Directory UMM :Data Elmu:jurnal:I:Information and Management:Authorlist E:

Information & Management 35 (1999) 331±344

Research

Surveying object technology usage and bene®ts:
A test of conventional wisdom1
Jane Fedorowicza,*, Alain O. Villeneuveb,2
a

Department of Accountancy, Bentley College, 175 Forest Street, Waltham, MA 02452, USA
Faculte d'administration, Universite de Sherbrooke, Sherbrooke (QueÂbec) J1K 2R1, Canada

b

Received 8 January 1998; accepted 27 September 1998

Abstract
Vendors and developers alike profess a profound shift in the paradigm of systems analysis, design, and programming based on
object-oriented techniques. A survey was sent to over 1200 IS professionals with an expressed interest in OO. The results of
the questionnaire provided descriptive information on their level of experience with OO in use in the ®eld and also to garnered
professional perceptions on the usefulness and bene®ts of various aspects of OO use. We found that many vendor-touted

bene®ts are upheld by professionals using these tools, yet not always to the extent predicted. In particular, the techniques were
harder to learn than expected and do not give a novice an anticipated edge in acquiring professional expertise. Overall,
however, respondents preferred to use OO for application development, as well as to support team-based activities such as
client communications, project team communications, and new team member familiarization. Professional users expected that
OO could require a greater time investment at the beginning of the system development life cycle, with time savings accruing
at the latter stages of a project's implementation and use. Expectations concerning the reusability and shareability of objects
also appeared to be met. The most favorable preferences and bene®ts were reported by those respondents who have used OO
most. # 1999 Elsevier Science B.V. All rights reserved.
Keywords: Object orientation; Systems development; Practitioner survey

1. Introduction
Object-oriented techniques (OOT) have been used
to help in understanding the relationship among entities in an enterprise's domain and also to provide
expanded programming languages that permit these

*Corresponding author. Tel.: +1-781-891-3153; fax: +1-781891-2896; e-mail: [email protected]
1
Preliminary results were presented at The First Informs
Conference on Information Systems and Technology, Washington,
DC, May 5±9, 1996.

2
Tel.: +1-819-821-7329; e-mail: [email protected]

relationships to be shared among programs and applications. Object orientation (OO) also encompasses
specialized analysis and design methodologies for
depicting and formalizing object entities.
OO has made its way into computer-aided software
engineering (CASE) tools, traditional programming
languages (e.g., C‡‡, OO-COBOL) and data base
management systems. The World Wide Web is said to
be the ideal vehicle for sharing objects and reusing
them, as well as a key bene®ciary of the technology
itself. A second generation of OO methodologies has
been developed; the best features of the leading
approaches of Rumbaugh, Booch and Jacobson are

0378-7206/99/$ ± see front matter # 1999 Elsevier Science B.V. All rights reserved.
PII: S-0378-7206(98)00098-6

332


J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

merged into a combined, universal OO method [20].
Examples of OO use in large-scale applications have
been published (e.g., [5]). Vendors and columnists
continue to laud the merits of the technology, although
others have begun to question the universal bene®ts
attributed to the collection of ideas, tools, methods and
languages claiming to exhibit OO characteristics. A
typical set of attributions appears in CACM: ``OOT
promotes a better understanding of requirements and
results in more modi®able and maintainable applications, providing other bene®ts such as reusability,
extensibility, robustness, reliability, and scalability.
OOT promotes better teamwork, good communication
among team members, and a way to engineer reliable
software systems and applications'' [12].
The intent of our work was to test the claims about
OO's in¯uence within professional practice. A survey
was sent to over 1200 IS professionals with an

expressed interest in OO. We hoped to obtain descriptive information on their experience with OO, and
also to gather their perception of OO usefulness and
bene®ts.

2. Conceptual framework
2.1. Perceptual and comparative assessment of OO
Given the large amount of press coverage that OO
has experienced in the past few years, it might be
expected that this technology has achieved positive
acceptance. The ®rst set of hypotheses aims to verify
this.
Several models have been proposed for determining
when an IS development process has achieved success. Although there is no consensus on the de®nition
of success, researchers have proposed a number of
constructs to measure both actual and perceived successful use, such as ease of use, usefulness, and
ef®ciency. Much of the literature on successful technology adoption focuses on technologies that have
been developed to support end users (e.g., [17, 18]). In
the case of OO, IS developers can be viewed simultaneously as users of systems (e.g., OO development
tools, languages and environments), adopters of methodologies (e.g., particular OO development methods),
and developers of other users' systems (the end product of the OO activity). Therefore, a study of OO


adoption must do more than incorporate the use constructs proposed for measurement of end-user reactions. In addition to understanding the perceived ®t
between the application development task itself and
OO tools and methods, comparisons of OO methods to
prior, traditional methods are needed to establish the
difference from traditional alternatives. These comparisons provide a baseline for assessing skill acquisition and productivity impacts within the application
development process.
Few researchers have ventured to study the impact
of alternative development methodologies on the
development process per se. Finlay and Mitchell
[15] discuss one company's experience with CASE
tools, noting that developers perceived improvements
in productivity, systems quality, and developer effectiveness. Although their study did not extend to OO,
the parallels between OO and CASE tools as embodiments of development and communication methods
provide preliminary validation for the productivity
bene®ts discussed in the popular press.
Thus, the ®rst set of hypotheses includes evaluative
criteria similar to those of most end-user studies,
including system usefulness and ef®ciency [9, 11].
It also covers issues that are peculiar to the role of

developer, comparing OO to traditional approaches in
job ef®ciency improvements (i.e., productivity
impact), and an overall comparison for application
development preferences.
H1a: OO is perceived to be useful and ef®cient.
H1b: OO is perceived to promote job ef®ciency more
than other approaches respondents have used.
H1c: OO is preferred for application development
over traditional approaches.
2.2. Learning and communication
A secondary feature of these ef®ciency bene®ts is
the reduced time and effort required to learn OOT. If
the users express a preference for OO on the dimensions comprising these constructs, then it is ®tting that
we try to understand the circumstances surrounding
the stated respondents' gains. Proponents of the OO
paradigm contend that it is more suitable for promoting professional-level performance, learning and the

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

transfer of expertise than more traditional approaches.

This contention is based on the notion that OO more
closely resembles the way humans store and retrieve
knowledge from mental models in memory than traditional approaches [13, 30]. This was corroborated in an
experiment by Wang [31], where student subjects using
an OO method produced analysis that more closely
matched the problem than those using structured analysis. The students also reported that the OO method was
easier to use than the structured analysis method.
In their experiment, Agarwal et al. [2] suggest that
performance using process-oriented and OO tools
depends on the cognitive ®t of the tool with the task,
with process-oriented tools outperforming or matching performance with OO tools in requirements analysis activities.
Vessey and Conger [28] noted that novices ®nd OO
harder to learn than other structured approaches and
that their initial performance suffers in comparison
with structured techniques. They identi®ed several
differences in the ability of novices to specify information requirements when using one of three development methodologies, with OO more dif®cult to
learn and resulting in poorer initial performance than
traditional structured techniques or the Jackson Systems Development approach. However, their subjects
were students without prior experience in any of the
techniques rather than practicing professionals, eliminating the potential for measuring user preferences

across methods and enjoining the generalizeablility of
their laboratory ®ndings to practitioners.
We hypothesize that OO will be perceived as superior to other approaches that respondents have used in
facilitating the transfer of expertise and will be easier
to learn for novices than traditional approaches. It is
also conjectured that the OO skill acquisition learning
curve will be shortened as a result of the similarities to
mental models as a representation of knowledge.
H2a: OO is perceived to be easy to use and easy to
learn.
H2b: OO is perceived to be easier to learn for novices
than traditional approaches.
H2c: OO is perceived to be easier to become skilled in
comparison to other approaches respondents have
used.

333

Coordination and communication have been found
to be determinants of systems development success [8,

10, 27] and provide the leverage to exploit OOT fully
[25, 26]. Both external and internal communication is
essential for the success of team-based projects. OO
promotes a formal, commonly understood communications vehicle, which is particularly useful in large,
complex projects [22]. Thus, we also examine the
usefulness of OOT as a communication mechanism.
H3a: OO is perceived to promote communication with
team members better than traditional approaches.
H3b: OO is perceived to promote communication with
users better than traditional approaches.
2.3. Time savings
Traditional systems development techniques have
been criticized for failing to produce successful systems, for producing error-ridden applications, and for
greatly exceeding expectations of budgeted time and
expenses. Most development effort (time and expense)
typically is spent on the programming and testing of
applications, with very little (perhaps 5%) of the effort
spent on requirements analysis. Yet, studies have
shown that most errors originate in the requirements
analysis or design phases of the systems development

life cycle (SDLC) but are not discovered until the
coding, testing, or maintenance phases [24]. One of
the supposed bene®ts of OOT is the emphasis placed
on understanding the application domain and performing requirements analysis. Hence, OO efforts that span
many phases of the SDLC should lead to improvements in the success rate, cost, and error rate of the
ensuing applications. Our fourth set of hypotheses test
the relationship between time savings and the use of
OOT. We further suggest that gains will be higher
when OO is used for more steps of the SDLC. Additionally, we expect that the time savings are not
universal across all stages of the SDLC, but rather
that OO is expected to take longer in the initial stages
and reap time saving bene®ts in the latter ones, a
savings which has yet to be documented for other
methodologies.
H4a: OO is perceived to save development time over
traditional approaches respondents have used.

334

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344


H4b: Larger time saving is perceived when OO is used
for more steps of the SDLC.
H4c: OO is perceived to take longer in the initial
stages of SDLC but to save time in later stages.
2.4. Project characteristics and reuse
Proponents of OO claim that there are major time
savings over traditional approaches; these should lead
to systems development productivity gains, which,
particularly in the area of code reuse, have already
been found to be associated with high levels of CASE
tool integration across several activities within the
SDLC [3]. OO proponents frequently discuss examples attributing reuse with dramatic increases in productivity.
Part of the leveraging effect of OO as a coordination
and communication mechanism is, like many other
modern methods, the ability to rely on a single model
throughout the SDLC. More diffusion will favor better
coordination amongst all of the individuals involved in
the development effort, as well as a common language
and representation for the objects themselves, hence
substantiating expected bene®ts in terms of reuse and
sharing of objects. We hypothesize that the more
diffused OO is throughout SDLC, the greater the
prevalence of object reuse and sharing. We also
hypothesize that more diffusion will lead to more
favorable impressions of OO, speci®cally those identi®ed in Hypotheses 1, 2, 3 and 4.
H5a: More object reuse is perceived when OO is used
for more steps of the SDLC.
H5b: More object sharing is perceived when OO is
used for more steps of the SDLC.
H5c: Usage in more steps of the SDLC is associated
with more favorable perceptions of OO.
Banker and Kemerer [4] found that economies of
scale frequently exist for large projects. In this vein,
we hypothesize that a scaling effect promotes reuse
and sharing of objects, in no small part due to larger
projects usually comprising more objects than smaller
ones. Since companies typically begin with smaller
yet mission-critical projects before embarking on

large projects, it is expected that libraries of objects
will not be available for initial applications, as reuse
will probably not be effective until a signi®cant library
has been assembled. Kraut and Streeter found that
formal, impersonal project communication mechanisms were used more frequently in larger projects,
especially once the project had ®nished the requirements and design stages of the SDLC. They also found
that these mechanisms were deemed more valuable:
they were more widely used.
H6a: Object reuse is more likely to be expected when
larger projects have been developed.
H6b: Object sharing is more likely to be expected
when larger projects have been developed.
H6c: Experience with larger projects is associated
with more favorable perceptions of OO.
2.5. Experience
Many of the bene®ts of OO are best achieved after
considerable experience with the techniques. Reuse is
frequently cited as exhibiting a growth curve based on
number of prior systems built with OOT. Finlay and
Mitchell noted an improvement in systems quality due
to the introduction of CASE tools, as well as perceived
initial improvement in productivity. Novice users were
more likely to perceive a positive impact on developer
productivity.
Therefore, we expect to ®nd a positive relationship
between OO experience with the approach and users'
perceptions, their preferences with respect to other
approaches, and their assessment of its contribution to
different aspects of SDLC. In addition to determining
whether experienced respondents are more favorable
in their evaluation of OO, these hypotheses will also
tell whether greater involvement leads to increased
expectations of the bene®ts and characteristics. In
contrast, other, non-OO experience, is not expected
to affect OO evaluation. Assuming a `paradigm shift'
inherent in a transition to OO, experience with other
methods should not affect respondent perceptions.
H7a: Experience with the OO approach is associated
with more favorable OO perceptions, preferences, and
assessments of bene®ts.

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

H7b: Non-OO experience does not demonstrate any
relationship with OO perceptions, preferences, or
assessment of bene®ts.
2.6. Tools and methods
OO properties are attributed to a broad range of
theoretical concepts and vendor products. Commercial products possess a range of OO `purity'. For
example, Smalltalk is perceived to exhibit pure OO
characteristics, while hybrid languages such as C‡‡
do not [1, 16]. Ease of use of available products also
varies from one to another. We expect to ®nd differences in perceptions concerning OO's bene®ts due to
the type of tool used [6, 7, 23]. We also expect that for
those using languages, e.g., programming in Smalltalk, will lead to higher perceptions of bene®ts than
those programming in C‡‡.
H8a: Usage of OO development environments is
associated with more favorable perceptions than
OO programming languages.
H8b: Usage of Smalltalk is associated with more
favorable perceptions of OO than usage of C‡‡.
Many argue that the use of an OO language does not
mandate that the application is necessarily OO, but
that it still may be developed using a traditional
methodology. Similar to CASE technology, adoption

335

of an object-based language does not guarantee many
of the expected bene®ts and must be accompanied by a
robust methodology [29]. Thus, we expect that respondents who have adopted one (or more) methodologies
to support OO development should expect to achieve
greater bene®ts than those who do not rely on a
methodology.
H9: Using formal OO methodologies is associated
with more favorable perceptions of OO.
Fig. 1 summarizes our research model.

3. Method
A survey was developed to ascertain OO usage
history, tool and product experience, perceptions
about OO and traditional development techniques,
overall perceptions of the usefulness of OO, assessments of reuse, shareability and similar claims, and
demographic data on respondents and their companies. Use of the survey research method provides a
broad-based mechanism for assessing the breadth of
penetration in a wide variety of companies and industries. Although some of the hypotheses could be
measured in a limited way in a laboratory setting,
the emphasis on perceptual data enables the collection
of a wide range of baseline data to lend generalizeability to subsequent experimental analysis. Several

Fig. 1. Research model.

336

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

common IS research scales on use and usability
formed the basis of the sections measuring perceptions
(e.g., [9]). Factual questions required checking one or
more applicable boxes or ®lling in a blank space. In
addition, respondents were asked to enumerate the
different OO tools and methods they have used.
The survey was pretested at two professional IS
conferences, one in a session on OO, and another at an
OO exposition, for a total of 77 pretest subjects.
Con®rmatory factor analysis was used to assess the
scales, which had been developed from our theoretical
model. All factors were clean and strong with no
cross-loading. No items were dropped from the pretest
version, three perceptual bene®ts items were added
and ®ve demographic items were transformed from
close-ended to open-ended.
Subsequently, a mailing list of practitioners with
interest in OO tools was obtained from a leading OO
tool company, and surveys were sent to over 1200
individuals. A total of 228 useable responses were
received.

hands-on experience with commercial OO tools,
50.3% of them had experience with more than one
tool. Many (40.0%) reported direct experience with
formal OO methodologies, 44% of those had experience with more than one methodology. Over 120
different tools and 30 methodologies were listed as
being used, reducing the potential for bias that might
be inferred from a narrow cross-section of tool and
method experience. Not all respondents reported using
commercial tools or formal methodologies. Ninetyseven had taken an OO course.
Most (60%) reported that OO is used by their
company for developing new applications, while
18% said that OO was used on existing applications.
When questioned about the use of OO for the different
steps in the SDLC, 30.7% of the respondents reported
using OO for understanding the domain, 55.3% for
analysis, 60.1% for design, 54.4% for implementation,
and 33.3% for maintenance.

3.1. Sample

Most perceptual data were captured on ®ve-point
self-anchoring Likert scales with a `6' representing no
opinion [19]. Perceptions of reusability, shareability,
understandability by novices, and usefulness for shortening the learning curve were three-point scales (yes/
no/no opinion). Because of the lack of available

Demographic data describing the sample is
included in Fig. 2. The respondents were a mix of
IS staff, IS managers, consultants, non-IS managers,
and upper level management. Most (70%) had direct,

3.2. Measures

Fig. 2. Sample demographics.

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

quanti®able analysis of reuse and shareability patterns
of objects at most sites at the time of the survey, a ®ner
demarcation was thought to be dif®cult to assess.
Several constructs and variables formed the basis
for the comparisons. Factor analysis was applied to
derive a construct comprising perceived ease of learning and ease of use to represent the perceived attributes of the technology (two-items, r ˆ 0.50). A
construct representing major outcomes of using the
technology, perceived usefulness and ef®ciency, was
also computed (two-items, r ˆ 0.68). A construct
re¯ecting the comparative ease of skills development
was derived (three items, ˆ 0.73), and another
regrouping different aspects of ef®ciency and productivity was devised (®ve items, ˆ 0.91).
Project size was self-reported from a set of choices
on a checklist, with the range covering pilot projects,
small projects only, large projects only, and all projects. Another variable, count of SDLC steps when OO
is used, encodes its breadth (or span) of activities
within the SDLC. Company size, years in computer
industry, years in current position, and years of experience with OO were transformed into logarithms due to
highly signi®cant skewness.
Each tool reported being used was coded either as
an environment, a language, or other (e.g., DBMSs,
object libraries) based on information available in the
trade literature ([21]; Lexis/Nexis; Company ads). A
variable was then devised to capture the respondent's
exposure to languages only, environment only, both
environment and language, or none. Another variable
was derived to capture those languages that the
respondent had used: C‡‡, C‡‡ and SmallTalk,
Smalltalk, or other.
4. Results
4.1. Assessment of OO and comparison to other
approaches
The ®rst set of hypotheses was supported (see
Table 1). In terms of its perceived characteristics,
OO was found to be useful and ef®cient. Overall,
respondents strongly prefer OO to other approaches
for application development and ®nd that it promotes
job ef®ciency more than other techniques.
The second set of hypotheses assesses the learning
bene®ts of OO. The results show that OO is not

337

considered to be easier for novices to learn than other
methods. Slightly less than half of the respondents
believe that objects are understandable by novices, a
non-signi®cant ®nding demonstrating a mix of opinion
on the ability of novices to immediately grasp OO
concepts. Contrary to the stated hypothesis, OO was
thought to take longer for an IS novice to learn and to
become an IS expert. These ®ndings are consistent with
Vessey and Conger's 1994 results, and also with published company reports in the trade press of a 12-month
learning period for getting developers up to speed with
OO. When speci®cally compared to other approaches
respondents have used, skill in OO was perceived to be
more dif®cult to acquire. However, once learned, OO
appears tobe the techniqueofchoice, based onthe results
of the ®rst two sets of hypotheses.
Once a project is underway, the bene®ts of OO as a
development environmentbegin to accrue. A majority of
the respondents believe that objects are useful for shortening the learning curve for new project team members.
Respondents strongly prefer OO to other approaches for
communicating with users and for communicating
among team members, demonstrating support for
Hypotheses 3a and 3b. Thus, any similarities of OO to
human knowledge representations may not assist in the
initial learning process but may help as a communications aid for team members and their clients.
The fourth set of hypotheses, examining the use of
OOT within SDLC stages, were all supported. Overall,
OO is perceived to save time in developing applications. Breadth of use throughout the SDLC signi®cantly correlates with time savings during
implementation and maintenance. When compared
to other approaches, OO was found to take about
the same time for understanding the domain and
design steps, take more time for analysis, and less
time for implementation and maintenance. A factor
analysis of the items pertaining to time savings
grouped understanding the domain with analysis
and design, and a second factor loaded implementation
with maintenance. It thus seems that an OO investment in the early aspects of the SDLC pays off later in
the cycle, especially for maintenance activities.
4.2. Reuse and sharing
The ®fth and sixth sets of hypotheses on object
reuse and sharing were all supported (see Table 2). A

338

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

Table 1
Perceptual and comparative assessments of OO. Note the lower the rating on scales, the more positive the perception of OO
na
Hypothesis 1
Factor for direct assessment of OO
Usefulness and efficiency (10 point scale)
Factors for comparative assessment to traditional methods
Efficiency in job (25)
Prefer OO for application development
Hypothesis 2
Easy to learn and use (10)
Faster for a novice to learn
Faster for a novice to become an expert
Ease of skills development (15)
Hypothesis 3
Shortens the learning curve of new team members (2)
Prefer OO for communicating with team members
Prefer OO for communicating with users
Hypothesis 4
Overall time savings (25)
Time required for
Understanding the domain
Analysis
Design
Implementation
Maintenance
Correlation between Count of SDLC steps and Time savings

Mb

zc

186

4.63

ÿ8.88 f

170
200

11.88
1.98

ÿ9.05 f
ÿ12.29 f

203
199
194
192

6.13
3.56
3.52
9.68

0.92
6.16 f
5.89 f
3.65 f

180
198
191

1.31
2.11
2.39

ÿ5.46 f
ÿ10.21 f
ÿ7.39 f

161

13.60

ÿ4.37 f

3.10
3.24
3.16
2.63
1.81

1.13
2.64 e
1.80
ÿ4.15 f
ÿ15.45 f

196
196
196
190
167
ÿ0.174 d

a

Different sample sizes due to missing values.
Mean value of construct on a ®ve-point scale unless indicated in parentheses after construct name. Two-point scales: 1 ˆ yes, 2 ˆ no.
c
z-Value re¯ects a mean value M which is signi®cantly different from the scale mean.
d
p < 0.05.
e
p < 0.01.
f
p < 0.001.
b

majority of respondents believe that objects are shareable and reusable. Both project size and breadth of use
throughout the SDLC correlate signi®cantly with perceptions of shareability and reusability of objects. This
may mean that OO is more bene®cial for larger
projects because the number of objects in the system
or the size of the development team is greater. Also,
when OO is used across more SDLC steps, the learning curve for OO may be shortened, due to more
uniform, formal communications throughout the
SDLC process.
Project size also signi®cantly correlates with presumed time savings during implementation and maintenance. Since diffusion of the approach across the
development cycle also leads to time savings during
the latter steps of the development, this may con®rm a
scaling effect due to project size and diffusion of the

approach. Larger projects may lead to more overall
bene®ts, as may the breadth of use of the approach
within the SDLC; larger diffusion of the approach
throughout the SDLC may lead to more usefulness and
ef®ciency.
4.3. Experience with OO
The tests of Hypothesis 7a found that respondents'
experience with OO has a positive relationship with
many, but not all, OO perceptions, preferences, and
bene®ts (see Table 3). Perceived ease of use is positively and signi®cantly associated with respondents'
work experience with OO, as is comparative ease of
skills development. This probably re¯ects a learning
process, where more experience leads to more mastery
of the technology and therefore to a more favorable

339

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344
Table 2
Reuse and sharability.

Note the lower the rating on scales, the more positive the perception of OO
na

Mb

Shareability (2)
Reuseability (2)

194
197

1.20
1.19

Hypothesis 5,6: Correlations with:
Sharing
Reuse

Count of SDLC steps
ÿ0.209 e
ÿ0.165 d

Project size
ÿ0.262 e
ÿ0.192 d

Easy to learn and use
Usefulness and efficiency
Ease of skills development
Efficiency in job

ÿ0.064
ÿ0.268 f
ÿ0.035
ÿ0.213 e

ÿ0.155
ÿ0.326 f
ÿ0.111
ÿ0.336 f

Prefer for development
Prefer for communicating with team
Prefer for communicating with users

ÿ0.260 f
ÿ0.307 f
ÿ0.205 e

ÿ0.357 f
ÿ0.293 f
ÿ0.205 e

Total time savings
Time to understand
Time for analysis
Time for design
Time for implementation
Time for maintenance

ÿ0.174 d
ÿ0.126
ÿ0.009
ÿ0.023
ÿ0.245 f
ÿ0.264 f

ÿ0.200 d
ÿ0.057
0.108
ÿ0.096
ÿ0.312 f
ÿ0.268 f

zc
ÿ10.65 f
ÿ10.90 f

a

Different sample sizes due to missing values.
Mean value of construct on a ®ve-point scale unless indicated in parentheses after construct name. Two-point scales: 1 ˆ yes, 2 ˆ no.
c
z-Value re¯ects a mean value M which is signi®cantly different from the scale mean.
d
p < 0.05.
e
p < 0.01.
f
p < 0.001.
b

perception of its characteristics. More work with the
approach may also aid in `unlearning' traditional
approach methods and favor a more positive perception of ease of skills development.
Perceived and comparative bene®ts of the approach
are associated with work experience with the
approach, size of projects, and the number of SDLC
steps covered with the approach. This means that the
more one works with the approach, the more OO's
contribution in terms of usefulness and ef®ciency is
realized. In addition, respondents' work experience
with OO also leverages time savings, reducing the
time required for understanding the domain, conducting the analysis, and implementing systems. Respondents from companies using OO for maintenance
activities report higher expectations for time saving
in the maintenance step than those from companies not
reporting maintenance activity (t-test under H of
unequal variance, p-value ˆ 0.0007).

Table 3
Experience
Hypothesis 7: Correlations with Years of OO Experience
Project size
Count of SDLC steps
Easy to learn and use
Usefulness and efficiency
Ease of skills development
Efficiency in job
Shareability
Reuseability
Shortens learning curve
Faster for novice to learn
Faster for novice to become expert
Prefer for development
Prefer for communicating with team
Prefer for communicating with users
a

p < 0.05.
p < 0.01.
c
p < 0.001.
b

0.381 c
0.265 c
ÿ0.156 a
ÿ0.194 b
ÿ0.261 c
ÿ0.248 b
ÿ0.103
ÿ0.092
0.089
0.037
ÿ0.166 a
ÿ0.279 c
ÿ0.241 c
ÿ0.175 a

340

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

More work experience with OO does not lead to
higher perception of shareability, reusability, and
accessibility to novices. These perceptions of experienced respondents did not differ signi®cantly from the
overall pool of subjects, suggesting that opinions
about OO use and bene®ts meet the initial expectations of developers. The only exception is that more
work experience is associated with the perception that
OO will allow novices to become expert faster.
In contrast, no difference in perception of bene®ts
or technique preferences were found due to company
size, years in computer industry, job tenure, gender
and age of the respondents. Therefore, Hypothesis 7b
was supported.
4.4. Tools and methods
The last hypotheses are intended to determine
which OO technologies provide the most perceived
bene®ts. There are a vast number of languages, CASElike methodologies, and other technologies that are
labeled OOT by their vendors. Our hypotheses divide
the respondents into those who use several technologies, and those who do not (see Table 4). Because

these variables contain categorical data, ANOVAs,
t-tests, multi-way frequency analysis (LOGIT) and
Tukey HSD tests were computed to assess signi®cance.
Hypothesis 8a looks at the added value of development environments over OO languages. For this, a
subset of the dataset was created: it includes only those
respondents reporting the use of a language (e.g.,
C‡‡ or Smalltalk) or an environment (e.g., STP/
OMT). Those reporting both or neither were eliminated, leaving a sample of 102. There were a few
notable, signi®cant differences between the language
only and environment only groups. The latter rated
OO higher on usefulness and ef®ciency and expressed
a preference for OO for communication with users.
The other results were not signi®cant.
When type of tool is ignored, the bene®ts of using
any type of tool becomes apparent. Those respondents
using any combination of languages and/or environments express a stronger preference for using OO for
application development than non-users, for communicating with users, and expect higher time savings,
especially for implementation and maintenance.
Those with tool experience see no difference with
other approaches in terms of time to understand the

Table 4
Use of tools Note: Only significant results are reported here
Hypothesis 8: ANOVA results
Preference for development
Ease of learning and use
Usefulness and efficiency
Efficiency in job
Ease of skills development
Preference for communication with team
Preference for communication with users
Shareable (n.s., Chi-squared tests)
Reusable (n.s., Chi-squared tests)
Shortens learning curve (n.s., Chi-squared tests)
Faster for novice to learn
Faster for novice to become expert
Time savings
Time to understand domain
Time for analysis
Time for design
Time for implementation
Time for maintenance
a

Subsample size ˆ 102.
Full sample size ˆ 228.
c
Subsample size for C‡‡ only ˆ 66, for Smalltalk ˆ 9.
b

Type of tools a

Use of (Any) tools b

OO Languages c

t-test (df)

F (df)

t-test (df)

6.42 (1 190)
2.10 (84)
3.25 (67)
2.49 (90)

5.51 (1 181)

5.80 (1 151)
3.99 (1 186)

4.92 (1.180)
6.25 (1 157)

341

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

domain; those who do not use tools believe that it will
take longer to understand the domain with OO than
with traditional approaches.
Hypothesis 8b was dif®cult to test, because there are
only 9 respondents who solely use Smalltalk (a percentage that is quite similar to the general population
of OO language users). Therefore, statistical signi®cance is suspect. The only variable exhibiting a signi®cant difference is Ease of Skills Development,
wherein Smalltalk users found OO less dif®cult to
learn and become skillful in than the pure C‡‡ users.
However, this ®nding con®rms the widespread perception that C‡‡ is a much more dif®cult skill to
acquire than Smalltalk.
Hypothesis 9 examines whether the adoption of a
formal methodology (e.g., Booch, Rumbaugh)
increases the perceptions of OO bene®ts. Signi®cance
was established for respondents' preference for OO
for application development, and its use as a communication tool with users and team members (see
Table 5). There were no signi®cant differences in
expectations for time savings.
Additional analysis was run to ascertain whether
methodologies were more effective when both tools

and methods were used. When OO methods are used
in conjunction with an OO development environment
or OO programming language, perceptions of time
savings in the area of maintenance become signi®cant.
Also, joint use of tools and methodologies was also a
prerequisite for signi®cant ®ndings for usefulness and
ef®ciency.

5. Discussion
The descriptive analysis reported here provides an
initial understanding of the bene®ts and experiences of
professionals practicing in the ®eld. Overall, our
hypotheses are supported by the data. We ®nd that
many of the vendors' claims are upheld by professionals using the tools, yet not always to the extent that
the vendors wish. In particular, the techniques are hard
to learn, and do not give a novice an anticipated edge
in acquiring expertise. Overall, however, respondents
prefer to use OO for application development, as well
as to support team-based activities, such as client
communications, project team communications, and
new team member familiarization. Professional users

Table 5
Use of methods note: only significant results are reported here
Hypothesis 9: ANOVA results

Preference for development
Ease of learning and use
Usefulness and efficiency
Efficiency in job
Ease of skills development
Preference for communication with team
Preference for communication with users
Shareable (n.s., Chi-squared tests)
Reusable (n.s., Chi-squared tests)
Shortens learning curve (n.s., Chi-squared tests)
Faster for novice to learn
Faster for novice to become expert
Time savings
Time to understand domain
Time for analysis
Time for design
Time for implementation
Time for maintenance
a

Full sample size ˆ 228
Subsample size ˆ 158.

b

Experience with
methods a

Experience with methods
(Tool users only) b

F (df)

F (df)

9.46 (1 190)

7.07 (1 146)
3.94 (1 131)

12.55 (1 188)
8.19 (1 181)

12.20 (1 143)
5.87 (1 138)

5.09 (1 118)

342

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

Table 6
Summary of significant findings
Overall, respondents prefer OO to other traditional analysis and design methods for application development. This preference is stronger when
OO is used in a greater number of SDLC steps, when project size is larger, and when the respondent has greater OO experience. Those using
any tools (environment, language, or both) express stronger preferences than those who do not, as do respondents using formal methodologies,
especially when methodologies are used alongside tools.
Respondents find OO to be more useful and efficient than other approaches to systems development. This opinion is stronger when OO is used
in a greater number of SDLC steps, when project size is larger, and when the respondent has greater OO experience. Respondents using
development environments are more favorable to OO than those relying on OO programming languages. Those using formal methodologies in
conjunction with any type of tool also express more favorable opinions of usefulness and efficiency.
When compared with other approaches, respondents find OO to be more difficult to acquire skills in, although experienced users find it less
difficult than those with less OO experience. In particular, they found it harder to learn and harder to become skillful in than other approaches.
Those with greater OO experience also found OO to be more easy to learn and use in general than those with less experience. Those using both
methods and tools also find OO easier to learn and use. The respondents perceive that OO will take longer for a novice to learn than traditional
techniques, and longer for a novice to become an IS expert. Those with more OO experience thought that IS expertise is harder to acquire than
those with less experience.
Compared to other approaches, OO is preferred because of the characteristics leading to job efficiency, with respondents considering OO to be
more efficient, more productive, more effective, easier to do your job, and improving job performance. These preferences are stronger when
OO is used in a greater number of SDLC steps, when project size is larger, and when the respondent has greater OO experience.
Respondents prefer OO for communicating with users and for communicating among team members. These preferences are stronger when OO
is used in a greater number of SDLC steps, when project size is larger, and when the respondent has greater OO experience. Client
communication rating is enhanced by the use of environments over languages, also when using any type of tool, and when employing a formal
methodology, especially when also using a tool. Likewise, team communication is improved when a formal methodology is used, especially in
conjunction with a tool. Respondents also find OO to be useful for shortening the learning curve for new project members.
OO leads to time savings when used across a greater number of SDLC steps, for larger projects, and when any tool is used. However, all
respondents expect OO to take longer to perform the analysis step of SDLC than traditional methods. Time savings are expected to accrue
from the implementation and maintenance steps, especially when used across a greater number of SDLC steps, for larger projects, when OO
tools are used, and when methodologies are used with tools. Tool users expect OO to take about the same time as traditional methods for
understanding the domain, yet those not using tools expect to take longer for this step.
Respondents think that objects they have worked with are or will be shareable and reusable. This opinion is stronger when OO is used in a
greater number of SDLC steps, when project size is larger, and when the respondent has greater OO experience.

expect that OO will require a greater time investment
at the beginning of SDLC, with time savings accruing
at the latter stages of a project's implementation and
use. Expectations concerning the reusability and
shareability of objects also appear to be met. Table 6
summarizes the main ®ndings from the analysis.
The most favorable preferences and bene®ts are
reported by those respondents who have used OO the
most. Those with OO experience, who have worked on
the largest projects, and who have employed OO in the
most SDLC steps have the most positive responses to
the survey. Those who adopt formal methodologies
and development environments also express more

favorable opinions than those who work without them.
Thus, in its users' eyes, OO appears to exceed expectations once the professional has invested considerable
time in learning the techniques and applying the tools
needed to develop systems effectively within the `OO
paradigm'.

6. Limitations
This study reports on data from a mailed questionnaire. As is the case with all questionnaires, this
research method trades off control and measurement

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

for a greater breadth of coverage and range of respondents. The survey was sent early in the adoption life
cycle of OO technology, so that the potential for
gathering pre-existing internal statistics for measurement of much of the data of interest was low. Given
our goal of encouraging response by employing a short
survey instrument, we elected to rely on perceptual
data rather than asking respondents to collect or
measure OO activity outside of their personal domain.
We anticipated the relatively low response rate to the
questionnaire, given the amount of unsolicited mail in
this ®eld, and the short average tenure of the typical
respondent added to an expected high rate of returned
or forwarded mail.
The respondents represent a cross-section of the IS
industry actually using OO techniques as re¯ected by
the range of reported job titles and companies. They
also are active practitioners, and may be more experienced with OO techniques and development tools than
the ``average'' IS developer. Because they, for the
most part, reported at least some experience with
OO, they may be overly enthusiastic about the technology. Cognitive dissonance theory would suggest
that developers who have spent considerable effort
learning OO techniques would report overly positive
perceptions to justify the invested effort [14]. Thus,
the positive ®ndings may not re¯ect the population
involved in IS development activities.

7. Conclusion
The result of this analysis is immediately useful for
®rms that are beginning to move toward an OO
environment. It recognizes that a long initial period
of transition will lead to substantial expected bene®ts,
but only after the company has invested considerable
time and effort. It is also useful for those ®rms and
professionals with some experience in OO, guiding
them to further pursue OO to garner expected bene®ts.
It substantiates claims in the trade press that the move
to the OO paradigm involves more than learning a new
programming language.
This study invokes concerns about the dif®culties
professionals report in moving to OO, in spite of
suggestions that link OO representations to the natural
knowledge representation schemes of experts. It leads
to questions about the properties of OO that best

343

promote reuse and sharing, as these bear out as the
keys to success for the approach.

References
[1] R. Adhikari, Adopting OO languages? Check your mindset
at the door, Software Magazine, November, (1995) 49±50,
54±55.
[2] R. Agarwal, A.P. Sinha, M.R. Tanniru, Cognitive ®t in
requirements modeling: A study of object and process
methodologies, Journal of Management Information Systems
13(2), 1996, pp. 137±162.
[3] R.D. Banker, R.J. Kauffman, Reuse and productivity in
Integrated Computer-Aided Software Engineering: An empirical study, MIS Quarterly 15(3), 1991, pp. 375±401.
[4] R.D. Banker, C.F. Kemerer, Scale economies in new software
development, IEEE Transactions on Software Engineering
15(10), 1989, pp. 1189±1205.
[5] W. Berg, M. Cline, M. Girou, Lessons learned from the OS/
400 Project, Communications of the ACM 38(10), 1995, pp.
54±64.
[6] B.A. Blake, P. Jalics, An assessment of object-oriented
methods and C‡‡, Journal of Object-oriented Programming
9(1), 1996, pp. 42±48.
[7] T. Bruckhaus, N.H. Madhavji, I. Janssen, J. Henshaw, The
impact of tools on software productivity, IEEE Software,
September, (1996) 29±38.
[8] B. Curtis, H. Kraemer, N. Iscoe, A ®eld study of the software
design process for large systems, Communications of the
ACM 31(11), 1988, pp. 1268±1287.
[9] F.D. Davis, Perceived usefulness, perceived ease of use, and
user acceptance of information technology, MIS Quarterly
13(3), 1989, pp. 319±340.
[10] S.M. Dekleva, The in¯uence of the information systems
development approach on maintenance, MIS Quarterly
16(3), 1992, pp. 355±371.
[11] W.H. DeLone, E.R. McLean, Information systems success:
the quest for the dependent variable, Information Systems
Research 3(1), 1992, pp. 60±95.
[12] M.E. Fayad, W.-T. Tsai, Introduction to section on objectoriented experiences, Communications of the ACM 38(10),
1995, pp. 50±53.
[13] M. Fayad, W.-T. Tsai, M.L. Fulghum, Transition to objectoriented software development, Communications of the
ACM 39(2), 1996, pp. 109±121.
[14] L. Festinger, A Theory of Cognitive Dissonance, Stanford,
CA: Stanford University Press, 1989.
[15] P.N. Finlay, A.C. Mitchell, Perceptions of the bene®ts from
the introduction of CASE: An empirical study, MIS
Quarterly 18(4), 1994, pp. 353±370.
[16] A. Goldberg, , Why Smalltalk? Communications of the ACM
38(10), 1995, pp. 105±106.
[17] D.L. Goodhue, Understanding user evaluations of information systems, Management Science 41(12), 1995, pp. 1827±
1844.

344

J. Fedorowicz, A.O. Villeneuve / Information & Management 35 (1999) 331±344

[18] D.L. Goodhue, R.L. Thompson, Task-technology ®t and
individual performance, MIS Quarterly 19(2), 1995, pp. 213±
236.
[19] E.M. Hufnagel, C. Conca, User response data: The potential
for errors and biases, Information Systems Research 5(1),
1994, pp. 48±73.
[20] D. Kara, Object-oriented analysis and design gears up for the
mainstream, Application Development Trends, March, 1996,
109±110.
[21] P. Korzeniowski, Breaking down the OO language barrier,
Software Magazine, September 1995, 43±45, 49±53.
[22] R.E. Kraut, L.A. Streeter, Coordination in software development, Communications of the ACM 38(3), 1995, pp. 69±
81.
[23] A.H. Lindsay, P.R. Hoffman, Bridging traditional and object
technologies: Creating transitional applications, IBM Systems Journal, 36(1) 1997.
[24] Object Management Group (1992), unpublished slide presentation on Hughes DoD Composite Software Error History.
[25] C.M. Pancake, The promise and the cost of object
technology: a ®ve-year forecast, Communications of the
ACM 38(10), 1995, pp. 33±49.
[26] Software Magazine (1996), The Business Case for Object
Technology, May, Editorial supplement.
[27] M. Tan, Establishing mutual understanding in systems
design: an empirical study, Journal of Management Information Systems 10(4), 1994, pp. 159±182.
[28] I. Vessey, S.A. Conger, Requirements speci®cation: learning
object, Process and Data Methodologies 37(5), 1994, pp.
102±113.
[29] I. Vessey, S.L. Jarvenpaa, N. Tractinsky, Evaluation of
vendor products: CASE tools as methodology companions,
Communications of the ACM 35(4), 1992, pp. 91±105.
[30] A.O. Villeneuve, J. Fedorowicz, Understanding expertise in
information systems, Or what's all the fuss about OO,
Decision Support Systems 21(2), 1997, pp. 111±131.
[31] S. Wang, Toward formalized object-oriented management
information systems analysis, Journal of Management
Information Systems 12(4), 1996, pp. 117±141.

Jane Fedorowicz is Associate Professor
of Accountancy at Bentley College
where she is teaching accounting and
information systems courses. She received her M.S. and Ph.D. degrees in
Systems Sciences from Carnegie-Mellon
University. She has previously taugh