omputers in
CComputers in Human Behavior 22 (2006) 427–447
www.elsevier.com/locate/comphumbeh
Human Behavior
Adding contextual specificity to thetechnology acceptance model
Daniel J. McFarland a,*, Diane Hamilton b
a Management Information Systems, College of Business, Rowan University, 201 Mullica
Hill Road, Glassboro, NJ 08028, USAb Management Information Systems, College of Business, Rowan University, 201 Mullica Hill Road,
Glassboro, NJ 08028, USA
Available online 11 November 2004
Abstract
This paper examines the influence of contextual specificity when describing technology
acceptance. Social cognitive theory provides the basis for adding several independent variables
(computer anxiety, prior experience, other�s use, organizational support, task structure, and
system quality) and one intervening variable (computer-efficacy) to the technology acceptance
model (TAM). This extended model was tested using a mail survey and the results are tabu-
lated using partial least squares. The results show that system usage is strongly influenced by
computer anxiety, prior experience, other�s use, organizational support, task structure, system
quality, and perceived usefulness. In addition, perceived usefulness is the strongest mediator in
determining system usage.
� 2004 Elsevier Ltd. All rights reserved.
Keywords: Technology acceptance; Self-efficacy; Computer efficacy; System usage; Partial least squares
1. Introduction
Researchers and practitioners alike strive to understand individuals� unwillingnessto accept systems that appear to promise substantial benefits. Davis, Bagozzi, and
0747-5632/$ - see front matter � 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.chb.2004.09.009
* Corresponding author. Tel.: +1 856 256 5426; fax: +1 856 256 4439.
E-mail address: [email protected] (D.J. McFarland).
428 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
Warshaw (1989, p. 587) conclude that, ‘‘understanding why people accept or reject
computers has proven to be one of the most challenging issues in IS research.’’ This
lack of understanding continues despite recent improvements in application usability
and ease of use (Hasan, 2003). With employees seemingly accepting and rejecting
systems unsystematically, many organizations are failing to achieve the benefitspromised to them by software manufacturers.
The technology acceptance model (TAM) is one of the most widely used models
for describing IT usage behaviors (Igbaria, Guimaraes, & Davis, 1995). The TAM
asserts that IT behaviors are based largely on users� perceptions of a system�s easeof use and usefulness. While the model ‘‘has been empirically proven to have high
validity’’ (Chau, 1996, p.187), it ‘‘only supplies general information on users� opin-ions about a system’’ (Mathieson, 1991, p. 173). Similarly, user evaluation measures,
such as perceived ease of use and perceived usefulness, encompass many differentuser meanings and theoretical constructs (Chau, 1996; Moore & Benbasat, 1991; Se-
gars & Grover, 1994). Goodhue (1995, p. 1828) concludes that ‘‘there are so many
different underlying constructs, it is probably not possible to develop a single general
theoretical basis for user evaluations.’’ Cognitive psychologists support arguments
opposing the mental averaging of an activity domain. ‘‘Combining diverse attributes
into a single index creates confusions about what is actually being measured and how
much weight is given to particular attributes in the forced summary judgment’’ (Ban-
dura, 1997, p. 11).
2. Technology acceptance models
Igbaria et al. (1995) conclude that the TAM is one of the simplest, easiest to use,
and most powerful computer usage models. Similarly, Chau (1996) described the
TAM as one of the most influential of the over 20 computer usage models that Saga
and Zmud (1994) reviewed.The theoretical foundation for the TAM is Fishbein and Ajzen�s (1975) theory of
reasoned action (TRA). ‘‘The TAM adapted the generic TRA model to the particu-
lar domain of user acceptance of computer technology, replacing the TRA�s attitu-dinal determinants, derived separately for each behavior, with a set of two variables
perceived ease of use (PEOU) and perceived usefulness (PERUSE)’’ (Igbaria et al.,
1995, p. 88). PEOU is defined as ‘‘the degree to which a person believes using a par-
ticular system would be free of effort’’ and PERUSE is ‘‘the degree to which a person
believes that using a particular system would enhance his or her job performance’’(Davis, 1989, p. 320). Fig. 1 shows a common operationalization of the TAM (Igbaria
et al., 1995). The TAM is also based on ‘‘the cost-benefit paradigm from behavioral
decision theory’’ (Davis, 1989, p. 321). In general, the cost-benefit paradigm posits
that human behavior is based on a person�s cognitive tradeoff between the effort re-
quired to perform an action and the consequences of completing the action (Jar-
venpaa, 1989). Within MIS, the TAM asserts that a person will use an application
if the performance benefits outweigh the effort of using the application (Davis,
1989). Davis (1989) assesses performance benefits by measuring a person�s anticipated
External Variables
Perceived Ease of Use
Perceived Usefulness
System Usage
Fig. 1. Simplified technology acceptance model.
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 429
consequences of using the system (a.k.a., PERUSE), and effort by assessing a per-
son�s belief that using an application is free of effort (a.k.a., PEOU). Several empir-
ical studies demonstrate the efficiency, effectiveness and validity of the TAM and the
superiority of the TAM to the TRA (Adams, Nelson, & Todd, 1992; Chau, 1996;
Davis, 1986; Davis et al., 1989; Hendrickson, Glorfeld, & Cronan, 1994; Hubona
& Cheney, 1994; Igbaria et al., 1995; Mathieson, 1991; Segars & Grover, 1994).
Although the TAM ‘‘has been empirically proven to have high validity’’ (Chau,1996, p. 187), the model explains only a fraction of the observed IT usage variance.
Davis et al. (1989) study showed one of the highest explained variances. These
researchers were able to explain between 45% and 57% of the variance associated
with volunteer students� intentions to use a non-required word-processing applica-
tion. However, studies with similar objectives that investigated usage in the field ex-
plained much less variance. The percent variance explained in field studies varies
between 4% (Adams et al., 1992) and 45% (Igbaria et al., 1995). Consider PEOU;
it is unclear if these evaluations are based on the ease with which users interact withthe system or if it is based on the ease with which users interact with the task them-
selves. For example, when attempting to solve a challenging statistical problem, peo-
ple may not consider the system�s usability if they believe that they do not possess the
requisite statistical expertise. Consequently, without regard to system differences, all
statistical applications may receive equally poor PEOU evaluations. Similarly, PER-
USE measures may evaluate task usefulness rather than technological usefulness. To
illustrate, consider a system designed to monitor critically ill patients in an emer-
gency room. Without considering the system�s effectiveness, people may evaluate itas being very useful.
3. Self-efficacy and computer-efficacy
A promising addition to behavioral research is the concept of self-efficacy (Com-
peau & Higgins, 1995; Gist & Mitchell, 1992; Hasan, 2003). ‘‘Self-efficacy refers to
the judgments an individual makes about his or her capability to mobilize the moti-vation, cognitive resources, and course of action needed to orchestrate future
430 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
performance on a specific task’’ (Martocchio & Dulebohn, 1994, p. 358). This defi-
nition emphasizes three critical characteristics of self-efficacy. First, self-efficacy is
one�s belief in his or her capability to produce an outcome rather than an assessment
regarding the impacts of the outcome. Next, self-efficacy�s focus is on overall results
rather than component level skills. Finally, self-efficacy is a judgment of �what onecan do� in the future rather than an assessment of �what one has done� in the past.
Empirical studies have shown self-efficacy to be a distinctive, valid, and significant
construct (Bandura, 1997; Gist & Mitchell, 1992; Igbaria & Ivari, 1995).
To predict human behavior, people�s belief that their actions contribute to suc-
cess, and a self-assessment of their capabilities to accomplish the activity should
both be considered (Bandura, 1997). Although these beliefs are separate, they
are not independent of one another. ‘‘Beliefs that outcomes are determined by
one�s own behavior can be either demoralizing or empowering, depending onwhether or not one believes one can produce the required behavior’’ (Bandura,
1997, p. 20). In combining these beliefs, people develop self-images of future suc-
cess or failure states (Markus & Nurius, 1986). These self-conceptualizations of
the future serve to guide and motivate behavior. As a result, peoples� beliefs in
their causative capabilities ‘‘influence the way they think, feel, motivate themselves
and act’’ (Bandura, 1995, p. 2). ‘‘Given the importance of self-efficacy for predict-
ing and improving work performance and behavior’’ (Igbaria & Ivari, 1995, p.
588), several investigators argue the need for further research to examine the roleof self-efficacy in computing behavior (Gist & Mitchell, 1992; Gist et al., 1989;
Igbaria & Ivari, 1995).
In borrowing self-efficacy from cognitive psychology, MIS researchers have
defined computer-efficacy as one�s general belief that he/she is capable of putting
computer technologies to use (Compeau & Higgins, 1995; Vankatesh & Davis,
1996). Fig. 2 graphically depicts Compeau and Higgins� computer-efficacy model.
Empirical studies show computer-efficacy influencing: technology adoption (Burk-
hardt & Brass, 1990; Igbaria & Ivari, 1995), system usage (Compeau & Higgins,1995; Igbaria & Ivari, 1995), system ease of use perceptions (Vankatesh & Davis,
1996), affective states (Compeau & Higgins, 1995; Igbaria & Ivari, 1995), and com-
puter training (Gist, Schwoerer, & Rosen, 1989; Hill, Smith, & Mann, 1987; Webster
& Martocchio, 1992).
Davis (1989) theorized that computer-efficacy was distinctive from PERUSE and
PEOU and subsequent empirical research has demonstrated this distinctiveness
(Igbaria & Ivari, 1995; Vankatesh & Davis, 1996). Additionally, research findings
have demonstrated computer-efficacy�s significance in explaining computing behav-ior (Fenech, 1998; Igbaria & Ivari, 1995; Vankatesh & Davis, 1996). Computer-
efficacy is defined as one�s belief that he or she is capable of using a computer to
complete a task, without regard to the task�s difficulty or consequences. For example,
while not improving my job performance, and not being an easy task, I am quite
confident in my ability to use C++ to develop a simulation depicting my monthly
earning and spending habits.
Although the results of existing computer-efficacy studies are encouraging, like
the TAM, computer-efficacy has been defined as a general construct, that is, one�s
Encouragement by others
Others’ Use
Support
Computer-efficacy
Outcome expectation
User affective state
User anxiety
System usage
Fig. 2. Compeau & Higgins� computer-efficacy model.
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 431
self-assessment of his/her ‘‘abilities to use information and computer technologies ingeneral’’ (Vankatesh & Davis, 1996, p. 452). Insofar as computer-efficacy is ‘‘system-
independent’’ (Vankatesh & Davis, 1996, p. 473), these models suggest that if a per-
son is confident using one application to complete a particular task, he or she will be
confident using any application to complete any task. However, Goodhue (1995)
concludes that specific task and system quality significantly affect user�s evaluationsof systems. Similarly, while investigators have suggested that system quality may im-
pact computer-efficacy, they did not investigate or explore the potential relationship
(e.g., Compeau & Higgins, 1995; Vankatesh & Davis, 1996).Empirical evidence suggests that general indices of efficacy ‘‘bear little to no rela-
tion either to efficacy beliefs related to particular activity domains or to behavior. . .When global efficacy beliefs are related to performance, evidence suggests that par-
ticularized efficacy beliefs account for the relation. Global beliefs lose their predic-
tiveness when the influence of particular efficacy beliefs is removed’’ (Bandura,
1997, p. 42).
Additionally, social cognitive theory posits that behavior is selective for a par-
ticular individual and the current environment (Bandura�s, 1986). As a result, it isproblematic to generalize behavioral responses. To illustrate, consider a general
efficacy measure for sports. If one considers him- or herself to be a talented golfer
and a terrible kick-boxer, it is problematic to define a single measure of �sportsefficacy� or to assume that in playing one sport well, an individual possesses the
propensity to play all sports well. Within the context of computers, a business sys-
tems analyst may have very high efficacy with regard to defining and developing a
complicated business program using a cryptic programming language. However,
the same individual may have very low efficacy regarding his or her ability to con-trol a nuclear power plant using a system with a friendly graphical user interface.
As a result, when defining computer-efficacy, one must consider contextually rel-
evant characteristics (e.g., the task and technology) and characteristics of the
individual (e.g., past experiences).
432 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
4. Research objective
The objective of this study is to examine the influence of contextual variables
on end-user IT acceptance behaviors. Self-efficacy theory and social cognitive
theory describe how contextual variables affect human attitude and behavior. Inparticular, before acting, an individual considers enactive experiences (i.e., experi-
ence dealing with this situation before), vicarious experiences (i.e., watching others
deal with this situation), and social persuasion (i.e., support and encouragement
received). Furthermore, a triadic relationship exists among the individual�s affec-
tive state, the environmental characteristics, and the individual�s behavior (Ban-
dura�s, 1986). MIS researchers have adapted each of these constructs as follows:
prior experience measures enactive experience (Igbaria, Parasuraman, & Baroudi,
1996), other�s use measures vicarious experience (Compeau & Higgins, 1995),organizational support measures social persuasion (Igbaria, Zinatelli, Cragg, &
Cavaye, 1997), computer anxiety measures affective state (Compeau & Higgins,
1995), and the computing environment is defined in terms of the task structure
and the system quality (Goodhue, 1995). Fig. 3 illustrates the model considered
in this study. Table 1 defines the study variables and describes the operationaliza-
tion of the variables. It further indicates the number of survey items (questions)
used to measure each of the variables.
We consider several sets of hypotheses in evaluating our model, as stated below.The hypotheses assess the extent to which the exogenous variables influence the
endogenous variables and the extent to which the endogenous variables influence
each other.
H1. Task structure will have a positive, direct relationship with (a) computer-effi-
cacy; (b) perceived ease of use evaluations, (c) perceived usefulness evaluations,
(d) system usage.
H2. Prior experience will have a positive, direct relationship with (a) computer-
efficacy, (b) perceived ease of use evaluations, (c) perceived usefulness evalua-tions, (d) system usage.
H3. Other�s use will have a positive direct relationship with (a) computer-efficacy,
(b) perceived ease of use evaluations, (c) perceived usefulness evaluations, (d)
system usage.
H4. Organizational support will have a positive, direct relationship with (a) compu-
ter-efficacy, (b) perceived ease of use evaluations, (c) perceived usefulness eval-uations, (d) system usage.
H5. Anxiety will have a negative, direct relationship with (a) computer-efficacy, (b)
perceived ease of use evaluations, (c) perceived usefulness evaluations, (d) sys-
tem usage.
Other’s Use
System Quality
Organizational Support
Prior Experience
Anxiety
Task Structure
Computer Efficacy
Perceived Ease of Use
Perceived Usefulness
System Usage
Fig. 3. Research model.
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 433
H6. System quality will have a positive, direct relationship with (a) computer-
efficacy, (b) perceived ease of use evaluations, (c) perceived usefulness evalua-
tions, (d) system usage.
H7. Computer-efficacy will have a positive, direct relationship with (a) system
usage, (b) perceived usefulness, (c) perceived ease of use.
H8. Perceived ease of use evaluations will have a positive, direct relationship with
(a) perceived usefulness evaluations, (b) system usage.
H9. Perceived usefulness evaluations will have a positive, direct relationship withsystem usage.
5. Research methodology
A mail survey was used to gather the data for this study. The survey included 41
statements such as:
� using a computer improves, or would improve, my overall job performance;
� my experience using computers at work has been successful;
Table 1
Variable definitions, number of survey items and theoretical support
Variables Definition Previous studies
Independent
Task structure Five items measuring the extent to
which task is non-routine and varied
(Goodhue and Thompson, 1995; Goodhue,
1995; Igbaria, 1998)
Prior experience Two items measuring the individual�spast experience
(Igbaria and Ivari, 1995; Igbaria et al., 1995;
Igbaria et al., 1996; Taylor and Todd, 1995;
Vankatesh and Davis, 1996)
Other�s use Three items assessing the degree to
which the individual observed others
using a computer
(Compeau and Higgins, 1995; Igbaria et al.,
1996)
Computer
anxiety
Five items measuring an individual�suneasiness or apprehension towards
computers
(Compeau, 1992; Compeau and Higgins,
1995; Igbaria and Chakrabarti, 1990; Igbaria
and Ivari, 1995; Miura, 1987; Raub, 1981;
Staples et al., 1998)
Organizational
support
Four items assessing management
encouragement and resource support
(Compeau and Higgins, 1995; Igbaria, 1990;
Igbaria and Ivari, 1995; Igbaria et al., 1995;
Igbaria et al., 1996; Igbaria et al., 1997;
Thompson et al., 1991)
System quality Five items assessing system
functionality, performance, and
interactivity
(Goodhue and Thompson, 1995; Goodhue,
1995; Igbaria et al., 1995; Igbaria et al., 1990;
Lucas, 1975; Lucas, 1978; Vankatesh and
Davis, 1996)
Mediating
Perceived
usefulness
(PERUSE)
Four items measuring the degree to
which a person believes system use
will enhance job performance
(Adams et al., 1992; Davis, 1989; Davis et al.,
1989; Igbaria et al., 1995; Igbaria et al., 1996;
Igbaria et al., 1997; Jackson et al., 1997;
Langford and Reeves, 1998; Vankatesh and
Davis, 1996)
Perceived ease
of USE
(PEOU)
Four items measuring the degree to
which a person believes system use
will be free of effort
(Adams et al., 1992; Davis, 1989; Davis et al.,
1989; Igbaria et al., 1995; Igbaria et al., 1997;
Jackson et al., 1997; Langford and Reeves,
1998; Vankatesh and Davis, 1996)
Computer-
efficacy
Seven items assessing the extent to
which an individual feels confident in
using a computer
(Compeau and Higgins, 1995; Fenech, 1998;
Igbaria and Ivari, 1995; Langford and Reeves,
1998; Vandenbosch and Higgins, 1995;
Vankatesh and Davis, 1996)
Dependent
System usage Two self-reported system usage items:
frequency of use and duration of use
(Adams et al., 1992; Blair and Burton, 1987;
Cheney and Dickson, 1982; Compeau and
Higgins, 1995; DeLone, 1988; Igbaria and
Ivari, 1995; Igbaria et al., 1996; Igbaria et al.,
1989; Igbaria et al., 1997; Lee, 1986)
434 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
� my immediate supervisor uses computers at work extensively;
� my boss supports and encourages me to use a computer; and
� for fear of making a mistake I cannot correct, I hesitate using computers at
work.
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 435
Respondents rated each statement on a seven-point Likert scale anchored at
Strongly Agree and Strongly Disagree. These statements were adapted from prior
studies (as listed in Table 1), to measure each of the contextual variables included
in the model. The questionnaire was mailed to end-users with a cover letter that
briefly described the study. Stamped, self-addressed return envelopes were providedfor the convenience of the respondents. We sought to survey end-users from mid to
large, for-profit organizations that are served by internal IS staff members. This was
accomplished by targeting organizations with between 25 and 100 internal IS mem-
bers. Since for the purposes of this study system usage was being viewed as volun-
tary, we targeted individuals in professional and managerial roles. Additionally,
end-users were selected from diverse industries throughout the US. The list of
end-user names was purchased from Applied Computer Research, an organization
specializing in contact list development and management. A pretest of the question-naire was conducted with people from both academia and industry. Each respondent
was asked to complete the questionnaire, and provide feedback regarding the process
and measures. Additionally, interviews were conducted to ensure that questionnaire
responses were consistent with the underlying constructs.
Of the 700 surveys mailed, 114 were completed and returned, representing a re-
sponse rate of 16%. Six surveys were incomplete and therefore not used. As a result,
108 completed surveys were analyzed, with a resulting response rate of 15%. Demo-
graphic information was also collected from each respondent regarding his/herorganizational level, functional area, educational level, gender, and age. Measures
for each variable in the proposed model were obtained using the questionnaires.
5.1. Tests for model consistency and validity
The statistical analysis consisted of two stages. The first stage assessed of the reli-
ability of the measures used to operationalize the variables in this study. This
involved assessing the contribution and reliability of multiple indicators for thelatent and manifest variables. The second stage tested the proposed conceptual
model. This involved assessing the contribution and reliability of multiple latent
variable and manifest variable path coefficients.
The research model, as seen earlier in Fig. 3, consists of several latent variables
(constructs), such as organizational support, computer-efficacy, perceived ease of
use, and system usage. Latent variables are theoretical constructs that are not di-
rectly observable. In turn, latent variables are measured through a set of manifest
(indicator) variables. Unlike latent variables, manifest variables are directly observ-able. Consequently, manifest variables provide a means by which the latent variables
can be assessed. This study employed latent variable partial least squares (LVPLS)
analysis (Lohmoeller, 1984). LVPLS simultaneously assesses the extent to which
the manifest variables measure the latent variables (the outer model, or measurement
model) and the extent to which the latent variable relationships match those pro-
posed in the research model (the inner model, or structural model). The composite
reliability assesses the internal consistency of the latent variables, and according to
436 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
the procedure suggested by Nunnally (1978), our instruments demonstrate a strong
to moderate level of reliability. Another assessment of reliability, proposed by Forn-
ell and Larcker (1981), is the average variance extracted. This measure also showed
strong to moderate levels of reliability. Therefore, we concluded that all latent var-
iable measures exhibited sufficient levels of internal reliability. Convergent validity ofthe manifest variables was assessed by analyzing the factor loading scores for each;
all individual loadings were in the range of 0.4–0.95. Hair, Anderson, Tatham, and
Black (1992) suggest that individual item loadings greater than 0.3 are significant.
Therefore, all items demonstrated convergent validity.
Since this study employed a single data collection method, we tested for common
method variance. Podsakoff and Organ (1986) suggest conducting an unconstrained,
single factor analysis for models that intend to measure multiple constructs. A dom-
inance of one factor would suggest that items were related due to common methodvariance. We conducted an unconstrained, single factor analysis;the total explained
variance by the 11 retained factors (using a minimum eigenvalue of one) was 75%
and the first factor accounted for 26% of the variance. Since multiple factors were
retained and the first factor did not dominate the variance, common method vari-
ance was not detected.
The discriminant validity of the analysis was assessed by comparing the inter-
correlations among the latent variables with the correlation the measures have
with their respective constructs (Fornell, Tellis, & Zinkhan, 1982). Discriminantvalidity is demonstrated when the measures are more strongly related to them-
selves than to the other latent variables in the model. Discriminant validity was
demonstrated for all constructs except the perceived ease of use construct. In this
case, the construct loaded more highly with the anxiety construct than it did with
itself. However, the violations were mild and according to Chin (1998) an item
should be dropped due to insufficient loading only if it is determined that the vio-
lation is a result of method variance or some other concept. Several other
researchers have found similar discriminant validity issues (e.g., Igbaria & Ivari,1995; Staples et al., 1998). In situations where the violations were minor and
the internal reliability measures were adequate, the authors disregarded the viola-
tions. As a result, since we were unable to detect common method variance, and
since our internal reliability measures were adequate, and since the cross-loading
was mild, all items were retained.
The PLS procedure simultaneously calculates factor loadings for the items and
the path coefficient variables. The item loadings are used to assess the strength
and reliability of the measures, as previously explained, and the path coefficient val-ues and loadings are used to evaluate the theoretical relationships posed in the con-
ceptual model (Igbaria et al., 1997). The exogenous variable path coefficients
represent the total effect that the variable has on the endogenous variable. This total
effect consists of a direct and an indirect effect on the endogenous variable. ‘‘An indi-
rect effect represents those effects interpreted by the intervening variables; it is the
product of the path coefficients along an indirect route from cause to effect via trac-
ing arrows in the headed direction only. When more than one indirect path exists, the
total indirect effect is their sum’’ (Igbaria et al., 1995, p. 99).
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 437
While the parsimonious data assumptions of PLS provide methodological
conveniences, they also limit the ability to assess the statistical significance of the
conceptual model�s path coefficients (Meznar & Nigh, 1995). As a result, the non-
parametric jackknifing technique (Fenwick, 1979) was used in conjunction with t-sta-
tistics to determine the statistical significance of the path coefficients. This practice isconsistent with prior studies using PLS (e.g., Igbaria et al., 1995; Igbaria et al., 1997;
Meznar & Nigh, 1995; Staples et al., 1998). The jackknife method repeatedly ana-
lyzes the statistic in question using a resampling methodology. Rather than making
a priori variability assumptions (as is done in the traditional parametric t- and z-
tests), jackknifing uses large numbers of computations to explore the empirical var-
iability of a statistic. To test the hypotheses, t-statistics were calculated using the PLS
path coefficients (direct and indirect effects), and the jackknifing path coefficient bias
estimates.
6. Results of the hypothesis testing
This study investigates the following six exogenous variables:
� prior experience,
� other�s use,� anxiety,
� system quality,
� task structure, and
� organization support, and their influence on the following four endogenous
variables:
� system usage,
� perceived ease of use,
� perceived usefulness and computer-efficacy.
See Fig. 4 for the path diagram that depicts the structural equations associated
with this study.
The PLS procedure is a distribution-free procedure, which separates the over-
all model into two sub-models. The measurement model assesses how well the
manifest variables describe the exogenous latent variables. The remaining model
is the structural model, which assesses how well the exogenous latent variables
describe the endogenous latent variables. The evaluation and assessment ofthese models is handled independently first, then together as a total fit of the
overall conceptual model. The structural model is depicted as a path diagram,
with the paths representing relationships between the exogenous and endog-
enous variables. Fig. 5 depicts the path diagram with the manifest variable
loading scores and the structural model standardized regression coefficients for
the instrument.
The results of the multivariate test of the structural model provide the stand-
ardized regression coefficients (i.e., the path loadings) for the conceptual model.
438 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
The path loadings represent the direct effect that the exogenous variables have
on the endogenous variables. An indirect effect exists when an exogenous varia-
ble influences an intervening variable, which in turn influences the endogenous
variable. An indirect effect is calculated as the product of the path loadings
for all coefficients on the indirect path. If multiple indirect paths exist, the totalindirect effect is the sum of the individual indirect effects. The total effect is the
sum of the direct effect and the total indirect effect. The statistical significance
for the path loadings were determined using the t-statistics and the nonparamet-
ric jackknifing procedure (Alwin & Hauser, 1975). Complete results of hypothe-
ses testing are shown in Table 2, according to the type of statistical test(s)
conducted.
6.1. Communality coefficient
The structural model deals with latent variables that are not directly observa-
ble. The measurement model provides the means by which these latent variables
can be observed. Specifically, the measurement model represents the relationships
and loadings of the questionnaire items into their respective latent variable. As
such, the assessment of the measurement model focuses on the degree to which
the indicator variables load with their respective constructs. Individual manifest
variable loadings are used to assess latent variable reliabilities and discriminant
Other’s Use
System Quality
Organizational Support
Prior Experience
Anxiety
Task Structure
Computer Efficacy
Perceived Ease of Use
Perceived Usefulness
System Usage
Q7 Q8 Q9 Q10 Q11 Q12
Q40Q41Q42Q43Q44
Q15Q16Q17Q18
Q5 Q6
Q32Q34Q35Q38Q39
Q19Q20Q21Q22Q23
Q24 Q25 Q26 Q27 Q28 Q29 Q30
Q31Q33Q36Q37
Q1Q2Q3Q4
Q13 Q14
Fig. 4. Path diagram for structural equations.
Other’s Use
System Quality
Organizational Support
Prior Experience
Anxiety
Task Structure
Computer Efficacy
Perceived Ease of Use
Perceived Usefulness
System Usage
-.03
.02
.12.38
Q7-.58 Q8-.78 Q9-.80 Q10-.84 Q11-.74 Q12-.68
Q19-.46 Q20-.53 Q21-.87 Q22-.85 Q23-.65
Q40-.86 Q41-.89 Q42-.87 Q43-.86 Q44-.81
Q15-.82Q16-.88Q17-.90Q18-.82
Q5-.82Q6-.88
Q32-.68Q34-.86Q35-.76Q38-.42Q39-.73
Q1-.73 Q2-.90Q3-.86 Q4-.81
Q13-.90Q14-.62
Q31-.51Q33-.85Q36-.70Q37-.75
Q24-.66 Q25-.65 Q26-.83 Q27-.78 Q28-.(-.80) Q29-.74 Q30-.56
.10
.15
.15-.26
-.04
-.07
.08.11
.30.09
.50
.34-.30 -.63
-.26
-.11
.01 .03
-.14
.12
.02
.07.11
-.03
.28
. 10
Fig. 5. Manifest variable loading scores and structural model standardized regression coefficients.
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 439
validity. However, when these loadings are analyzed in total, they provide an
assessment of the overall fit of the measurement model. The communality coeffi-
cient provides an overall assessment of how well the manifest variables describe
their respective latent variables. Falk and Miller (1992) suggest that values less
than 0.30 are considered to be too low to be acceptable. For this study, the com-
munality coefficient was 0.58. Since the measure exceeded Falk and Miller�s rec-ommendations, we conclude that the measurement model provides adequate
reliability.
6.2. Root mean square of covariance
Since the PLS procedure simultaneously determines the structural and meas-
urement model loadings, we seek to assess the reliability of the combined model.
Falk and Miller (1992) suggest using root mean square of the covariance be-tween the manifest variable residuals and the latent variable spans to evaluate
the overall fit of the conceptual, and underlying measurement model. The root
mean square of the covariance between the manifest variable residuals and the
latent variable residuals (RMS Cov(E,U)) represents the correlation between
the variance of the manifest and latent variables that are not accounted for
by the model relationships. A RMS Cov(E,U) equal to zero would indicate that
the model perfectly described the relationships between the manifest and latent
variables. Falk and Miller (1992) suggest that RMS Cov(E,U) values of 0.02 rep-resent superior models and RMS Cov(E,U) values above 0.20 are evidence of an
Table 2
Results of hypothesis testing
Hypothesis Standardized regression
coefficient
p-value
H1 – High task structure will have a positive, direct relationship
with:
(a) computer efficacy 0.01 Not sig
(b) ease of use perceptions 0.03 0.01
(c) usefulness perceptions �0.14 0.001a
(d) system usage 0.12 0.001
H2 – Prior experience will have a positive, direct relationship with:
(a) computer efficacy 0.30 0.001
(b) ease of use perceptions 0.09 0.001
(c) usefulness perceptions 0.50 0.001
(d) system usage 0.34 0.001
H3 – Other�s use will have a positive, direct relationship with:
(a) computer efficacy �0.030 0.05a
(b) ease of use perceptions 0.02 Not sig
(c) usefulness perceptions 0.12 0.001
(d) system usage 0.38 0.001
H4 – Organizational support will have a positive, direct
relationship with:
(a) computer efficacy �0.04 0.01a
(b) ease of use perceptions �0.07 0.001a
(c) usefulness perceptions 0.08 0.001
(d) system usage 0.11 0.001
H5 – Anxiety will have a negative, direct relationship with:
(a) computer efficacy �0.3 0.001
(b) ease of use perceptions �0.63 0.001
(c) usefulness perceptions �0.26 0.001
(d) system usage �0.11 0.001
H6 – System quality will have a positive, direct relationship with:
(a) computer efficacy 0.10 0.001
(b) ease of use perceptions 0.15 0.001
(c) usefulness perceptions 0.15 0.001
(d) system usage �0.26 0.001a
H7 – Computer efficacy will have a positive, direct relationship
with:
(a) ease of use perceptions 0.11 0.001
(b) usefulness perceptions 0.07 0.001
(c) system usage 0.02 Not sig
H8 – Perceived ease of use evaluations will have a positive, direct
relationship with:
(a) usefulness perceptions 0.28 0.001
(b) system usage �0.03 Not sig
H9 – Perceived usefulness evaluations will have a positive, direct
relationship with system usage
0.10 0.001
a Significant, however, the effect was in the opposite direction from what was expected.
440 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
inadequate model. The RMS Cov(E,U) value was 0.07. This measure indicates
that overall the model provides strong evidence that the data adequately sup-
ports the model.
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 441
As is shown in Table 2, we found substantial support that contextual variables do
indeed directly affect IT acceptance. Specifically, system usage was directly and signif-
icantly affected by task structure, prior experience, other�s use, organizational support,anxiety, and system quality.
However, a few variables impacted the endogenous variables in directions oppo-site from what was expected. These variables are listed below along with possible rea-
sons for the opposite effects:
High task structure was found to reduce system usefulness perceptions. Perhaps the
respondents felt that computers are more useful for less structured tasks since less
structured tasks are more difficult to solve.
Other�s use was found to lower computer efficacy. Perhaps the frequency of peer
observations has an inverse relationship with one�s confidence in using a system; thatis, if a person is not confident in his/her ability to use a system he/she may spend
more time observing others using it before trying it him/herself.
Similarly, organizational support was found to lower computer efficacy and per-
ceived ease of use assessments. Respondents may believe that organizations provide
more support for those systems that are more difficult to use.
Lastly, high quality had a negative affect on system usage. Since respondents indi-
cated that high quality systems improve efficacy, ease of use, and usefulness, one
would expect that they would use a high quality system more often. Since this resultwas not found, perhaps the respondents felt that the systems they use most often
have poor quality.
7. Discussion of results
With the hopes of maintaining or improving competitiveness, organizations are
investing significantly in information technology. Unfortunately, these investments
do not guarantee that people will actually use the systems. In fact, studies show that
people sometimes choose not to use potentially beneficial systems. As well as repre-
senting a large lost investment, the unrealized potential of a system can be monu-
mental. In extreme cases, unused information systems may impact the viability of
the organization, if the particular information system is deemed a necessity. These
factors help explain why information technology acceptance is one of the top con-cerns for IS managers. Unfortunately, while IT acceptance has attracted the atten-
tion of researchers, the topic continues to be one of the most challenging and least
understood areas of MIS research.
The TAM is one of the most powerful and influential IT acceptance models. How-
ever, researchers suggest that the model may be too general. In addition, investiga-
tors suggest that the TAM does not fully consider or appreciate the impacts of
contextual variables. Self-efficacy in particular has been investigated as a potential
extension to the TAM. However, similar to the generalization concerns of the
442 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
TAM, based on a review of social cognitive theory, the MIS operationalization of
self-efficacy, computer-efficacy, may be overly generalized. This is in keeping with
Bandura�s (1986) suggestion that efficacy assessments must be particularized for a
specific situation. Furthermore, studies have shown that generalized efficacy meas-
ures bear little or no relation with particularized measures.The objective of this study was to add contextual variables to the TAM. The inde-
pendent variables considered in this study were chosen based on the antecedents of
self-efficacy and social cognitive theory. Specifically, Bandura�s (1986) suggests thatself-efficacy is formed through enactive experiences, vicarious experiences, social per-
suasion, and affect.The operationalization of these constructs was based on prior
MIS studies. Enactive experience was assessed by measuring prior experience. Vicar-
ious experience was assessed by measuring others� use. Social persuasion was as-
sessed by measuring organizational support and affect was assessed by measuringanxiety. To capture the environmental characteristics as suggested by social cogni-
tive theory, we measured task structure and system quality.
The mediating variables and dependent variable considered in this study were
chosen based on the technology acceptance model with a computer-efficacy exten-
sion. As a result, three mediators were measured, namely, computer-efficacy, per-
ceived ease of use, and perceived usefulness. Consistent with other TAM studies, a
single dependent variable, system usage, was measured. The operationalization of
these constructs was based on prior information system studies.A field research design was employed with data gathered by means of mailed
questionnaires. The mailing list was purchased from a market research organization,
Applied Computer Research, Incorporated (Phoenix, AZ). The hypotheses gener-
ated from the conceptual model were tested using partial least squares. The tests
of the hypotheses show that all the contextual variables (computer anxiety, prior
experience, other�s use, organizational support, and system quality) significantly af-
fect computer-efficacy. (One contextual variable, task structure, did not.) Further-
more, the contextual variables directly affected system usage.These findings suggest that while the TAM is valid, substantially stronger results
may be obtained if researchers particularize their research instruments. Furthermore,
in support of social cognitive theory, the influence of contextual variables should not
be overlooked or trivialized.
8. Research contribution
While the merits of the TAM are notwithstanding, the findings of this study sup-
port providing greater specificity when analyzing computing behaviors. As a result,
we proposed that the particularization of the TAM would provide overall better re-
sults. These findings are consistent with social cognitive theory and the Theory of
Reasoned Action.
Prior studies report that the explained variance of IT usage has been somewhat
inconsistent and not necessarily strong. The explained variance of system usage
(28%) in our study was consistent with the findings of prior studies.
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 443
Another important finding of this study is the significance of the contextual var-
iables. We found that contextual variables significantly impacted the mediating var-
iables, as well as the dependant variable, system usage. While these findings are
consistent with social cognitive theory, they are in conflict with the TAM and
TRA which posit that contextual variables only influence behavior indirectly,through mediating variables. As a result, the findings of this study suggest that
researchers should not indiscriminately disregard or trivialize the role of contextual
variables as they relate to IT behaviors. As suggested by social cognitive theory, it
appears that one�s behavior is a function of his or her characteristics and experiences
as they relate to the specific situation.
9. Research limitations
Although this study found many significant relationships between the latent var-
iables, causality should not be implied (Falk & Miller, 1992). The findings of this
study are appropriate for predictions and/or confirmation of theoretical constructs.
As a result, results of this study justify the conjecture that self-efficacy theory is appli-
cable to MIS research. Furthermore it is reasonable to predict, based on social cog-
nitive theory and these empirical results, that particularized instruments will provide
more insights into computing attitudes, perceptions, and behaviors. However, it isnot appropriate or justified to suggest that the presence or absence of one construct
will cause a change in another construct.
Another limitation of the methodology employed in this study is that, due to their
nature, we were unable to manipulate the independent variables. Since experimental
manipulation was not possible, we were limited to the extent to which we could con-
trol the study (Kerlinger, 1986). As a result, it is with less certainty that we may con-
clude that a true relationship exists between variables. This is due to our inability to
rule-out the potential that the observed relationship was not the result of one ormore unmeasured variables. Additionally, the present study used a single data col-
lection procedure. If multiple data collection methods were employed, the overall
convergent and discriminant validities could have been improved by using a multi-
trait-multimethod analysis (Campbell & Fiske, 1959).
Self-reported instruments were used to measure the research variables. As a result,
the items may suffer from several types of response biases, such as the halo effect
when the responses are influenced by the respondent�s overall impression of the ob-
ject (e.g., I like computers, so I will respond positively to all questions) and the errorof central tendency when the respondent avoids the extremes on the response scale
(Kerlinger, 1986).
In regard to the data analysis method, PLS requires multiple indicators for each
latent variable (Lohmoller, 1984; Chin, Marcolin, & Newsted, 1996; Falk & Miller,
1992). While there is no empirical evidence to determine the ideal number of indica-
tors, two may be too few. Since the PLS procedure gives preference to fitting the struc-
tural model, at the expense of the measurement model (Falk & Miller, 1992), a larger
number of manifest variables will allow the procedure to account for a greater portion
444 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
of the structural model�s variance. In the present study two latent variables had fewer
than four manifest variables, namely prior experience and system usage. As a result,
additional indicators may have improved the data analysis.
10. Suggestions for future research
Beyond addressing the limitations of this study, there are several research areas
and/or procedures that would confirm or enhance the findings in this study.
While the nature of this research does not allow for manipulation of the inde-
pendent variables, varying the research design can increase control. For example,
multiple instruments can be administered to the same people. By matching the
data sets, this design can significantly improve control by factoring out the influ-ences of several significant independent variables and other unmeasured extrane-
ous variables.
Additionally, the realm of potentially significant contextual variables is large.
Researchers can investigate a host of other contextual variables such as math anxi-
ety, math aptitude, reading aptitude, organizational structure, management style,
prior task training, and prior training regarding the specific technology. While con-
textual variable selection should be guided by theory, in consideration with the find-
ings of this study, one could also consider the influences of the specific situation inselecting appropriate and potentially significant variables.
Several constructs could be defined to a greater level of detail. For instance, the
prior experience and other�s use constructs were limited to the use of a computer. Fu-
ture studies could measure experience performing a specific task and/or experience
using a specific application.
Similarly, in the present study we asked the respondents to describe the types of
tasks they typically addressed when using information technology. Future studies
could describe and investigate the role of a particular task. Based on the findingsof this study, we suspect that these further particularizations would improve the
overall results of the study.
Finally, in consideration of the findings of this study, the roles of the various con-
textual variables could differ by application and situation. As a result, the search for
the optimal set of contextual and mediating variables may be fruitless it could all
depend on the unique mix of individuals, tasks, applications, and organizations.
However, finding a theory that would help one chose and weigh contextual variables
based on the particulars of a situation would provide substantial value and insightfor IS researchers and IS practice.
References
Adams, D. A., Nelson, R. R., & Todd, P. A. (1992). Perceived usefulness, ease of use, and usage of
information technology: A replication. MIS Quarterly, 16(2), 227–247.
Alwin, D. E., & Hauser, R. R. (1975). Decomposition of effects in path analysis. American Sociological
Review, 40, 37–47.
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 445
Bandura�s, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs,
NJ: Prentice-Hall.
Bandura, A. (1995). Exercise of personal and collective efficacy in changing societies. In A. Bandura (Ed.),
Self-efficacy in changing societies (pp. 1–45). New York, NY: Cambridge University Press.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York, NY: W.H. Freeman & Company.
Blair, E., & Burton, S. (1987). Cognitive processes used by survey respondents to answer behavioral
frequency questions. Journal of Consumer Research, 14, 280–288.
Burkhardt, M. E., & Brass, D. J. (1990). Changing patterns or patterns of change: The effects of a
change in technology on social network structure and power. Administrative Science Quarterly, 35(1),
104–127.
Campbell, D. J., & Fiske, D. (1959). Convergent and discriminant validation by the multitrait-
multimethod matrix. Psychological Bulletin, 54, 81–105.
Chau, P. Y. K. (1996). An empirical assessment of a modified technology acceptance model. Journal of
Management Information Systems, 13(2), 185–204.
Cheney, P. H., & Dickson, G. B. (1982). Organizational characteristics and information systems success:
An exploratory investigation. Academy of Management Journal, 25(1), 170–184.
Chin, W. W. (1998). Issues and opinion on structural equation modeling. MIS Quarterly, 22(1), 7–16.
Chin, W. W., Marcolin, B. L., & Newsted, P. R. (1996). A partial least square latent variable modeling
approach for measuring interaction effects: Results from a Monte-Carlo simulation study and voice
mail emotion/adoption study. In Proceedings of the 17th international conference on information systems
(pp. 21–41). Cleveland, OH.
Compeau, D. R. (1992). Individual reactions to computing technology: A social cognitive theory
perspective. Doctoral dissertation. The University of Western Ontario.
Compeau, D. R., & Higgins, C. A. (1995). Computer self-efficacy: Development of a measure and initial
test. MIS Quarterly, 19(2), 189–211.
Davis, F. D. (1986). A technology acceptance model for empirically testing new end-user information
systems: Theory and results. Doctoral dissertation. MIT Sloan School of Management.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information
technology. MIS Quarterly, 13(3), 319–340.
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: A
comparison of two theoretical models. Management Science, 35(8), 982–1003.
DeLone, W. H. (1988). Determinants of success for computer usage in small business. MIS Quarterly,
12(1), 51–61.
Falk, R. F., & Miller, N. B. (1992). A primer for soft modeling. Akron, OH: The University of Akron.
Fenech, T. (1998). Using perceived ease of use and perceived usefulness to predict acceptance of the World
Wide Web. Computer Networks & ISDN Systems, 30(1-7), 629–630.
Fenwick, I. (1979). Techniques in market measurement: The jackknife. Journal of Marketing Research,
16(3), 410–414.
Fishbein, M., & Ajzen�s, I. (1975). Belief, attitude, intention and behavior: An introduction to theory and
reason. Reading, MA: Addison-Wesley.
Fornell, C., & Larcker, D. F. (1981). Structural equation models with unobservable variables and
measurement error: Algebra and statistics. Journal of Marketing Research, 18(3), 382–387.
Fornell, C. R., Tellis, G. L., & Zinkhan, G. M. (1982). Validity assessment: A structural equation
approach using partial least squares. In American marketing association educators� proceedings (pp.
405–409). Chicago, IL.
Gist, M. E., & Mitchell, T. R. (1992). Self-efficacy: A theoretical analysis of Its determinants and
malleability. Academy of Management Review, 7(2), 183–211.
Gist, M. E., Schwoerer, C., & Rosen, B. (1989). Effects of alternative training methods on self-efficacy and
performance in computer software training. Journal of Applied Psychology, 74, 884–891.
Goodhue, D. L., & Thompson, R. L. (1995). Task-technology fit and individual performance. MIS
Quarterly, 19(2), 213–236.
Goodhue, D. L. (1995). Understanding user evaluations of information systems. Management Science,
41(12), 1827–1844.
446 D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447
Hair, J. F, Anderson, R. E, Tatham, R. L., & Black, W. C. (1992).Multivariate data analysis with readings
(3rd ed.). New York, NY: MacMillan.
Hasan, B. (2003). The influence of specific computer experiences on computer self-efficacy beliefs.
Computers in Human Behavior, 19(4), 443–450.
Hendrickson, A. R., Glorfeld, K., & Cronan, T. P. (1994). On the repeated test-retest reliability of the end-
user computing satisfaction instrument: A comment. Decision Sciences, 25(4), 655–667.
Hill, T., Smith, N. D., &Mann, M. F. (1987). Role of efficacy expectations in predicting the decision to use
advanced technologies: The case of computers. Journal of Applied Psychology, 72(2), 307–313.
Hubona, G. S., & Cheney, P. H. (1994). System effectiveness of knowledge-based technology: The
relationship of user performance and attitudinal measures. In Proceedings of the 27th annual Hawaii
international conference on system sciences (pp. 532–541), Hawaii.
Igbaria, M. (1998). Personal communication.
Igbaria, M. (1990). End-user computing effectiveness: A structured equation model. Omega International
Journal of Management Science, 18(6), 637–652.
Igbaria, M., & Chakrabarti, A. (1990). Computer anxiety and attitudes towards microcomputer use.
Behavior and Information Technology, 9(3), 229–241.
Igbaria, M., & Ivari, J. (1995). The effects of self-efficacy on computer usage. Omega International Journal
of Management Science, 23(6), 587–605.
Igbaria, M., Guimaraes, T., & Davis, G. B. (1995). Testing the determinants of microcomputer usage via a
structural model. Journal of Management Information Systems, 11(4), 87–114.
Igbaria, M., Parasuraman, S., & Baroudi, J. J. (1996). A motivational model of microcomputer usage.
Journal of Management Information Systems, 13(1), 127–143.
Igbaria, M., Parasuraman, S., & Pavri, F. (1990). A path analytic study of the determinants of
microcomputer usage. Journal of Management Systems, 2(2), 1–14.
Igbaria, M., Pavri, F., & Huff, S. (1989). Microcomputer application: An empirical look at usage.
Information and Management, 16(4), 187–196.
Igbaria, M., Zinatelli, N., Cragg, P., & Cavaye, A. L. M. (1997). Personal computing acceptance factors in
small firms: A structural equation model. MIS Quarterly, 21(3), 279–305.
Jackson, C. M., Chow, S., & Leitch, R. A. (1997). Toward an understanding of the behavioral intention to
use an information system. Decision Sciences, 28(2), 357–389.
Jarvenpaa, S. L. (1989). The effect of task demands and graphical format on information processing
strategies. Management Science, 35(3), 285–303.
Kerlinger, F. N. (1986). Foundations of behavioral research (3rd ed.). New York, NY: Harcourt Brace
Jovanovich College Publisher.
Langford, M., & Reeves, T. E. (1998). The relationships between computer self-efficacy and personal
characteristics of the beginning information systems student. Journal of Computer Information Systems,
38(4), 41–45.
Lee, D. S. (1986). Usage patterns and sources of assistance to personal computer use. MIS Quarterly,
10(4), 313–325.
Lohmoeller, J. B. (1984). LVPLS 1.6 Program manual: Latent variable path analysis with partial least-
squares estimation. Universitaet zu Koehn, Zentralarchiv fuer Empirische Sozialforschung, Colgne,
Germany.
Lucas, H. C. (1975). Performance and the use of an information system. Management Science, 21(8),
908–919.
Lucas, H. C. (1978). Empirical evidence for a descriptive model of implementation. MIS Quarterly, 2(2),
27–41.
Markus, H., & Nurius, P. (1986). Possible selves. American Psychologist, 41, 954–969.
Martocchio, J. J., & Dulebohn, J. (1994). Performance feedback effects in training: The role of perceived
controllability. Personnel Psychology, 47(2), 357–373.
Mathieson, K. (1991). Predicting user intentions: Comparing the technology acceptance model with the
theory of planned behavior. Information Systems Research, 2(3), 173–191.
Meznar, M. B., & Nigh, D. (1995). Buffer or bridge. environmental and organizational determinants of
public affairs activities in american firms. Academy of Management Journal, 38(4), 975–997.
D.J. McFarland, D. Hamilton / Computers in Human Behavior 22 (2006) 427–447 447
Miura, I. T. (1987). The relationship of computer self-efficacy expectations to computer interest and course
enrollment in college. Sex Roles, 16, 303–311.
Moore, G. C., & Benbasat, I. (1991). Developement of an instrument to measure the perceptions of
adopting an information technology innovation. Information Systems Research, 2(3), 192–222.
Nunnally, J. (1978). Psychometric methods (2nd ed.). New York, NY: McGraw-Hill.
Podsakoff, P., & Organ, M. (1986). Self-reports in organizational research: Problems and prospects.
Journal of Management, 12(4), 531–544.
Raub, A. C. (1981). Correlates of computer anxiety in college students. Doctoral dissertation. University
of Pennsylvania, Philadelphia, PA.
Saga, V., & Zmud, R. (1994). The nature and determinants of IT acceptance, routinization, and infusion.
IFIP Transactions A (Diffusion, Transfer, and Implementation of Information Technology), 45, 67–86.
Segars, A. H., & Grover, V. (1994). Re-examining perceived ease of use and usefulness: A confirmatory
factor analysis. MIS Quarterly, 17(4), 517–525.
Staples, D. S., Hulland, J. S., & Higgins, C. A. (1998). A self-efficacy theory explanation for the
management of remote workers in virtual organizations. Journal of Computer-Mediated Communica-
tion, 3(4), 1–36.
Taylor, S., & Todd, P. A. (1995). Assessing IT usage: The role of prior experience. MIS Quarterly, 19(4),
561–570.
Thompson, R. L., Higgins, C. A., & Howell, J. M. (1991). Personal computing: Toward a conceptual
model of utilization. MIS Quarterly, 15(1), 125–143.
Vandenbosch, B., & Higgins, C. A. (1995). Executive support systems and learning: A model and empirical
test. Journal of Management Information Systems, 12(2), 99–130.
Vankatesh, V., & Davis, F. D. (1996). A model of the antecedents of perceived ease of use: Development
and test. Decision Sciences, 27(3), 451–481.
Webster, J., & Martocchio, J. J. (1992). Microcomputer playfulness: Development of a measure with
workplace implications. MIS Quarterly, 16(2), 201–226.