Upload
romika-jain
View
216
Download
0
Embed Size (px)
Citation preview
8/2/2019 _Chen___Hirschheim_2004_
1/39
Info Systems J(2004) 14, 197235
2004 Blackwell Publishing Ltd
197
Blackwell Science, LtdOxford, UKISJInformation Systems Journal1350-1917Blackwell Publishing Ltd, 200414
3197235
Original Article
Examination of IS researchW Chen & R Hirschheim
A paradigmatic and methodologicalexamination of information systems researchfrom 1991 to 2001
WenShin Chen* & Rudy Hirschheim
*Bauer College of Business, University of Houston, Houston, TX 77204-6282, USA, and
Ourso College of Business Administration, Louisiana State University, Baton Rouge, LA
70803, USA, email: [email protected]
Abstract. The field of information systems (IS) has evolved for more than threedecades. Although many schools of thought have emerged and even become well
established, few historical analyses of research paradigms and methodologies
have been undertaken. One of the rare exceptions is Orlikowski & Baroudi (1991).
Yet, the IS research community has evolved substantially since 1991 in many
aspects. A variety of journal outlets have emerged and become well established.
More attention has been paid to paradigmatic and methodological issues. Political
and professional contexts have also changed noticeably. Therefore, it should be an
opportune time for the field to ask: What changes are manifested in journal pub-
lications? Is the field making progress regarding pluralism in IS research? How
will the fields publications practices change in the future? The purpose of thispaper is to investigate these questions and, in turn, reflect on the paradigmatic and
methodological progress made since 1991. We examined 1893 articles published
in eight major IS publication outlets between 1991 and 2001. Our findings suggest
that the long-term endeavours of interpretivist researchers might need to continue
because the paradigmatic progress appears somewhat inconsequential; positivist
research still dominates 81% of published empirical research. In particular, US
journals, as opposed to European journals, tend to be more positivist, quantitative,
cross-sectional and survey oriented. With respect to research design, survey
research is still the most widely used method (41%), although case studies have
gained substantial recognition (36%). Further, the increase of qualitative research
(30%), empirical studies (61%) and longitudinal cases (33%) at the expense of lab-
oratory experiments (18%) might suggest that IS researchers have become more
interested in obtaining scientific knowledge in real world settings. In summary, we
suggest that the field has been dominated by the positivist paradigm, despite calls
to the contrary. Indeed, if the field was to truly embrace pluralism, it would have to
find ways to fundamentally change the publication practices of the journal system,
including the current tenure and promotion system, which pose considerable
obstacles for the acceptance of alternative paradigms.
8/2/2019 _Chen___Hirschheim_2004_
2/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14
, 197235
198
Keywords:
IS research, paradigms, research methodology, positivism,
interpretivism
INTRODUCTION
There appears a growing interest as to how the information systems (IS) research community
evolves with respect to philosophical and methodological assumptions. Considerable attention
has been paid to the issues of research diversity and methodological pluralism during the past
two decades (e.g. Hirschheim & Klein, 1989; 1992; Galliers, 1991; Klein et al., 1991; Olaisen,
1991; Walsham, 1995a; Hirschheim et al., 1996; Wicks & Freeman, 1998; Cook & Brown,
1999). Specifically, alternative paradigms such as interpretivism have emerged in the field
(Walsham, 1995b) and have become more widely accepted even in traditionally positivistori-
ented journals such as MIS Quarterly
(
MISQ
) (Trauth & Jessup, 2000 ). Qualitative researchmethods, as opposed to the dominant quantitative methods, have also become more popular
in the field (Lee, 1989; Walsham, 1995a; Silverman, 1998).
Yet, the belief that alternative paradigmatic research exists within the IS field remains largely
anecdotal and conceptual (Hassard, 1988; Hirschheim & Klein, 1989; Klein et al
., 1991;
Olaisen, 1991). Little effor t has been made to empirically examine the philosophical and meth-
odological assumptions underlying IS journal publications. This is not to say that there are no
visible examples of alternative paradigmatic publications adopting Burrell & Morgans (1979)
four paradigms (functionalism, interpretivism, radical structuralism and radical humanism);
their differences with respect to ontological and epistemological assumptions have been widely
discussed in IS literature (cf. Hirschheim & Klein, 1989; Olaisen, 1991; Hirschheim et al
., 1995;
Walsham, 1995a).
While conceptual discussion brings attention to the issues surrounding alternative para-
digms, an empirical examination would help to reveal how such attention is reflected in journal
publications. After all, to discuss is one thing, to reflect on real
research practice is another. As
such, we extend the analysis of Orlikowski & Baroudi (1991) which is one of the rare empirical
examinations of IS researchs paradigmatic and methodological base. Orlikowski and Baroudi
examined 155 articles published between 1985 and 1989 in MISQ
, Communications of the
ACM
(
CACM
), Management Science
and Proceedings of the International Conferences on
Information Systems
(
ICIS
). Their findings indicated that the positivist paradigm overwhelm-
ingly dominated the IS research community (96.8%) whereas little attention was paid to the
interpretive paradigm (3.2%) and no empirical research work was done using a critical para-
digm (0%). With such an extreme positivist dominance, it was not surprising that methods such
as survey research (49.1%) and laboratory experiments (27.1%) that aim for deductive, hypo-
thetical reasoning and generalization, were the dominant research designs.
Such paradigmatic and methodological skewing in a research community might not be
healthy. Orlikowski & Baroudi (1991), thus, ar ticulated the essence of each paradigm and spec-
ified the appropriateness of employing other research methodologies to better investigate
research questions of interest. While the positivist paradigm could aim for replicability and gen-
8/2/2019 _Chen___Hirschheim_2004_
3/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14
, 197235
199
eralizability, the interpretivist paradigm could enhance the in-depth understanding of the phe-
nomenon examined. Each paradigm provides unique attributes for different purposes of
scientific inquiry; the same argument could also be applied to each methodology. The issue,
thus, should not be simply about paradigm wars (Tashakkori & Teddlie, 1998) but about theresearch questions investigated. Alternative paradigms or methodologies such as interpretiv-
ism and qualitative methods should be welcomed and encouraged because they provide dif-
ferent dimensions for research investigation that the positivist paradigm and survey methods
would not be able to accomplish.
Although Orlikowski & Baroudi (1991) brought attention to the paradigmatic and method-
ological issues of IS research to the community, their examination was conducted more than
a decade ago. Much has changed in the IS research community since 1990. First, many new
journals have been established to provide more research outlets and different perspectives.
For example, Information Systems Research
(
ISR
) launched its debut issue in 1990 and
became one of the top journals in the field.Accounting, Management, and Information Tech-nology
(
AMIT
) (
Information and Organization
after 2001), Information Systems Journal
(
ISJ
)
and European Journal of Information Systems
(
EJIS
) all started in 1991 and became some of
the major research outlets, especially in Europe. The recently established electronic journals,
Communications of the AIS
and Journal of the AIS
, are only a few years old and just finding
their respective niches.
Second, more attention has been paid to paradigmatic and methodological pluralism. Landry
& Banville (1992), for example, highlighted the issue of the IS field being more of a fragmented
adhocracy than a unified discipline. Hirschheim & Klein (2003) took this viewpoint one step fur-
ther to explain why the fragmented adhocracy occurred and how the IS community could pos-
sibly build a body of knowledge which might help establish the identity of IS as a discipline. In
addition, alternative research paradigms have been considered more acceptable in the field.
For example, Wicks & Freeman (1998) challenged the traditional dominant positivist position;
Walsham (1995a; 1995b) indicated the emergence of interpretivism; Klein et al
. (1991) advo-
cated the necessity of pluralism for any serious IS research agenda. Other discussions in rela-
tion to paradigmatic or methodological issues in recent years can also be found in Cook &
Brown (1999), Farhoomand & Drury (1999), Goles & Hirschheim (2000), Hirschheim & Klein
(1992), Iivari & Hirschheim (1996), Iivari et al
. (1998), Silverman (1998) and Trauth & Jessup
(2000).
Furthermore, the political and professional context has also changed significantly since
1990. For example, the mainstream journal, MISQ
, publicly announced its acceptance of alter-
native research approaches in 1993. Allen Lee, a qualitative researcher and pluralist, was later
appointed as its editor-in-chief. His effort to change the political context and advocate research
pluralism should not go unnoticed. The debate between Benbasat & Weber (1996) and Robey
(1996) indicates the variety of interests in the field, even among some of the most well known
researchers. While Benbasat and Weber argue for a unified discipline, Robey advocated the
necessity of research diversity and suggested specific ways to achieve it. Their argument
should also provoke deeper discussion in the community as to how the field should be estab-
lished. Moreover, a special issue of MISQ
in March 1999 gathered a group of well-known
8/2/2019 _Chen___Hirschheim_2004_
4/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14
, 197235
200
researchers (Applegate & King, 1999; Benbasat & Zmud, 1999; Davenport & Markus, 1999;
Lee, 1999; Lyytinen, 1999; Markus & Lee, 1999) to debate the issue of rigour vs. relevance.
Communications of the AIS also dedicated an entire issue to a similar discussion in March
2001.All of this suggests that it has been an exciting and progressive period for the IS research
community since 1990. The communitys interest in paradigmatic and methodological issues
has widened considerably. With an expanded variety of recently established journal outlets,
alternative voices should be expected to be heard. As such, it would be reasonable to expect
that these changes of paradigmatic and methodological understanding would be reflected in
journal publications. Our research purpose, thus, is to uncover whether the journal publications
do indeed reflect the changes occurring since 1990. To this end, we undertook an empirical
analysis of 1893 articles published in eight major IS journals from 1991 to 2001. The selection
of journals four European and four US and the period of investigation should help us under-
stand the progress that has been made in recent years, including:
the contrast between the traditionally dominant paradigm, positivism (Orlikowski & Baroudi,
1991), and an emerging paradigm, interpretivism (Walsham, 1995b), in each journal;
a cross-continental comparison between European and North American journals, as con-
ventional wisdom has long perceived that European journals are more receptive to interpretive
approaches while North American journals tend to be positivist; and
four areas of comparison of methodology: empirical vs. non-empirical, qualitative vs. quan-
titative, cross-sectional vs. longitudinal, and the comparison of five research designs (survey,
laboratory experiment, field experiment, case study and action research).
Such an empirical examination would not just continue the research tradition of Orlikowski &
Baroudi (1991) but also extend the empirical understanding of the current state of the IS dis-
cipline. Conceptual arguments, regardless of how convincing they might appear, would not
reflect the state of concurrent research practice in the journal publications. The empirical result
of the study could, thus, help us to understand better how the IS research community has
evolved with respect to its paradigmatic and methodological assumptions.
In order to achieve this goal, the paper is organized as follows. First, we articulate our clas-
sification of paradigms and methodologies. Then, our research procedure is described espe-
cially as it relates to the journals chosen for study and the paradigms and methodologies used
in the analysis. The section that follows outlines the findings of the empirical examination of 1893
articles. The discussion of the results and the implication of the analysis are then presented in
relation to our research questions. Finally, we present some concluding remarks, which reflect
upon our history, and we also offer some suggestions for future research endeavours.
PARADIGMATIC
CLASSIFICATIONS
Considerable variation exists when it comes to classifying paradigms. For instance, Landry &
Banville (1992) categorize researchers into three groups: (1) mainstream navigators
who
8/2/2019 _Chen___Hirschheim_2004_
5/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14
, 197235
201
inherit a logical positivism tradition; (2) unity advocates
who favour methodological monism;
and (3) knights of change
who advocate methodological pluralism. This categorization broadly
differentiates those who endorse one single paradigm (mainstream navigators and unity advo-
cates) and those who advocate methodological pluralism (knights of change). Similar to Burrelland Morgans typology, Landry and Banvilles grouping differentiates functionalists (main-
stream navigators and unity advocates) from non-functionalists (interpretivist, radical humanist
and radical structuralist). According to Hirschheim & Klein (1992), neohumanism and radical
structuralism have not been well developed in the IS research community.
1
Orlikowski &
Baroudi (1991) and Goles & Hirschheim (2000) indicate that positivism dominates IS research
while other paradigms are relatively small in number. In fact, the only real alternative paradigm
observable in any numbers in IS research is interpretivism (Walsham, 1995a; 1995b; Nandha-
kumar & Jones, 1997; Trauth & Jessup, 2000). Thus, the paradigmatic comparison in our anal-
ysis will simply focus on positivism vs. interpretivism.
The major differences between positivism and interpretivism concerning research arethreefold. Ontologically
, positivists believe that reality exists objectively and independently
from human experiences while interpretivists emphasize the subjective meaning of the real-
ity that is constructed and reconstructed through a human and social interaction process
(Burrell & Morgan, 1979). Epistemologically
, positivists are concerned with the hypothetic-
deductive testability of theories. Scientific knowledge should allow verification or falsification
and seek generalizable results. As such, a causal relationship is usually presented and a
tight coupling among explanation, prediction and control is expected (Orlikowski & Baroudi,
1991). Interpretivists, by contrast, assume that scientific knowledge should be obtained not
through hypothetic-deductive reasoning but through the understanding of human and social
interaction by which the subjective meaning of the reality is constructed (Walsham, 1995a).
Methodologically
, positivists contend that, to test hypothetic-deductive theory, research
should take a value-free position and employ objective measurement to collect research evi-
dence. A quantitative method such as the survey is a typical positivist instrument. Interpre-
tivists, on the other hand, argue that to understand the meaning embedded in human and
social interaction, researchers need to engage in the social setting investigated and learn
how the interaction takes place from the participants perspective. Field studies that engage
researchers in the real social setting would be more appropriate for generating interpretive
knowledge (Orlikowski & Baroudi, 1991).
In brief, positivist research is mostly depicted through: (1) the formulation of hypotheses,
models, or causal relationships among constructs; (2) the use of quantitative methods,
although not always necessary, that test theories or hypotheses; and (3) researchers objec-
tive, value-free interpretation. A classic example of positivist research listed in Hirschheim &
Klein (1992) is the Dickson et al
. (1977) discussion of the Minnesota experiments. Interpretivist
studies, in contrast, could be observed through: (1) evidence from a non-deterministic (free
1
Alternative paradigmatic classification could be found in Chua (1986), in which the three basic paradigms are positivism,
interpretivism and critical theory (Hirschheim & Klein, 2003). Critical theory, however, is not yet well established in the IS
field (Orlikowski & Baroudi, 1991).
8/2/2019 _Chen___Hirschheim_2004_
6/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14
, 197235
202
will) perspective; (2) researchers engagement in the specific social and cultural setting inves-
tigated; and (3) an analysis based on participants viewpoints (Orlikowski & Baroudi, 1991;
Walsham, 1995a). Examples of interpretive studies that include these attributes are Orlikowski
(1993) and Orlikowski & Robey (1991).
METHODOLOGICAL
CLASSIFICATIONS
Various schemes have been applied among researchers to classify research methodology. For
example, Cash & Nunamaker (1989; 1990; 1991) compare qualitative research methods to
experimental and survey research methods. They discovered that only 17% of the papers
included in their analysis employed qualitative research methods. Alavi et al
. (1989) analyzed
empirical and non-empirical articles and found that empirical articles constitute only 46.5% of
publications in the period between 1968 and 1988. They further divide empirical studies intoeight categorizes: laboratory experiment, field experiment, field study, case study, survey,
development of MIS instruments, ex post descriptions and others. Orlikowski & Baroudi (1991)
surveyed 155 articles in four major North American journals. The major categories in their find-
ing are surveys (49.1%), laboratory experiments (27.1%), case studies (13.5%), mixed meth-
ods (3.2%), instrument development (2.6%), protocol analysis (1.3%) and action research
(0.6%). Galliers & Land (1987) and Galliers (1991) created a taxonomy for IS research
approaches within which 10 modes were discussed: theorem proof, laboratory experiment,
field experiment, case study, survey, forecasting and future research, simulation and game/role
playing, subjective/argumentative, descriptive/interpretive and action research. One of the
more recent examinations is by Farhoomand & Drury (1999) in which the methodology of the
studies they analyzed were categorized as non-empirical (39%), survey (32%), case study
(17%), laboratory experiment (10%) and field experiment (2%).
Among these different taxonomies the most consistent comparisons are: empirical vs. non-
empirical (Alavi et al
., 1989) and quantitative vs. qualitative (Cash & Nunamaker, 1989).
Detailed examinations of research designs appear diverse and ambiguous. For example,
although Galliers (1991) included more research methodologies than others, forecasting/future
research, simulation and game/role playing, and theorem/proof, never gained much attention
in the IS community. Additionally, although Alavi et al
. (1989) distinguish between case study
and field study, this distinction is not well articulated. Most case studies are conducted in the
field; hence, the attempt to separate field study from case study appears unnecessary.
Our methodological classification will primarily follow the classification of Orlikowski &
Baroudi (1991) for two reasons. First, Orlikowski & Baroudi (1991) are concerned not just with
methodology but also with paradigmatic trend. Thus, with a similar purpose, both studies could
provide a better comparison if the same classification is employed. Second, their sample was
collected between 1985 and 1989 from four major US research outlets, while ours was col-
lected between 1991 and 2001 from eight major journals. A better reflection, as to how the field
has progressed, could be made if the same classification is used. However, three categories
have been modified for the following reasons:
8/2/2019 _Chen___Hirschheim_2004_
7/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14
, 197235
203
1
Mixed method is not a clear category on their list. We compare mixed method with quanti-
tative and qualitative methods. Articles that use more than one research design (e.g. laboratory
experiment, case study) will be counted in each category and will also be considered as mixed
method.
2
Protocol analysis and instrument development are more practitioner oriented. As this is not
the main interest of our research, we place them in the others category.
3
With respect to research evolving over time, we distinguish between repeated measure
design and cross-sectional, multiple snapshots. While repeated measure design involves
multiple occasions, its focus is on understanding how phenomena change as time passes,
which makes it more dynamic than cross-sectional, multiple snapshots.
RESEARCH
PROCEDURE
Our research procedure involved two stages. The first stage was conducted from fall 2000 to
spring 2001. We collected and analysed articles published in MISQ andAMITto capture and
compare overall trends, if any, in US and European journals. Based on the results of the initial
analysis, we extended article collection to an additional six research outlets and modified our
categorization approach in the second stage, which was from fall 2001 to spring 2002. The
comparison of research duration was extended to four categories: cross-sectional, multiple
snapshots, longitudinal and repeated measure designs. We believe that this comparison pre-
sents a better picture of the different research endeavours involved. Repeated measure design
is of particular interest because it emphasizes the time factor and, thus, can provide a richer
understanding of the researched phenomenon.
Journal representations
For the purpose of illustrating the research trend in IS and comparing the traditionally domi-
nant positivist approach (Orlikowski & Baroudi, 1991) and the emerging interpretive
approach (Walsham, 1995a; 1995b), we decided to focus on mainstream journals which, we
thought, would reflect changes or progress in the field. As such, we examined four main-
stream US publications MISQ
, ISR
, Journal of Management Information Systems
(
JMIS
)
and ICIS
and four main European journals
2
AMIT
,
3
ISJ
,
4
Journal of Information Technol-
ogy
(
JIT
) and EJIS.
The rationale behind these choices has been well explained byWalsham (1995a) and Orlikowski & Baroudi (1991). In brief, these US journals
5
represent
the largest forums of IS publications and arguably, the most influential IS researchers. The
2
This distinction between European and US is simply based on where the journal originates, i.e. the location of the pub-
lisher. To a large extent, it also reflects the composition of the editorial board (Galliers & Meadow, 2003).
3
In 2001,AMITchanged its name to Information and Organization.4ISJwas called Journal of Information Systems until 1997.5We are, of course, aware that ICIS is not a journal but its acceptance rate (traditionally around 10%) is much like the top
IS journals. It also has a pre-eminent position with respect to other conferences in the scholarly community.
8/2/2019 _Chen___Hirschheim_2004_
8/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
204
four selected European journals, on the other hand, embrace a more non-traditional, non-
positivist nature and attract IS researchers with non-traditional orientations. The contrast
and comparison between these two groups should reflect IS research trends and, thus,
illustrate any differences within the IS research community.
6
The only difference betweenour list and Walshams (1995a) is that we have replaced CACMwith JMIS. Our reasoning is
that CACM has changed so as to appeal more to general readers and members of the
Association for Computing Machinery (ACM), thereby reducing the scholarly nature of its
publication. Thus, the journals desire to become practical to meet the needs of the larger
practitioner community makes it inappropriate for this study.
Paradigmatic representations
As Goles & Hirschheim (2000) indicate, Although the proportion of positivist to non-
positivist articles varies somewhat from study to study, the indisputable consensus is thatpositivism dominates information systems research (p. 254). Hirschheim & Klein (1992)
further explain that examples of IS research in radical structuralism and neohumanism
(radical humanism) are rarely found. What might gain increasing attention in IS research,
according to Walsham (1995a), is interpretive approach. As such, comparing and contrast-
ing the trend of positivist and interpretive paradigmatic approaches might better reflect IS
research.
Positivist
The criteria for categorizing positivist articles are the indications of hypotheses, propositions,model formation, quantifiable measures of variables and the inferences drawn from samples to
populations (Orlikowski & Baroudi, 1991), all of which demonstrate the belief that objective
data could be collected to predict the relationship among factors and to test hypotheses or the-
ories (Walsham, 1995a).
Interpretive
The criteria for selecting interpretive articles are threefold. First and foremost, articles
should not involve any positivist indicators as described above, that is, no deterministic
perspectives imposed by the researchers. Second, participants perspectives are taken as
the primary sources of understanding and investigating the phenomena. Third, the phe-
nomena are examined with respect to cultural or contextual circumstances (Walsham,
1995a).
6Alavi et al. (1989) also include journals that are not in our list such as Data Base, Decision Science, Harvard Business
Review, Management Science and Sloan Management Review. They are either practitioner oriented (Harvard Business
Reviewand Sloan Management Review), not IS focus (Management Science and Decision Science) or according to Hard-
grave & Walstrom (1997), not considered part of the of the top-ranking journal list ( Data Base).
8/2/2019 _Chen___Hirschheim_2004_
9/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
205
Methodological representations
Our analysis of research methodology focuses on the following areas: empirical vs. non-
empirical; quantitative vs. qualitative; cross-sectional vs. longitudinal; and the comparisons
among five research designs (survey, laboratory experiment, field experiment, case study and
action research).
Empirical vs. non-empirical
While empirical studies rely on observations and data, non-empirical studies emphasize ideas
and concepts (Alavi et al., 1989). Each of which makes different contributions to the research
community. While non-empirical studies help to develop concepts and build theory, empirical
studies provide concrete evidence for testing theories. As a discipline becomes more mature,
one might expect theory-testing to outweigh theory-building efforts. Therefore, a comparison
between empirical and non-empirical studies serves as a meaningful indication of the progressmade in the field.
Our criterion for categorizing studies as empirical or non-empirical is determined by whether
the articles obtain real data or observations, which could be gathered through quantitative,
qualitative, or a mixed approach, including archival data.7
Quantitative vs. qualitative
Although qualitative research methods such as case studies have gained increasing attention
(Benbasat et al., 1987; Lee, 1989; Klein et al., 1991; Trauth & OConnor, 1991; Visala, 1991;
Walsham, 1995b; Darke et al., 1998; Silverman, 1998; Klein & Myers, 1999; Markus & Lee,1999), the comparison between quantitative and qualitative research methods in IS research
has been treated somewhat inadequately. While the former typically uses numerical analysis
to illustrate the relationship among factors in the phenomenon studied, the latter emphasizes
the description and understanding of the situation behind the factors. To make the comparison
meaningful, both methods should be categorized as empirical studies. As non-empirical studies
typically involve describing or arguing for theories and frameworks, they could be categorized
as a qualitative research method but this would muddy the distinction of quantitative methods.
As such, our criterion for categorizing research methods as quantitative or qualitative is based
on whether studies use a statistical or numerical approach to collect and analyse data. There
is a possibility that research could use both quantitative and qualitative methods in differentstages of the study. On such occasions, we categorize them as a mixed research method.
Cross-sectional vs. longitudinal
As described previously, in contrast to Orlikowski & Baroudi (1991), our categorization of the
duration of a research project distinguishes between repeated measure design and multiple
7There are occasions when we locate articles with data/observations in the non-empirical category. In such cases, articles
were merely descriptive without scholarly research questions, practitioner oriented for systems development, or mathe-
matic simulation and analysis.
8/2/2019 _Chen___Hirschheim_2004_
10/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
206
snapshots. Four categories are designated: longitudinal, repeated measure design, multiple
snapshot and cross-sectional each requiring different research efforts and time commitment.
We define longitudinal as research that evolves over an uninterrupted period of time and
focuses on process. Cross-sectional, on the other hand, is research that collects data throughone snapshot at a particular point of time (Orlikowski & Baroudi, 1991). Multiple snapshots are
also cross-sectional in nature but involve more than one single data-collection period or setting,
which could mean different experiments, treatments or subjects and so on. Repeated measure
designs are those that conduct studies at various time periods to examine how a phenomenon
evolves at various periods of time.
Six major research designs
Six research designs were identified, as per Orlikowski & Baroudi (1991):
Survey: Studies that fall into this category gather data by means of questionnaires. Case study: Studies that are involved with a single site or a few sites over a certain period oftime are located in this category. The case study inquiry usually relies on multiple sources of
evidences (Yin, 1994, p. 13).
Laboratory experiment: These include studies that take place within a designed, controlledenvironment and usually involve special treatments of different groups to contrast the precise
relationships among variables (Galliers, 1991).
Field experiment: As with laboratory experiments, special treatments are used to study twoor more controlled groups. Field experiments differ in that they are conducted in a real-world
setting. Action research: Often embodied in a case study, action research distinguishes itself in thatthe researchers are an integral part of the phenomenon under study. The researchers input
often influences the outcomes of the phenomenon and his/her role could change from
researcher to subject (Galliers, 1991).
Others: this category includes articles that are practitioner oriented (systems or tools devel-opment), non-empirical pieces, or descriptive/argumentative as noted in Gallierss (1991) clas-
sification. Research with secondary data such as public records or existing datasets is also
located in this category.
The following section presents our empirical findings. Our analysis discusses overall trends,
which combines the results of all eight journals from 1991 to 2001 and a total of 1893 articles.
The detailed categorization of each article is listed in the Appendix wherein readers can see
how each article was located with respect to the classification criteria: (1) positivist vs. inter-
pretive; (2) empirical vs. non-empirical; (3) quantitative vs. qualitative vs. mixed; (4) longitudinal
vs. repeated measure vs. multiple snapshots vs. cross-sectional; and (5) survey vs. lab exper-
iment vs. field experiment vs. case study vs. action research vs. others. 8 These detailed cat-
8Includes subcategories of conceptual, existing (archival) data, documents/reports, meta-analysis, systems design, news
data, simulation/mathematic calculation, Delphi techniques, observation and tutorials.
8/2/2019 _Chen___Hirschheim_2004_
11/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
207
egorizations are listed by journals, and then summarized in terms of the number of articles by
year and categories. More details for each journal are listed in the Appendix under the title
Summary of Categorization of Journal Articles.
FINDINGS
Whilst a detailed analysis of 1893 articles might appear overwhelming, in this section we focus
our analysis on three major areas that better fit our research interests: (1) overall trend; (2) the
cross-continental comparison; and (3) the trend of individual journals. Table 1 displays a sum-
mary of all articles as they relate to our classification criteria.
Overall trend: all journals
Positivist vs. interpretive
Figure 1 shows that positivist research maintains a clearly dominant position while the number
of interpretive studies has increased slightly over the years. As can be seen from Table 1, the
overall percentages of positivist and interpretive studies are 81% (917) and 19% (214) respec-
tively. What is perhaps somewhat surprising is the fact that the overall trend has remained
remarkably constant over the past 10 years.
Empirical vs. non-empirical
Figure 2 illustrates the trend of empirical vs. non-empirical studies. In the early 1990s, non-
empirical studies were in the majority in IS research. This is consistent with the finding of Alavi
et al. (1989). Since 1993, however, the number of empirical-research studies has exceeded
non-empirical studies and the gap has widened over the years. Moreover, among the non-
empirical studies there are many articles that contain empirical data but we locate them in the
non-empirical category as they are typically practitioner-oriented studies. As such, the real
number of empirical-research studies, i.e. those which contain empirical data, is more than
what Figure 2 illustrates. In Table 1, the overall percentages of empirical and non-empirical
studies are 60% (1131) and 40% (762) respectively. This exactly mirrors the findings of Far-
hoomand & Drury (1999) who examined 2098 articles published between 1985 and 1996 and
noted that the breakdown between empirical and non-empirical studies was 61% and 39%
respectively. Even though the journal selections are significantly different in the two studies, the
results are the same. So the dominant trend of empirical research in IS is demonstrably visible.
Quantitative vs. qualitative
Figure 3 shows that quantitative research still dominates the IS discipline but the number of
qualitative research papers has been increasing over the years. The gap between quantitative
8/2/2019 _Chen___Hirschheim_2004_
12/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
208
and qualitative research in 1997 and 1998 was very narrow. Subsequently, the gap has wid-
ened. Mixed method research, on the other hand, has remained steady but not substantial from
1991 to 2001. As noted in Table 1, the overall percentages of quantitative, qualitative and
mixed methods are 60%, 30% and 10% respectively.
Longitudinal vs. cross-sectional
Figure 4 shows that there has been a steady, continuous popular trend of cross-sectional stud-
ies although there has been a considerable growth in longitudinal studies as well. The growth
of longitudinal research corresponds well with the trend toward qualitative methods, as shown
in Figure 3. This should be expected because the qualitative research method usually involves
case studies, which requires a long-term research process. Repeated measure design and
multiple snapshots, on other hand, continue to be flat and few in number, indicating a very lim-
ited growth over the decade. Figure 5 displays that the overall percentage of longitudinal,
repeated measure design, multiple snapshots and cross-sectional as 33%, 4%, 4% and 59%
respectively.
Major research designs
Figure 6 illustrates that survey research and case studies are, by far, the most popular research
designs while action research and field experiments are rather insignificant. Laboratory exper-
iments, on the other hand, maintain a somewhat steady position throughout the years, albeit
with occasional fluctuations. While the survey method remains a steady research force, case
studies demonstrate a noticeable growth throughout the years. In the years 1997, 1998, 2000,
the number of case studies actually exceeded that of survey research. This could be an indi-
cation that IS researchers are increasingly interested in gathering rich data through the use of
the case study method.
Table 1. Summary of journals in all categories
Year Positivist Interpretive Empirical Non-empirical Quantitative Qualitative Mixed Cross-sectional
1991 63 3 66 82 50 6 10 53
1992 60 4 64 72 49 7 8 45
1993 73 12 85 64 49 22 14 55
1994 92 10 102 79 73 23 6 70
1995 89 10 99 76 63 29 7 66
1996 97 20 117 70 65 36 16 65
1997 79 32 111 66 54 47 10 50
1998 89 31 120 48 59 52 9 63
1999 100 24 124 72 75 34 15 65
2000 92 42 134 60 70 54 10 69
2001 83 26 109 73 73 31 5 66
Total 917 214 1131 762 680 341 110 667
8/2/2019 _Chen___Hirschheim_2004_
13/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
209
Figure 7 offers the overall percentages of survey, case study, laboratory experiment, action
research and field experiment as 41%, 36%, 18%, 3% and 2% respectively. This result differs
somewhat from that of Orlikowski & Baroudi (1991) where the survey method embodied 49.1%
of IS research production followed by laboratory experiments (27.1%) and case studies
(13.5%). It is, however, more in line with Farhoomand & Drury (1999) where the survey, case
study and laboratory experiment share was 52%, 28% and 16% respectively. This indicates
that while surveys maintain their steady popular position, case studies have gained consider-
able interest over the decade. It must be noted, however, that research categorized as others
makes up 46% of the entire research production. In another words, the IS research community
still relies heavily on non-empirical studies such as conceptual/framework development.
longitudinal
Multiple
snapshots
Repeated
measures Survey
Case
study
Lab
experiment
Action
research
Field
experiment Others
11 0 2 30 15 20 0 3 90
12 5 2 38 16 15 0 1 81
23 2 5 35 30 23 0 0 75
22 4 6 52 25 15 3 2 94
30 3 0 51 30 15 1 1 87
40 6 6 44 42 28 7 2 87
51 7 3 42 52 16 6 2 74
52 3 2 41 56 11 1 2 64
41 12 6 47 47 27 3 3 88
56 5 4 35 53 21 8 2 85
36 1 6 39 36 13 3 1 96
374 48 42 454 402 204 32 19 921
Figure 1. Overall paradigmatic trend.
0
20
40
60
80
100
120
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
Nu
mberofArticles
Positivist Interpretive
8/2/2019 _Chen___Hirschheim_2004_
14/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
210
Cross-continental comparisons
US vs. European journals: positivist vs. interpretive
The focus of our second analysis is on the cross-continental comparison between US and
European journals. Conventional wisdom suggests that US journals are more positivist ori-
ented and use more quantitative methods while European journals tend to be more inter-
pretive with more qualitative methods. Here, we empirically examine whether this
assumption holds true. Figure 8 illustrates that US positivist research clearly dominates with
58% while US interpretive articles account for only 7%. This clearly indicates that US
research is overwhelmingly dominated by the positivist paradigm. On the other hand, the
Figure 2. Overall trend of empirical vs. non-empirical studies.
0
20
40
60
80
100
120
140
160
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
NumberofArticles
Empirical Non-Empirical
Figure 3. Overall trend of quantitative vs. qualitative studies.
0
10
20
30
40
50
60
70
80
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
NumberofArticles
Quantitative Qualitative Mixed
8/2/2019 _Chen___Hirschheim_2004_
15/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
211
difference between European positivist (23%) and European interpretive studies (12%) is
considerably less significant. Moreover, as Figure 9 shows US positivist research not only
remains a clearly dominating force but it has also maintained its predominant position over
the years. The same, although to a somewhat lesser extent, can be seen in Europe where
approximately two-thirds of the papers are positivist. Interpretive studies in both continents,
while showing a modest increase, are still quite small in number. Interpretive researchers,
expecting a more equitable publication distribution may or may not be surprised by this
result.
US vs. European journals: quantitative vs. qualitative
As seen in Figure 10, US quantitative research is vastly dominant (46%), but the second most
popular type of publication is qualitative research in European journals (17%). It slightly edges
Figure 4. Overall trend of cross-sectional vs. longitudinal studies.
0
10
20
30
40
50
60
70
80
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
NumberofArticles
Cross-sect ional Longitudinal Multiple Snapshots Repeated Measures
Figure 5. Percentage of cross-sectional vs. longitudinal studies.
Cross-Sectional
59%
Longitudinal33%
MultipleSnapshots
4%
RepeatedMeasures
4%
8/2/2019 _Chen___Hirschheim_2004_
16/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
212
out qualitative research in the USA (13%) and quantitative research in European journals
(14%). Mixed methods in both US and European journals are still relatively small, comprising
only 6% and 4% respectively.
From Figure 11, one notes that as far as the overall trend is concerned, the only method
that clearly stands out throughout the years is the quantitative method in the USA. It has
remained consistent over the past 11 years and even increased slightly, in recent years. The
gradual increase of European qualitative research is also noteworthy. No other trend is
discernible.
Figure 7. Percentage of five research designs.
Survey
41%
Casestudy
36%
Labexperiment
18%
Actionresearch
3%
Field
experiment2%
Figure 6. Overall trend of five major research designs.
0
10
20
30
40
50
60
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001Year
NumberofArticles
Survey Case Study Lab Experiment Action Research Field Experiment
8/2/2019 _Chen___Hirschheim_2004_
17/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
213
Trends of individual journals
To simplify our analysis of the individual journals, we restrict our examination to paradigmatic
trend and research design. For each journal, we will present an overall paradigmatic trend and
a pie chart indicating the percentage of the five major research designs employed in the jour-
nal. The sequence of presentation is MISQ,AMIT/IO, EJIS, JIT, ISJ, ISR, JMIS and ICIS, with
no particular priority.
MISQ trend
The total number ofMISQ articles included in our analysis is 232, broken down into 166 empir-
ical and 66 non-empirical studies. The paradigmatic trend of MISQ shown in Figure 12 very
Figure 8. Paradigmatic comparison: US vs. European journals.
Positivist
(US)
58%
Interpretive
(European)
12%Interpretive
(US)
7%
Positivist
(European)
23%
Figure 9. Paradigmatic trend: US vs. European journals.
0
10
20
30
40
50
60
70
80
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
NumberofArticles
US Positivist European Positivist
US Interpretive European Interpretive
8/2/2019 _Chen___Hirschheim_2004_
18/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
214
closely corresponds to the overall paradigmatic trend with positivism dominating interpretivism
throughout the years. Never in any given year has interpretive studies produced more than five
articles in MISQ. Occasionally (e.g. 1991, 1992 and 1995), no interpretive research appears in
the journal. With respect to research design (Figure 13), survey research is still the most pop-
ular data collection method with 46% while case studies make up 34% of the research output.
The other three methods appear much less significant. Although MISQ has publicly announced
Figure 10. Quantitative and qualitative percentages: US vs. European journals.
Mixed
(US)
6%
Quantitative
(European)
14%
Qualitative
(US)
13%
Quantitative
(US)46%
Mixed
(European)
4%
Qualitative(European)
17%
Figure 11. Quantitative vs. qualitative trend: US vs. European journals.
0
10
20
30
40
50
60
70
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
NumberofArticles
US Quantitative European Quantitative US Qualitative
European Qualitative US Mixed European Mixed
8/2/2019 _Chen___Hirschheim_2004_
19/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
215
its interest in accepting alternative research, overall, MISQ remains a positivist (91%), quan-
titative (64%), cross-sectional (60%) and survey research (46%) dominated journal.
AMIT/IO trend
We examined 122 articles in AMIT/IO, of which 65 were empirical and 57 non-empirical.
Figure 14 depicts an interesting paradigmatic trend in AMIT/IO. The two paradigmatic lines
are intertwined particularly in the mid-1990s, until interpretive studies became the main-
stream in 1997. In 2001, IO published zero positivist research papers thus establishing
Figure 12. Paradigmatic trend of MISQ.
0
2
4
6
8
10
12
14
16
18
1991 1993 1995 1997 1999 2001
Year
NumberofArticles
Positivist
Interpretive
Figure 13. Percentage of research designs in MISQ.
Survey46%
Labexperiment
15%
Fieldexperiment
4%
Casestudy34%
Actionresearch
1%
8/2/2019 _Chen___Hirschheim_2004_
20/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
216
interpretive research as the paradigm of choice. The overall percentage of interpretive and
positivist studies in AMIT/IO are 63% and 37% respectively. Among the journals examined,AMIT/IO is the only one in which interpretive research is the mainstream research
production.
Figure 15 demonstrates that the most popular method inAMIT/IO is the case study making
up 64% of the publications. Positivist research methods such as surveys, laboratory experi-
ments and field experiments have relatively marginal representation with 14%, 15% and 2%
respectively. Action research appears to have gained increasing attention in this outlet with a
modest total of 5% production. This should be encouraging for action researchers because
most journals pay little attention to this type of research (e.g. EJIS). Clearly, this journal is
receptive to alternative research approaches.
Figure 14. Paradigmatic trend ofAMIT/IO.
0
1
2
3
4
5
6
7
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
NumberofArticles
Positivist Interpretive
Figure 15. Percentage of research designs in AMIT/IO.
Survey
14%
Case
study64%
Actionresearch
5%
Field
experiment
2%
Lab
experiment15%
8/2/2019 _Chen___Hirschheim_2004_
21/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
217
EJIS trend
We examined 233 EJIS articles with a breakdown of 118 empirical and 115 non-empirical stud-
ies. Figure 16 indicates that EJISs paradigmatic trend looks similar to that of MISQ a clear
indication of the dominance of positivist research, the past 2 years notwithstanding. In 1991,
1992, 1995 and 1996, there were no interpretive articles published in EJIS. The percentage of
positivist studies in EJIS is 88%, which closely resembles that ofMISQ (91%). If conventional
wisdom is that European journals are more interpretive and qualitatively oriented, EJIS does
not follow this trend. In terms of research design, Figure 17 shows that EJIS is clearly domi-
nated by survey methods (51%) while action research (2%) and field experiment (1%) are
insignificant. Case studies also gain wide acceptance in EJIS although looking at its paradig-
matic trend (Figure 16), it is clear that these are likely to be positivist case studies.
JIT trend
We examined 198 articles in JITwith 120 empirical and 78 non-empirical studies. As Figure 18
indicates, our study only involves articles published from 1994 onwards. This was unfortunately
due to our inability to obtain the earlier issues. We note that the gap between interpretive and
positivist studies has narrowed noticeably since 1996. In fact, if one excludes 1994 and 1995,
the difference between the number of positivist and interpretive studies is, at best, marginal.
The percentages of positivist and interpretive research in JITare 60% and 40% respectively.
With respect to research design, Figure 19 reveals that JITis receptive to qualitative research
as case studies and action research share relatively high percentages of 52% and 7% respec-
tively. Only 39% of the articles used survey methods, a somewhat low percentage relative toother outlets.
Figure 16. Paradigmatic trend of EJIS.
0
2
4
6
8
10
12
14
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
Numbe
rofArticles
Positivist Interpretive
8/2/2019 _Chen___Hirschheim_2004_
22/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
218
ISJ trend
We examined 172 articles in ISJ, of which 90 were empirical and 82 non-empirical.Figure 20 demonstrates an interesting paradigmatic pattern of ISJ. Both lines of the para-
digmatic trends punctuate around each other, whereas interpretive studies occasionally
exceed positivist ones (e.g. 1997 and 2000). The number of interpretive studies has also
shown, by and large, an increasing trend through the years. The overall percentages of pos-
itivist and interpretive studies in ISJ are 60% and 40% respectively. The percentage of
research designs in ISJalso appears noticeably different. While case studies dominate the
journal, a relatively high number of action research papers (12%) indicates its traditional
European qualitative orientation (Figure 21). Survey research, comprising only 23%, is the
lowest of any examined outlet.
Figure 18. Paradigmatic trend of JIT.
0
2
4
6
8
10
12
14
16
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
NumberofArticles
Posi tivist Interpretive
Figure 17. Percentage of research designs in EJIS.
Survey51%
Casestudy40%
Actionresearch
2%
Labexperiment
6%
Fieldexperiment
1%
8/2/2019 _Chen___Hirschheim_2004_
23/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
219
ISR trend
Out of the 213 ISR articles examined, 108 were empirical and 105 were non-empirical.
Figure 22 indicates that interpretive studies in ISR are virtually non-existent. In 6 out of
11 years, ISR published no interpretive studies. The gap between positivist and interpretive
research has, if anything, increased over the years. The total percentages of positivist and
interpretive studies in ISR are 94% and 6% respectively. In the 11-year period, ISR produced
only seven empirical, interpretive studies. Quantitative studies make up 87% while cross-
sectional studies are similarly high at 74%. Like MISQ, ISR is positivist, quantitative and cross-
sectional oriented. Figure 23 further indicates that ISR is a traditional positivist outlet where
Figure 19. Percentage of research designs in JIT.
Survey
39%
Case
study
52%
Action
research
7%
Field
experiment
0%
Lab
experiment
2%
Figure 20. Paradigmatic trend of ISJ.
0
1
2
3
4
5
6
7
8
9
10
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
N
umberofArticles
Positiv ist Interpretive
8/2/2019 _Chen___Hirschheim_2004_
24/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
220
laboratory experiments (45%) and survey research methods (38%) dominate more than 80%
of the published research.
JMIS trend
We examined 370 JMIS articles comprising 234 empirical and 136 non-empirical studies.
Figure 24 illustrates that the paradigmatic trend ofJMIS also shows the extremely dominating
position of positivist research throughout the years. As in the case of ISR, the overall percent-
ages of positivist and interpretive studies in JMIS are 94% and 6% respectively. With respect
to research design, Figure 25 demonstrates that surveys are the preferred instrument (50%)
while qualitative approaches comprise less than one fourth of research publications (22%). If
Figure 21. Percentage of research designs in ISJ.
Survey
23%
Labexperiment
14%Field
experiment1%
Casestudy50%
Actionresearch
12%
Figure 22. Paradigmatic trend of ISR.
0
2
4
6
8
10
12
14
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
NumberofArticles
Positivist Interpretive
8/2/2019 _Chen___Hirschheim_2004_
25/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
221
traditional wisdom considers US journals to be more positivist and quantitative, then MISQ, ISR
and JMIS certainly bear this out.
ICIS trend
We examined 351 ICIS articles comprising 230 empirical and 121 non-empirical studies. Sim-
ilar to the other US outlets, ICIS also strongly tends to be positivist (83%). As illustrated by
Figure 26, positivist research clearly dominates interpretive studies with the number of its pub-
lications remaining high. Indeed, the gap was largest in 2001. With respect to research design,
ICIS papers appear to be more equally divided among surveys (39%), case studies (36%) and
laboratory experiments (22%) (Figure 27).
Figure 23. Percentage of research designs in ISR.
Survey
38%
Labexperiment
45%
Fieldexperiment
3%
Casestudy14%
Actionresearch
0%
Figure 24. Paradigmatic trend of JMIS.
0
5
10
15
20
25
30
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
NumberofArticles
Positivist Interpretive
8/2/2019 _Chen___Hirschheim_2004_
26/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
222
DISCUSSION
The empirical evidence gathered from 1893 articles in eight leading IS publication outlets
between 1991 and 2001 demonstrates that even after years of advocacy of paradigmatic plu-
ralism, not much has really changed. The gap between positivist and interpretive research is
still significant. The increase of interpretive studies is still marginal and a considerable dis-
crepancy can be easily found in MISQ, ISR, JMIS, ICIS and EJIS. In other words, five out of
eight journals are dominated by the positivist approach. This is even more apparent in ISR and
JMIS where positivist research comprises 94% of the publications. The only journal in which
Figure 25. Percentage of research designs in JMIS.
Survey
50%
Lab
experiment
25%
Casestudy
22%
Field
experiment
3%
Action
research
0%
Figure 26. Paradigmatic trend of ICIS.
0
5
10
15
20
25
30
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001
Year
Numb
erofArticles
Posi tivist Interpretive
8/2/2019 _Chen___Hirschheim_2004_
27/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
223
interpretive research exceeds positivist research isAMIT/IO. This exception might be due to
the editorial policy advocated in its debut issue:
The journal is open to all forms of scholarship and analysis concerning information
systems . . . . No reference discipline, research method or theoretical perspective will be pre-
ferred over another. The journal is open to established topics and approaches . . . as well as
to more openly interpretive and critical analysis. (Boland & OLeary, 1991, p. 1)
With respect to the cross-continental comparison, US journals tend to be dominated by the
positivist approach. Despite calls for research methodological pluralism and the appointment
of a sympathetic editor-in-chief ofMISQ, no significant change is reflected in the mainstream
research publications. A recent paper by Galliers & Meadow (2003) offers a possible expla-
nation. While examining articles published in two mainstream North American (ISR and MISQ)
and two European journals (ISJand JSIS) from 1994 to 2000, Galliers and Meadow discovered
that the ISR and MISQ editorial board members were primarily North American researchers
(between 75% and 87.5% forISR and between 51.2% and 86.8% forMISQ). These journals
also tended to publish articles written by North American authors (74% for ISR and 83% for
MISQ). A similar situation was found in European journals where the editorial board members
were largely European as were the papers published. Galliers and Meadow concluded that theIS field maintained a strong home grown perspective.
While these results may be of little surprise to some, the research community might wish
to assess what progress, if any, has been made over the last decade towards pluralism.
The analysis of Orlikowski & Baroudi (1991) noted the overwhelming dominance of positiv-
ism (96.8%). Our analysis shows that the positivist paradigm still accounts for 81% of all
empirical publications, and almost 86% of the US journal empirical publications (661 out of
738 articles). Clearly, paradigmatic progress over the past 11 years has been marginal. If
diversity is to be encouraged to avoid an unnecessarily restrictive and insular perspective
for studying IS phenomena (Orlikowski & Baroudi, 1991; Robey, 1996; Hirschheim & Klein,
Figure 27. Percentage of research designs in ICIS.
Survey
39%
Lab
experiment
22%
Case
study36%
Action
research
3%
Field
experiment
0%
8/2/2019 _Chen___Hirschheim_2004_
28/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
224
2002a), IS researchers need to ask: What can the IS research community do to accept
alternative research perspectives?
Consistent results could also be observed in the comparison between quantitative and qual-
itative research and between longitudinal and cross-sectional studies. It is quite noticeable thatquantitative and cross-sectional studies enjoyed more popular positions throughout the
decade. This might be due to the publish or perish research publication games (Walsham,
1995a). Quantitative research usually employs survey questionnaires or laboratory experi-
ments. Researchers time commitment and engagement in the research process are, relatively
speaking, minimal. As pressures converge for tenure evaluation, researchers will predictably
choose an easier approach to ensure that their research works are published in time for the
promotion. Quantitative and/or cross-sectional studies, therefore, appear to be a better solution
for that purpose.
Another interesting observation is illustrated by the comparison between empirical and
non-empirical studies. Our analysis demonstrates that IS researchers have increased theiruse of empirical data collection. Descriptive, conceptual and analytical types of research are
still popular but are in decline (Alavi et al., 1989). Such a result is consistent with Far-
hoomand & Drury (1999) in which 61% of research is empirical in nature. This might be
due to long-term advocacy and discussion of rigour and relevance. For example, Communi-
cations of AIS devoted an entire issue to the subject in March 2001 while MISQ did some-
thing similar in March 1999. As a field matures, theoretical and conceptual developments
become less appealing. Empirical studies, in contrast, become more popular because of the
need for theory testing and practical relevance. The growing number of empirical studies
might be an indication that the IS field has become more mature and is continuing to
evolve.
With respect to research design, our findings show some change. Orlikowski & Baroudi
(1991) noted that the most frequently adopted methods were surveys (49.1%), laboratory
experiments (27.1%) and case studies (13.5%). Our analysis, in contrast, demonstrates that
while survey (41%) is still the most widely used research design, the number of case studies
(36%) has increased substantially. Farhoomand & Drury (1999) found a similar situation with
case studies comprising 28% and surveys 52% of all empirical publications. Such results might
indicate that researchers are not just engaging in more empirical studies but rather paying
more attention to how the phenomenon evolves. [This is based on the belief that the case
study, normally conducted through an uninterrupted period, might better reveal the detailed
process of the phenomenon investigated and/or even be incorporated into positivist, theory-
testing research (Lee, 1991; Sarker & Lee, 1998).]
Implications
Our analysis indicates that whilst there has been some paradigmatic change in the IS research
community since 1990, this change has not significantly manifested itself in journal publica-
tions. In most research journals, positivism maintains its prevailing, dominant position. This
could have implications with respect to pluralists and general researchers.
8/2/2019 _Chen___Hirschheim_2004_
29/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
225
First, if the field were to become a pluralist discipline (Klein et al., 1991; Robey, 1996), the
long-term advocacy for pluralism would need continuous and stronger endeavours. It is rea-
sonable to claim, a single research perspective for studying information systems phenomena
is unnecessarily restrictive (Orlikowski & Baroudi, 1991, p. 1). It could also be true that thereare signs that interpretivism is gaining ground (Walsham, 1995b, p. 376) and There has been
an increase in recent years in the number of in-depth case studies which focus on human
actions and interpretations surrounding the development and use of computer-based informa-
tion systems (IS) (Walsham, 1995a, p. 74). The empirical evidence shown in this study, how-
ever, suggests that while the emergence of interpretivism in IS research and the wider use of
case studies might be true, their influences are still rather insignificant, particularly given that
case studies have been widely used in a positivist fashion (e.g. Swanson, 1992; Jones, 1994;
Kirsch & Cummings, 1996; Nelson et al., 1997; Reeves-Ellington & Anderson, 1997). As Lan-
dry & Banville (1992) claimed: it is relatively easy to get in, but getting in the field is quite afar
from exerting a meaningful and lasting influence (p. 86). Although plentiful activities haveevolved in the IS community since 1990, to assure a long lasting influence, pluralist research-
ers will need a stronger commitment and continuous effort.
Particular attention could be paid to the publication system and the tenure and promotion
system in the field. Traditional wisdom has long recognized that positivist research is more eas-
ily accepted because its research tradition has been more successfully established (Hir-
schheim & Klein, 2002a). Positivist research is generally considered more rigorous, scientific,
generalizable and reliable, and therefore it is encouraged and widely accepted (Tornatzky &
Klein, 1982). As such, it is safer for authors to stick to positivist orthodoxy (Walsham, 1995b,
p. 391). Walsham (1995b) further indicates:
[P]ositivist research is typically less time consuming than interpretive methods such as eth-
nography, certainly in elapsed time terms, and thus authors are likely to choose the former in
response to our publish or perish age. (p. 391)
While under the pressure of tenure and promotion, researchers would tend to choose a less
time consuming approach positivism to avoid perishing. This might help explain why inter-
pretive influences are still marginal and likely to continue to be this way. As such, for a pluralist
research tradition to be established, both the publication and tenure and promotion systems
might need some modifications. The reason that most researchers employ a less time-
consuming approach is because for tenure and promotion, the quantity of publications counts.
The less time consuming the approach, the more productions are possible; the more produc-
tions, the better the chance for tenure and promotion. While the tenure and promotion system
values the quality of publications as well, it is difficult to get tenure at a good school without
suitable numbers of publications. In the existing system, researchers are more concerned
about how to get a research project published rather than asking significant research ques-
tions. Many research papers fail to address meaningful issues as they are too difficult and take
too long to publish.
On the publication system side, pluralists need to develop ways of educating journal editors
so that they would be more receptive to alternative research approaches. First, editors should
8/2/2019 _Chen___Hirschheim_2004_
30/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
226
move beyond conservatism and parochialism (Galliers & Meadow, 2003, p. 6) and be more
receptive to authors from other continents. Failing to do so could indicate that the development
of the discipline itself could well be retarded by our failure to appreciate the work of colleagues
elsewhere in the world . . . journal editors may well be the unwitting mouthpieces of unwittingresearchers, publishing results of partial, culturally biased accounts (Galliers & Meadow, 2003,
p. 6). In addition, editors traditionally play the role of gatekeeper rather than facilitator to
ensure the rigour of research publications. However, such a view of rigour typically means
applying a positivist approach and thus excludes other scholarly research (Hirschheim & Klein,
2002a). As Hirschheim and Klein argue, there are many scholarly vehicles for knowledge cre-
ation and they need to be recognized as rigorous as long as they employ sound forms of rea-
soning and evidence giving (p. 13). While positivist research might serve the purpose of
generalizability, interpretivist studies aim for an in-depth understanding of meaning embedded
in the phenomenon investigated (Orlikowski & Baroudi, 1991). They provide an alternative
dimension for scientific inquiries and should supplement rather than exclude each other.In other words, the editors evaluation of a research article should not be based upon the
approach employed, but on the reasoning and evidence provided. Editors should not reject
interpretive research simply because there is no objective criterion for the evaluation. Following
Habermass notion of the force of the better argument, Hirschheim & Klein (2002a) suggest
three aspects for the objective evaluation of research: intelligibility, noveltyand believability.
Intelligibilityrelates to the question how well the research approach and results are compre-
hensible, i.e. how closely others can follow them with similar qualification. Novelty can be
judged in at least three ways: (1) by the amount of new insight added; (2) by the significance
of the research reported in terms of the implications it has for seeing important matters in a
new light and/or provide a new way of thinking about the phenomenon under study; (3) by the
completeness and coherence of the research report(s) . . . Believability, on the other hand,
relates to how well the research arguments make sense in light of our total knowledge. The key
question for believability is how well the research in method and results fits with other ideas and
arguments that are taken for granted within the current stage of knowledge (p. 13). These cri-
teria could then help journal editors to evaluate research employing alternative approaches,
more reasonably.
If the tenure and promotion system would better incorporate the quality of articles into its
evaluation criteria, it would encourage researchers to be more committed to conduct research
projects that require in-depth, long-term endeavour. If the publication system could be more
receptive to alternative research approaches and evaluate the research based on reasoning
and argument instead of the traditional form of rigourgeneralizability, it could encourage
researchers to ask different research questions requiring alternative research approaches. For
a pluralist research tradition to be established, the collaboration of these two systems would
then be essential.
Third, our analysis could help general researchers better allocate their research endeavours
in terms of their publication outcomes. For quantitative researchers, better publication outlets
appear to be MISQ, ISR, EJIS, ICIS and JMIS. The survey method is prevalent in almost every
journal but particularly dominates the research in these five outlets. For qualitative researchers,
8/2/2019 _Chen___Hirschheim_2004_
31/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
227
on the other hand, the best publication opportunities are found in Information and Organiza-
tion, JITand ISJalthough MISQ, ICIS and EJIS appear increasingly receptive to alternative
methods such as the case study.
There are three implications of these findings. First, for authors, recognizing the political andprofessional preferences of each journal could help them allocate their resources with respect
to the reviewing and publication process. In responding to the publish or perish game, authors
could then better ensure their survival chances. Second, for members of tenure and promotion
committees, a variety of journal outlets such as Information and Organization, JITand ISJ,
among others that publish high quality research using alternative approaches, should be rec-
ognized and taken into consideration in tenure and promotion evaluations. Third, for publish-
ers, an attempt to enhance the availability and visibility of alternative journals is critical. For
example, Information and Organization is not yet electronically available which reduces its
exposure to the general research community.
It is, therefore, not surprising that Information and Organization, JITand ISJdo not appearon the journal-ranking list of ISWorld. High quality research that involves a long-term commit-
ment and is published in these journals appears not to be seriously recognized and appreci-
ated. This may lead researchers to shun alternative research approaches because such
research is not recognized in the tenure and promotion system. Clearly, this serves to dampen
any drive toward pluralism in the IS field.
Limitations
While this paper provides a reflection of IS research, some limitations need to be noted to avoid
possible misinterpretation. First, our goal has been reflection rather than generalization. For
researchers who are interested in generalizing results, the scope of data collection would need
to be broadened. Management Science and Organization Science, for example, are not
included in our survey due to their non-IS orientation. Their inclusion might, however, have
some effect on the field. The most cited articles between 1986 and 1995, according to Wal-
strom & Leonard (2000), include many influential works that are published in those journals:
DeSanctis & Gallupe (1987), Daft & Lengel (1986), Swanson (1974) in Management Science;
and DeSanctis & Poole (1994) in Organization Science. Therefore, for the purpose of gener-
alization, a broader survey that includes a greater variety of journals should be encouraged.
Second, while categorizing the paradigmatic position of articles, we adopted the approach
of Walsham (1995a) and Orlikowski & Baroudi (1991). As Walsham (1995a) indicated, even
within the interpretive paradigm, there is no homogeneous school of thought with respect to
knowledge claims. Different interpretive researchers might apply different categorizing
schemes with different scaling justifications, resulting in different outcomes. Given this poten-
tial subtle bias, we can only claim that our findings serve as a reflection of IS research based
on the Walsham (1995a) and Orlikowski & Baroudi (1991) interpretive school of thought.
Finally, our analysis is based on quantity evidence. One may argue that the quantity of arti-
cles does not accurately reflect their influence. Alternative approaches, such as citation anal-
ysis, that emphasize quality should also be examined. If so, we would then conduct an analysis
8/2/2019 _Chen___Hirschheim_2004_
32/39
W Chen & R Hirschheim
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
228
based on the results of Walstrom & Leonard (2000). However, none of the most cited articles
on their list was published between 1991 and 1995, which is the overlapping period of their and
our data collection. Therefore, it would not appropriately reflect the influence of articles pub-
lished between 1991 and 2001. A parallel analysis of classic articles published between 1991and 2001 might be needed to better illustrate the influence of research, with respect to quality.
CONCLUSION
Although many schools of thought have emerged and even become well established, reflective
studies that trace such historical trends appears inadequate. Hirschheim & Klein (2002b) once
stated:
We are not sure whether the field considers itself too young to need such a reflection or
whether there simply are not enough old timers around who could provide such a view.
Whatever the case, we believe this to be a serious shortcoming of the IS discipline.
This paper, therefore, aims to reduce such a deficiency. We have first stated the needs of
conducting such a historical reflective piece, described some prior works related to paradig-
matic and methodological issues, and explained our rationale of classifying paradigms and
methodology. Our intent is to continue the research tradition of Orlikowski & Baroudi (1991) and
reflect on the progress made in the IS discipline since 1990. We have argued and illustrated
that many activities have evolved in the IS research community since the paper of Orlikowski
& Baroudi (1991) was published. We then ask, Are those changes reflected in journal publi-
cations? Our empirical analysis of 1893 articles reveals a result similar to that of Orlikowski &
Baroudi (1991), particularly in relation to the paradigmatic trend. The influence of long-term
advocacy of paradigmatic pluralism, thus, appears negligible. In other words, changes that
have occurred and efforts that have been devoted since 1990 still have not manifested them-
selves in journal publications. This could be good news for mainstream navigators (traditional
positivists) and unity advocates (Landry & Banville, 1992) because a more unified, positivist
tradition may have led to a better-established IS field thus far.
Such a result might be due to the existing tenure and promotion and publication systems.
The tenure system overly emphasizes the number of publications in promoting a faculty mem-
ber and, thus, encourages researchers to engage in less time-consuming projects. Positivist
and quantitative research, as opposed to interpretive or qualitative research, generally suits
this purpose better, particularly when researchers are under tremendous publish or perish
pressure. In addition, on the publication system side, journal editors and reviewers tend to be
less receptive to interpretive and qualitative research because the evaluation criteria are less
objective (Hirschheim & Klein, 2002a). Consequently, researchers may be more reluctant to
conduct time-consuming research projects particularly in the domain of interpretivism where
the reward of publication is less likely.
As such, if the IS discipline believes that a monistic, Kuhnian view of IS research is inad-
equate (Hirschheim et al., 1996; p. 4), then a continuous commitment to paradigmatic plural-
8/2/2019 _Chen___Hirschheim_2004_
33/39
Examination of IS research
2004 Blackwell Publishing Ltd, Information Systems Journal14, 197235
229
ism is critical. This would likely involve the need to modify the current tenure and publication
systems. While the former could consider incorporating the quality of publications into the eval-
uation criteria, the latter could develop a more receptive attitude toward alternative research
approaches. In doing so, IS researchers would not just conceptually embrace the possibility ofalternative assumptions and interests but also empirically allow the notions to be manifested in
journal publications. Only when alternative approaches are empirically reflected in the journal
publications would the IS discipline establish itself in a pluralist research tradition that allows
alternative approaches to help build its body of knowledge.
REFERENCES
Alavi, M., Carlson, P. & Brooke, G. (1989) The ecology of
MIS research: a twenty year status review. In: The Pro-
ceedings of the Tenth International Conference on Infor-
mation Systems, DeGross J.I., Hendersonm, J.C. &
Konsynski, B.R. (eds), pp. 363375. Boston, MA.
Applegate, L.M. & King, J.L. (1999) Rigor and relevance:
careers on the line. MIS Quarterly, 23, 1718.
Benbasat, I., Goldstein, D.K. & Mead, M. (1987) The case
research strategy in studies of information systems. MIS
Quarterly, September, 369386.
Benbasat, I. & Weber, R. (1996) Research commentary:
rethinking diversity in information systems research.
Information Systems Research, 7, 389399.
Benbasat, I. & Zmud, R.W. (1999) Empirical research ininformation systems: the practice of relevance. MIS
Quarterly, 23, 316.
Boland, R.J. Jr & OLeary, T. (1991) Editorial Technolo-
gies of inscribing and organizing. emerging research
agendas. Accounting, Management and Information
Technology, 1, 17.
Burrell, G. & Morgan, G. (1979) Sociological Paradigms
and Organizational Analysis. Heinemann, London, UK.
Cash, J. & Nunamaker, J (1989) The Information Systems
Research Challenge, Vol. I: Qualitative Research Meth-
ods. Harvard Business School Press, Boston, MA, USA.
Cash, J. & Nunamaker, J (1990) The Information Systems
Research Challenge, Vol. II: Experimental Research
Method. Harvard Business School Press, Boston, MA,
USA.
Cash, J. & Nunamaker, J. (1991) The Information Systems
Research Challenge, Vol. III: Survey Research Method.
Harvard Business School Press, Boston, MA, USA.
Chua, W.F. (1986) Radical developments in accounting
thought. The Accounting Review, 61, 601632.
Cook, S.D.N. & Brown, J.S. (1999) Bridging epistemolo-
gies: the generative dance between organizational
knowledge and organizational knowing. Organization
Science, 10, 381400.
Daft, R.L. & Lengel, R.H. (1986) Organizational informa-
tion requirements media richness and structural design.
Management Science, 20, 178188.
Darke, P., Shanks, G. & Broadbent, M. (1998) Success-
fully completing case study research: combining rigour,
relevance and pragmatism. Information Systems Jour-
nal, 8, 273289.
Davenport, T.H. & Markus, M.L. (1999) Rigor vs. relevance
revisited: response to Benbasat and Zmud. MIS Quar-
terly, 23, 1923.
DeSanctis, G. &