Computers in

Human Behavior
Computers in Human Behavior 23 (2007) 2804–2822
www.elsevier.com/locate/comphumbeh

Theory of planning behavior (TPB) and

customer satisfaction in the continued use of
e-service: An integrated model
Chechen Liao a, Jain-Liang Chen a, David C. Yen b,*

a
Department of Information Management, College of Management,
National Chung Cheng University, Ming-Hsiung, Chia-Yi 621, Taiwan
b
Department of Decision Sciences and MIS RTF School of Business, Miami University,
Oxford, OH 45056, United States

Available online 30 June 2006

Abstract

This study has devoted much effort to developing an integrated model designed to predict and
explain an individual’s continued use of online services based on the concepts of the expectation dis-
confirmation model and the theory of planned behavior. Empirical data was collected from a field
survey of Cyber University System (CUS) users to verify the fit of the hypothetical model. The mea-
surement model indicates the theoretical constructs have adequate reliability and validity while the
structured equation model is illustrated as having a high model fit for empirical data. Study’s find-
ings show that a customer’s behavioral intention towards e-service continuance is mainly determined
by customer satisfaction and additionally affected by perceived usefulness and subjective norm. Gen-
erally speaking, the integrated model can fully reflect the spirit of the expectation disconfirmation
model and take advantage of planned behavior theory. After consideration of the impact of systemic
features, personal characteristics, and social influence on customer behavior, the integrated model
had a better explanatory advantage than other EDM-based models proposed in prior research.
Ó 2006 Elsevier Ltd. All rights reserved.

Keywords: Confirmatory factor analysis; Expectation confirmation model; Expectation disconfirmation model;
Technology acceptance model; Theory of planning behavior; Structured equation model

*
Corresponding author. Tel.: +1 513 529 4826; fax: +1 513 529 9689.
E-mail address: yendc@muohio.edu (D.C. Yen).

0747-5632/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.chb.2006.05.006
 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822 2805

1. Introduction

In the promotion and development of Internet and information technology, e-commerce

has not only ushered in new applications, business models and economics, but has also
gradually changed the way that enterprisers conduct their business. Forced by the
advancement of e-commerce, e-service has increasingly become a critical channel through
which customer services can be automatically rendered over Internet throughout the con-
sumption life cycle. E-service can provide not only basic commerce functions such as online
catalogues, online transactions, and order fulfillment but also a series of customer-oriented
activities such as online help, configuration and customization, and security mechanisms to
enhance/improve customer satisfaction.
Examining the development cycle of e-service, the first phase focused on initiating inno-
vative online services and exploring new customers despite the fact that enterprises do not
substantially gain profit from their investment in the early stage. However, upon noticing
the Internet bubble burst and the scenario that profit was out of proportion to the huge
investment, the second phase of e-service revolution should focus on the customer’s real
needs and also the increase in actual revenue generated from the retention and satisfaction
of current customers. According to an investigation by InformationWeek into the Top 100
E-Business rankings of the most innovative practitioners of electronic business, these com-
panies do not primarily set their goals simply on either revenue growth or increasing prof-
its via launching e-business ventures. Instead, they pay more attention to improving
customer satisfaction and enhancing service quality (Violino, 1999). Nowadays, E-Busi-
nesses are confronting many new challenges and one among these challenges is to satisfy
the customers with high quality e-service over the Internet with an aim to retain them in
the future. For service-oriented e-venders, providing customer-centered and quality-
oriented online services is more sensitive to gaining profit than e-venders who are prod-
uct-oriented. Understanding how system design, personal characteristics and social norm
affect customer’s satisfaction and loyalty as well as applying this knowledge to designing
an appropriate online system for a better e-service provision are critical issues to be taken
into serious consideration.
Many researchers have devoted themselves to clarifying the impact of customer satis-
faction on the individual’s repurchase behavior. This notion is generally referred to as
the ‘‘expectation disconfirmation model’’ (Oliver, 1993; Spreng, MacKenzie, & Olshavsky,
1996; Van Montfort, Masurel, & Van Rijn, 2000; Spreng & Chiou, 2002), ‘‘expectation
disconfirmation paradigm’’ (McKinney, Yoon, & Zahedi, 2002; Patterson, Johnson, &
Spreng, 1997) or ‘‘expectation confirmation theory’’ (Bhattacherjee, 2001). According to
expectation disconfirmation model, a customer’s repurchase intention is preceded by cus-
tomer satisfaction; whereas customer satisfaction is directly affected by disconfirmation
resulted between a customer’s pre-purchase expectations and post-purchase performance
of a product or service. Applying the theory to the continued use of an information sys-
tem, Bhattacherjee (2001) integrated the concept of the technology acceptance model
(TAM) with the expectation disconfirmation model to reflect the impact of a customer’s
expectation of system-specific attributes on customer satisfaction and the intention of
continuous usage. The expectation disconfirmation model can properly reflect a cus-
tomer’s repurchase behavior from the viewpoint of an information system, since all func-
tionalities of Internet service are fully dependent on the facilitations of information
technologies (Koufaris, 2002). However, customers’ adoption of certain products or
2806 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822

services is considered to have a strong relationship with the communication channels and
social systems (Rogers, 1995). Besides, the individual volition regarding the capabilities for
demonstrating interested behavior as well as usable resources and facilitating conditions
can seriously affect a person’s intention to adopt online services. From the above discus-
sion, the expectation disconfirmation model lacks extensive consideration to carefully
evaluate the impact of social norm and personal characteristics on an individual’s behav-
ior towards continued use.
To overcome the incapability of the expectation disconfirmation model, proper exten-
sions of the model can quoted from the concepts of the theory of planning behavior
(TPB) (Ajzen, 1991), which is s generic model most widely used in explaining and analysis
human behavior. According to this aforementioned theory, human behavior is guided by
behavioral intentions as well as perceived behavioral controls, while behavioral intention
is jointly determined by attitude towards the behavior, subjective norm, and perceived
behavioral control. As a result, the research purpose of the current study is to propose
an integrated model that fully synthesizes the concepts of the expectation disconfirmation
model and TPB in a complementary manner to better understand a customer’s behavioral
intention towards continued use of online services. The hypothetical model was primarily
based on the expectation disconfirmation model to monitor the effect of customer satisfac-
tion, while TPB is provided to reflect the influence of personal characteristics and subjective
norm. Some interrelationships between the expectation disconfirmation model and TPB are
also identified to clarify the synthetic effects of the factors introduced in these two models.
A ‘‘Cyber University’’ e-learning system constituted by National Sun Yat-Sen Univer-
sity of Taiwan was chosen as the research context. Cyber University System (CUS) is a
full-functional e-learning system, which can remotely provide asynchronous online teach-
ing and learning service for registered students. Formal students can take credit courses
from CUS and then, learn from online teaching materials. To improve the learning perfor-
mance, CUS not only provides students with the following features such as online discus-
sion boards, online exercises, and real time chatting rooms to perform interactive learning,
but also supports faculty with many administrative functions such as course, homework,
examination, and grade management to monitor their students. Up until the year 2005,
CUS has recruited more than 2000 formal students from different counties to take credit
courses and consequently it becomes one of the most famous and successful e-learning sys-
tems in Taiwan. Since CUS is one of the most representative e-learning systems in Taiwan,
it is authors’ belief that choosing CUS as research context can faithfully reflect the real
characteristics of e-learning system’s users in Taiwan.

2. Literature review

2.1. Expectation disconfirmation model

The study of customer repurchase behavior continues to catch the attention from
researchers since a large portion of consumer purchases are second or n-time purchases
rather than initial acceptances (Oliver, 1993). To clarify how customer satisfaction can
affect customer retention and loyalty, an application of the expectation disconfirmation
model explains the customer consumption decision in the post-purchase process and con-
stantly dominates academic research & managerial practice (Spreng et al., 1996). According
to the model, customer satisfaction is the only one requirement that determines a cus-
 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822 2807

tomer’s intention to repurchase, and hence it leads to the researcher’s interest in studying
the factor of customer satisfaction/dissatisfaction (CS/D).
Customer satisfaction is considered to have a strong relationship with disconfirmation
(D), which is defined as the difference between a consumer’s pre-purchase expectations (E)
and the post-purchase performance (P) of a product or service (formulated as D = P  E)
(Oliver, 1980). Accordingly, customer satisfaction refers to ‘‘the summary psychological
state resulting when the emotion surrounding disconfirmed expectations is coupled with the
consumer’s prior feelings about the consumption experience’’ (Oliver, 1981). Because a cus-
tomer’s expectation and perception of performance can vary from one to another, discon-
firmation can be positive when actual performance is higher than pre-expectation and the
consumer is satisfied; or disconfirmation can be negative when perceived performance falls
short of pre-expectations and thus, the consumer will be dissatisfied. More specifically,
pre-expectations should have a negative impact on disconfirmation since lower pre-expec-
tations are more likely to be exceeded by perceived performance and cause positive discon-
firmation. On the contrary, perceived performance should have positive impact on
disconfirmation since higher perceived performance is more likely to surpass pre-expecta-
tion and therefore lead to positive disconfirmation.
The hypothetical relationship between disconfirmation and customer satisfaction has
become a key point in the area of consumer research and has been examined in a great
deal of empirical studies (for a review, see Oliver, 1997; Yi, 1990). Moreover, many
researchers suggest that expectations should have a direct effect on satisfaction (Oliver,
1981) and perceived performance (Churchill & Suprenant, 1982; Oliver & Burke, 1999;
Spreng et al., 1996). These relationships can result from an assimilation effect whereby
expectations establish the baseline or standard of comparison for the consumers to make
evaluative judgments about the product or service (Oliver, 1993). Prior researchers also
found that perceived performance has a significant effect on customer satisfaction (Chur-
chill & Suprenant, 1982; Tse & Wilton, 1988). In addition, Oliver (1997) calls this aug-
mented model as the ‘‘expectancy disconfirmation with performance model’’.

2.2. Ambiguity of the expectation construct

Although many researchers provide adequate evidence to support the primary assump-
tions of the expectancy-confirmation paradigm (Oliver, 1997; Yi, 1990), some empirical
studies have revealed mixed findings with regard to the relationships among expectations,
disconfirmation, performance, and customer satisfaction (Halstead, Hartman, & Schmidt,
1994). Some scholars suggest that such a paradox might result from the ambiguous defi-
nitions of expectation (Boulding, Kalra, Staelin, & Zeithaml, 1993; Teas, 1993; Van Dyke,
Kappelman, & Prybutok, 1997). Summarizing from the various definitions of expectation,
there are three diverse types of expectations which have been discussed: the ‘‘will expecta-
tion’’, the ‘‘should expectation’’, and the ‘‘ideal expectation’’. The ‘‘will expectation’’
focuses on forecasting or predicting future performance and is referred to as a customer’s
beliefs of what will happen in the post-purchase duration. The ‘‘should expectation’’ estab-
lishes a normative standard for performance and is referred to as what a customer believes
should happen in the next service encounter. The ‘‘ideal expectation’’ figures out the opti-
mal performance and is referred to as what a customer wants in an ideal sense.
In the classic ideal view of the expectation confirmation model, disconfirmation is
expected to have a positive relationship with customer satisfaction only when perceived
2808 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822

performance is less than or equal to ‘‘ideal expectations’’ (Parasuraman, Zeithami, &

Berry, 1994). This specification could be problematic since a customer’s ideal expectation
is at a finite level and thus, disconfirmation would be expected to have a negative relation-
ship with customer satisfaction when perceived performance exceeds expectation (Teas,
1993; Van Dyke et al., 1997). On the other hand, perceived performance can be influenced
by ‘‘will expectation’’ and ‘‘should expectation’’ but in an opposite manner (Boulding
et al., 1993; Van Dyke et al., 1997). Customers with higher ‘‘should expectation’’ would
more likely have lower perception of performance whereas higher ‘‘will expectation’’ could
increase the ultimate perception of performance. Moreover, some studies even argue that
‘‘will expectation’’ cannot be discriminated from the perceived performance in the tradi-
tional expectation disconfirmation model (Teas, 1993).
Customers’ expectation concerning the desired attributes of a product or service may
change with experience (Halstead & Page, 1992). It may cause a strong correlation
between expectations and perceived performance when both are measured in a post-pur-
chases stage (Page & Spreng, 2002). Oliver and Burke (1999) suggest that expectations be
measured before consumption since expectations influenced by performance are received
after consumption. Therefore, in the original design of the expectation confirmation
model, a customer’s expectation is measured before purchase whereas perceived perfor-
mance is measured after the consumption experience so that the correlation should be
low enough to discriminate expectation from performance.

2.3. Application of EDM in IS continuance

In the application context of information system (IS) continuance, researchers have

integrated the concepts of TAM (Davis, 1989) into the expectation disconfirmation model
to explain and analyze IS continuance behavior based on users’ expectations of systemic
attributes (Bhattacherjee, 2001). TAM posits that a user’s behavioral intention to initially
adopt an innovative information system is jointly determined by his/her attitude towards
using the system and perceived usefulness, whereas attitude is a direct function of per-
ceived usefulness and perceived ease of use. Perceived usefulness refers to ‘‘people’s salient
beliefs that using the technology will enhance his or her job performance’’. Perceived ease of
use refers to ‘‘people’s salient beliefs that using the system will be free of effort’’. Likewise,
perceived usefulness and perceived ease are in fact two important motivators affecting a
user’s post-adoption decision since the user should reevaluate his/her early acceptance
decision during post-adoption stage and make their judgment on IS continuance (Rogers,
1995). Thus, some researchers view continuance as an extension of initial acceptance
behavior and propose that both initial acceptance and continued use should employ
the same set of variables to explain the decision process and criterion (Bhattacherjee,
2001).
According to the assumptions of the expectation disconfirmation model, disconfirma-
tion is calculated as the difference between pre-consumption expectation and experienced
performance (Oliver, 1980). However, when expectation and performance are both mea-
sured in the post-purchase stage, post-consumption expectation should be influenced by
perceived performance and thus it will cause a strong correlation between these two con-
structs (Oliver & Burke, 1999; Page & Spreng, 2002). Hence, in the revised expectation dis-
confirmation model, the construct of pre-consumption variables are removed from the
theoretical model since the effects of pre-consumption expectation are already captured
 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822 2809

within the disconfirmation and satisfaction constructs, whereas the post-consumption

variables are recruited to capture the effects of both post-consumption expectation and
perceived performance (Bhattacherjee, 2001). Besides, rational users may try to regulate
their perception to remedy cognitive dissonance or psychological tension caused by dis-
confirmation of pre-acceptance expectation. Disconfirmation during actual use should
affect post-consumption expectation such as perceived usefulness and perceived ease of
use based on the cognitive dissonance theory (Bhattacherjee, 2001; Festinger, 1957).

2.4. Theory of planned behavior (TPB)

The development of TPB is originally based on the theory of reasoned action (TRA)
(Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975) which is designed to explain almost
any human behavior and has been proven successful in predicting and explaining human
behavior across various application contexts (Davis, Bagozzi, & Warshaw, 1989). Accord-
ing to TRA, a person’s actual behavior in performing a certain action is directly guided by
his or her behavioral intention, which in turn is jointly determined by the subjective norm
and attitude toward the behavior. By definition, behavioral intention is a measure of the
strength of one’s willingness to try while performing certain behaviors (Ajzen, 1991).
Grounded on the effort of TRA, TPB is proposed to eliminate the limitations of the ori-
ginal model when dealing with behavior over which people have incomplete volitional con-
trol. In fact, TPB differs from TRA in its addition of perceived behavior control, which
potentially has a direct effect on behavioral intention (Ajzen, 1991).
Attitude (A) refers to ‘‘the degree of a person’s favorable or unfavorable evaluation or
appraisal of the behavior in question’’ (Fishbein & Ajzen, 1975). Ajzen (1991) further
described that a favorable or unfavorable attitude has a direct proportion to the strength
of the behavioral beliefs about likely consequences and can be formulated with an expec-
tancy-value model. Accordingly, attitude (A) is equated with the attitudinal belief (Bi) link-
ing the behavior to a certain outcomeP as weighted by an evaluation of the desirability of
the outcome (ei) in question A = Biei (Ajzen, 1991).
Subjective norm refers to ‘‘the perceived social pressure to perform or not to perform the
behavior’’ (Ajzen, 1991). In other words, subjective norm is related to the normative beliefs
about the expectation from other people. It can be formed as the individual’s normative
belief (nbj) concerning a particular reference
P weighted by the motivation to comply with
the referent (mcj) in question SN = nbjmcj. Although subjective norm can be separated
into informational and normative influences (Karahanna, Straub, & Chervany, 1999), typ-
ical applications of TPB consider subjective norm to include only the normative influence.
This limitation may cause the insignificant relationship between the factors of subjective
norm and behavioral intention (Mathieson, 1991; Taylor & Todd, 1995b). Besides, Venk-
atesh and Davis (2000) found that behavioral intention would positively influence the sub-
jective norm in a mandatory usage context, whereas the effect was insignificant in
voluntary contexts.
Perceived behavioral control refers to ‘‘people’s perception of ease or difficulty in per-
forming the behavior of interest’’. It is associated with the beliefs about the presence of con-
trol factors that may facilitate or hinder the performance of the behavior (Ajzen, 2002).
Thus, control beliefs about resources and opportunities are associated with an underlying
perceived behavioral control, which can be formed as the Pcontrol beliefs (cbi) are weighted
by the perceived power of the control factor (pi) PBC = cbjpj. Control factors can be fur-
2810 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822

ther categorized into internal and external constraining factors, while internal control
related to knowledge/self-efficacy and external control relates to the environment (Ajzen,
1991).

2.5. Synthesis of EDM and TPB

The relationship between EDM and TPB can be examined from the perspectives of atti-
tude, subjective norm, and perceived behavior control. While some researchers consider
satisfaction to be synonymous with attitude (LaTour & Peat, 1979), the conceptual defi-
nitions of satisfaction and attitude are pragmatically different. For example, Hunt
(1977) defines attitude as an emotion but satisfaction as an evaluation of that emotion.
Oliver (1980, 1981) regarded satisfaction as a transient and experience-specific effect with
the evaluation of pre-consumption attitude, while attitude is a more enduring effect.
Therefore, satisfaction should causally precede post-consumption attitude. For synthesiz-
ing the concepts of TPB into EDM, the theoretical construct of attitude should be utilized
instead of satisfaction, since satisfaction in post-consumption stage is an evaluation of pre-
consumption attitude.
Subjective norm is proposed as having a direct influence on the perceived usefulness
through internalization mechanisms (Venkatesh & Davis, 2000). Internalization refers to
how an individual incorporates a referent’s belief into his/her own belief structure (Kel-
man, 1958; Warshaw, 1980). Based on internalization, if a superior or peers suggest that
an information system might be useful, a person may tend to believe that it is useful. On
the other hand, perceived behavioral control has a strong relation with perceived ease of
use, since users’ judgments of ease or difficulty of use of a system will correlate to both
internal and external control factors (Ajzen, 1991; Venkatesh, 2000). In the IS usage con-
text, internal control usually refers to computer self-efficacy whereas external control is
conceptualized as the individual perception of technology and resource facilitating condi-
tions (Taylor & Todd, 1995b). Much empirical evidence has indicated that computer self-
efficacy is a significant motivator of system-specific perceived ease of use (Venkatesh &
Davis, 1996). Moreover, an individual’s general perception of technology and resource
facilitating conditions serves as situational anchors in the formation of perceived ease of
use (Venkatesh, 2000). Thus, perceived behavior control will be an important motivator
in the formation of system-specific perceived ease of use.

3. Research model

A customer’s behavioral intention towards continued use of online services should

incorporate customer satisfaction of system features and the influence of social system
and personal characteristics. Although the expectation disconfirmation model can prop-
erly reflect customer satisfaction in terms of the disconfirmation between the expectation
of system features and perceived performance, there are no prior works to clarify the
impact of an individual’s psychological state and normative beliefs on his or her behav-
ioral intention and perception of system performance. Therefore, this study attempts to
synthesize the essence of the expectation disconfirmation model and TPB to predict and
explain a customer’s behavioral intention toward online service continuance. Subjective
norm and perceived behavioral control constructs are recruited from TPB to capture
the effects of normative beliefs in a social system and personal psychological state respec-
 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822 2811

tively. The expectation disconfirmation model can be applied to indicate the effect of cus-
tomer satisfaction on behavioral intention, while perceived usefulness and perceived ease
of use are recruited from TAM to reflect a customer’s expectation of systemic features.
However, this study focuses on measuring a user’s post-consumption expectation
instead of pre-consumption expectation since the effects of pre-consumption expectation
are already found within the disconfirmation and satisfaction constructs (Bhattacherjee,
2001). In addition, post-consumption expectation is more convenient to measure with
other constructs in the same cross-section. Also, perceived performance constructs are
omitted for this proposed research model, since post-consumption expectation should
have strong correlation with perceived performance (Oliver & Burke, 1999; Page & Spr-
eng, 2002). To enhance the correlation between post-consumption expectation and per-
ceived performance, ‘‘will expectation’’ is measured in this study with the understanding
that ‘‘will expectation’’ cannot be separated from the perceived performance (Teas,
1993). The research model with all theoretical constructs is provided in Fig. 1, whereas
the hypothetical relationships among theoretical constructs are summarized in Table 1.

H8
PU SN

H1
H9 H5
H2

H6 H3
H10 DIS SAT INT

H11 H7
H4
H12
EOU PBC

Fig. 1. Research model.

Table 1
Research hypotheses
Hypothesis 1 Subject norm will positively affect behavioral intention
Hypothesis 2 Perceived usefulness will positively affect behavioral intention
Hypothesis 3 Satisfaction will positively affect behavioral intention
Hypothesis 4 Perceived behavior control will positively affect behavioral intention
Hypothesis 5 Perceived usefulness will positively affect satisfaction
Hypothesis 6 Disconfirmation will positively affect satisfaction
Hypothesis 7 Perceived ease of use will positively affect satisfaction
Hypothesis 8 Subjective norm will positively affect perceived usefulness
Hypothesis 9 Disconfirmation will positively affect perceived usefulness
Hypothesis 10 Perceived easy of use will positively affect perceived usefulness
Hypothesis 11 Disconfirmation will positively affect perceived ease of use
Hypothesis 12 Perceived behavior control will positively affect perceived ease of use
2812 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822

4. Research design

To examine the synthetic effect of the expectation confirmation model and TPB on an
individual’s intention towards online service continuance, a survey of the experience in
usage of Cyber University System constituted by National Sun Yat-Sen University of Tai-
wan was employed to test out the research model. The instrumentation, sampling, and
scale validation process are described in the subsections below.

4.1. Instrumentation

The questionnaire is composed of two parts; the first part used nominal scales to collect
basic information pertaining to the respondents while the second part was comprised of
the theoretical constructs of EDM and TPB to measure respondents’ perception regarding
the continued use of online services. All constructs were originally adopted from prior
research, whereas some appropriate revisions were applied to make the measurement fit
the research context. The measure of each theoretical construct consisted of multiple quiz-
zes utilizing a 7-point Likert scale ranging from one (strongly disagree) to seven (strongly
agree). The measurement items are presented in Table 2.

4.1.1. Basic information

Basic information including gender, age, education, occupation, experience in using
CUS, total number of courses taken, and total hours of CUS usage a week were collected
to realize the characteristics of respondents. All characteristics of respondents were mea-
sured with a single quiz formalized in a nominal or ordinal scale.

4.1.2. Theoretical constructs of EDM and TPB

This part of the questionnaire is comprised of the measurements of EDM and TPB. The
measurement of EDM was developed on the constructs of perceived usefulness, perceived
ease of use, disconfirmation, satisfaction, and intention towards online service continu-
ance. As shown in Table 2, perceived usefulness and perceived ease of use were both mea-
sured using a 4-item scale adapted from Venkatesh and Davis (1996, 2000).
Disconfirmation, satisfaction, and behavioral intention were, however, adopted from
Bhattacherjee (2001). In specific, these aforementioned three parts were measured using
a 3-item scale, 4-item scale, and 3-item scale respectively. On the other hand, the measure-
ment of TPB was based on the constructs of subjective norm and perceived behavior con-
trol. Subjective norm and perceived behavior control were adapted from Taylor and Todd
(1995a) and Bhattacherjee (2000), both of which include three items.

4.2. Sample organizations and respondents

Cyber University System (CUS) founded by National Sun Yat-Sen University is a full-
functional, famous, and successful e-learning system in Taiwan and thereby it was chosen
as the research context. All of the 2014 formal students were included in the sample frame
and invited to participate in the e-mail survey. The students were asked to complete a ques-
tionnaire recording their experience in usage of the online e-learning system. Because the
application data of a student is private and forbidden to distribute, the administration of
CUS provided a survey service for academicians to collect the empirical data via e-mail.
 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822 2813

Table 2
Items used to measure research constructs
Usefulness
Definition: People’s salient beliefs that using the technology will enhance his or her job performance (Davis, 1989)
PU1 Using the CUS improves my performance in my learning
PU2 Using the CUS improves my productivity in my learning
PU3 Using the CUS enhances my effectiveness in my learning
PU4 I find the CUS to be useful in my learning
Easy of use
Definition: People’s salient beliefs that using the system will be free of effort (Davis, 1989)
PEOU1 My interaction with the CUS is clear and understandable
PEOU2 Interaction with the CUS does not require a lot of my mental effort
PEOU3 I find it easy to get the CUS to do what I want it to do
PEOU4 I find the CUS to be easy to use
Disconfirmation
Definition: The difference between a consumer’s pre-purchase expectations and the post-purchase performance of
a product or service (Oliver, 1980)
DIS1 My experience with using CUS was batter than what I expected
DIS2 The service level provide by CUS was better than what I expected
DIS3 Overall, most of my expectations from using CUS were confirmed
Satisfaction
Definition: The summary psychological state resulting when the emotion surrounding disconfirmed expectations
is coupled with the consumer’s prior feelings about the consumption experience (Oliver, 1981)
SAT1 My overall experience of CUS use was very satisfied
SAT2 My overall experience of CUS use was very pleased
SAT3 My overall experience of CUS use was very contented
SAT4 My overall experience of CUS use was absolutely delighted
Subjective norm
Definition: The perceived social pressure to perform or not to perform the behavior (Ajzen, 1991)
SN1 People important to me supported my use of e-brokers
SN2 People who influence my behavior wanted me to use e-brokers instead of any alternative means
SN3 People whose opinions I valued preferred that I use e-brokers for managing investments
Perceived behavioral control
Definition: People’s perception of ease or difficulty in performing the behavior of interest (Ajzen, 1991)
PBC1 I would be able to use e-brokers well for managing personal investment
PBC2 Using e-brokers was entirely within my control
PBC3 I had the resources, knowledge, and ability to use e-brokers
Continuance intention
Definition: The individual’s intention to continued use the information system (Bhattacherjee, 2001; Davis, 1989)
INT1 I intend to continue using CUS rather than discontinue it use
INT2 My intentions are to continue using CUS than use any alternative means
INT3 If I could, I would like to continue using CUS as much as possible

The survey process is further described below. First, the investigator gets approval from the
administrator of CUS and sends an invitation letter out through e-mail to express the need
for collection of empirical data concerning students’ experience in using the online service.
The administrator of CUS, then, forwarded the message to registered formal students via e-
mail and instructed the receivers to click a hyperlink and redirected them to an online ques-
tionnaire system. Consequently, 2014 invitation letters were sent to registered formal stu-
dents through e-mail. In order to improve the return rate, another follow-up invitation
letter was sent to non-responding students with the same aforementioned procedure after
2814 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822

2 weeks. Finally, 469 students had finished and returned the questionnaire. The demo-
graphic data of respondents in comparison with the sample frame was shown in Table 3.
Since some students did not respond to the invitation for some reasons, the actual sam-
ple received might not match perfectly the sample frame from which the actual sample was
originally drawn. Moreover, non-response of the elements might induce selection bias or

Table 3
Profile of respondents and sample frame
Characteristics Respondents Sample frame
Frequency Percent Frequency Percent
2
Gender (v : 0.097, p-value: 0.756)
Male 294 62.69 1232 61.17
Female 175 37.31 782 38.83
Age (v2: 0.438, p-value: 0.932)
<20 15 3.20 80 3.97
20  30 217 46.27 891 44.24
30  40 145 30.92 667 33.12
>40 92 19.62 376 18.67
Education (v2: 0.994, p-value: 0.911)
High school 9 1.92 39 1.94
Technical 52 11.09 214 10.63
College 278 59.28 1281 63.60
Master 120 25.59 438 21.75
Doctoral 10 2.13 42 2.09
Occupation (v2: 4.177, p-value: 0.759)
Student 164 34.97 626 31.08
Academic 69 14.71 302 15.00
Manufacturing 59 12.58 316 15.69
Profession 56 11.94 222 11.02
Business 27 5.76 200 9.93
Self-employed 16 3.41 74 3.67
Retiree 3 0.64 23 1.14
Other 75 15.99 251 12.46
Experience in CUS (v2: 2.213, p-value: 0.529)
<1 Semester 69 14.71 412 20.46
1–3 Semesters 286 60.98 1177 58.44
3–5 Semesters 73 15.57 270 13.41
>5 Semesters 41 8.74 155 7.70
Courses in CUS (v2: 1.194, p-value: 0.879)
1 course 154 32.84 738 36.64
2 courses 104 22.17 395 19.61
3 courses 87 18.55 401 19.91
4 courses 43 9.17 179 8.89
>4 courses 81 17.27 301 14.95
CUS usage a week (v2: 2.701, p-value: 0.609)
<3 h 248 52.88 1222 60.675
3–6 h 133 28.36 495 24.578
6–9 h 43 9.17 142 7.051
9–12 h 19 4.05 68 3.376
>12 h 26 5.54 87 4.320
 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822 2815

systematic non-response bias in the sampling procedure. To ensure the online question-
naire had not induced selection bias, the statistic data of a sample frame provided by
CUS administrator was compared to the demographic data of respondents. Each charac-
teristic was compared using a chi-square test in terms of its frequency distribution. The
results as shown in Table 3 which demonstrated that all characteristics of the respondents
had no significant differences with the sample frame. In addition, in order to test for sys-
tematic non-response bias, the respondent samples were divided into early and late seg-
ments, and these two groups of samples were correlated with their demographic data.
The results as shown in Table 4 also indicated that there was no significant systematic
non-response bias in the responding sample. Since this study placed an emphasis on the
continued use of online services and only people who had more than one-semester expe-
rience in using CUS were chosen as survey respondents. Therefore, 400 students were
finally chosen as valid respondents for further analyses. In summary, the response rate
was 23% and valid response rate was 20%.

4.3. Scale validation

The construct validity of the research instrument was assessed via Confirmatory Factor
Analysis (CFA) using AMOS for Windows. To perform a CFA, all of the constructs and
reflective indicators were depicted and composed as a measurement model in which all
constructs were allowed to correlate with each other. Generally speaking, the process of
validation should comprise of three steps as described below.
First, the measurement model should demonstrate goodness of fit for the empirical data
and meet the requirements of certain indexes. For example, chi-square normalized by
degrees of freedom (v2/df) should be less than five (Bentler, 1989); adjusted goodness-
of-fit index (AGFI) should be larger then 0.8, goodness-of-fit index (GFI), normed fit
index (NFI), and comparative fit index (CFI) should all exceed 0.9, and root mean square
error (RMSE) should be less than 0.10 (Henry & Stone, 1994; Scott, 1994).
Secondly, convergent validity should fulfill the following criteria simultaneously: (1)
factor loading (k) for an item should exceed 0.7 and significant in T-test, (2) construct reli-
ability should be greater than 0.8, and (3) average variance extracted (AVE) for a con-
struct should be larger than 0.5 (Fornell & Larcker, 1981).
Finally, discriminant validity can be assessed using a series of v2 difference tests (Bago-
zzi & Phillips, 1982). Each test must involve two separate analyses: the unconstrained
model and the constrained model. In the unconstrained model, all constructs are freely
correlated. However, in the constrained model, the covariance of a certain two constructs
is set equal to one while other constructs remain freely correlated. Two constructs are
claimed as having well discriminant validity if the v2 difference between the two models
is significant.

Table 4
Correlation tests for systematic non-response bias
Gender Age Education Occupation Experience Courses CUS usage
Correlation 0.049 0.045 0.085 0.031 0.032 0.004 0.041
P-value 0.457 0.491 0.193 0.638 0.631 0.958 0.529
N 234 234 234 234 234 234 234
2816 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822

Table 5
Confirmatory factor analysis for convergent validity
Construct Item loading Construct reliability AVE Factor correlations
INT SN SAT PBC PU EOU DIS
INT 0.80–0.94 0.85 0.66 1.00
SN 0.74–0.91 0.84 0.64 0.57 1.00
SAT 0.96–0.97 0.97 0.90 0.79 0.52 1.00
PBC 0.92–0.94 0.94 0.84 0.63 0.42 0.65 1.00
PU 0.86–0.93 0.93 0.78 0.67 0.48 0.63 0.60 1.00
EOU 0.84–0.89 0.90 0.70 0.58 0.44 0.66 0.68 0.63 1.00
DIS 0.91–0.93 0.92 0.79 0.77 0.60 0.88 0.62 0.62 0.62 1.00
v2/df = 1.71, GFI = 0.93, AGFI = 0.91, NFI = 0.97, CFI = 0.99, RMSE = 0.040.
All item loadings (k) in CFA model were significant at p = 0.001 level.

Table 6
Chi-square difference tests for discriminant validity
Construct INT SN SAT PBC PU EOU
SN 505.07
SAT 431.31 562.98
PBC 635.47 618.51 937.55
PU 587.24 579.78 1248.79 977.09
EOU 685.08 613.89 884.25 812.87 892.36
DIS 416.56 485.35 389.57 896.60 885.70 864.97
None-constrained model: v2 = 396.49, degrees of freedom = 231.
Constrained model: degrees of freedom = 232.
All v2 differences were significant at p = 0.001 level.

In the current CFA model, the all fit index (v2/df = 1.71, GFI = 0.93, AGFI = 0.91,
NFI = 0.97, CFI = 0.99, RMSE = 0.040) suggested an adequate model fit for the empir-
ical data. Furthermore, all constructs have also shown good convergent validity as all of
the criteria for convergence were met. As shown in Table 5, all factor loadings (k) of the
reflective indicators were significant at p = 0.001 and exceeded 0.7 while construct reliabil-
ities ranged from 0.84 to 0.97, and AVE ranged from 0.64 to 0.90.
For examining discriminant validity, a series of v2 difference tests are shown in Table 6.
All of the v2 differences tests were significant at p = 0.001 level. Accordingly, the results
showed that the unconstrained CFA model was significantly better than any constrained
models with the possible combination of two constructs. In other words, the results dem-
onstrated adequate discriminant validity for all constructs.

5. Hypotheses testing

The structured equation model (SEM) was used to test the 12 hypotheses proposed in
this study. The hypothetical model (see Fig. 2) was depicted using visual tools provided by
AMOS. Each indicator was connected to its theoretical construct in a reflective manner as
well as linked accordingly to the hypothesis. Exogenous constructs including disconfirma-
tion, subjective norm, and perceived behavioral control were freely correlated. The vari-
ance of each exogenous construct and the error variance of each endogenous construct
 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822 2817

PU 0.13*
SN
R2=0.50
0.16**
0.32** 0.06
0.21**

0.74** SAT 0.50** INT

0.38** DIS
R2=0.79 R2=0.70

0.32** 0.17**
0.12**
EOU 0.49**
PBC
R2=0.53

Fig. 2. Standardized solution of SEM model. Note: v2/df = 1.80, GFI = 0.92, AGFI = 0.90, NFI = 0.97,
CFI = 0.98, RMSE = 0.042. *p < 0.01, **p < 0.001.

were initially set to one. Model estimation was executed through use of the maximum like-
lihood technique.
The evaluation of the estimative model solution shell comprised of three steps. First, the
theoretical model shell has goodness of fit to the empirical data while the same fit indexes
applied in assessing measurement model were met. Second, the path significance and stan-
dardized path coefficient corresponding to each hypothesis of the research model were
examined. Finally, the total variance in each exogenous construct was explained in terms
of its dependent constructs.
For the hypothetic SEM model, v2/df is 1.80, GFI is 0.92, AGFI is 0.90, NFI is 0.97,
CFI is 0.98, and RMSE is 0.042. All fit indexes have suggested adequate model fit between
the research model and the empirical data. Furthermore, all the paths are significant at
p = 0.01 except the relationship between perceived usefulness and disconfirmation which
is not supported. The results actually sustain the EDM and TPB quite well.
Behavioral intention towards online service continuance is predicted by satisfaction
(b = 0.50), perceived usefulness (b = 0.21), subjective norm (b = 0.16), and perceived
behavior control (b = 0.12), which jointly explained 70% of the variance in intention.
As exhibited in prior research, satisfaction is the most significant motivator of behavior
intention. Comparing the results with prior EDM-based studies in IS continuance, the
impact of perceived usefulness and satisfaction on intention are similar to what had been
indicated in Bhattacherjee (2001), whereas the explanatory power of the current research
model is higher than the previous model with R2BI ¼ 0:41. It is possibly because the subjec-
tive norm and perceived behavioral control devote its direct effect to explain the error var-
iance in intention.
Customer satisfaction is predicted by disconfirmation (b = 0.74) and perceived ease of
use (b = 0.17) with 79% of the total variance explained. The effect of disconfirmation on
satisfaction is much greater than perceived ease of use. It reveals that disconfirmation is an
important indicator of customer satisfaction. However, the influence of perceived useful-
ness on satisfaction is not significant in this study. This result is obviously inconsistent
with the finding of Bhattacherjee (2001), which indicated that perceived usefulness should
2818 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822

have a significant impact on customer satisfaction. The reason for this inconsistency may
lie in the research design and context and can be explained by two various perspectives.
First, as Oliver (1993) postulated that pre-consumption expectation should have a direct
effect on satisfaction since people tend to make evaluative judgments with what they
expect to belief. Nevertheless, the effect of post-expectation should become weaker as dis-
confirmation assimilates the likely consequence of evaluative judgments in the post-con-
sumption stage. Secondly, perceived usefulness can directly affect behavioral intention
because people believe adopting certain information systems will increase their job perfor-
mance and organizational rewards (Davis, 1989). Thus, people might engage in a behavior
to increase rewards without adjusting their attitude. More specifically, people might intend
to perform certain behavior continuously in the hope of obtaining rewards from an orga-
nization, even though the behavior precipitates displeasure (Mathieson & Chin, 2001). In
this study, adopting e-learning is obviously a job requirement and has a strong relation-
ship with organizational rewards. Hence, perceived usefulness can directly determine
behavioral intention but not necessary be mediated by satisfaction. Accordingly, the insig-
nificant effect of perceived usefulness on satisfaction is quite reasonable in the current
research design and context.
Perceived usefulness is determined by perceived ease of use (b = 0.38), disconfirmation
(b = 0.32), and subjective norm (b = 0.13), which jointly explain 50% of error variance in
perceived usefulness. This result is consistent with prior studies (Bhattacherjee, 2001;
Davis, 1989) which indicate perceived ease of use and disconfirmation are two significant
motivators of perceived usefulness. Moreover, subjective norm is significant but is a
weaker indicator of perceived usefulness. It means that the effect of ‘‘word of mouth’’ still
has a significant impact on perceived usefulness in the post-consumption stage. Finally,
perceived behavior control (b = 0.49) and disconfirmation (b = 0.32) can jointly determine
perceived ease of use and explain 53% of error variance. It reveals that control factors con-
cerning computer self-efficacy and facilitating conditions should be the first consideration
of system engineers/developed to build system ease of use.

6. Research findings

This study attempts to harmonize the essence of EDM and TPB to propose an inte-
grated model for explaining customers’ continued use of online services in terms of sys-
temic, personal, and social factors. Some interesting findings that were not discussed in
prior research are also discovered from the current research model.
First, subjective norm and perceived behavioral control are identified as two significant
motivators of behavioral intention towards continuous use of online services. With subjec-
tive norm and perceived behavioral control, the total variance explained in behavioral
intention increases from 41% (in Bhattacherjee, 2001) to 70%. This reveals that synthesiz-
ing the theoretical constructs of TPB with EDM can definitely increase the accuracy of an
integrated model to predict and explain customers’ behavioral intentions.
Second, perceived usefulness should have a direct effect on behavioral intention since a
customer believes that continuing use of online services will increase their job performance
and organizational reward. Moreover, continued use of online services may not necessarily
have a connection with customer satisfaction in certain application contexts. It might indi-
cate that the first consideration of system design is to promote a customer’s intention
towards online service continuance by focusing on practicability and profitability to
 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822 2819

increase customers’ job performance and organizational rewards. For e-learning systems
constituted by educational organizations, improving students’ professional skill and pro-
viding certain academic degrees should be the most important motivators to promote the
continuance intention.
Thirdly, a customer’s behavioral intention is mainly determined by satisfaction, which
in turn is motivated by disconfirmation and perceived ease of use. Moreover, disconfirma-
tion and perceived ease of use can jointly explain 79% of error variance in satisfaction. It
reveals that the hypothetical model can fully represent the spirit of EDM despite that pre-
consumption expectation was used instead of post-consumption expectation and perceived
performance was omitted from the hypothetical model. With these revisions, the current
integrated model is fit for cross-sectional research since all theoretical constructs are able
to be measured in the same time.
Fourth, though prior research had indicated that perceived ease of use should have
weaker or no influence on attitude in the later stages of adoption (Davis et al., 1989; Kara-
hanna et al., 1999), perceived ease of use is a significant but weaker motivator of satisfac-
tion in this study. This mainly because satisfaction is an experience-specific effect in the
evaluation of pre-consumption attitude (Oliver, 1980, 1981), ease of use is still an impor-
tant systemic feature concerning the customers’ evaluative judgment on IS continuance.
Moreover, since perceived behavioral control is a strong indicator of perceived ease of
use, the system designer should focus efforts on improving facilitating conditions of online
services in order to promote customers’ perception of ease of use.

7. Research limitations

This empirical study has several limitations existing in the research design and research
context. First, for screening suitable respondents who were involved in the decision of
whether to continue use of online services, a single-item measure with an ordinal scale
was employed to examine a user’s experience in using CUS. Students who had more then
one-semester experience in using CUS were chosen as valid respondents. This might create
problems since no explicit boundary had been proposed to discriminate the initial adopters
from continued users. However, some referable researches concerning the longitudinal
studies of a user’s intention towards IS acceptance can support the criterion for discrim-
ination. According to the longitudinal TAM-based studies by Davis et al. (1989) and
Venkatesh and Davis (2000), 14 weeks and six months were used to distinguish the
long-term behavior of adoption from initial acceptance. Because continuance occurs in
the post-consumption stage and the product or service performance should be perceived
after adoption, an appropriate boundary to discriminate between initial acceptance and
continued use should be greater than the latest approved time of adoption. Thus, using
six month or one semester (two semesters a year in Taiwan) as a boundary to discriminate
continued users from initial adopters is quite reasonable.
Second, about 63% of the respondents were male in this empirical study. It reveals that
the gender distribution is not symmetric between men and women. Therefore, the results
of the current empirical study might tend to model the specific behavior of men rather than
general behavior of the customer. Much evidence has shown that gender differences can
cause discrepancies in the effects of attitude, perceived behavioral control, and subjective
norm on a user’s behavioral intention (Armitage, Norman, & Conner, 2002). Other refer-
ences also show that the effect of ease of use and perceived usefulness on behavioral
2820 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822

intention can be moderated by gender difference (Gefen & Straub, 1997). Accordingly,
further research may be needed to examine the moderating effect of gender difference
on the behavioral performance of an online user.

8. Conclusions

This study has devoted much effort on developing an integrated model to predict and
explain an individual’s continued use of online services based on the concepts of the expec-
tation disconfirmation model and the theory of planned behavior. Empirical data was col-
lected from a field survey of CUS users to verify the fitness of the hypothetical model. The
measurement model indicates the theoretical constructs have adequate reliability and
validity, while the structured equation model is testified to having a high model fit for
the empirical data. Study’s findings show that a customer’s behavioral intention towards
e-service continuance is mainly determined by customer satisfaction and additionally
affected by perceived usefulness and subjective norm. Generally speaking, the integrated
model can fully reflect the spirit of the expectation disconfirmation model and take advan-
tage of the planned behavior theory. Upon considering the impact of systemic features,
personal characteristics, and social influence on customer behavior, the integrated model
had a better explanatory power than other EDM-based models proposed in prior research.
Some suggestions derived from the empirical studies are described as follows. First, per-
ceived usefulness can directly affect behavioral intention without mediating customer sat-
isfaction in certain application contexts. Therefore, providing useful services that can
improve job performance and organizational rewards is the first consideration in promot-
ing a customer’s continuance intention. Second, though some researchers indicate that the
effect of perceived ease of use should become weaker in the later stages of IS adoption, it
still has a significant impact on customer satisfaction. Since behavioral intention and per-
ceived ease of use are determined by perceived behavioral control, some consideration
concerning certain facilitating mechanisms to help customers use online services with ease
might be useful for promoting customers’ retention. Third, subjective norm is verified as
having a significant influence on behavioral intention. Thus, advertisement, propaganda,
and periodic review should be applicable tactics for encouraging continued use of online
services. However, to apply this integrated model to a more generalized application con-
text, further studies such as the impact of gender difference on the variation of users’
behavioral intention are needed to examine the appropriateness of the hypothetical model.

References

Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Process, 50,
179–211.
Ajzen, I. (2002). Perceived behavior control, self-efficacy, locus of control, and the theory of planned behavior.
Journal of Applied Social Psychology, 32, 1–20.
Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ:
Prentice-Hall.
Armitage, C. J., Norman, P., & Conner, M. (2002). Can the theory of planned behavior mediate the effects of age,
gender and multidimensional health locus of control? British Journal of Health Psychology, 7, 299–316.
Bagozzi, R. P., & Phillips, L. W. (1982). Representing and testing organizational theories: A holistic construal.
Administrative Science Quarterly, 27, 459–489.
Bentler, P. M. (1989). EQS structural equations program manual. Los Angeles: BMDP Statistical Software.
 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822 2821

Bhattacherjee, A. (2000). Acceptance of E-commerce services: the case of electronic brokerages. IEEE
Transactions on System, Man, and Cybernetics—Part A: Systems and Humans, 20(4), 411–420.
Bhattacherjee, A. (2001). Understanding information system continuance: an expectation-confirmation model.
MIS Quarterly, 25(3), 351–370.
Boulding, W., Kalra, A., Staelin, R., & Zeithaml, V. A. (1993). A dynamic process model of service quality: from
expectations to behavioral intentions. Journal of Marketing Research, 30(1), 7–27.
Churchill, G. A., & Suprenant, C. (1982). An investigation into the determinants of customer satisfaction. Journal
of Marketing Research, 19, 491–504.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology.
MIS Quarterly, 13, 319–339.
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: a comparison of
two theoretical models. Management Science, 35, 982–1002.
Festinger, L. A. (1957). A theory of cognitive dissonance. Evanston, IL: Row and Peterson.
Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior: An introduction to theory and research.
MA: Addison-Wesley.
Fornell, C., & Larcker, D. F. (1981). Evaluating structural equations with unobservable variables and
measurement error. Journal of Marketing Research, 18, 39–50.
Gefen, D., & Straub, D. W. (1997). Gender differences in the perception and use of email: An extension to the
technology acceptance model. MIS Quarterly, 21(4), 389–400.
Halstead, D., Hartman, D., & Schmidt, S. L. (1994). Multi-source effects on the satisfaction formation process.
Journal of the Academy of Marketing Science, 22(2), 114–129.
Halstead, D., & Page, T. M. Jr., (1992). The effects of satisfaction and complaining behavior on consumer
repurchase intentions. Journal of Satisfaction, Dissatisfaction, and Complaining Behavior, 5, 1–11.
Henry, J. W., & Stone, R. W. (1994). A structural equation model of end-user satisfaction with a computer-based
medical information systems. Information Resources Management Journal, 7(3), 21–33.
Hunt, H. K. (1977). CS/D—overview and future research directions. In H. K. Hunt (Ed.), Conceptualization and
measurement of consumer satisfaction and dissatisfaction (pp. 455–488). Cambridge, MA: Marketing Science
Institute.
Karahanna, E., Straub, D. W., & Chervany, N. L. (1999). Information technology adoption across time: a cross-
sectional comparison of pre-adoption and post-adoption beliefs. MIS Quarterly, 23(2), 183–213.
Kelman, H. C. (1958). Compliance, identification, and internalization: three processes of attitude change. Journal
of Conflict Resolution, 2, 51–60.
Koufaris, M. (2002). Applying the technology acceptance model and flow theory to online consumer behavior.
Information Systems Research, 13(2), 205–224.
LaTour, S. A., & Peat, N. C. (1979). Conceptual and methodological issues in consumer satisfaction research. In
W. L. Wilkie (Ed.). Advances in consumer research (vol. 8, pp. 431–437). Ann Arbor, MI: Association for
Consumer Research.
Mathieson, K. (1991). Predicting user intentions: comparing the technology acceptance model with the theory of
planned behavior. Information System Research, 2(3), 173–191.
Mathieson, K., & Chin, W. W. (2001). Extending the technology acceptance model: the influence of perceived
user resources. Database for Advance in Information Systems, 32(3), 86–102.
McKinney, V., Yoon, K., & Zahedi, F. M. (2002). The measurement of Web-customer satisfaction: an
expectation and disconfirmation approach. Information System Research, 13(3), 296–315.
Oliver, R. L. (1980). A cognitive model for the antecedents and consequences of satisfaction. Journal of
Marketing Research, 17, 460–469.
Oliver, R. L. (1981). Measurement and evaluation of satisfaction processes in retail settings. Journal of Retailing,
57(3), 25–48.
Oliver, R. L. (1993). Cognitive, affective, and attribute bases of the satisfaction response. Journal of Consumer
Research, 20, 418–430.
Oliver, R. L. (1997). Satisfaction: a behavioral perspective on the consumer. New York, NY: McGraw-Hill.
Oliver, R. L., & Burke, R. R. (1999). Expectation process in satisfaction formation: a field study. Journal of
Service Research, 1(3), 196–214.
Page, T. J., Jr., & Spreng, R. A. (2002). Difference scores versus direct effects in service quality measurement.
Journal of Service Research, 4(3), 184–192.
Parasuraman, A., Zeithami, V. A., & Berry, L. L. (1994). Reassessment of expectations as a comparison in
measuring service quality: implication for further research. Journal of Marketing, 58(1), 111–124.
2822 C. Liao et al. / Computers in Human Behavior 23 (2007) 2804–2822

Patterson, P. G., Johnson, L. W., & Spreng, R. A. (1997). Modeling the determinants of customer satisfaction for
business-to-business professional services. Journal of the Academy of Marketing Science, 25(1), 4–17.
Rogers, E. M. (1995). Diffusion of innovations (4th ed.). New York: Free Press.
Scott, J. E. (1994). The measurement of information systems effectiveness: evaluating a measuring instrument.
Proceedings of the Fifteenth International Conference on Information Systems (pp. 111–128). Vancouver,
British Columbia.
Spreng, R. A., & Chiou, J. S. (2002). A cross-cultural assessment of the satisfaction formation process. European
Journal of Marketing, 36(7/8), 829–839.
Spreng, R. A., MacKenzie, S. B., & Olshavsky, R. W. (1996). A reexamination of the determinants of consumer
satisfaction. Journal of Marketing, 60(July), 15–32.
Taylor, S., & Todd, P. (1995a). Assessing IT usage: The role of prior experience. MIS Quarterly, 19(4), 561–570.
Taylor, S., & Todd, P. A. (1995b). Understanding information technology usage: a test of competing models.
Information System Research, 6(2), 144–176.
Teas, R. K. (1993). Expectations, performance evaluation, and consumers’ perceptions of quality. Journal of
Marketing, 57(4), 18–34.
Tse, D. K., & Wilton, P. C. (1988). Models of consumer satisfaction formation: an extension. Journal of
Marketing Research, 25(May), 204–212.
Van Dyke, T. P., Kappelman, L. A., & Prybutok, V. R. (1997). Measuring information system quality: concerns
on the use of the SERVQUAL questionnaire. MIS Quarterly, 21(2), 195–208.
Van Montfort, K., Masurel, E., & Van Rijn, I. (2000). Service satisfaction: an empirical analysis of consumer
satisfaction in financial services. The Service Industries Journal, 20(3), 80–94.
Venkatesh, V. (2000). Determinants of perceived easy of use: Integrating control, intrinsic motivation, and
emotion into the technology acceptance model. Information Systems Research, 11(4), 342–365.
Venkatesh, V., & Davis, F. D. (1996). A model of the antecedents of perceived ease of use: development and test.
Decision Science, 27(3), 451–481.
Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the technology acceptance model: four
longitudinal field studies. Management Science, 46(2), 186–204.
Violino, B. (1999). The leaders of e-business tools take charge of the new economy. InformationWeek, 62–83.
Warshaw, P. R. (1980). A new model for predicting behavioral intentions: an alternative to Fishbein. Journal of
Marketing Research, 17, 153–172.
Yi, Y. (1990). A critical review of consumer satisfaction. In V. A. Zeithaml (Ed.). Review of marketing (Vol. 4,
pp. 68–123). Chicago, IL: American Marketing Association.