the web, web sites are continually changing and being
updated. Hence, web sites that were content analysed before were not excluded, but
included in the process as it was assumed a degree of change would occur in the coding.
The web sites analysed are presented in Table B2. A total of 37 unique web sites and a
total of 413 web pages (i.e., a web page sampling ratio of 16.8%), averaging 5 pages per
234
Five pages were considered for each site. To ensure consistency, three of these were
standardized: the home page, a FAQ or Help page, and a Privacy and or Security Policy
page. The two other pages ranged from shopping pages to news and information pages.
WS/CA-0006
www.fairfax.com.au* Media 6 - - -
WS/CA-0007
www.news.com.au** Media 7 5 - -
235
WS/CA-0030 www.property.com.au Real Estate - - 13 20
Manifest coding (MC) was the main form of coding used, however some latent coding
(LC) was used to increase the propensity of the item counted as being a true
representative of the item analysed. The list of terms and features was identified
Despite the fact that two or more raters are often used in content analyses (Kolbe and
Burnett 1991), only one coder was involved in the data collection process. To increase
objectivity and decrease error in the data collection process, a coding sheet and set of
rules and procedures was used. Data were entered into a database for manageability and
Tables B3 and B4 present the final summation scores in descending order from most-to-
least frequent use of terms and features. The top five terms were: links, home, cookie/s,
browser and download (Table B3). The top five features were: textual links, page
sections, graphical links, still graphics, and other logos (Table B4).
how and what they use to navigate the web. Respondents were asked to talk the
236
interviewer through their navigation process when given two tasks to complete using
the web. This method of exploratory data collection was used to ascertain the type of
students was recruited using email recruitment. The sample consisted of four male and
one female interviewees that were between the ages of 22 and 28 years of age. All
participants self-reported: low to medium use of internet and/or web-based chat services;
some purchasing experience; medium to high level of experience with computers, the
web overall, and using the web for information search; and a high level of email use
expertise with the web. Respondents were asked to indicate their preferred browser and
browser default page so as to align the surfing observed during the interview closely
After completing a consent form, and survey on background questions, the interviewee
was given a task description and asked to complete the task. No specified time period
was allocated for task completion and it was not required that the student complete the
task. A time period of 20 minutes was allocated to each task. The tasks were used as a
tasks students were asked to verbalise what they were doing and were asked a number
of pre-set questions by the interviewer to aid their verbalisation of the behaviour being
conducted. The first task consisted of finding information about suitable accommodation
in a specified region for an up and coming ‘weekend away’ and the second task required
interviewees to locate information about a certain book title they were interested in
purchasing.
237
The interviewer played a very passive role listening to the interviewees and only where
various terms, features and web use. At the conclusion of both tasks the interviewees
were debriefed as to the purpose of the study and what in fact the interviewer was
observing (most interviewees thought the purpose was to elicit task specified
information).
B.6.3 Results
238
Table B3: Frequency of Web and Web Site Terminology
Expert Survey (n=12) Observational Study (n = 15) Help Files CA (n = 3 files) Web Sites CA (n = 413pgs) In-depth Interviews (n = 5)
Most –to Least Frequent Frequency of Text Units Frequency of Text Units Most –to Least Frequent Frequency of Text Units
Java Script Boolean (19) URL (605) Link (625) URL (27)
HTML Internet (16) Http (396) Home (584) Home Page (24)
Hyper/links Domain (11) Browser (278) Cookie (409) Bookmark (16)
Plug-in Browser (10) Link (203) Browser (403) Link (14)
CGI Scripts Web (9) Internet (178) Download (318) Frame (13)
E-commerce URL (5) Web (121) FAQ (235) Search Engine (12)
Banner Ads Home Page (4) Home Page (47) Directory (190) Browser (11)
Bandwidth Http (3) java (46) Real Audio (148) Download (11)
Meta-tags Web Page (3) Jpeg (45) Cache (144) Directory (10)
Frames Link (3) Gif (34) Server (132) Server (8)
Search Engine ISP (3) Email (31) Upload (82) FAQ (5)
JPEG Hyperlink (1) Encryption (20) Bookmark (59) Plugin (4)
Flash Surfing (1) Bandwidth (18) Java (56) Boolean (3)
Internet Favourites/Bookmarks (1) Search Directory (16) URL (56) Portal (3)
Email Email (1) Virus (13) Domain (50) World Wide Web (2)
Browser Flamming (1) Cookies (9) Frame (46) Webmaster (2)
URL Spamming (1) Domain (8) Flash (44) Shockwave (2)
Cookies Hypertext (6) Streaming (38) Streaming (2)
Bookmarks Search Engine (4) Search Engine (37) Java (2)
Homepage Hypermedia (3) Metaword (36) Encryption (1)
WWW ftp (1) SSL (27) SSL (1)
Web Ring Encryption (27) Domain (1)
FTP Banner Ad (27) Banner Ad (1)
GIF Bandwidth (25) Crawler (0)
Download Plugin (12) Hit (0)
HTTP Boolean (12) Bandwidth (0)
Shockwave (12) Metaword (0)
Crawler (9) Flash (0)
Hit (8) Upload 0)
Cache (0)
Cookie (0)
239
Expert Survey (n=12) Observational Study (n = 15) Help Files CA (n = 3 files) Web Sites CA (n = 413pgs) In-depth Interviews (n = 5)
Most –to Least Frequent Frequency of Text Units Frequency of Text Units Most –to Least Frequent Frequency of Text Units
Text Search Directory (11) Html (508) Textual Links (21176) Link/s (46)
HTML Page Title (8) Text (111) Page Sections (2189) Button (Back, Stop etc) (28)
Banner Ad Hyperlink (7) Bookmarks (110) Graphical Links (2150) Frames (17)
Nav/Menu Bars Search Engine (6) Toolbar Buttons (103) Still Graphics (1725) Tabs (12)
Frames Bookmark (5) Buttons (39) Other Logo (933) Shopping Cart (11)
Graphics Email (5) Frames (36) Drop Down Menu (785) Search Engine (11)
Java/script Default Page (3) Plug-ins (31) Vendor Logo (684) Drop Down Menu (10)
FAQ Button (3) FAQ Feedback Link (550) Hyperlink (9)
Feedback Browser Icon (3) URL Components (21) Navigation Bar Top (522) Pop-up Ads (7)
What’s New Graphics (3) Graphics (12) Animated Graphics (446) Video (4)
Audio Menu Bar (2) Security Information (11) Action Button (443) URL (6)
Animation Chat (2) Video (10) Help Link (430) FAQ (2)
Search Engines Text (1) Sound (10) Navigation Bar Left (307) Graphics (2)
Video Web Site (1) Security Indicator (7) Privacy & Security Policy (297) Http (2)
Cascading Style Sheets (CSS) Tool Bar (1) Internal / External Images (7) Ad Button Right (289) Banner Ad (1)
Buttons Address Bar (1) Whatʹs New (7) Navigation bar Bottom (209) Text Only (1)
Site Maps Scroll Bar (1) Dynamic Html (6) FAQ (209) Site Map (1)
Sound Drop-down Menu (1) Bandwidth Settings (6) Search Site (187) What’s New (1)
Forms Site Design (0) Mailto Links (4) Banner Ad Top (182) Audio (1)
Payment Facilities Web Phone (0) Search Engine (4) Ad Button Top (173) Other Logo (1)
Hyper/links Links (3) Ad Button Bottom (140) Advertising (0)
https (3) Whatʹs New (134) Web Ring (0)
Location Field (1) Ad Button Left (127) Vendor Logo (0)
History Items (1) Email Entry (90) Page Sections (0)
Search Tool (1) Text Only Option (89) Search Web (0)
Find Tool (1) Site Map Option (88) Email Entry (0)
Scroll Bar (1) Audio (82) Search Site (0)
Animation (0) Frames Top (77) Ad Button (0)
Video (63) Privacy & Security Policy (0)
Frames Bottom (56) Help Link (0)
Search Web (54) Animation (0)
Banner Ad Bottom (39) Navigation Bar (0)
Navigation bar Right (30) Feedback Link (0)
Frames Left (23) Https (0)
No Frames Option (13)
Error Message (11)
Https (4)
Popup Menu (3)
240
Table B5: Frequency of Web and Web Site Procedures (Behaviours)
Expert Survey (n=12) Observational Study (n = 15) Help Files CA (n = 3 files) Web Sites CA (n = 413pgs) In-depth Interviews (n = 5)
Most –to Least Frequent Frequency of Text Units Frequency of Text Units Most –to Least Frequent Frequency of Text Units
Situational Characteristics Behaviour (s) Behaviour (s) Behaviour (s)
Connection Speed Search (48) Using Bookmarks (143) Scrolling (33)
Response Times Using Search Terms (20) Using Links (103) Editing URL (21)
Computer Literacy Adding Favourites (19) Editing Bookmarks (64) Clicking (19)
Cost of Access Using Boolean Operators (11) Stopping Transfer (57) Retracing Steps (Back) (18)
Who Pays Drag & Drop (9) Using a Search Engine (57) Search Engine Use – Boolean (17)
User Aim/Motive Clicking (6) Saving Web Pages (45) Navigation (14)
Computer System Deleting Favourites (6) Download (43) Browser Use (13)
Graphics/Size Composing Email (6) Using Cache (39) Shopping (12)
Place of access Internet Use (5) Setting Bookmark Preferences (35) Adding Favourites (11)
Location Surfing (4) Using History Menu (34) Search Engine Use – Search Terms (9)
Category Search (4) Using the Location Field (33) Page Navigation (8)
Video Conferencing (3) Forward & Back (29) Link Visitation (8)
Encrypt/Decrypt (3) Using Home Button (26) Spamming (4)
Domain Purchase (2) Using Reload Button (25) Find in Page (4)
Close Browser (2) Accessing the Internet (25) Smart Searching (1)
Editing Address Bar (1) Using Frames (23)
Link Visitation (1) Using Tool Bar and Menu Links (23) Additional System Characteristics
Back/Forward (1) Printing Web Pages (20) Speed (71)
Browser icon Movement (1) Turning Images Off (16) Email (32)
Go To (1) Using a URL (13) Cursor Change (28)
Open Browser Window (12) URL Contents (21)
Searching History List (12) Browser Icon (18)
Viewing Menu List (11) System Transfer (10)
Navigating a Page (11) Error – Page Not Found (6)
Filling in Forms (10) Status Bar (6)
Selecting Default Home Page (9) Error – Connection (4)
Searching for information (8)
Open a Web Page (8)
Auto Scroll (8)
Status Message Area (7)
Using Multiple Browser Windows (7)
241
Viewing the Component Bar (7)
Displaying Previously Viewed Pages (7)
Displaying Pop-up Menu (7)
Change Page Background (7)
Finding a Bookmark (7)
Accepting Cookie (7)
Reporting Error Message (6)
Clicking (5)
Find in Page (5)
Identifying Used Links (4)
Browser Icon Animation (3)
Progress bar (3)
Using the Component Bar (3)
Learn about Browser (3)
Using Navigation Tool Bar (3)
Viewing Bookmarks and History (3)
Scrolling (2)
Automatic Update Pages (2)
Using the Search Button (2)
Using Guide Button (2)
Using Images Button (2)
Using Security Button (2)
Changing Mouse Cursor (1)
Find and Return to Pages (1)
242
APPENDIX C: VARIABLE CONCEPTUALISATION AND OPERATIONALISATION (PRE/POST TEST)
Operationalisation Operationalisation Questionnaire
ID Scale Conceptualisation
(Pre-item Testing) (Post-item Testing) (Section/Item)
Current Web Session Usage
Current Web Session Usage 1-item, 8-category 1-item, 8-category
WSUF Frequency
How often the web is accessed within a certain time frame
Measure Measure
D15
Current Web Session Usage Number of new and/or different web sites and search tools 4-item, 7-point Likert Scale (1=SA, 3-item, 7-point Likert Scale
WSUEB Extent (Breadth) accessed 7=SD) (1=SA, 7=SD)
D1, D3, D8
Current Web Session Usage 4-item, 7-point Likert Scale (1=SA, 4-item, 7-point Likert Scale
WSUED Extent (Depth)
Total number of web sites and search tools accessed
7=SD) (1=SA, 7=SD)
D4, D6, D9, D11
243
Skilled and/or extraordinary static information of facts, 28 -scale items: 24 x 3-category 10 -scale items: 10 x 3-category
Actual Specialised Declarative B1, B3, B5, B7, B25, B29,
ASDWK Web Knowledge Content
terms, attributes (what) of X, required to perform skilled (T/F/DK) and (T/F/DK) and
B31, B33, B35, B36
domain related tasks successfully 4 x 5- category (MC) 1 x 5- category (MC)
Perceived Web Knowledge
An individual’s personal judgement of the level of
Perceived Procedural Web 8-items, 7-point Likert Scale (1=SA, 4-items, 7-point Likert Scale
SWPK Knowledge Content
knowledge stored in their memory about how to use certain
7=SD) (1=SA, 7=SD)
A5, A8, A21, A43
features and/or terms of the web
An individual’s personal judgement of the level of
Perceived Declarative Web 9-items, 7-point Likert Scale (1=SA, 7-items, 7-point Likert Scale A12, A14, A19, A25,
SWDK Knowledge Content
knowledge stored in their memory about what certain
7=SD) (1=SA, 7=SD) A32, A38, A52
features and/or terms of the web are
Perceived Overall Web An individual’s personal judgement of the overall level of 3-items, 7-point Likert Scale (1=SA, 2-items, 7-point Likert Scale
SWOK Knowledge Content knowledge content about the web stored in their memory 7=SD) (1=SA, 7=SD)
A23, A48
a: After item testing and purification an additional item was added to measure WSUVS. Thus WSUVS after item-testing and purification would include 2-items.
244
APPENDIX D: SCALE DEVELOPMENT - ITEM FACTOR LOADINGS (STUDENT SAMPLE ONE: N=128)
Web Session Usage Extent – Breadth Web Session Usage Extent – Depth
SS1: Q40 .910 SS1: Q27 .900
SS1: Q43 .713 SS1: Q45 .892
SS1: Q71 .960 SS1: Q41 .941
Dimension Site Search Engine SS1: Q70 .928
Dimension Variance 46% 36% Dimension Link Search Engine Site
Dimension Reliability/Correlation r = 0.4 NA Dimension Variance 43% 26% 25%
Scale Variance 82% Dimension Reliability/Correlation r = 0.8 NA NA
Scale Reliability 0.7 Scale Variance 94%
Scale Reliability 0.8
Perceived Web Usefulness
Perceived Ease of Web Use
SS1: Q48 .878
SS1: Q47 .853 SS1: Q53 .789
SS1: Q67 .756 SS1: Q55 .756
SS1: Q39 .673 SS1: Q54 .725
SS1: Q38 .650 SS1: Q60 .699
SS1: Q21 .785 SS1: Q59 .671
SS1: Q37 .780 SS1: Q61 .657
SS1: Q49 .763 SS1: Q62 .605
SS1: Q35 .688 SS1: Q29 .784
SS1: Q51 .721 SS1: Q31 .776
SS1: Q19 .720 SS1: Q30 .764
SS1: Q34 .659 SS1: Q58 .694
SS1: Q52 .658 SS1: Q32 .647
SS1: Q50 .630 SS1: Q18 .820
Dimension Comm. Purchase Information Quality SS1: Q33 .692
Dimension Variance 23% 22% 19% 10% Dimension Behavioural Informational Transactional
Dimension Reliability/Correlation 0.9 0.9 0.8 r = 0.6 Dimension Variance 29% 29% 13%
Scale Variance 75% Dimension Reliability/Correlation 0.9 0.9 r = 0.7
Scale Reliability 0.9 Scale Variance 73%
Scale Reliability 0.9
245
APPENDIX E: SCALE DEVELOPMENT - ITEM FACTOR LOADINGS (STUDENT SAMPLE TWO: N=153)
Actual Common Procedural Actual Specialised Declarative Actual Common Declarative
SS2: Q20 .819 SS2: Q63 .791 SS2: Q71 .742
SS2: Q75 .754 SS2: Q65 .701 SS2: Q26 .734
SS2: Q77 .666 SS2: Q70 .629 SS2: Q25 .714
SS2: Q76 .888 SS2: Q40 .609 SS2: Q44 .713
SS2: Q78 .686 SS2: Q22 .603 SS2: Q42 .534
SS2: Q37 .968 SS2: Q28 .686 SS2: Q33 .793
Dimension Speed of Use Web features Updates SS2: Q49 .587 SS2: Q52 .720
Dimension Variance 31% 26% 18% SS2: Q26 .581 SS2: Q60 .708
Dimension Reliability 0.7 r = 0.5 NA SS2: Q59 .509 SS2: Q61 .547
Scale Variance 75% SS2: Q46 .834 SS2: Q34 .540
Scale Reliability 0.8 SS2: Q17 .706 Dimension Standards Tools & Terms
Dimension Tools & Terms Standards Cookies Dimension Variance 30% 27%
Actual Specialised Procedural Dimension Variance 24% 19% 16% Dimension Reliability 0.8 0.8
Dimension Reliability 0.8 0.6 r =0.6 Scale Variance 57%
SS2: Q21 .790 Scale Variance 59% Scale Reliability 0.9
SS2: Q18 .734 Scale Reliability 0.9
SS2: Q73 .652 Perceived Procedural Web Knowledge
SS2: Q39 .564 Perceived Declarative Web Knowledge
SS2: Q74 .809 SS2: Q5 .937
SS2: Q36 .777 SS2: Q13 .915 SS2: Q2 .879
SS2: Q79 .714 SS2: Q4 .912 SS2: Q8 .876
SS2: Q57 .853 SS2: Q3 .891 SS2: Q7 .815
SS2: Q51 .680 SS2: Q18 .867 Dimension 1
SS2: Q23 .525 SS2: Q11 .860 Scale Variance 77%
SS2: Q19 .486 SS2: Q10 .818 Scale Reliability 0.9
Dimension Web Features Speed of Use Quality SS2: Q14 .805
Dimension Variance 21% 20% 18% Dimension 1 Perceived Overall Web Knowledge
Dimension Reliability 0.7 0.7 0.7 Scale Variance 75% SS2: Q1
r =.0.8
Scale Variance 59% Scale Reliability 0.9 SS2: Q12
Scale Reliability 0.8
246
APPENDIX F: WEB SITE AND WEB SURVEY DESIGN
this survey. The objective of the design was to present a credible and non-commercial
image of the research project to prospective respondents and to do so in a way that made
it easy for them to complete the survey. Thus, a very simple site structure was devised,
with limited external links to encourage survey participation and decrease respondent
Figure F1: Phd Web Audience Study - Simple Web Site Structure
The web site was hosted on the main web server (http://www.unsw.edu.au) within the
UNSW University Wide Network (UWN). A domain name for the web site was
ease of respondent access to the site through URL recall and to increase the perceived
credibility of the study (Nielsen, 1999). Given these hosting decisions and the use of
official university symbols on the web site, the UNSW ‘Electronic Identity Standards
Policy’ (as at May 2000) was adhered to and the web site design was also approved by
247
F.2 WEB SURVEY DESIGN
The foundations for designing the interactive web-based survey comprised: the standard
Internet protocol (hypertext mark-up language, HTML), FileMaker Pro™ , and Claris
Dynamic Markup Language (CDML) and additional common gateway interface (CGI)
The web survey was designed for ease of completion. The large number of questions
asked (Q=180) was broken up into 6 separate sections, including an introduction with
screening questions. For example, as the study was targeting Australian residents,
respondents were first asked to indicate their residency status, this was followed with a
question asking the respondents for the source of study awareness. Following the
categories that included questions covering ‘Web Perceptions’, ‘Web Knowledge’, ‘Web
Respondents navigated these sections using navigational tools provided in the survey
design (i.e., hot-linked buttons) and upon survey completion used a ‘Submit’ button to
execute web-survey and database interaction. The items used in the survey (as discussed
in Chapter 7) used differing question response formats and thus differing field types
were used to correspond to each question response format. For example, interval-
measured Likert scales were measured using ‘radio buttons’; ordinal-measured multiple-
minimise this potential problem, questions were intermixed and the ordering was varied
248
Data use & treatment
Link to Terms & and researcher
The design of the web-site, and the survey instrument, are shown in Figures F2, F3 and
249
F4.
Competition terms & conditions as approved by state and territory Gaming License Hotlink button to survey
gaming regulations
Title & Tool bar
Figure F3: Web Site - Competition Terms and Conditions (terms.html)
250
Survey introduction and screening questions Section A: Perceptions of the World Wide Web
Figure F4: Web Site - Survey (Questionnaire.html)
251
Section A: Continued
252
Section B: Actual Knowledge Content of the World Wide Web
253
Section B: Continued Section C: Exposure to and Provision of Communication about the Web
254
Section C: Continued
255
Section D: Current and Past Web Session Usage Experience
256
Section D: Continued
257
Section E: Respondent Descriptive Information
258
Section E: Continued
259
APPENDIX G: WEB SURVEY ADVERTISING AND PUBLICITY
260
G.2 OFFLINE PUBLICITY
Official UNSW Media Release Newspaper Article (The Australian IT (Tuesday 14/11/00; p3)
261
Magazine Review: Australian NetGuide (January, 2001, p16).
262
APPENDIX H: WEB SITE PERFORMANCE STATISTICS
Web Site Performance Statistics: Pilot Study (10-12th Oct) and Main Study (18th Oct 2000 to 31st Jan 20001)
October (Pilot) October November December January
Main Study
10-12B1 18-31B1 01-15B3 16-30B3 1-15B2 & B3 16-31B3 1-15 16-31
Total
Hits Entire Site (Successful) 22,275 59317 14881 9817 41664 4020 5867 4922 140,488
Average Per Day 7,425 4236 992 654 2777 251 391 307 1,325
Home Page 1,167 3608 692 426 2214 191 276 248 7,650
Page
Page Views (Impressions) 1,927 5618 1265 792 3460 361 491 420 12,407
Views
Document Views 1,927 5618 1265 792 3460 361 491 420 12,407
Visits Visits 1,642 4244 883 561 3083 292 383 328 9,774
Average Visit Length 00:03:38 00:03:21 00:02:48 00:03:17 00:02:36 00:03:15 00:02:28 00:02:34 00:03:00
Median Visit Length 00:01:28 00:01:28 00:01:22 00:01:27 00:01:23 00:01:40 00:01:20 00:01:33 00:01:26
International Visits 24.23% 23.68% 22.42% 16.57% 25.49% 26.02% 23.75% 21.34% 23.72%
Visits of Unknown Origin 21.43% 21.22% 24.23% 19.25% 20.27% 17.8% 17.49% 17.07% 20.69%
Visits from Australia 54.32% 55.08% 53.34% 64.17% 54.23% 56.16% 58.74% 61.58% 55.57%
Visitors Unique Visitors 1,232 2643 601 385 1929 210 251 232 5,104
Who Visited Once 1,053 2177 492 309 1593 169 197 195 4,170
Who Visited More Than Once 179 466 109 76 336 41 54 37 934
B1: Banner Ad Campaign 1 (DoubleClick Network Placement) – 22nd to 29th October 2000; B2: Banner Ad Campaign 2 (DoubleClick Network Placement) – 05th to 12th
December 2000; B3: Banner Ad Campaign 3 (Small Network Placement) – 30th October to 31st December 2000
263
APPENDIX I: DOUBLECLICK™ BANNER AD CAMPAIGN REPORT/S
Banner Ad Placement (1) (Oct 2000) Banner Ad Placement (2) (Dec 2000)
Total number of unique users who saw ads 478 095 389 522
seek.com.au 266 632 750 0.28% seek.com.au 235 351 655 0.28%
goeureka.com.au 446 665 593 0.13% goeureka.com.au 314 593 401 0.13%
investorweb.com.au
quicken.com.au 118 004 230 0.19% 43 262 151 0.35%
disney.au
au.mirror.nasdaq.com 35 833 215 0.60% 12 629 132 1.05%
cdnow.au
melb.tradingpost 101 915 202 0.20% 39 685 125 0.31%
melb.tradingpost
property.com.au 82 629 144 0.17% 86 674 101 0.12%
264
APPENDIX J: SCALE VALIDATION - ITEM FACTOR LOADINGS (WEB SAMPLE: N=2077)
Web Session Usage Extent - Breadth Web Session Usage Extent - Depth Web Session Usage Variety - Situational
WS1-D8: WSUEB1M .863 WS1-D6: WSUED1M .885 WS1-D16: WSUVS1
r = 0.6
WS1-D1: WSUEB2M .824 WS1-D11: WSUED2M .843 WS1-D17: WSUVS2
WS1-D13: WSUEB3M .979 WS1-D9: WSUED3M .967
Dimension Site Search Engine WS1-D4: WSUED4M .977
Dimension Variance 49% 34% Dimension Link Search Engine Site
Dimension Reliability/Correlation r = 0.5 NA Dimension Variance 38% 26% 25%
Scale Variance 83% Dimension Reliability/Correlation r = 0.5 NA NA
Scale Reliability 0.7 Scale Variance 89%
Scale Reliability 0.6
265
Actual Specialised Procedural Web Knowledge Actual Common Procedural Web Knowledge Perceived Overall Web Knowledge
266
APPENDIX K: SCALE PERFORMANCE COMPARISON (STUDENT SAMPLES AND WEB SAMPLE)
Scale Development Scale Validation
Scale Scales (Student Samples) (Web Sample)
Dimensions VAR% α Dimensions VAR% α
267
APPENDIX L: VARIABLE DISTRIBUTION
L.1 HISTORGRAMS
Figure L1: WSUF Normality Test (-) Figure L2: WSUEDUR Normality Test (+) Figure L3: WSUVMNO1 Normality Test
600
800
600
500
600
400
400
300
400
200
200
200
Std. Dev = 1.21 Std. Dev = 1.48 100 Std. Dev = 2.50
Mean = 6.8 Mean = 3.6 Mean = 5.7
N = 2077.00 0 N = 2077.00 N = 2077.00
0 0
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0
Web Session Use Freqency Duration of Session Use No. of Use Motivations (Sum)
Figure L4: WSUVS Normality Test (+) Figure L5: ACPWK Normality Test (-) Figure L6: ASPWK Normality Test (-)
1000
800
800
800
600
600 600
400 400
400
Situational Variety - (Sum) 0.0 2.0 4.0 6.0 8.0 10.0 12.0
268
Figure L7: ACDWK Normality Test (-) Figure L8: ASDWK Normality Test (-)
1400 700
1200 600
1000 500
800 400
600 300
400 200
hypothesized that each variable will have a normal distribution to be representation of the population. The results presented in Table L1. It is
worth noting, however, that due to the large sample size (n=2077) the test of significance is very sensitive. Thus both graphical plots and statistical
tests were used to assess normality, however for brevity in this report only the statistics are reported here.
269
Table L1: Variable/s –Test of Normality
Label No. Shape Normality Test Description
Skew Kurtosis Statistic Sig.
Current Web Session Usage Extent
WSEB 2077 .656 .216 5.258 0.000 Since the significance is 0.000 the null hypothesis is rejected The distribution of WSUEB is different from the assumed distribution (a
normal distribution) and there is virtually no chance that this difference is due to sampling error. It is thus highly probable that WSUEB
is not normally distributed. After further examination WSUEB has a positive skewed relatively peaked distribution (PS / LD).
WSUED 2077 -.350 -.025 3.826 0.000 Since the significance is 0.000 the null hypothesis is rejected The distribution of WSUED is different from the assumed distribution (a
normal distribution) and there is virtually no chance that this difference is due to sampling error. It is thus highly probable that
WSUED is not normally distributed. After further examination WSUED has a negative skewed relatively flat distribution. (NS / PD)
Perceived Ease of Web Use and Web Usefulness
PEWU 2077 -.712 .513 3.562 0.000 Since the significance is 0.000 the null hypothesis is rejected The distribution of PEWU is different from the assumed distribution (a
normal distribution) and there is virtually no chance that this difference is due to sampling error. It is thus highly probable that PEWU
is not normally distributed. After further examination PEWU has a negative skewed relatively peaked distribution (NS / LD)
PWU 2077 -.490 .263 2.319 0.000 Since the significance is 0.000 the null hypothesis is rejected The distribution of PWU is different from the assumed distribution (a
normal distribution) and there is virtually no chance that this difference is due to sampling error. It is thus highly probable that PWU is
not normally distributed. After further examination PWU has a negative skewed relatively peaked distribution (NS / LD)
Perceived Web Knowledge
SWPK 2077 -1.055 .870 6.022 0.000 Since the significance is 0.000 the null hypothesis is rejected The distribution of SWPK is different from the assumed distribution (a
normal distribution) and there is virtually no chance that this difference is due to sampling error. It is thus highly probable that SWPK
is not normally distributed. After further examination SWPK has a negative skewed relatively peaked distribution (NS / LD)
SWDK 2077 -.650 -.065 3.792 0.000 Since the significance is 0.000 the null hypothesis is rejected The distribution of SWDK is different from the assumed distribution (a
normal distribution) and there is virtually no chance that this difference is due to sampling error. It is thus highly probable that SWDK
is not normally distributed. After further examination SWDK has a negative skewed relatively flat distribution (NS / PD)
SWOK 2077 -.693 .021 5.718 0.000 Since the significance is 0.000 the null hypothesis is rejected The distribution of SWOK is different from the assumed distribution (a
normal distribution) and there is virtually no chance that this difference is due to sampling error. It is thus highly probable that SWOK
is not normally distributed. After further examination SWOK has a negative skewed relatively peaked distribution (NS / LD)
b NS = Negative Skew; PS = Positive Skew; ND = Normal Distribution; LD = Leptokurtic (Peaked) Distribution; PD = Platykurtic (Flat) Distribution
270
APPENDIX M: MULTIPLE REGRESSION – RESIDUAL PLOTS
Figure M-1a: MRA1 Residual Plot Figure M-2a: MRA2 Residual Plot
Figure M-3a: MRA3 Residual Plot
WSDMEXP2: 0 No Experience
2
WSDMEXP2: 0 No Experience
5
WSDMEXP2: 0 No Experience
1
3
Regression Standardized Residual
0 4
2
-2 2 0
-3 1 -1
0 -2
-4
-3
-5 -1
-4 -3 -2 -1 0 1 2 3
-4
-2 -4 -3 -2 -1 0 1 2 3
Regression Standardized Predicted Value -4 -3 -2 -1 0 1 2 3
Regression Standardized Predicted Value
Regression Standardized Predicted Value
Figure M-1b: MRA1 Residual Plot Figure M-2b: MRA2 Residual Plot Figure M-3b: MRA3 Residual Plot
WSDMEXP2: 1 Experience WSDMEXP2: 1 Experience
2 3
WSDMEXP2: 1 Experience
4 2
0
Regression Standardized Residual
-2
2 0
1 -1
-4
0
-2
-6
-1
-4 -3 -2 -1 0 1 2 3 -3
-5 -4 -3 -2 -1 0 1 2 3
Regression Standardized Predicted Value -2
-5 -4 -3 -2 -1 0 1 2 3 Regression Standardized Predicted Value
271
Figure M-4a: MRA4 Residual Plot Figure M-5a: MRA5 Residual Plot Figure M-6a: MRA6 Residual Plot
WSDMEXP2: 0 No Experience
WSDMEXP2: 0 No Experience 4
WSDMEXP2: 0 No Experience 4
4 3
3
1
2
0
0
1 -1
-1
-2
0
-2
-3
-4 -3 -2 -1 0 1 2 3
-1 -3
Regression Standardized Predicted Value
-4
-2 -4 -3 -2 -1 0 1 2 3
-3 -2 -1 0 1 2 3 4
Figure M-4b: MRA4 Residual Plot Figure M-5b: MRA5 Residual Plot Figure M-6b: MRA6 Residual Plot
WSDMEXP2: 1 Experience
WSDMEXP2: 1 Experience WSDMEXP2: 1 Experience
4
3 4
3 2 3
Regression Standardized Residual
0 1
1
-1 0
0
-2 -1
-1 -3 -2
-4 -3
-2
-4 -3 -2 -1 0 1 2 3 -4 -3 -2 -1 0 1 2 3
-3 -2 -1 0 1 2 3 4 5
272
Figure M-7a: MRA7 Residual Plot Figure M-8a: MRA8 Residual Plot
WSDMEXP2: 0 No Experience
WSDMEXP2: 0 No Experience
3
4
2
1
Regression Standardized Residual
-1
-2
-2
-3
-4
-4
-6
-5
-3 -2 -1 0 1 2 3
-4 -3 -2 -1 0 1 2
Figure M-7b: MRA7 Residual Plot Figure M-8b: MRA8 Residual Plot
2
2
Regression Standardized Residual
0
0
-2
-2
-4
-4
-6
-5 -4 -3 -2 -1 0 1 2 3 -6
-5 -4 -3 -2 -1 0 1 2 3
Regression Standardized Predicted Value
Regression Standardized Predicted Value
273
APPENDIX N: MULTIPLE REGRESSION – NORMALITY P-P PLOT
Figure N-1a: MRA1 Normality P-P Plot Figure N-2a: MRA2 Normality P-P Plot
Normal P-P Plot of Regression Standardized Residual Normal P-P Plot of Regression Standardized Residu
Figure N-3a: MRA3 Normality P-P Plot
Normal P-P Plot of Regression Standardized Residual
WSDMEXP2: 0 No Experience WSDMEXP2: 0 No Experience
1.00 WSDMEXP2: 0 No Experience
1.00
1.00
.75
.75
.75
.50
Expected Cum Prob
.50
0.00
0.00 .25 .50 .75 1.00 0.00
0.00 .25 .50 .75 1.00 0.00
Observed Cum Prob 0.00 .25 .50 .75 1.00
Figure N-1b: MRA1 Normality P-P Plot Figure N-2b: MRA2 Normality P-P Plot Figure N-3b: MRA3 Normality P-P Plot
Normal P-P Plot of Regression Standardized Residua
Normal P-P Plot of Regression Standardized Residual WSDMEXP2: 1 Experience Normal P-P Plot of Regression Standardized Residual
1.00
WSDMEXP2: 1 Experience WSDMEXP2: 1 Experience
1.00 1.00
.75
.75
.75
.50 .50
Expected Cum Prob
.25
.25 .25
0.00
0.00 0.00 0.00 .25 .50 .75 1.00
0.00 .25 .50 .75 1.00
0.00 .25 .50 .75 1.00
Observed Cum Prob
Observed Cum Prob
Observed Cum Prob
274
Figure N-4a: MRA4 Normality P-P Plot Figure N-5a: MRA5 Normality P-P Plot Figure N-6a: MRA6 Normality P-P Plot
Normal P-P Plot of Regression Standardized Residual Normal P-P Plot of Regression Standardized Residual
WSDMEXP2: 0 No Experience
WSDMEXP2: 0 No Experience WSDMEXP2: 0 No Experience 1.00
1.00 1.00
.75
.75 .75
.50
.50
Expected Cum Prob
.50
0.00 0.00
0.00
0.00 .25 .50 .75 1.00 0.00 .25 .50 .75 1.00
0.00 .25 .50 .75 1.00
Figure N-4b: MRA4 Normality P-P Plot Figure N-5b: MRA5 Normality P-P Plot Figure N-6b: MRA6 Normality P-P Plot
Normal P-P Plot of Regression Standardized Residual
WSDMEXP2: 1 Experience Normal P-P Plot of Regression Standardized Residua WSDMEXP2: 1 Experience
1.00 1.00
WSDMEXP2: 1 Experience
1.00
.75 .75
.75
.50 .50
Expected Cum Prob
.50
Expected Cum Prob
275
Figure N-7a: MRA7 Normality P-P Plot Figure N-8a: MRA8 Normality P-P Plot
.75 .75
.50 .50
Expected Cum Prob
0.00 0.00
0.00 .25 .50 .75 1.00 0.00 .25 .50 .75 1.00
Figure N-7b: MRA7 Normality P-P Plot Figure N-8b: MRA8 Normality P-P Plot
.75 .75
.50
.50
Expected Cum Prob
.25
.25
0.00
0.00 .25 .50 .75 1.00 0.00
0.00 .25 .50 .75 1.00
Observed Cum Prob
Observed Cum Prob
276
APPENDIX O: MULTIPLE REGRESSION ASSUMPTION CHECK
Linearity 3 ? 3 2 3 2 2 3
1 Some collinearity may be present between the independent variables given the close nature of the area of context investigated (i.e., perception & knowledge)
277
APPENDIX P: SAMPLE & VARIABLE DESCRIPTION
Overall 10 37 53 21 53 26 2 25 73
Procedural 9 36 55 13 39 48 1 14 85
Declarative 6 25 69 18 53 29 1 23 76
Web Perceptions
Ease of Use 3 38 59 3 38 60 4 38 59
Usefulness 2 42 56 2 41 57 3 43 55
Current Web Session Usage
Frequency 4 2 29 69 2 38 60 1 22 77
Situational
72 24 4 80 18 2 66 29 5
Variety5
Motivational
33 53 14 45 48 7 25 57 19
Variety6
Extent Breadth7 62 33 8 62 34 4 63 32 5
Extent Depth8 7 56 37 9 63 28 4 52 44
Extent
41 52 7 47 50 3 37 53 10
Duration9
1: Low = 0-2 correct out of 6; Medium = 3-4 correct out of 6; High = 5-6 correct out of 6.
2: Low = 0-3 correct out of 10; Medium = 4-6 correct out of 10; High = 7-10 correct out of 10.
3: Low = 0-3 correct out of 11; Medium = 4-7 correct out of 11; High = 8-11 correct out of 11.
4: Low = Once a month to Once every two weeks; Medium = Once a week to 4-6 times a week; High =
Once a day to 5 or more times a day.
5: Low = 1-2 locations & 0-3 location types; Medium = 3-4 locations & 4-7 locations types; High = 5 or
more locations & 8-11 location types
6: Low = 0-3 use motivations; Medium = 4-8 use motivations; High = 9-12 use motivations.
7: Low = Use the same search engines, web sites and type of sites in a web session; High = Use the
different search engines, web sites and types of sites in a web session
8: Low = Low number of bookmarks/favourites saved, use only a couple of search engines, visit few
web sites; High = High number of bookmarks/favourites saved, use a large number of search
engines, visit a large number of search engines
9: Low = Less than 1 hour in a web session; 1-6 hours during a web session; 7-13 hours during a web
session
278
Figure P1: Actual Comm Dec. Knowledge Figure P2: Actual Comm Proc. Knowledge
Figure P3: Actual Spec. Dec. Knowledge Figure P4: Actual Spec. Proc. Knowledge
279
Figure P5: Perceived Overall Knowledge Figure P6: Perceived Proc. Knowledge
Figure P7: Perceived Dec. Knowledge Figure P8: Perceived Ease of Web Use
Figure P9: Perceived Web Usefulness Figure P10: Web Use Frequency
280
Figure P11: Web Use: Sit. Variety Figure P12: Web Use: Motive Variety
Figure P13: Web Use Extent: Breadth Figure P14: Web Use Extent: Depth
281
P.2 SAMPLE MEAN COMPARATIVE ANALYSIS
Table P2: Mean Scores and Standard Deviations (SD)
Web Site Design &/or Maintenance (WSD/M) Experience
Without WSD/M (n=900) With WSD/M (n = 1177)
Mean SD Mean SD
Actual Knowledge
Common Procedural 4.77 1.433 5.51 .895
Common Declarative 7.52 2.470 9.08 1.562
Specialised Procedural 6.95 2.675 9.03 1.762
Specialised Declarative 4.72 2.956 7.69 2.275
Perceived Knowledge
Overall 8.74 2.700 11.64 2.037
Procedural 19.21 5.254 24.01 3.470
Declarative 30.25 8.524 39.62 6.524
Web Perceptions
Ease of Use 54.11 10.380 57.98 9.248
Usefulness 69.87 12.194 72.33 11.404
Current Web Usage
Frequency 6.52 1.286 7.07 1.082
Situational Variety 3.57 1.973 4.49 2.107
Motivational Variety 4.89 2.430 6.32 2.382
Extent Breadth 9.00 3.373 8.89 3.548
Extent Depth 17.76 4.400 19.64 3.964
Extent Duration 3.31 1.373 3.81 1.521
282
APPENDIX Q: BIVARIATE ANALYSIS - CONVERGENT
VALIDATION
Q.1 INTRODUCTION
In this appendix the results of bivariate analyses are reported. Nonparametric
bivariate statistics and contingency table analysis (i.e., cross tabulations) are used
to further validate the findings reported in Chapter 10. Further discussion as to the
treatment of variables and the motivation for the use of nonparametric bivariate
For each hypothesis tested, the results are presented and discussed stating the
observed differences between those users without, and those with, web site design
and maintenance experience (WSD/M Experience). These two groups are labelled:
‘Web User Group A (No WSD/M Experience)’ and ‘Web User Group B (With
WSD/M Experience)’. The descriptive profile of each user group was presented in
and a person’s current web session usage? To examine this question more
relationship between a user’s perceived usefulness and ease of use of the web and
that person’s current web session usage frequency, variety and extent. Results are
presented below.
283
H1A: Perceived ease of web use (PEWU) will have a curvilinear (L/H/M)
H0A: Perceived ease of web use (PEWU) will have no relationship with current web
coefficients (i.e., linear with non-linear) and secondly by examining a 3x3 cross-
The chi-squared based correlation coefficients for H1A are reported in Table Q1.
Measures) H1A: Perceived Ease of Web Use & Web Session Usage Frequency
Web Site Design and Asymp.
Approx.
Maintenance Statistic Value Std.
Sig.
Experience Errora
No Experience Nominal by Nominal Goodman and Kruskal’s tau .002 .002 .605b
(non-linear) Cramer’s V .052 .297
Ordinal by Ordinal Kendall’s tau-b .007 .032 .818
(linear) Gamma .015 .065 .818
No. of Valid Cases 900
Experience Nominal by Nominal Goodman and Kruskal’s tau .000 .001 .953b
(non-linear) Cramer’s V .043 .370
Ordinal by Ordinal Kendall’s tau-b -.017 .028 .553
(linear) Gamma -.040 .067 .553
No. of Valid Cases 1177
a. Not assuming the null hypothesis
b. Based on chi-squared approximation
From looking at the non-linear correlation coefficients in Table Q1, it is evident that
a small positive association exists between PEWU and WSUF for users with no
284
linear correlation coefficients reported in Table Q1 tell the same story. To assess the
with Gamma. As Goodman and Kruskal’s tau is less than Gamma, we can infer
that there is no curvilinear relationship for web users without WSD/M experience.
When examining the non-linear correlation coefficients for users with WSD/M
experience in Table Q1, it is evident that a very weak positive association exists
between PEWU and WSUF. But, again, this result is not statistically significant.
This very weak positive result is inconsistent with the negative linear correlation
coefficients reported in Table Q1, however these also are not statistically
Kruskal’s tau and Gamma are compared. For users with WSD/M experience
Goodman and Kruskal’s tau is in fact greater than Gamma. Thus, for this user
Q.2.1.2 Cross-tabulation
Although the findings for both web user groups are not statistically significant,
the relationship between PEWU and WSUF for users with WSD/M experience –
between PEWU and WSUF for users who have no WSD/M experience. Thus H1A
285
Table Q2: 3x3 Cross-tabulation H1A:
Web Session Use Freqency (L/M/H) * Perceived Ease of Web Use (L/M/H) Crosstabulation
Consistent with the linear correlation coefficients reported in Table Q1, for users
who have experience with WSD/M, it seems that an extremely weak negative
(linear) relationship might exist between PEWU and WSUF. However as these
findings are not statistically significant, and the relationship detected is linear, not
WSUF. This was the case for web users with and web users without WSD/M
experience. Thus H1A is rejected and H0A is accepted for both user groups.
286
Q.2.2 H2A: PWU & WSUF
H2A proposes a positive (L/M/H) relationship between perceived web usefulness
H2A: Perceived web usefulness (PWU) will have a positive relationship with
H0A: Perceived web usefulness (PWU) will have no relationship with current web
As the variables examined in H2A are measured using different types of data, and
the distribution of variable scores for both PWU and WSUF are negative,
Spearmanʹs rho (rs) is the most suitable statistic to use (see Chapter 8). H2A is a
directional hypothesis and a 1-tailed test is conducted. This is reported for H2A in
Table Q3.
There is a weak positive relationship between PWU and WSUF for users with no
WSD/M experience, and this is significant. Thus, H2A is accepted and H0A
287
There is a positive relationship between PWU and WSUF for users with WSD/M
experience, and this is significant. Thus, H2A is accepted and H0A rejected for this
user group.
H2A is supported for both web user groups, indicating a positive relationship
exists between PWU and WSUF. For both groups this relationship is fairly weak,
however it is weakest for users without web site design and maintenance
experience.
use (PEWU) and current web session usage variety - situational (WSUVS).
H3A: Perceived ease of web use (PEWU) will have a curvilinear (L/H/M)
H0A: Perceived ease of web use (PEWU) will have no relationship with current web
The Chi-squared correlation coefficients for H3A are reported in Table Q4.
288
Table Q4: Chi-square Correlation Coefficients (Symmetric & Directional)
H3A: Perceived Ease of Web Use & Web Session Usage Variety - Situational
Web Site
Asymp.
Design and Approx.
Statistic Value Std.
Maintenance Sig.
Errora
Experience
No Nominal by Nominal Goodman and Kruskal’s tau .000 .001 .959b
Experience (non-linear) Cramer’s V .016 .979
Ordinal by Ordinal Kendall’s tau-b -.005 .033 .870
(Linear) Gamma -.017 .101 .870
No. of Valid Cases 900
Experience Nominal by Nominal Goodman and Kruskal’s tau .002 .003 .393b
(non-linear) Cramer’s V .035 .589
Ordinal by Ordinal Kendall’s tau-b .011 .029 .714
(Linear) Gamma .028 .077 .714
No. of Valid Cases 1177
a. Not assuming the null hypothesis
b. Based on chi-squared approximation
statistically significant association exists between PEWU and WSUVS for users
without WSD/M experience. This result is also consistent with the linear
correlation coefficients reported in the Table Q4 for this user group. To assess the
compared with Gamma. As Goodman and Kruskal’s tau is slightly greater than
Gamma, we can surmise that a very weak curvilinear relationship may be present.
When examining the non-linear correlation coefficients for users with WSD/M
experience in Table Q4, it is evident that a small association that is not statistically
significant exists between PEWU and WSUVS. This result is consistent with the
linear correlation coefficients reported in Table Q4. To assess the possible presence
compared. For users with WSD/M experience, Goodman and Kruskalʹs tau was
less than Gamma. Thus, for this user group a curvilinear relationship may not exist
289
Q.2.3.2 Cross-tabulation
Although the above findings reported for both web user groups are not statistically
confirm the nature of the relationship between PEWU and WSVS for users with no
WSD/M experience.
Table Q5 that a very weak u-shaped curvilinear relationship exists between PEWU
relationship being found, this result is not statistically significant and thus, H3A is
Table Q5 that a very weak inverted u-shape curvilinear relationship exists between
PEWU and WSUVS for users who have WSD/M experience. Despite a curvilinear
relationship being found, this result is not statistically significant and thus, H3A is
290
Table Q5: 3x3 Cross-tabulation - H3A: Perceived Ease of Web Use & Web
Session Usage Variety – Situational
Situational Variety - (L/M/H) * Perceived Ease of Web Use (L/M/H) Crosstabulation
H3A proposed that a curvilinear relationship would exist between PEWU and
evident that in fact a very weak curvilinear relationship might exist for both web
users groups. For web users with no WSD/M experience, a u-shaped relationship
may be present, and for web users with WSD/M experience an inverted u-shaped
significant for either user group, and the strength of the curvilinear relationships is
extremely weak. Thus, H3A is rejected and H0A is accepted for both groups.
use (PEWU) and current web session usage variety - motivational (WSUVMNO1).
291
H3B: Perceived ease of web use (PEWU) will have a curvilinear (L/H/M)
H0: Perceived ease of web use (PEWU) will have no relationship with current web
The Chi-squared correlation coefficients for H3B are reported in Table Q6.
H3B: Perceived Ease of Web Use & Web Session Usage Variety - Motive
Web Site
Asymp.
Design and Approx.
Statistic Value Std.
Maintenance Sig.
Errora
Experience
No Nominal by Nominal Goodman and Kruskal’s tau .005 .004 .078b
Experience (non-linear) Cramer’s V .064 .119
Ordinal by Ordinal Kendall’s tau-b .036 .032 .261
(Linear) Gamma .067 .060 .261
No. of Valid Cases 900
Experience Nominal by Nominal Goodman and Kruskal’s tau .000 .001 .954b
(non-linear) Cramer’s V .018 .948
Ordinal by Ordinal Kendall’s tau-b .015 .027 .589
(Linear) Gamma .027 .050 .589
No. of Valid Cases 1177
a. Not assuming the null hypothesis
b. Based on chi-squared approximation
association is seen between PEWU and WSUVMNO1 for users with no WSD/M
This result is consistent with the linear correlation coefficients reported in Table
292
Kruskal’s tau is compared with Gamma. As Goodman and Kruskal’s tau is less
From examining the non-linear correlation coefficients for users with WSD/M
consistent with the linear correlation coefficients reported in Table Q6. To assess
the possible presence of a curvilinear association, Goodman and Kruskalʹs tau and
Gamma are compared. For users with WSD/M experience it was identified that
Goodman and Kruskalʹs tau was less than Gamma and thus a curvilinear
Q.2.4.2 Cross-tabulation
the 3x3 cross-tabulation is warranted to further assess the nature of the relationship
moderate positive pattern was found to exist between PEWU and WSUVMNO1 in
Table Q7 for web users without WSD/M experience. Therefore, H3B is rejected and
weak positive pattern was found to exist between PEWU and WSUVMNO1 in
Table Q7 for web users with WSD/M experience. Therefore, H3B is rejected and
293
TableQ7: 3x3 Cross tabulation - H3B: Perceived Ease of Web Use & Web Session
WSUVMNO1 in H3B. However, this was not found for either user group. In fact,
accepted for users both with and without web site design and maintenance
experience.
294
H4A: Perceived web usefulness (PWU) will have a positive relationship with
H0: Perceived web usefulness (PWU) will have no relationship with current web
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
PWU and WSUVS for users with no WSD/M experience. Therefore, H4A is
between PWU and WSUVS for users with WSD/M experience. Thus H4A is
H4A: Perceived Web Usefulness & Web Session Usage Variety - Situational
Web Site Design and
Maintenance Situational
Experience (0/1) Variety (Sum)
No Experience Spearman’s Perceived Web Correlation Coefficient .061 *
Rho Usefulness Sig. (1-tailed) .035
(Sum) N 900
Experience Spearman’s Perceived Web Correlation Coefficient .054 *
Rho Usefulness Sig. (1-tailed) .033
(Sum) N 1177
*. Correlation significant at the .05 level (1-tailed)
H4A is supported for both web user groups, indicating a statistically significant
295
Q.2.6 H4B: PWU & WSUVMNO1
H4B proposes a positive relationship between perceived web usefulness (PWU)
H4B: Perceived web usefulness (PWU) will have a positive relationship with
H0B: Perceived web usefulness (PWU) will have no relationship with current web
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
H4B: Perceived Web Usefulness & Web Session Usage Variety - Motive
Web Site Design and
Maintenance No. of Use
Experience (0/1) Motivations
No Experience Spearman’s Perceived Web Correlation Coefficient .322 **
Rho Usefulness Sig. (1-tailed) .000
(Sum) N 900
Experience Spearman’s Perceived Web Correlation Coefficient .211 **
Rho Usefulness Sig. (1-tailed) .000
(Sum) N 1177
*. Correlation significant at the .05 level (1-tailed)
**. Correlation significant at the .01 level (1-tailed)
As shown in Table Q9, according to the Spearman rho correlation coefficient, there
WSUVMNO1 for users with no WSD/M experience. Thus H4B is accepted and
296
Web User Group B: With WSD/M Experience
As shown in Table Q9, according to the Spearman rho correlation coefficient there
WSUVMNO1. Thus H4B is accepted and H0B is rejected for this user group.
H4B is supported for both web user groups indicating that a positive relationship
relationship is weaker for users with web site design and maintenance experience
than those without this experience. Thus, as perceived usefulness of the web
increases users without web site design and maintenance experience have a
stronger tendency to use the web for a larger number of reasons than those with
use (PEWU) and current web session usage extent - breadth (WSUEB).
H5A: Perceived ease of web use (PEWU) will have a curvilinear (L/H/M)
H0: Perceived ease of web use (PEWU) will have no relationship with current web
297
Table Q10: Chi-squared Correlation Coefficients (Symmetric and Directional)
H5A: Perceived Ease of Web Use & Web Session Usage Extent - Breadth
Web Site Design
Asymp.
and Approx.
Statistic Value Std.
Maintenance Sig.
Errora
Experience
No Experience Nominal by Nominal Goodman and Kruskal’s tau 0.007 0.005 0.019b
(non-linear) Cramer’s V 0.079 0.023
Ordinal by Ordinal Kendall’s tau-b -0.085 0.033 0.009
(Linear) Gamma -0.167 0.062 0.009
No. of Valid Cases 900
Experience Nominal by Nominal Goodman and Kruskal’s tau 0.000 0.001 0.883b
(non-linear) Cramer’s V 0.037 0.512
Ordinal by Ordinal Kendall’s tau-b 0.015 0.028 0.580
(Linear) Gamma 0.030 0.055 0.580
No. of Valid Cases 1177
a. Not assuming the null hypothesis
b. Based on chi-squared approximation
A very weak positive association exists between PEWU and WSUEB for users with
and WSUEB, Goodman and Kruskal’s tau is compared with Gamma. As Goodman
and Kruskal’s tau is slightly greater than Gamma, a very weak curvilinear
evident that this relationship is not statistically significant. This result is consistent
with the linear correlation coefficients reported in the Table Q10. To assess the
are compared. It was identified that Goodman and Kruskalʹs tau was less than
Gamma and thus for a curvilinear relationship does not exist between PEWU and
298
Q.2.7.2 Cross-tabulation
nature of the relationship between PEWU and WSUEB, especially for users with no
WSD/M experience as a weak curvilinear relationship was identified for this group
Web Session Use Extent - Breadth (L/M/H) * Perceived Ease of Web Use (L/M/H) Crosstabulation
Table Q11 that a very weak u-shaped relationship might exist between PEWU and
299
Web User Group B: With WSD/M Experience
With respect to the nature of the relationship proposed in H5A, as evident in Table
Q11, no clear pattern exists between PEWU and WSUEB for web users with
WSD/M experience. Thus, H5A is rejected and H0A is accepted for this user group.
users with no web site design and/or maintenance experience. Thus, H5A is
accepted and H0A is rejected for this user group. However, no clear relationship
was identified between PEWU and WSUEB for users with WSD/M experience.
Thus, for this user group H5A is rejected and H0A accepted.
use (PEWU) and current web session usage extent - depth (WSUED).
H5B: Perceived ease of web use (PEWU) will have a curvilinear (L/H/M)
H0B: Perceived ease of web use (PEWU) will have no relationship with current web
The Chi-squared correlation coefficients for H5B are reported in Table Q12.
300
Table Q12: Chi-squared Correlation Coefficients (Symmetric and Directional)
H5B: Perceived Ease of Web Use & Web Session Usage Extent - Depth
Web Site Design
Asymp.
and Approx.
Statistic Value Std.
Maintenance Sig.
Errora
Experience
No Experience Nominal by Nominal Goodman and Kruskal’s tau .010 .006 .001b
(non-linear) Cramer’s V .082 .016
Ordinal by Ordinal Kendall’s tau-b -.065 .033 .048
(Linear) Gamma -.123 .062 .048
No. of Valid Cases 900
Experience Nominal by Nominal Goodman and Kruskal’s tau .001 .002 .492b
(non-linear) Cramer’s V .036 .558
Ordinal by Ordinal Kendall’s tau-b .021 .028 .461
(Linear) Gamma .039 .053 .461
No. of Valid Cases 1177
a. Not assuming the null hypothesis
b. Based on chi-squared approximation
significant very weak positive association exists between PEWU and WSUED for
users with no WSD/M experience. This result is not consistent with the linear
and Kruskal’s tau is compared with Gamma. As Goodman and Kruskal’s tau is
slightly greater than Gamma , a very weak curvilinear relationship may be present
By examining the non-linear correlation coefficients in Table Q12 for web users
association exists between PEWU and WSUED. This result is consistent with the
be curvilinear, Goodman and Kruskalʹs tau and Gamma are compared and it was
identified that Goodman and Kruskalʹs tau was less than Gamma. Thus for this
user group a curvilinear relationship does not exist between PEWU and WSUED.
301
Q.2.8.2 Cross-tabulation
nature of the relationship between PEWU and WSED for users with no WSD/M
Perceived Ease of Web Use & Web Session Usage Extent - Depth
Web Session Use Extent - Depth (L/M/H) * Perceived Ease of Web Use (L/M/H) Crosstabulation
Table Q13 that a weak positive relationship exists between PEWU and WSUED for
this user group. As this finding is not consistent with H5B, that a curvilinear
relationship will exist between PEWU and WSUED, H5B is rejected and H0B
302
Web User Group: Experience
Consistent with the results reported in Table Q12, it is evident in Table Q13 that no
clear pattern exists between PEWU and WSUED for users with WSD/M experience
and thus H5B is rejected and H0B is accepted for this user group.
PEWU and WSUED for users with no WSD/M experience as opposed to the
of Spearman’s Rho (rs) correlation coefficient for this web user group further
relationship between PEWU and WSUED. As this result does not support the
relationship hypothesised, H5B was rejected and H0B was accepted for this user
group.
H5B: Perceived Ease of Web Use & Web Session Usage Extent - Depth
Web Site Design and Web Session
Maintenance Usage Extent –
Experience (0/1) Depth (Sum)
No Experience Spearman’s Perceived Ease of Correlation Coefficient .372 **
Rho Web Use Sig. (1-tailed) .000
(Sum) N 900
Experience Spearman’s Perceived Ease of Correlation Coefficient .211 **
Rho Web Use Sig. (1-tailed) .000
(Sum) N 1177
*. Correlation significant at the .05 level (1-tailed)
**. Correlation significant at the .01 level (1-tailed)
In comparison, for users with web site design and/or maintenance experience, no
statistically significant or clear relationship was found to exist between PEWU and
WSUED for this user group. Thus, H5B was rejected and H0B was accepted for this
user group.
303
Q.2.9 H5C: PEWU & WSUEDUR
H5C proposes a curvilinear (L/H/M) relationship between perceived ease of web
use (PEWU) and current web session usage extent - duration (WSUEDUR).
H5C: Perceived ease of web use (PEWU) will have a curvilinear (L/H/M)
H0C: Perceived ease of web use (PEWU) will have no relationship with current web
The Chi-square correlation coefficients for H5C are reported in Table Q15.
H5C: Perceived Ease of Web Use & Web Session Usage Extent - Duration
Web Site Design
Asymp.
and Approx.
Statistic Value Std.
Maintenance Sig.
Errora
Experience
No Experience Nominal by Nominal Goodman and Kruskal’s tau .001 .001 .897b
(non-linear) Cramer’s V .034 .726
Ordinal by Ordinal Kendall’s tau-b -.005 .032 .866
(Linear) Gamma -.011 .063 .866
No. of Valid Cases 900
Experience Nominal by Nominal Goodman and Kruskal’s tau .001 .002 .700
(non-linear) Cramer’s V .026 .815
Ordinal by Ordinal Kendall’s tau-b -.020 .028 .459
(Linear) Gamma -.038 .051 .459
No. of Valid Cases 1177
a. Not assuming the null hypothesis
b. Based on chi-squared approximation
No clear association exists between PEWU and WSUEDUR for users without
significant. This result is consistent with the linear correlation coefficients reported
304
in Table Q15. To assess the possibility of a curvilinear association Goodman and
Kruskal’s tau is compared with Gamma. As Goodman and Kruskal’s tau is slightly
A small positive association exists between PEWU and WSUEDUR, but this
association is not statistically significant. This result is not consistent with the linear
association, Goodman and Kruskalʹs tau and Gamma are compared identifying
that Goodman and Kruskalʹs tau was slightly greater than Gamma. Thus for this
user group a very weak curvilinear relationship may exist between PEWU and
WSUEDUR.
Q.2.9.2 Cross-tabulation
nature of the relationship between PEWU and WSUEDUR, for both user groups as
With respect to the nature of the relationship proposed in H5C, as evident in Table
Q16, no clear pattern is exists between PEWU and WSUEDUR. Thus H5C is
With respect to the nature of the relationship proposed in H5C, as evident in Table
WSUEDUR for web users with WSD/M experience. However, this is not
statistically significant - H5C is rejected and H0 is accepted for this user group.
305
Table Q16: 3x3 Cross-tabulation - H5C: Perceived Ease of Web Use & Web
For web users with no web site design experience, no statistically significant
pattern was found to occur between PEWU and WSUEDUR. For those with web
site design and maintenance experience, a very weak u-shaped relationship was
found to exist between PEWU and WSUEDUR, however this association was not
statistically significant. Thus, H5C was rejected and H0C accepted for both user
groups.
306
H6A: Perceived web usefulness (PWU) will have a positive relationship with
H0A: Perceived web usefulness (PWU) will have no relationship with current web
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
H6A: Perceived Web Usefulness & Web Session Usage Extent – Breadth
Web Site Design and Web Session
Maintenance Usage Extent –
Experience (0/1) Breadth (Sum)
No Experience Spearman’s Perceived Web Correlation Coefficient -.086 **
Rho Usefulness Sig. (1-tailed) .005
(Sum) N 900
Experience Spearman’s Perceived Web Correlation Coefficient -.056 *
Rho Usefulness Sig. (1-tailed) .028
(Sum) N 1177
*. Correlation significant at the .05 level (1-tailed)
**. Correlation significant at the .01 level (1-tailed)
between PWU and WSUEB for users with no WSD/M experience. Thus H6A is
between PWU and WSUEB for users with WSD/M experience. Thus H6A is
In summary, H6A is not supported for either web user group as the results
307
and WSUEB, a statistically significant and extremely weak negative relationship
exists. Thus H6A is rejected and H0A accepted for both web user groups.
H6B: Perceived web usefulness (PWU) will have a positive relationship with
H0B: Perceived web usefulness (PWU) will have no relationship with current web
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
PWU and WSUED for users with no WSD/M experience. Thus H6B is accepted
WSUED for users with WSD/M experience. Thus H6B is accepted and H0B is
308
Table Q18: Nonparametric Correlation Coefficient: Spearman Rho (rs)
H6B: Perceived Web Usefulness & Web Session Usage Extent Depth
Web Site Design and Web Session
Maintenance Usage Extent –
Experience (0/1) Depth (Sum)
No Experience Spearman’s Perceived Web Correlation Coefficient .360 **
Rho Usefulness Sig. (1-tailed) .000
(Sum) N 900
Experience Spearman’s Perceived Web Correlation Coefficient .190 **
Rho Usefulness Sig. (1-tailed) .000
(Sum) N 1177
**. Correlation significant at the .01 level (1-tailed)
H6B is supported for both web user groups as the results indicate a statistically
significant positive relationship exists between PWU and WSUED for both web
user groups. However this relationship is stronger for web users without WSD/M
experience than it is for those users with this experience. Thus H6B is accepted and
H6C: Perceived web usefulness (PWU) will have a positive relationship with
H0C: Perceived web usefulness (PWU) will have no relationship with current web
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
309
Table Q19: Nonparametric Correlation Coefficient: Spearman Rho (rs)
H6C: Perceived Web Usefulness & Web Session Usage Extent - Duration
Web Site Design and Duration of Web
Maintenance Session Use
Experience (0/1) (Sum)
No Experience Spearman’s Perceived Web Correlation Coefficient .176 **
Rho Usefulness Sig. (1-tailed) .000
(Sum) N 900
Experience Spearman’s Perceived Web Correlation Coefficient .112 **
Rho Usefulness Sig. (1-tailed) .000
(Sum) N 1177
**. Correlation significant at the .01 level (1-tailed)
WSUEDUR for users with no WSD/M experience. Thus H6C is accepted and H0C
WSUEDUR for users with WSD/M experience. Thus H6C is accepted and H0C is
H6C is supported for both web user groups. The results indicate a statistically
significant positive relationship exists between PWU and WSUEDUR for both user
310
Table Q20: Web User Group A: RQ1 Hypothesis Result Summary
the web and a person’s perceived usefulness of the web? To examine this question
311
Q.3.1 H7A: ACPWK & PWU
H7A proposes a curvilinear relationship between actual common procedural web
H7A: Actual common procedural web knowledge (ACPWK) will have a curvilinear
Chi-square based correlation coefficients for H7A are reported in Table Q22.
H7A: Actual Common Procedural Web Knowledge & Perceived Web Usefulness
Web Site Design and Asymp.
Approx.
Maintenance Statistic Value Std.
Sig.
Experience Errora
No Experience Nominal by Nominal Goodman and Kruskal’s tau .004 .004 .177b
(non-linear) Cramer’s V .046 .421
Ordinal by Ordinal Kendall’s tau-b -.013 .032 .692
(Linear) Gamma -.027 .069 .692
No. of Valid Cases 900
Experience Nominal by Nominal Goodman and Kruskal’s tau .003 .003 .142b
(non-linear) Cramer’s V .042 .378
Ordinal by Ordinal Kendall’s tau-b -.030 .028 .291
(Linear) Gamma -.105 .100 .291
No. of Valid Cases 1177
a. Not assuming the null hypothesis
b. Based on chi-squared approximation
A negligible association exists between ACPWK and PWU for users with no
not consistent with the linear correlation coefficients reported in Table Q22 as they
312
compared with Gamma and as Goodman and Kruskalʹs tau is slightly greater than
ACPWK and PWU. This result is not consistent with the linear correlation
Goodman and Kruskalʹs tau and Gamma are compared and it was identified that
Goodman and Kruskalʹs tau was greater than Gamma. Thus, for this user group a
Q.3.1.2 Cross-tabulation
the 3x3 cross-tabulation is warranted to further assess the nature of the relationship
With respect to H7A, it is evident from Table Q23 that a very weak u-shaped
relationship exists between ACPWK and PWU. As this result is not statistically
significant, H7A is rejected and H0A is accepted for this user group.
With respect to H7A, it is evident from Table Q23 that a very weak u-shaped
relationship might exist between ACPWK and PWU for this user group. As this
relationship is not statistically significant, H7A is accepted and H0A is rejected for
313
Table Q23: 3x3 Cross-tabulation - H7A: Actual Common Procedural Web
It is evident that for both user groups a statistically insignificant but very weak u-
shaped relationship was found between ACPWK and PWU. Thus, H7A was not
supported for either user group. (Reject H7A and Accept H0).
314
H8A: Actual common declarative web knowledge (ACDWK) will have a positive
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
Usefulness
Web Site Design and Perceived Web
Maintenance Usefulness
Experience (0/1) (Sum)
No Experience Spearman’s Actual Common Correlation Coefficient .221 **
Rho Declarative Web Sig. (1-tailed) .000
Knowledge (Sum) N 900
Experience Spearman’s Actual Common Correlation Coefficient .092 **
Rho Declarative Web Sig. (1-tailed) .000
Knowledge (Sum) N 1177
**. Correlation significant at the .01 level (1-tailed)
and PWU for users with no WSD/M experience. Thus, H8A is supported for this
ACDWK and PWU for users with WSD/M experience. Thus, H8A is supported for
315
Q.3.2.2 H8A: Summary
between ACDWK and PWU for both user groups. It is also evident that this
relationship is stronger for those without web site design and maintenance
experience than those with this experience. Thus, H8A is accepted and H0A is
H9A: Actual specialised procedural web knowledge (ASPWK) will have a positive
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
Usefulness
Web Site Design and Perceived Web
Maintenance Usefulness
Experience (0/1) (Sum)
No Experience Spearman’s Actual Specialised Correlation Coefficient .182 **
Rho Procedural Web Sig. (1-tailed) .000
Knowledge (Sum) N 900
Experience Spearman’s Actual Specialised Correlation Coefficient .049 *
Rho Procedural Web Sig. (1-tailed) .000
Knowledge (Sum) N 1177
*. Correlation significant at the .05 level (1-tailed)
**. Correlation significant at the .01 level (1-tailed)
316
Web User Group A: No WSD/M Experience
and PWU for users with no WSD/M experience. Thus, H9A is supported for this
ASPWK and PWU for users with WSD/M experience. Thus, H9A is supported and
for both groups. It is also evident that this relationship is stronger for those without
web site design and maintenance experience than those with this experience. Thus,
H10A: Actual specialised declarative web knowledge (ASDWK) will have a positive
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
317
Table Q26: Nonparametric Correlation Coefficient: Spearman Rho (rs)
Usefulness
Web Site Design and Perceived Web
Maintenance Usefulness
Experience (0/1) (Sum)
No Experience Spearman’s Actual Specialised Correlation Coefficient .156 **
Rho Declarative Web Sig. (1-tailed) .000
Knowledge
N 900
(Sum)
Experience Spearman’s Actual Specialised Correlation Coefficient .038 N.S
Rho Declarative Web Sig. (1-tailed) .096
Knowledge
N 1177
(Sum)
*. Correlation significant at the .05 level (1-tailed)
**. Correlation significant at the .01 level (1-tailed)
N.S = Not Statistically Significant.
and PWU for users with no WSD/M experience. Thus, H10A is supported and
between ASDWK and PWU for users with WSD/M experience. Thus, H10A is
and PWU for users without web site design and maintenance experience. In
318
H11A: Perceived procedural web knowledge (SWPK) will have a positive (+)
H0A: Perceived Procedural web knowledge (SWPK) will have no relationship with
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
SWPK and PWU for users with no WSD/M experience. Thus, H11A is supported
SWPK and PWU for users with WSD/M experience. Thus, H11A is supported and
319
Q.3.5.2 H11A: Summary
H11A is supported for both web user groups indicating that a statistically
significant moderate positive relationship exists between SWPK and PWU. This
relationship is slightly stronger for those users without WSD/M experience, than
H0: Perceived declarative web knowledge (SWDK) will have no relationship with
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
320
Web User Group A: No WSD/M Experience
SWDK and PWU for users with no WSD/M experience. Thus, H12A is supported
SWDK and PWU for users with WSD/M experience. Thus, H12A is supported and
H12A is supported for both user groups indicating that a statistically significant
moderate positive relationship exists between SWDK and PWU. In addition, this
relationship is stronger for those without WSD/M experience, than those with
experience.
H13A: Perceived overall web knowledge (SWOK) will have a positive (+)
H0: Perceived overall web knowledge (SWOK) will have no relationship with
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
321
Table Q29: Nonparametric Correlation Coefficient: Spearman Rho (rs)
SWOK and PWU for users with no WSD/M experience. Thus, H13A is supported
SWOK and PWU for users with WSD/M experience. Thus, H13A is supported and
positive relationship exists between SWOK and PWU. In addition, this relationship
is stronger for those without WSD/M experience, than those with this experience.
In summary, as shown below in Table Q30, 6 out of 7 hypotheses are supported for
322
Table Q30: Web User Group A: RQ2 Hypothesis Result Summary
the web and a person’s perceived ease of web use? To examine this question more
between a users knowledge content of the web and their perceived ease of web
323
H14A: Actual common procedural web knowledge of the web will have a positive
H0: Actual common procedural web knowledge of the web will have no
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
Usefulness
Web Site Design and Perceived Ease
Maintenance of Web Use
Experience (0/1) (Sum)
No Experience Spearman’s Actual Common Correlation Coefficient .233 **
Rho Procedural Web Sig. (1-tailed) .000
Knowledge (Sum) N 900
Experience Spearman’s Actual Common Correlation Coefficient .150 **
Rho Procedural Web Sig. (1-tailed) .000
Knowledge (Sum) N 1177
**. Correlation significant at the .01 level (1-tailed)
and PEWU for users with no WSD/M experience. Thus, H14A is accepted and H0A
and PEWU for users with WSD/M experience. Thus, H14A is accepted and H0A is
324
Q.4.1.2 H14A: Summary
H14A is supported for both user groups, indicating that a statistically significant
weak positive relationship exists between ACPWK and PEWU. This relationship is
stronger for users without web site design and maintenance experience than those
H15A: Actual common declarative web knowledge (ACDWK) of the web will have a
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
325
Web User Group A: No WSD/M Experience
and PEWU for users with no WSD/M experience. Thus, H15A is supported and
and PEWU for users with WSD/M experience. Thus, H15A is supported and H0A
H15A is supported for both user groups, indicating that a statistically significant
weak positive relationship exists between ACDWK and PEWU. This relationship is
stronger for users without web site design and maintenance experience than those
H16A: Actual specialised procedural web knowledge (ASPWK) will have a positive
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
326
Web User Group A: No WSD/M Experience
and PEWU for users with no WSD/M experience. Thus, H16A is supported and
ASPWK and PEWU for users with web WSD/M experience. Thus, H16A is
H16A is supported for both user groups, indicating that a statistically significant
stronger for users without web site design and maintenance experience than those
327
H17A: Actual specialised declarative web knowledge (ASDWK) will have a positive
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
ASDWK and PEWU for users with no WSD/M experience. Thus, H17A is
ASDWK and PEWU for users with WSD/M experience. Thus, H17A is rejected and
328
Q.4.4.2 H17A: Summary
H17A is supported only for web users without web site design and maintenance
relationship exists between ASDWK and PEWU for this group. No statistically
significant relationship was found to exist between ASDWK and PEWU for users
with experience. Thus for web users without experience H17A is accepted and
H0A rejected, and for web users with experience H17A is rejected and H0A is
accepted.
H0A: Perceived procedural web knowledge (SWPK) will have no relationship with
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
H18A: Perceived Procedural Web Knowledge & Perceived Ease of Web Use
Web Site Design and Perceived Ease
Maintenance of Web Use
Experience (0/1) (Sum)
No Experience Spearman’s Perceived Procedural Correlation Coefficient .658 **
Rho Web Knowledge Sig. (1-tailed) .000
(Sum) N 900
Experience Spearman’s Perceived Procedural Correlation Coefficient .463 **
Rho Web Knowledge Sig. (1-tailed) .000
(Sum) N 1177
**. Correlation significant at the .01 level (1-tailed)
329
Web User Group A: No WSD/M Experience
and PEWU for users with no WSD/M experience. Thus, H18A is supported and
SWPK and PEWU for users with WSD/M experience. Thus, H18A is supported
positive relationship exists between SWPK and PEWU for both groups. This
relationship is stronger for users without web site design and maintenance
H0A: Perceived declarative web knowledge (SWDK) will have no relationship with
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
330
Table Q37: Nonparametric Correlation Coefficient: Spearman Rho (rs)
H19A: Perceived Declarative Web Knowledge & Perceived Ease of Web Use
Web Site Design and Perceived Ease
Maintenance of Web Use
Experience (0/1) (Sum)
No Experience Spearman’s Perceived Declarative Correlation Coefficient .629 **
Rho Web Knowledge Sig. (1-tailed) .000
(Sum) N 900
Experience Spearman’s Perceived Declarative Correlation Coefficient .418 **
Rho Web Knowledge Sig. (1-tailed) .000
(Sum) N 1177
**. Correlation significant at the .01 level (1-tailed)
and PEWU for users with no WSD/M experience. Thus, H19A is supported and
SWDK and PEWU for users with WSD/M experience. Thus, H19A is supported
positive relationship exists between SWDK and PEWU for both groups. This
relationship is stronger for users without web site design and maintenance
331
H20A: Perceived overall web knowledge (SWOK) will have a positive relationship
H0A: Perceived overall web knowledge (SWOK) will have no relationship with
Spearmanʹs rho (rs) is the most suitable statistic to use here. This is reported for
H20A: Perceived Overall Web Knowledge & Perceived Ease of Web Use
Web Site Design and Perceived Ease
Maintenance of Web Use
Experience (0/1) (Sum)
No Experience Spearman’s Perceived Overall Correlation Coefficient .535 **
Rho Web Knowledge Sig. (1-tailed) .000
(Sum) N 900
Experience Spearman’s Perceived Overall Correlation Coefficient .366 **
Rho Web Knowledge Sig. (1-tailed) .000
(Sum) N 1177
**. Correlation significant at the .01 level (1-tailed)
and PEWU for users with no WSD/M experience. Thus, H20A is supported and
SWOK and PEWU for users with WSD/M experience. Thus, H20A is supported
332
Q.4.7.2 H20A: Summary
positive relationship exists between SWOK and PEWU for both groups. This
relationship is stronger for users without web site design and maintenance
In summary, 7 out of 7 hypotheses are supported for users with no web site design
In summary, 6 out of 7 hypotheses are supported for users with experience in web
333
Q.5 EMPIRICAL BIVARIATE RESULT SUMMARY
For the first research question investigating the relationship between a user’s
perceptions of the web and current web session usage it was found that of the 6
hypothesised curvilinear relationships only one was supported for users with no
WSD/M experience (PEWU & WSUEB). Furthermore, all 6 were rejected for users
with WSD/M experience. In essence, the relationships were either linear or not
linear relationships were supported for both web users with and without WSD/M
experience. This gives strong support for the hypothesized positive relationship
For the second research question investigating the relationship between a user’s
knowledge of the web and levels of perceived usefulness of the web, 6 out of the 7
hypotheses proposed were supported for users with no WSD/M experience. This
perceived web usefulness. For users with this experience, 5 of the 7 hypotheses
were supported.
For the third and final research question that investigates the relationship between
a user’s knowledge of the web and levels of perceived ease of web use, 6 out of the
7 hypotheses were supported for both web users with and without WSD/M
presented here in this appendix provide further validation for the results of the
334
APPENDIX R: NONPARAMETRIC CORRELATION COEFFICIENTS: SPEARMAN RHO
WSUF WSUVS WSUVMNO1 WSUEB WSUED WSUEDUR PEWU PWU ACPWK ACDWK ASPWK ASDWK SWOK SWPK SWDK
WSUF 1.00 .18** .24** .02 .13** .04** .14** .11** .12** .19** .22** .28** .34** .34** .34**
WSUVS 1.00 .25** -.03 .12** .06** .09** .07** .16** .18** .20** .20** .22** .20** .22**
WSUVMNO1 1.00 .08** .30** .24** .27** .27** .27** .30** .32** .34** .37** .37** .38**
WSUEB 1.00 .06** .03 -.06** -.07** .02 .01 .01 .01 -.07** -.06** -.06**
WSUED 1.00 .24** .31** .28** .21** .17** .19** .21** .33** .36** .35**
WSUEDUR 1.00 .19** .15** .16** .11** .13** .15** .22** .24** .23**
PEWU 1.00 .74** .22** .24** .20** .17** .47** .56** .52**
PWU 1.00 .15** .17** .14** .12** .36** .41** .38**
ACPWK 1.00 .45** .47** .48** .40** .41** .42**
ACDWK 1.00 .59** .66** .50** .50** .50**
ASPWK 1.00 .69** .52** .51** .52**
ASDWK 1.00 .57** .54** .58**
SWOK 1.00 .89** .89**
SWPK 1.00 .88**
SWDK 1.00
* Correlation significant at the .05 level (2-tailed)
** Correlation significant at the .01 level (2-tailed)
335
Table R2: Nonparametric Correlation Coefficients: Spearman Rho - Users with (n=1177) and Users without (n=900) WSD/M Experience
WSUF WSUVS WSUVMNO1 WSUEB WSUED WSUEDUR PEWU PWU ACPWK ACDWK ASPWK ASDWK SWOK SWPK SWDK
WSUF 1.0 -.01 .07* -.01 .12** -.07* .09** .06 -.04 .06 .06 .14** .20** .23** .20**
WSUVS 1.00 .24** -.02 .13** .10** .06 .06 .17** .22** .21** .19** .13** .10** .09**
WSUVMNO1 1.00 .10** .36** .27** .30** .32** .28** .28** .32** .29** .25** .28** .24**
WSUEB 1.00 .07 .03 -.06 -.09* .03 .03 .04 .06 -.10** -.10** -.12**
WSUED 1.00 .28** .37** .36** .24** .19** .24** .22** .35** .38** .37**
WSUEDUR 1.00 .21** .18** .20** .14** .14** .12** .15** .18** .16**
With No PEWU 1.00 .76** .23** .24** .21** .17** .54** .66** .63**
WSD/M PWU 1.00 .19** .22** .18** .16** .40** .48** .45**
Experience ACPWK 1.00 .51** .51** .5** .34** .34** .33**
ACDWK 1.00 .64** .69** .42** .43** .40**
ASPWK 1.00 .70** .45** .42** .40**
ASDWK 1.00 .45** .41** .40**
SWOK 1.00 .85** .85**
SWPK 1.00 .85**
SWDK 1.00
WSUF 1.00 .23** .26** .04 .08 .07* .10** .11** .11** .13** .18** .22** .28** .27** .28**
WSUVS 1.00 .18** -.03 .03 -.03 .03 .05 .06 .01 .04 .06 .12** .11** .13**
WSUVMNO1 1.00 .08** .19** .19** .18** .21** .16** .19** .17** .22** .30** .29** .31**
WSUEB 1.00 .08** .07 -.05 -.06 .03 .03 .03 .01 -.04 -.03 -.04
WSUED 1.00 .17** .21** .19** .09** .04 .02 .06** .20** .25** .21**
WSUEDUR 1.00 .14** .11** .08** .02 .04 .09** .19** .21** .20**
With PEWU 1.00 .73** .15** .15** .08** .04 .37** .46** .42**
WSD/M PWU 1.00 .10** .10** .05 .04 .33** .37** .34**
Experience ACPWK 1.00 .24** .29** .28** .26** .31** .30**
ACDWK 1.00 .37** .46** .32** .31** .31**
ASPWK 1.00 .51** .33** .33** .34**
ASDWK 1.00 .39** .38** .43**
SWOK 1.00 .80** .85**
SWPK 1.00 .84**
SWDK 1.00
* Correlation significant at the .05 level (2-tailed); ** Correlation significant at the .01 level (2-tailed)
336
APPENDIX S: ANOVA REPORTED MEAN SCORES
Table S1: Mean Scores: Effect of Level of Actual [4] and Perceived [3] Web
Table S2: Mean Scores: Effect of Level of Actual [4] and Perceived [3] Web
337