You are on page 1of 23

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/223327265

Software engineering article types: An analysis of the literature. Journal of


Systems and Software, 81(10), 1694-1714

Article  in  Journal of Systems and Software · October 2008


DOI: 10.1016/j.jss.2007.11.723 · Source: dx.doi.org

CITATIONS READS

34 1,143

2 authors:

Michela Montesi Patricia Lago


Complutense University of Madrid Vrije Universiteit Amsterdam, Amsterdam, Netherlands
53 PUBLICATIONS   223 CITATIONS    272 PUBLICATIONS   4,172 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

HAPPYNESS: Emotion-aware sustainable service quality assurance View project

Wise/EU View project

All content following this page was uploaded by Michela Montesi on 05 April 2019.

The user has requested enhancement of the downloaded file.


This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy

Available online at www.sciencedirect.com

The Journal of Systems and Software 81 (2008) 1694–1714


www.elsevier.com/locate/jss

Software engineering article types: An analysis of the literature


Michela Montesi a, Patricia Lago b,*
a
Department of Archives and Information Science, University of Amsterdam, Turfdraagsterpad 9, 1012XT Amsterdam, The Netherlands
b
Department of Computer Science, VU University Amsterdam, De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands

Received 9 February 2007; received in revised form 19 November 2007; accepted 25 November 2007
Available online 3 December 2007

Abstract

The software engineering (SE) community has recently recognized that the field lacks well-established research paradigms and clear
guidance on how to write good research reports. With no comprehensive guide to the different article types in the field, article writing and
reviewing heavily depends on the expertise and the understanding of the individual SE actors.
In this work, we classify and describe the article types published in SE with an emphasis on what is required for publication in journals
and conference proceedings. Theoretically, we consider article types as genres, because we assume that each type of article has a specific
function and a particular communicative purpose within the community, which the members of the community can recognize. We draw
on written sources available, i.e. the instructions to authors/reviewers of major SE journals, the calls for papers of major SE conferences,
and previous research published on the topic.
Despite the fragmentation and limitations of the sources studied, we are able to propose a classification of different SE article types.
Such classification helps in guiding the reader through the SE literature, and in making the researcher reflect on directions for
improvements.
Ó 2007 Elsevier Inc. All rights reserved.

Keywords: Article types; Article genres; Requirements for publication; Software engineering; Survey

1. Introduction the final stage of the research process and summarize it


as a whole. Typically, the methodology used to carry out
In this study, we classify the types of articles or article the research and the type or genre of the corresponding
genres commonly published in software engineering (SE) article coincide, such as in the circumstance of case studies.
journals and conferences, such as surveys, case studies or Therefore, when classifying the article genres published in a
experience reports. The SE community has recently field, one takes also into account the methodological pat-
stressed the need for clearer criteria of evaluation and more terns of research in that field.
precise guidance on how to write different types of articles By using the expression ‘‘article genres”, we conceive of
(Shaw, 2003; Wieringa et al., 2005). Our wish is to make a research articles as standardized documents. Intended
contribution to this problem, and thus improve our under- readers have certain expectations about their form, con-
standing of what makes each type of SE research article a tent, and communicative purpose, in the same way as
good SE research article. film-goers have precise expectations for film genres such
Defining the types of articles published within a scien- as thrillers, westerns, or comedies. By considering research
tific community is not easy, since it implies dealing with articles as genres, we also want to stress that they play a
epistemological issues as well. Research reports represent determined role in the communication among the members
of the scientific community. Kwasnik and Crowston (2005)
*
Corresponding author. Tel.: +31 20 5987745; fax: +31 20 5987728.
point out that, for a document to be defined as a genre,
E-mail addresses: M.Montesi@uva.nl (M. Montesi), patricia@cs.vu.nl ‘‘social acceptance” is needed. In other words, research
(P. Lago). articles can be considered genres because the scientific

0164-1212/$ - see front matter Ó 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jss.2007.11.723
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1695

community recognizes them as such, and assign them spe- Communication with other members of the community
cific communicative purposes. ‘‘Article genres” and ‘‘types who are not primarily involved in research is a subject of
of articles” are synonyms in this study, and so are the debate in other disciplines as well, such as education or
words ‘‘article” and ‘‘paper”. With these premises, we are library and information science (Haddow and Klobas,
now able to illustrate what a classification of article genres 2004). It is also especially burning in SE, as some consider
can mean for SE. it crucial for the progress of the discipline (Glass et al.,
2002). Some SE researchers wonder whether it would be
possible to strengthen the impact of research in industry
1.1. Article genres reflect the patterns of research and settings, by increasing empirical research and experimenta-
communication tion (Tichy, 1998; Budgen and Kitchenham, 2005; Zelko-
witz et al., 2003). Research based on experimentation
Article genres published in a discipline reflect both the would allow practitioners to find project solutions on a val-
nature of research and the patterns of communication idated basis of knowledge (Basili, 1996). Others propose
between different actors in that discipline. In a previous that technology adoption decisions by practitioners may
study, Montesi and Mackenzie-Owen (2008) compared be supported or improved by Evidence-Based Software
the article genres published in SE, Biology, and Education Engineering (EBSE) (Dyba et al., 2005). In this perspective,
scientific research journals, with results showing two inter- it would be the empirical or experimental type of article the
related aspects. Firstly, it was found that SE journals do most likely to influence practice. But there is as yet no evi-
not attach to the methodological section the importance dence to support such claim. In fact, Punter (2003) found
that education and biology journals do. In most cases, that the practitioners using a specific SE portal were inter-
SE journals consider a three-section structure for major ested in various types of information, including but not
articles (Introduction – Main Body – Conclusion), whereas limited to experience reports and results of empirical stud-
biology and education journals consider rather a four-sec- ies about specific techniques, methods and tools. They also
tion structure that includes a method section as well. Some- valued current research directions, references to further
thing similar was observed by Posteguillo (1999). On an sources of information, or overviews of SE techniques,
epistemological level, the omission of any reference to the methods and tools, with no specific preference for any of
methodological section seems to connect with the claims these items.
of those who speak of a scant recourse to proper validation
and experimentation in the field (e.g. Fenton, 1993; Tichy 1.2. Article genres play a unique role in each discipline
et al., 1995; Glass, 1995; Zelkowitz and Wallace, 1998).
Secondly, the cited study (Montesi and Mackenzie-Owen, In our opinion and following Hjørland (1998) and
2008) revealed that in SE and education, the publication Hjørland (2002), article genres are shaped to conform to
of certain types of articles, such as the tutorial articles, the unique needs of each discipline. This can be argued
accounted for different patterns of communication than to detriment of those who often rely on comparisons with
in biology. In particular, it pointed to a more composite older disciplines such as medicine, to contend that more
audience, more likely to include practitioners besides experimental and/or empirical research should be under-
researchers, and to differentiate more evidently between taken (Pickard and Kitchenham, 1998; Kitchenham et al.,
specialist and non-specialist readers. 2002; Dyba et al., 2005). This is not a unanimous position,
If article genres carry in fact information on the nature however. Some, for instance, claim that SE needs ‘‘intelli-
of research in a field and on the preferential audience they gent debate” more than experimentation (Fletcher, 1995).
address, we could make the following assumption: by con- Fletcher argues that computer science has never been con-
sidering the type of article, it is possible to assess the impact cerned with natural phenomena and empirical predictions
of research in the community, including, with certain limi- such as physics or medicine (Fletcher, 1995). Marcos
tations, its degree of applicability. In this hypothetical per- (2005) stresses this point as well. In her opinion, formal
spective, certain article genres can be regarded as more and empirical sciences deal with the study of existing phe-
applicable because they are more explicitly addressed to nomena and objects, whereas SE deals with the study of
practitioners and/or non-specialists (such as tutorial arti- methods and techniques necessary to create new objects.
cles, for instance). By contrast, we could consider some Thus, SE research methods resemble software development
others less applicable because they are not based on a methods, and tend to be mostly of a qualitative and crea-
sound methodology (such as those lacking a method sec- tive nature. Finally, Miller (2004) criticizes the often advo-
tion, for example). This, of course, should first be proved. cated recourse to statistical significance testing in SE.
However, some authors already claim that some genres, Statistical significance testing, in his opinion, would be
such as case studies and the field studies, have more poten- unsuitable for SE experiments and studies for several rea-
tial to connect theory and practice (Ardimento et al., 2003; sons, and in particular for being intended for hard sciences.
Segal et al., 2005). Being conducted in industry settings, With regard to this issue, it is significant that, even when
they imply an interaction between researchers and practi- methodologies and relative article names are adopted from
tioners which should make them more relevant for practice. other disciplines, the meaning is different in SE. As far as
Author's personal copy

1696 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

we know, for example, a case study in SE is different from a (1) Published, relevant papers, such as, for example,
case study in medicine or law, as reported by Perry et al. Zelkowitz and Wallace (1998) or Sjøberg et al.
(2004), and it is probably different from a case study in (2005). For the sake of clarity, we intentionally avoid
medicine or psychology. In short, research patterns and taking into account some studies published on the
methodologies are different from a discipline to another topic. Glass et al. (2002), for instance, provide a clas-
because each discipline tackles different problems, aims at sification of research papers in SE, based on several
different objectives, and deals with different objects. With- criteria, among which topic, research approach, and
out denying that more experimentation and/or empiricism research method. Nonetheless, we have the impres-
may increase the degree of applicability of research, we sion, corroborated by Glass et al.’s results, that their
would like however to focus on traditional patterns of com- scheme of classification doesn’t suit SE research com-
munication between researchers and practitioners in SE, pletely well, because it has been designed for all the
and on how different article genres reflect them. three computing disciplines: SE, Information Systems
and Computer Science.
(2) The SE journals included in the Journal Citation
1.3. Article genres help deciding how a good article should be
Reports (2006) in May 2006. The JCR are elaborated
based on citation and article counts, and thus list the
Besides the lack of rigor in conducting empirical
most frequently cited and highest impact journals in a
research, and the absence of the well-established research
field. This means that the journals reported in the
paradigms that characterize other fields of science and
JCR are highly cited and used by the researchers of
engineering (Tichy et al., 1995), some claim that software
a certain field. At the time we consulted it, the JCR
engineers do not have clear guidance on how to write good
for ‘‘Computer Science (category) – Software Engi-
reports (Shaw, 2003). These three problems are indeed
neering (sub-category)” listed 76 journal titles. In par-
interrelated as the research report reflects the different
ticular, we studied the corresponding ‘‘Information
stages of the research work. But, in addition, a fourth prac-
to Authors”, which is intended to offer submitting
tical dimension may be involved. Reviewers and program
authors relevant guidance on the writing of proper
committee members are short of means for properly judg-
SE papers. At the time we consulted them, not all
ing the work of other members of the community. For
journals provided information relevant to the present
example, they can qualify research as good or bad for very
study. Those that are cited here with their abbrevi-
different reasons, because there aren’t agreed-on criteria for
ated names are reported in Appendix A. Some of
judging the quality of research reports (Wieringa et al.,
the titles reported in the JCR correspond, strictly
2005). The problem of assessing the quality of scientific evi-
speaking, to magazines more than to journals. How-
dence affects practitioners also, even though they seem to
ever, we use the term journal indifferently, as all are
be more worried about the relevance for practice than
included in this list. Similarly, some readers could
about the rigor of published research (Dyba et al., 2005).
argue that certain titles do not strictly belong to the
Defining articles as article genres may contribute to this
field of SE but to other close fields. As a matter of
problem as well because we assume that the requirements
fact, some consider the JCR, which are elaborated
of quality established for each type of article should com-
by the Institute for Scientific Information (ISI), inad-
ply with the communicative purpose of the same articles.
equate for the computing disciplines. According to
In Shaw’s words, ‘‘It seems reasonable to assume that if
Sidiropoulos and Manolopoulos (2005), for example,
authors were more consciously aware of typical paper
the scientific areas in which each field is divided by
types, they would find it easier to write papers that pre-
the ISI remain static over the years, and are unable
sented their results and supporting evidence clearly”
to follow the dramatic evolution of computer science.
(Shaw, 2002). In this perspective, a first step to take would
Others point out that the JCR should take in confer-
be to answer questions such as these: What is a good
ence publications, in order to reflect properly citation
research report? How should it be? How can one judge
and publication patterns in the computing disciplines
it? It is very difficult to give a satisfactory and definitive
(Moed and Visser, 2007). With their limitations, the
answer to such questions without the consensus of a great
JCR are all the same taken here as a point of refer-
part of the community, which requires time and great
ence, because they are elaborated on the basis of cita-
effort. A possible alternative is to summarize and organize
tion patterns within a certain field. Complementary
the written information currently available on the matter.
information is provided in Section 3.
And this is precisely the aim of this study.
When available, we also studied the corresponding
information for Peer Reviewers. Among the tools
2. Methodology available for reviewers, the Genre Submission Guide
of the journal IEEE Software (IEEE Software Guide)
The present description of the major types of research stands out (IEEE, 2006). The Guide describes in detail
papers published in SE is based exclusively on existing writ- ten different manuscript types, providing guidance in
ten sources, and in particular on the following: the following aspects: purpose, expected readers, sub-
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1697

mitted word length, potential review criteria, and, in Table 1


most cases, also the standard outline. When journal Overview of the types of articles
titles cited in this study are followed by the letter R III.A. Publishable papers: general
(Reviewers), then the information originates from III.B. Extended versions of conference papers
III.C. Empirical research reports
the corresponding information for reviewers. III.C.(1) Observational studies
(3) The instructions to authors of other relevant journals III.C.(2) Case studies
not included in the list of the Journal Citation Reports. III.C.(3) Field studies
These relevant journals have been pointed by a group III.C.(4) Experimental papers
of three interviewed SE researchers. Examples are III.C.(5) Meta-analyses
III.D. Experience papers
ACM Computing Surveys, and the Knowledge Engi- III.E. Theoretical papers
neering Review, for the description of special types III.F. Tutorial articles
of papers such as surveys and tutorial articles, and III.G. Surveys
Journal on Software and System Modeling (SoSyM) III.H. Short papers
and Software Process: Improvement and Practice. III.I. Papers oriented towards practice
III.I.(1) Educational papers
(4) The calls for papers of major SE conferences, as sug- III.I.(2) ‘‘How to”/ tool report papers
gested by the three interviewed SE researchers. Calls III.I.(3) Application papers
for papers of 2005, 2006 and a few of 2007 were stud- III.I. Opinion papers
ied, depending on the information available over the
Internet. All the conferences whose calls for papers  The readers and searchers can use Table 1 as a reading
were studied are reported in Appendix B. As for the guide to better determine how certain article types can
journals, not all conferences provide relevant infor- answer their expectations. This is especially true if they
mation for this study. come to Software Engineering as outsiders, e.g. if they
are working in multi- or interdisciplinary fields, or if
As commented above, article genres tend to coincide they are young researchers, approaching publications
with the methodology adopted in the research. This is for the first time. Moreover, our classification provides
why we will mention sometimes the methodological an overview of the SE publication types, and guidance
options available in SE. Different classifications of SE about how to read and write them, and how to put their
methods are available, the one of the Software Engineering reported type of research into context.
Body of Knowledge (SWEBOK, (Abran et al., 2004)),  Experienced researchers can get inspiration for further
Zelkowitz and Wallace’s classification of empirical valida- reflection and further improvements. To this aim, the
tion methods (Zelkowitz and Wallace, 1998), or Glass summarizing box attached to each section about the
et al.’s categories (Glass et al., 2002). Nevertheless, the article types is named ‘‘reflection box”.
focus in this paper is on article genres, and questions
related to methodology will be addressed only if relevant
3.1. General requirements for publishable papers
to this focus.

3. Article genres published in SE Box 1 Reflection box on: general requirements for
publishable papers
The article genres reported here are described in the
order shown in Table 1. A section introducing the general – Regular papers may be shaped as different genres,
requirements for publication precedes the detailed discus- such as, for example, experimental papers, case
sion of each article type. The space devoted to each type studies, or surveys.
depends on the information available and is consequently – A typical SE paper has a three-section structure:
biased in that sense. For example, long sections are devoted Introduction – Main Body – Conclusion.
to empirical and experimental research due to the abun- – Appropriate style must be comprehensible.
dance of recent contributions to the topic. It is also possible – Relevance for practice, novelty, and originality are
that some article genres are not mentioned at all because the most important requirements for publication.
not reported significantly in the written sources studied.
In order to help the reader follow the development of the Papers published in journal main sections, usually called
subject, a box with a few summarizing sentences precedes ‘‘regular papers” or sometimes also ‘‘original archival
each section. In the discussion and conclusion sections, papers”, may include different genres. IEEE Software, for
the reader will find a summary of the major points dis- example, mentions ‘‘industry experience reports”, ‘‘case
cussed for each article genre. We also provide a few refer- studies”, and ‘‘tool reports”. Computer speaks also of ‘‘opin-
ences to published papers, which we consider as good ion pieces” and ‘‘solutions to technical problems”. Other
examples of the genres commented on. journals include in this group of papers ‘‘tutorials”, ‘‘sur-
Our classification of article genres may be of help to var- veys”, ‘‘review papers”, ‘‘theoretical papers”, ‘‘standards
ious actors in SE, and in various ways: discussions” (IEEE Micro), and ‘‘experimental papers”.
Author's personal copy

1698 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

In the context of conferences, research reports are often text or bullet points to better visualize the message of the
called ‘‘technical papers” or ‘‘technical solution papers” text (IEEE Software Guide).
(ESEC/FSE 2007, RE 2006, for example), but still the range The first requirement for a paper to be accepted and
of article genres is fairly broad. At EASE 2006, for example, published is to conform to the individual journal policy
technical papers may deal with ‘‘any aspect of product and about organization and style. In addition, authors should
process evaluation and assessment, both qualitative and also meet other requirements. The J ACM and other
quantitative, including: experiments (laboratory and field), ACM journals, for example, put forward three general cri-
replication of empirical studies, case studies, surveys, obser- teria for acceptance:
vational studies, field studies, action research, evaluation
methodology, systematic reviews and meta-analysis”. On ‘‘(1) the paper must be among the best papers of the year
the other hand, at ASE 2006 the only distinguishing feature in its area,
of technical papers is simply their describing innovative (2) the paper must be of interest to the broad
research. Other conferences speak rather of full papers in community,
contrast with short papers (ISESE 2006 and CSMR 2006, (3) the presentation must be effective”.
for example), or of scientific papers in contrast with experi-
ence papers (MoDELS 2006). Full papers are still widely A paper to be published in the ACM T Softw Eng Meth
comprehensive in terms of topic and approach. should focus on ‘‘important research topics” and propose
Regarding the layout of standard SE papers, findings of ‘‘reproducible, extensible, scalable” results, which are also
a previous study (Montesi and Mackenzie-Owen, 2008), relevant to practice. Relevance to practice is particularly
revealed that the structure of SE regular papers may be stressed in connection with theoretical papers, but it is also
slightly simpler than in other fields. Following the instruc- a desirable requirement for other sorts of papers in many
tions to authors of the SE journals reported in the JCR, a journals. It is crucial that papers indicate clearly what is
paper in SE consists generally of only three clearly distin- ‘‘new and significant” of the work presented, and that they
guishable parts: an introduction, a result section, some- go into as much details as possible, so that they can teach
times called simply ‘‘main body” (IBM Syst J), and a something to the readers (Dr Dobbs J). Among the desirable
conclusion section. The method section is almost never attributes of a good paper, two are especially stressed by
mentioned. This three-section structure is rarely described most journals, i.e. originality and novelty: ‘‘A paper may
further, although the headings that mark it are reflected describe original work, discuss a new technique or applica-
and confirmed in the elements of content to be reported tion, or present a survey of recent work in a given field”
in abstracts. Sometimes, however, the instructions to (IBM Syst J). Other attributes are: importance, significance
authors furnish indications on the content of the different and relevance; correctness; technical soundness and preci-
sections. The introduction, for instance, should tell why sion; effective presentation, understandability and readabil-
the topic is important, and define a thesis or a problem. ity. Authors should be able to address a broad audience,
In Computer, ‘‘Articles must have sufficient introductory and never forget about their potential readers while writing
material; at least 15 percent of article length should be tuto- an article (Commun ACM). ICSE 2006 and other confer-
rial [. . .]. A brief literature survey does not satisfy this ence guidelines put some emphasis on the ability to address
requirement”. In the conclusions, on the other hand, a software engineering audience with different levels and
authors outline future goals and lessons learned. Supple- areas of expertise, and on the quality of the presentation
mentary material may complete the article, and can be pub- (ASWEC 2007). Technical papers at ESEC/FSE 2007 must
lished exclusively online. In IEEE T Software Eng, for clearly identify ‘‘their contribution to the state of the art
example, supplementary material takes in proofs, code, and/or their potential usefulness for industry”.
experimental data, short movies, appendices, animations The instructions of IEE Proc-Softw, differently from most
and audio files, and it is published on the Digital Library journals, explain what the major attributes and requirements
with a pointer in the printed version of the paper. of good papers should mean, shown in Table 2.
Much emphasis is often placed on style and comprehen- The importance of some of these requirements such as
sibility. ‘‘Articles should be understandable to a broad comprehensibility and applicability is confirmed in
range of developers who require information they can Anthony’s study (Anthony, 1999). Comparing SE intro-
apply in their daily work. Writing should be down-to- ductions to the standard model of a research article in
earth, practical, original, and comprehensible to all read- other disciplines, he found that characterizing features of
ers, regardless of their specialty” (IEEE Software). To SE introductions are, among others, a section with exam-
improve comprehensibility, contributors should make use ples and definitions, and an evaluation of the research in
of examples and visual material, such as figures, tables, dia- terms of application and novelty of the results.
grams, charts and photographs, define technical terms, and Journals may also address peer reviewers, and provide
avoid both jargon and academic language. For the same them with information in order to judge submitted articles.
reason, journals may recommend the use of appendixes Such information is usually very similar to that provided to
for detailed mathematical discussions, long intricate authors. The ACM T Softw Eng Meth -R asks reviewers to
proofs, or theorems (J Math Imaging Vis), as well as boxed check if acceptable papers contain: (a) motivation and clear
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1699

Table 2
Excerpt from the instructions of IEE Proc-Softw
Originality: is the work scientifically rigorous, accurate and novel? Does the work contain significant additional material to that already published?
Has its value been demonstrated?
Relevance: is the material appropriate to the scope of the journal to which it is submitted?
Motivation: does the problem considered have a sound motivation? Does the paper clearly demonstrate the scientific interest of the results?
Referencing: has reference been made to the most recent and most appropriate work? Is the present work set in the context of the previous work?
Clarity: is the English clear and well written? Poorly written English may obscure the scientific merit of your paper. Are the ideas expressed clearly
and concisely? Are the concepts understandable?
Length: unless previously agreed with the journals Managing Editor, all submissions must conform to the IEE Proceedings Length Policy
document

objectives, (b) demonstration of novelty and superiority A special type of articles are those based on one or
compared to related work, (c) evidence of effectiveness, more papers previously presented at a workshop, a sym-
reproducibility, scalability, and practical relevance. posium or a conference. For instance, van Vliet (2006)
Reviewers should also ascertain if papers meet quality cri- is an extended version of van Vliet (2005). In the study
teria such as conciseness and technical precision, the con- previously mentioned (Montesi and Mackenzie-Owen,
sistent use of terminology, and the definition of key 2008), comparing article genres in SE, education and
concepts. Focus and limitation of length must often be biology, it was found that extended versions of confer-
evaluated together, as the ACM T Database Syst R sug- ence papers played a role only in SE. A distinguishing
gests: ‘‘TODS would like to discourage excessively long feature of extended papers is the double round of
papers (longer than 50 doubled-spaced pages including fig- reviews they need to pass before journal publication
ures, references, etc.) and unnecessary digressions even in (i.e. conference and journal reviews). In a survey with
shorter papers. This should help authors to focus on the authors of extended versions and journal editors, Mon-
most important aspects of their submission [. . .]”. For the tesi and Mackenzie-Owen (accepted for publication)
rest, practically the same quality criteria as for authors found that the extension process may change depending
are brought up, such as clarity, readability and the ability on the type of conference paper to extend (empirical,
to target as broad an audience as possible. theoretical, . . .).
Studying the abstracts of papers published at ICSE Some journal instructions to authors regulate the pub-
2002, Shaw (2003) describes how accepted research papers lication of these articles. Among the other conditions, one
answered questions such as: What was your contribution? third of the content approximately must be new and ori-
What is your new result? How is your result different from ginal, and sometimes authors must submit both versions,
and better than prior work?, etc. According to Shaw, when a description of the differences between the submitted
it comes to evaluating submitted papers, programme com- manuscript and the preliminary version, and reference
mittees look for interesting, novel, and exciting results. In the previous version in the paper (SoSyM). Papers pub-
addition, as SE knowledge grows incrementally, like in lished as such must be ‘‘important” (Theor Pract Log
other areas of science and engineering, programme com- Prog), ‘‘outstanding” (ACM T Database Syst), and ‘‘of
mittees are very interested in the authors’ interpretation particularly high quality” (J Comput Sci Technol). New
of prior work. material takes in explorations of further research, theo-
rems, proofs and/or implementation details (ACM T
3.2. Extended versions of conference papers Graphic).
If journals address reviewers the general tone doesn’t
change, though they offer some more clues on what to look
Box 2 Reflection box on: extended versions of confer- for. In the ACM T Database Syst -R, for instance, extended
ence papers versions of conference papers ‘‘should thoroughly consoli-
date the material, should extend it to be broader, and
– Extended versions of conference papers are a typi- should more carefully cover related research”. In addition,
cal genre of SE. the novel approach should be described more deeply, and
– Extended versions have passed through a double alternatives should also be considered more comprehen-
round of reviews. sively. In IEEE T Vis Comput Gr -R, while no new results
– Many journals require that approximately one third are required, key ideas, examples, and further elaboration
of the content be new. are expected. This confirms Shaw’s observation that in con-
– Publication on a journal allows discussing the ferences time and space limitations do not allow authors to
research beyond the time and space limitations set explore all the possible results of their research (Shaw,
by conferences. 2003).
Author's personal copy

1700 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

3.3. Empirical research papers or expanded upon”, but doesn’t specify further. In the
IEEE Software Guide, by contrast, we find extensive, yet
Box 3 Reflection box on: empirical research papers not completely clear, information. Here a report of empir-
ical results resembles so much a case study that the two
types coincide in almost all respects. There are practically
– Empirical research includes different research strat-
no major differences, except perhaps a couple of possibly
egies such as case studies, experimental research, or
significant divergences. The object of the research is the
qualitative research.
‘‘project” in case studies, and the ‘‘participants” in empiri-
– Clear and comprehensive guidelines on how to
cal reports. In addition, case studies should report on tech-
write a proper empirical research paper are not
niques, practices and tools, whereas empirical reports could
yet available.
simply explain the practices. This blurred differentiation
– The experimental context is considered important
between the two types may be the sign that historically
for all types of empirical research.
what has been understood as empirical research in SE are
– Empirical research is increasing in SE.
case studies.
Comparing the 20 most cited SE articles of 1994 with
those of 1999, Wohlin (2005) found that a larger number
As defined by Basili et al. (1999), ‘‘an empirical study, in of empirical articles were included in the top list of 1999
a broad sense, is an act or operation for the purpose of dis- as compared to that of 1994. By contrast, the topics prac-
covering something unknown or of testing a hypothesis, tically had not changed. He mentions the ‘‘need for more
involving an investigator gathering data and performing scientific approach to software engineering, including mea-
analysis to determine what the data mean. This covers var- surement and empiricism”.
ious forms of research strategies, including all forms of Good examples of this article genre are published in e.g.
experiments, qualitative studies, surveys, and archival Empirical Softw Eng.
analyses”.
Kitchenham et al. (2002) provide some preliminary
guidelines for SE empirical research, which are based 3.3.1. Observational studies
mostly on published medical literature on the topic, being
the authors’ own experience as researchers and reviewers Box 4 Reflection box on: observational studies
the only direct connection with SE. These guidelines are
intended to cover all types of SE empirical research, such – Observational studies include all empirical research
as case studies, surveys, and formal experiments, whether strategies that impose little control over the devel-
conducted in the field, in a laboratory or in a classroom. opmental process. Item–Contextual factors, such
In Kitchenham et al.’s opinion, clear guidance would help as the type of industry in which the products are
both to assess the quality of published research and to used, or the experience and skills of software staff,
increase the likelihood that empirical study results can be affect the utility and generality of the results.
used in meta-analyses (see Section 3.3.5). Six basic topic
areas are treated in detail: (a) the experimental context of
the research, (b) the experimental design, (c) the conduct
of the experiment and the collection of data, (d) the analy- ‘‘An observational method collects relevant data as a
sis, (e) the presentation of results, and (f) their interpreta- project develops. There is relatively little control over
tion. The description of the experimental context, the development process other than through using the
comprising industrial circumstances, research hypotheses, new technology that is being studied” (Zelkowitz and
and information on related research, is especially empha- Wallace, 1998). In the classification of Zelkowitz and
sized, for either observational studies or experiments. This Wallace (1998), observational studies include: field stud-
information should make the purpose of the research ies, case studies, project monitoring, and assertion. How-
understandable to all possible readers, whether researchers ever, in contrast with Zelkowitz and Wallace, we regard
or practitioners, besides helping to explain conflicting observational studies as different from case studies and
research results. field studies, where observation may be (but is not neces-
Typically, journals do not comment in depth on this sarily) undertaken. Assertion consists in the ‘‘developers
type of research papers. The J Syst Software, for example, being both experimenters and subjects of study”. In the
encourages ‘‘theoretical and pragmatic developments with sample of 612 SE papers they analyzed, assertion was
emphasis on those that have been tested by government, the most common validation method in the period
industry, or university empirical research”, but does not 1985–1995. In a more recent paper, Zelkowitz et al.
specify the types of empirical research. In similar terms, (2003) classify assertion as an ‘‘informal method” of
Empirical Softw Eng accepts different types of empirical validation.
research (controlled experiments, field studies, qualitative, According to Kitchenham et al. (2002), empirical studies
etc.), giving preference to studies ‘‘that can be replicated based on observing or measuring some aspect of software
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1701

development in a particular company, should identify sev- the results and future developments are addressed. In other
eral factors that might affect the generality and utility of instructions to authors, case studies are difficult to distin-
the conclusions. Among the others, they mention: (1) the guish from what is often called first-hand experience arti-
type of industry in which the products are used; (2) the cles. However, they seem to entail a more objective
nature of the software development organization; (3) the perspective, whereas the term itself first-hand experience
skills and experience of software staff; (4) the type of papers points to a more subjective report. Perry et al.
software products used; and (5) the software processes (2004) explanation comes as useful to clarify this issue:
being used. For instance, Wu et al. (2003) present the Experience reports focus on a retrospective experience that
results of an observational-study of collaboration in soft- was illuminating of a problem, whereas case studies start
ware design undertaken at a large software development out with a research question, and answer that question
company. by systematically collecting data. Further, Perry et al.
Several observational studies can form the basis for (2004) defer the interested reader to exemplary research
meta-analyses (Pickard and Kitchenham, 1998). case studies.
The IEEE Software Guide and several journal instruc-
3.3.2. Case studies tions, such as Computer, insist on the importance of the
environment or the context of case studies (‘‘type of appli-
cation domain, hardware and software platform, types of
Box 5 Reflection box on: case studies people building and using the systems, etc.”). Only if the
environment is described in detail, can the readers tell
– Case studies’ most distinguishing feature is their whether the findings obtained are relevant/applicable to
descriptive character. their situation. When it comes to describing a typical case
– It is difficult to distinguish case studies from experi- study outline, the IEEE Software Guide makes the ‘‘study
ence reports or first-hand experience reports. details” section change depending on whether a controlled
– The environment of the research is especially experiment or a non-controlled experiment was performed.
important to evaluate case studies. This clashes with Zelkowitz and Wallaces classification of
– They allow for different data collection procedures, case studies as observational research reports, which as
including controlled experiments. such do not admit controlled experiments. However,
– Their major advantage is that they reflect real world Kitchenham et al. (1995) and Wohlin et al. (2000), too,
situations. assign case studies properties that one would regard as
– Some authors argue that, for this reason, they can more typical of quantitative research methods. For exam-
communicate better to practitioners. ple, the need for formulating previous hypotheses, or for
specifying certain design criteria such as external and inter-
nal validity. According to Wohlin et al. (2000) (p. 12), this
is possible because case studies afford diverse data collec-
A distinguishing feature of case studies is their descrip- tion procedures.
tive character. The Commun ACM, for example, comments From a practical perspective, Zelkowitz and Wallace
that: ‘‘Case studies should maintain an objective perspec- (1998) stress that a case study’s major advantage consists
tive on the systems they describe, and should be both ana- in the fact that it needs little additional costs to project
lytical and descriptive”. Similarly, ICSE 2006 guidelines costs. Researchers can monitor certain attributes, such
insist on the descriptive nature of case studies, which as reliability and costs, in the development of a project
‘‘describe the application of one or more software engineer- itself, which is going to happen regardless of the need to
ing practice(s) in an industrial or organizational setting”, collect experimental data. On the other hand, a weakness
and provide ‘‘a detailed description of how the practice of this method is that ‘‘each development is relatively
was applied and why”. Despite the distinctly descriptive unique”, and comparing a development profile with others
character of case studies, however, it is still difficult to dis- is not always possible. Dyba et al. (2005), all the same,
tinguish them clearly from other reports, especially from point out that an advantage of their uniqueness is that
experience reports. In the IEEE Software Guide, for exam- they reflect real world situations. As in medicine medical
ple, we find that case studies resemble ‘‘industry experience treatments must be tested on real patients to obtain com-
reports”. Here the few differences between the two genres pelling evidence, so can observational studies, case studies,
look formal rather than substantial (length, the nature of surveys and field experiments in SE ‘‘avoid the limited rel-
references, etc.), because no epistemological reason justifies evance of small-scale, artificial software engineering
them. Similarly, case studies and industry experience experiments”.
reports share all but one of the review criteria, which seems The empirical SE community often looks at case studies
insufficient to fully explain their respective functions. In with suspicion, because they lack objectivity and generaliz-
industry experience reports, reviewers should establish if ability, and because replication is practically impossible.
the experience is likely to benefit the journal readers, Segal (2005), however, argues that the lack of objectivity
whereas in case studies they should see if implications of is a threat to any type of research that, in the circumstance
Author's personal copy

1702 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

of case studies, can be minimized with recourse to different 3.3.4. Experimental papers
sources of information. Furthermore, according to the
same author, an advantage of case studies is that they Box 7 Reflection box on: experimental papers
may communicate more successfully to practitioners than
traditional experiments. In fact, there are indications, not – The terms ‘‘experiment” and ‘‘experimentation” are
really supported by evidence, that case studies might be often used improperly in SE.
more suitable than other genres for bridging the gap – Many researchers complain about the lack of stan-
between theory and practice, as mentioned in the introduc- dardization regarding collection measures for
tion. Despite the low degree of external validity and gener- experimental research.
alization, case studies would be sought by the industry – Some also argue that current SE experimental
community particularly because carried out in real settings research is insufficiently representative of industrial
and tailored to individual needs. Zelkowitz et al. (2003) settings.
speak of a ‘‘self-interest of industry” which is concerned – It is particularly difficult to carry out true experi-
with choosing the right methods for its own environment mentation in SE real settings.
more than with benefiting competitors through generaliza-
tion. Kitchenham et al. (1995), comparing case studies with
formal experiments, note that case studies are important
for industrial evaluation of software engineering methods Both Zelkowitz and Wallace (1998) and Sjøberg et al.
and tools because they can avoid the scale-up problems (2005) comment that the terms ‘‘experimentation” and
of formal experiments. ‘‘experiment” are often used incorrectly and/or inconsis-
The industry’s interest for this genre may explain why tently in SE. Zelkowitz and Wallace (1998) observe that
case studies are sometimes assigned a persuasive function. what is called experimentation in SE hardly ever involves
The instructions to authors of the J Visual Comp Animat, a collection of data to confirm previously defined models
for example, which is specialized in computer animation, or theories. Sjøberg et al. (2005), on the other hand, note
encourages ‘‘submission of case studies to demonstrate that in SE the term ‘‘experiment” is often used as a syno-
to film makers what techniques have previously been nym for ‘‘empirical study”. Three definitions of experiment
drawn on and their results”. As another example of indus- are reported in Table 3.
trial interest, case studies presented at CAiSE 2006 pro- Kitchenham et al. (2002), who define general guidelines
vide the rationale for key decisions made at different for empirical research, comment upon experimentation as
stages, including purchase and usage of information well. They stress one more time the importance of contex-
systems. tual factors in designing formal experiments, which will
have to consider a rich and complex industrial setting. Fur-
ther recommendations are provided for the experimental
3.3.3. Field studies
design. In particular, a proper experimental design should
describe the population under investigation, the rationale
Box 6 Reflection box on: field studies and technique for sampling from that population, the pro-
cess for allocating and administering the treatments (or
– Neither the function of field studies nor their ‘‘intervention”), and the methods used to reduce bias and
difference from case studies is clear in most determine sample size. Nonetheless, because data collection
sources. measures are not standardized in SE, Kitchenham et al. con-
clude that it is difficult to know whether different studies are
really measuring the same thing and whether they are doing
Field studies resemble case studies with a less intrusive that in the same way: ‘‘Still, much work remains in making
component (Zelkowitz and Wallace, 1998). In a case study, sure we mean the same things when we discuss and measure
the subjects themselves often collect the data, whereas in a variables in empirical studies”. The lack of standardization
field study it is an outside group that monitors each subject of measurements in SE research is also stressed in Kitchen-
group. This approach is for instance well described by ham et al., 2001. Wohlin et al. (2000) argue that the informa-
Zhang et al. (2003), which report on a field study carried tion furnished in experimental reports must be detailed
out in an insurance company. However, some other times enough to allow other researchers to replicate the experi-
the expression ‘‘in the field” seems to point to an involve- ment. They also describe in detail the different phases of
ment of the reporting subjects, such as in ‘‘we seek to the experimental process and provide a standard outline
include articles from practitioners working in the field for reports of experiments (pp. 116–117). A good example
(including the user community)” (J Softw Maint Evol). of experimental paper can be found in (Babar et al., 2006).
Even if field studies are often mentioned in the instructions Sjøberg et al. (2005) focus specifically on controlled
to authors of major journals or in conferences guidelines, it experiments in SE. Their study relies on an analysis of
is not clear what their function is. The difference with case 103 conference and journal articles published in SE major
studies is not clear-cut either. journals and conferences between 1993 and 2002, and
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1703

Table 3
Definitions of experiment in SE
Zelkowitz and Wallace (1998): Basili et al. (1999): Sjøberg et al. (2005):
‘‘A controlled method provides for multiple ‘‘An experiment is a form of empirical study where ‘‘Controlled experiment in software engineering
instances of an observation in order to the researcher has control over some of the (operational definition): A randomized experiment
provide for statistical validity of the conditions in which the study takes place and or a quasi-experiment in which individuals or teams
results” control over the independent variables being (the experimental units) conduct one or more
studied; an operation carried out under controlled software engineering tasks for the sake of
conditions in order to test a hypothesis against comparing different populations, processes,
observation” methods, techniques, language, or tools (the
treatments)”

reporting on controlled experiments. Their aim was to increased considerably in the 1990s. Jedlitschka and Pfahl
measure to what extent the experiments were representative (2005) summarize and complement the guidelines on the
of the industrial settings in terms of subjects, tasks, appli- conduction of experimental research in SE published since
cations, and environment. They concluded by questioning 1999. They insist on the fact that more research involving
the fact that experimental research, the way it is now, different stakeholders must be carried out to achieve stan-
may be representative of the industrial settings, and by crit- dardization. Tichy (2000) lists a series of dont”s for review-
icizing several other aspects. The subjects taking part in the ers of experimental research reports, inviting them to major
experiments, for example, were in 87% of the cases stu- flexibility in judging the quality of such studies. He sug-
dents, and in only 9% professionals; the applications, in gests, among the other things, neither expecting perfection
75% of the cases, were constructed for the purpose of the nor decisive answers; not rejecting ‘‘obvious” results, and
experiment or constituted student projects, etc. They also not being casual about asking authors to redo their
found that the reporting was often vague and unsystematic, experiment.
lacking generally consistent terminology. Not to mention Authors are invited to submit experimental papers along
that both internal validity and external validity were with other genres in the instructions to authors of many
reported in a vague way, demanding for specific guidelines journals, though they are not provided with any specific
for the community. Finally, they recommended that exper- information. The ACM T Softw Eng Meth only reminds
imental papers report rigorously on the following aspects: its contributors that results must be interpreted in term
‘‘the type and number of subjects, including the mortality of practice.
rate; context variables such as general software engineering
experience and experience specific to the task of the exper- 3.3.5. Meta-analyses
iments; how the subjects were recruited; the application
areas and type of tasks; the duration of the tasks; and inter- Box 8 Reflection box on: meta-analyses
nal and external validity of the experiments, including
being specific about the sample and target population of – Meta-analyses allow to combine results of various
the experiments”. Sjøberg et al. (2005) also identified a rela- empirical studies.
tion between experimentation and SE topics: most experi- – The lack of standardization in reports of empirical
mental papers deal with ‘‘Software lifecycle/engineering” research makes it difficult to apply meta-analyses in
and ‘‘Methods/Techniques”, using Glass et al. classifica- SE.
tion scheme (Glass et al., 2002), and ‘‘Code inspections
and walkthroughs” and ‘‘Object-oriented design methods”,
using the IEEE Keyword Taxonomy (IEEE, 2002). Meta-analysis is a technique that allows combining the
An explication of the poor representativeness of SE results of various empirical studies. Its aim is ‘‘to provide
experimental research may be the fact that true experimenta- a quantitative and objective procedure for combining the
tion is often hard to carry out in SE. Laitenberger and Rom- information from different studies” (Pickard and Kitchen-
bach (2003) suggest that quasi-experimental designs would ham, 1998). Such procedures should increase the confi-
fit better industrial settings than true experimental designs. dence of results that one obtains from only one study,
A major problem with true experimental designs in indus- and, in general, make the process of reviewing the literature
trial settings is that a random allocation of subjects is almost more ‘‘scientific” (Djokic et al., 2001). For instance, Succi
impossible for schedule, availability, etc. Besides, the treat- et al. (2005) apply meta-analysis techniques on 200 projects
ment cannot be withheld. Quasi-experimental designs, to study the distribution of selected object-oriented metrics.
which do not require the allocation of subjects to be random, According to Pickard and Kitchenham (1998), meta-
may be a valid alternative, provided researchers indicate and analyses follow three stages. First the analyst must decide
discuss explicitly all the possible threat to internal validity. which studies to include. Individual studies should be of
Zendler (2001) analyses briefly the history of SE experi- the same nature (all case studies, or all cohort studies, etc.)
mental research since the 1970s, highlighting that it has all exploring the same variables, for instance, or having
Author's personal copy

1704 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

comparable measures. Then, in a second phase, the analyst Design] in practice” (Comput Aided Design), describe ‘‘the
must estimate an effect size for each individual study in order experience of applying functional programming to real
to make each of them comparable with the others. What problems” (J Funct Program), or ‘‘how particular tech-
makes this stage difficult is that often raw data of individual niques, practices, or tools were used in a practical setting”
studies is missing, and the analyst cannot estimate the effect (IEEE Software Guide). ICSE 2006 guidelines stress the
size correctly. Finally, the results from individual studies are need for the report to provide ‘‘a critical review of experi-
combined. This technique is quite popular in medicine ences during one or more phases of a software development
where researchers tend to compare replicated experiments; project”, and to describe all factors influencing the experi-
in SE, on the other hand, it would suit better observation ence in a way that readers can understand its impact in
studies (Pickard and Kitchenham, 1998). terms of success or failure. A typical example of ICSE
Miller (1999) is pessimistic about the applicability of experience report is provided in (Best et al., 2007). In some
meta-analysis to SE research. According to him, in general, sources, practitioners are indicated as the most likely
the major obstacle to the introduction of meta-analyses in authors of experience reports. According to Software Pract
SE would be the excessive variation within the several com- Exper, for example, the aim of experience reports is ‘‘to
ponents of the experiments in this field. Brooks (1997), on his encourage practitioners to share their experiences with
part, points out that there are not enough replications in the design, implementation and evaluation of techniques and
SE literature, in order to perform meta-analysis routinely. tools for software and software systems”. It is in conference
guidelines, however, where we find a major emphasis on
practitioners as the most likely authors of experience
3.4. Experience report papers
reports. At OOPSLA 2006, for example, ‘‘reports from
the field describing real-world experience” are called pre-
Box 9 Reflection box on: experience report papers cisely ‘‘practitioner reports”. At METRICS 2005, one of
– Experience reports are called by different names the authors should come from industry or government.
(‘‘lessons learned reports”, ‘‘industry experience At WICSA 2007 ‘‘practitioners are asked to describe best
reports”, etc.), and it is not clear if different names and worst experiences from the trenches, and researchers
imply variations. are asked to describe areas where research progress is slo-
– Experience reports appear to be connected with wed because of lack of access to real world problems”. This
application and practice. preferential authorship could account for the connection of
– In some sources, practitioners are the expected experience, practice and application mentioned at the
authors of experience reports. beginning. While expected authors may often be practitio-
– Soundness of the report and objectivity are impor- ners, expected readers, according to IEEE Software, are
tant publication requirements. software managers, maintainers, developers, applied soft-
ware researchers, and executives of companies interested
in investing in software technology.
The papers classified under this heading have in com- The Empirical Softw Eng journal offers detailed indica-
mon the notion of ‘‘experience”, remarked by some jour- tions on what and how to write in a proper ‘‘industrial
nals as first-person or first-hand experience. Other names experience report”: ‘‘Describe the context from which the
can also be used, such as ‘‘lesson learned reports”, ‘‘indus- results are reported [. . .]. Describe the software technolo-
try experience reports”, ‘‘application papers”, etc., and it gies [. . .]. Provide effectiveness and efficiency data leading
remains unclear whether they are all to be considered as to the assessment of the technologies. Interpret the results,
the same type of experience reports. In the instructions to explain their consequences, and draw conclusions. Discuss
authors, application, practice, and experience are interre- the limitations of the results and conclusions”. Whatever
lated, and especially the last two. With the name ‘‘applica- the topic they may deal with, practice and experience
tion papers”, however, different articles may be papers do not need to make a novel contribution (J Funct
understood, as explained further on. The IEEE Comput Program). The paper should give a well-informed, up to
Graph instructions on submissions to the ‘‘Applications date and accurate account of the application area by an
department” are a good example of how experience, appli- expert in that field. In fact, specific evaluation criteria
cation and practice are in fact really close to each other: may apply. As opposed to regular research papers, they
‘‘Its existence is inspired by consistent feedback from the need neither to discuss novel ideas nor to discuss related
readership reflecting a strong desire for more practical con- published work in the literature (Empirical Softw Eng).
tent, more real world applications, more practice and usage, The J Funct Program also reminds that these papers are
more implementation details, more experience-related difficult to be published in conferences due to their length.
information, more practical application of research results, This fact is confirmed by Kajiya (1993), who explains that
and bridging the gap between theory and application”. ‘‘long papers” were introduced at SIGGRAPH to allow a
Despite their apparently less academic content, experi- ‘‘systems and application category papers”. Such papers
ence reports are widely sought for by some journals. They needed more space in order to ‘‘describe the experience that
may present ‘‘the results of using a CDA [Computer Aided the builders have had with the system”. A feature of
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1705

‘‘industry experience reports” in the IEEE Software Guide lished only if they clearly show how the results contribute
is a section with up to 10 brief bullet points summarizing to Software Engineering practice (ACM T Softw Eng
the ‘‘lessons learned”. RE 2006 guidelines also insist that Meth), or when an understanding of the theory can lead
the focus of these contributions is on the ‘‘what” and the to better practical systems” (Software Pract Exper). In Sim-
‘‘lessons learned”, whereas explanations of the ‘‘why” are ulat Pract Theory, for example, theoretical articles should
better analyzed in other types of paper. supply information on ‘‘potential benefits; any experience
Some journals explain the criteria for acceptance of gained; any innovation achieved through a specific
experience reports. To be published in the ACM T Softw approach”. Then again, in Simul Model Pract Th, theoret-
Eng Meth, they ‘‘must provide thoughtful insights about ical articles must say: ‘‘Why the theory is relevant and how
the development world or the application of a technology, it can be applied, what is the novelty of the approach and
that result in the identification of new important challenges what are the benefits and objectives of a new theory,
for software engineering research”. In IEEE Comput Graph method or algorithm; what experience has been obtained
those articles are not refereed. They are accepted if the con- in applying the approach”. It is also necessary to include
tribution is unique, and it gives some result such as, for applications and examples that demonstrate that any theo-
example, ‘‘how graphics was used to solve a previously retical assumption is applicable to real world problems. In
unsolved problem” or ‘‘why this solution worked, what Sci Comput Program, theoretical papers can deal with
others were tried, and why this was better”. ICSE 2006 philosophical and sociological issues on all aspects of com-
evaluation criteria for experience reports are the same as puter software production and usage, including ethics.
for case studies. Soundness of the report, in this case, At CAiSE 2006, theoretical and conceptual papers are
involves describing the context of the experience (problem classified as proper ‘‘research papers”, in contrast with
domain, size of project, size of system, etc.), showing that other types, such as case studies or experimental reports.
the experience is reported on the basis of observation and They should describe ‘‘the problem tackled, the state of
analysis, and not of opinion, and measuring as accurately the art with respect to the problem, the solution suggested
as possible the important phenomena. and the potential – or even better, the evaluated – benefits
In the IEEE Software Guide, it remains undetermined of the contribution”. Typical examples of SE theoretical
whether Industry experience reports imply a historical or papers are published in the IEEE T Software Eng.
observational approach, i.e. whether the collection of data Relevance for practice as a condition for publication is
takes place after the techniques, practices, and tools ana- also stressed when journals address reviewers. Stricter crite-
lyzed were implemented, or in the course of their imple- ria than for other papers may be applied to evaluate theo-
mentation. Industry experience reports are considered retical papers, especially for what regards originality and
comparable to case studies, but the fact that they summa- impact (ACM T Database Syst -R).
rize lessons learned and experience, however, points to a
historical nature. In Zelkowitz and Wallace (1998) experi- 3.6. Tutorial articles
ence reports are not mentioned as such. These authors
speak of ‘‘legacy data” and”lesson learned reports”. They
are both classified as historical methods, since both take Box 11 Reflection box on: tutorial articles
place when a project is completed, not depending on any – Tutorials are mainly educational, and introduce the
deadlines, but differ in the type of data they analyze. Leg- work done in a field in a highly comprehensible
acy data reports focus on quantitative data, including the manner.
artifacts involved in the product, whereas lesson learned – Two types of tutorial are considered in the IEEE
reports reveal qualitative aspects of a project. Software Guide: one focusing on applications and
the other on research activities.
3.5. Theoretical papers – Tutorials address audiences with different back-
ground and levels of experience.
Box 10 Reflection box on: theoretical papers
The educative inclination of many SE publications
– Theoretical papers are often of a methodological or materializes in a particular type of article, the tutorial.
computational nature. Despite being published in most journals, tutorials are
– Relevance for practice is the most important rarely described. According to the ACM Comput Surv,
requirement for publication. which publishes exclusively tutorials and surveys, a tutorial
is ‘‘a paper that organizes and introduces work in the
Following the instructions to authors of the journals field”. A characterizing feature of it is the fact that basic
analyzed, SE theoretical papers tend to be methodological, concepts of the field are emphasized and represented by
but they also include computational papers, i.e. papers pre- concrete examples. On its part, Presence-Teleop Virt sug-
senting computational methods and algorithms. In most gests that they ‘‘should develop to an appropriate level of
cases, theoretical or methodological papers can be pub- detail a significant topic that is of interest to, but not well
Author's personal copy

1706 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

known to, a significant portion of the Presence readership”. In most instructions to authors of SE journals, the name
In Software Pract Exper, tutorials accomplish a second ‘‘survey” corresponds to what other disciplines, such as
function, that of documenting ‘‘well-timed techniques not education and biology, call ‘‘review” or ‘‘literature review”.
previously documented in computing literature”. Finally, However, the terms are not exactly synonyms, because sur-
IEEE T Reliab highly regards tutorials that centre around veys present some peculiarities. SE journals do not place
‘‘the chemistry & physics of failure”. any special emphasis on the need for an exhaustive litera-
The IEEE Software Guide counts two types of tutorials, ture review, which is anyway required, but stress rather
the ‘‘research tutorial” and the ‘‘practice tutorial”. Each of other aspects. SE surveys include, for example, ‘‘material
them develops one of the major aspects dealt with in a ‘‘sta- that is difficult or impossible to find elsewhere” (Pres-
tus report paper” (another genre described in the Guide). ence-Teleop Virt); they may focus ‘‘on emerging technolo-
Status reports discuss a given technology or a real-world gies or new aspects of existing technologies” (IEEE
problem, providing both an explanation of its applications, Internet Comput), or summarize and organize recent
and a description of research trends and major barriers to research results in a novel way (ACM T Math Software).
overcome. On the other hand, practice tutorials ‘‘Given a Effective presentation is highly expected (ACM T Math
specific technology (or real-world problem), explain how Software). According to the ACM T Database Syst, two
the technology is applied (or how the real-world problem types of surveys are possible: (a) those that summarize pre-
is solved) today in industry”. And research tutorials ‘‘Given vious literature on a theoretical or systems research topic;
a specific technology (or real-world problem), explain the (b) and those that explain approaches implemented in com-
most promising research activities underway or proposed mercial systems. The former type presents a new way of
that expand our knowledge of the technology (or of how understanding how the papers in the field fit together,
to solve the problem)”. Both types of tutorials and status whereas the latter reports the best industrial art, and is
reports are allowed the same length (4300–5400 words) acceptable even if it presents no new contribution. In a cer-
and the same number of references and/or pointers to pri- tain sense, for their educational and informative nature, SE
mary and secondary sources. Thus one expects a deeper surveys are somehow similar to tutorials. The ACM Com-
analysis and a major degree of specificity in tutorials. The puting Surveys explains that: ‘‘Basically a Computing Sur-
potential readership is practically the same in the three cases: veys article, whether a survey or a tutorial, answers the
software developers, software managers, high-tech (corpo- questions, ‘‘What is currently known about this area, and
rate, for tutorials) executives, and applied software research- what does it mean to researchers and practitioners?” It
ers. In all cases it is implied that readers are not experts in the should supply the basic knowledge to enable new research-
technology/real-world problem. Research tutorial and prac- ers to enter the area, current researchers to continue devel-
tice tutorial have a very similar outline, but the content of opments, and practitioners to apply the results”. Such
the tutorial section differs correspondingly to their purpose. definition retakes Goldberg (1982) and Wasserman (1984)
In research tutorials, it deals with the major research activi- distinction between the two genres. According to these
ties that are expected to expand or refine the technology or authors, tutorials assume their audience to be inexpert
the problem, as well as their impact in practice. In practice and emphasize basic concepts of the field providing con-
tutorials, the tutorial part provides taxonomy of applica- crete examples. By contrast, surveys assume their audience
tions, indicating in each case how the technology is applied to have a general knowledge of the field, and emphasize an
or how the problem is addressed. overall view of the literature.
Evaluation criteria of IEEE Comput Graph for tutorials Sometimes the variant more traditional designation of
take in the following aspects, among the others: they must ‘‘review” is preferred in the journal instructions to authors,
fall within the scope of the journal, be interesting, well writ- as can other alternative names, such as ‘‘analyses” (Knowl
ten, and well organized, be understandable to most readers, Eng R), or ‘‘state-of-the-art-report” (Comput Graph Form -
present background and historical material, current prob- R). Then again, the term ‘‘survey” in empirical areas of SE
lems and issues, and assess applicability and limitations. can refer to methods of research, in which participants are
requested to answer a questionnaire or give interviews
(Wohlin et al., 2000, pp. 10–12). This acceptation of the term
3.7. Surveys survey is closer to the understanding that the social sciences
have. In this way it is also understood by the instructions to
Box 12 Reflection box on: surveys authors of the Commun ACM. Probably due to the ambigu-
ity of the term survey (meaning both literature review and
– Surveys complement tutorials, but address a more the questionnaire type of research), Zelkowitz and Wallace
experienced audience. (1998) refer to this research strategy as ‘‘literature search”.
– Surveys do not need to be novel, but should make a Not many journals provide evaluation criteria for
contribution to the understanding of an area. reviewers who may have to evaluate surveys. The ACM
– They help researchers as well as practitioners to T Database Syst -R qualifies good surveys as focused, deep,
keep up to date. quite narrow, and educational for the journal readers. They
should make a contribution to the understanding of an
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1707

important area or sub-area. When surveys summarize the Different names are used to refer to this type of articles in
best industrial art, they can be accepted even if they give SE journal instructions to authors: Pearls, Short articles,
‘‘no new contribution over what has been used in industry Short communications, Short contributions, Short papers,
for years”, as long as ‘‘the papers content is not to be found Short submissions, etc. Short papers, like full-length papers,
in the published literature”. Surveys should also put recent allow for different methodological approaches and virtually
progress into a broader perspective and accurately assess any kind of topic included in the scope of the journals,
the limits of existing theories (IEEE T Vis Comput Gr -R). although they tend to centre around some sort of practical
Dyba et al. (2005) comment that surveys – that they call and/or technical application. In particular, short papers
‘‘systematic reviews” help keep up to date on the available may be ‘‘practitioner-oriented” (IEEE Software), describe
evidence regarding specific phenomena. In so doing, they ‘‘clever programming ideas” (J Funct Program) and ‘‘novel
foster evidence-based SE research, and contribute to clos- and/or topical applications” (J Res Pract Inf Tech), or pres-
ing the gap between practice and research. ent ‘‘problems on selected topics of computer science” (Algo-
Budgen and Brereton (2006), on their part, discuss the rithmica). Sometimes they may also report ‘‘on rapidly
relevance of systematic reviews for the academic commu- developing new topics” (Software Pract Exper). In the con-
nity. Systematic literature reviews supply an objective sum- text of conferences, short papers have an updating function.
mary of research evidence once a clear set of procedures is At ICSE 2006 they are presented in an ‘‘informal and highly
established to review background material. Typical exam- interactive” section, which allows authors to get critical feed-
ples of SE surveys are those published on ACM Comput back that they can use for their final paper. Attendees, on
Surv. Dyba et al. (2005) further argue that SE research is their part, keep abreast of emerging ideas and technologies
bound to remain fragmented without ‘‘a research culture in the area. ICSE 2006 short papers, while dealing with the
that strongly advocates systematic reviews and replica- same topics as those of the research track papers, cover work
tion”. To achieve this, the community needs to develop with ‘‘solid technical basis” but not yet fully validated. At
and adopt specific guidelines for systematic reviews. The QoSA 2006, short papers ‘‘can describe experience, ongoing
best search methodologies should also be indicated in order work, and new ideas”.
to reduce bias and guarantee as complete a coverage as The maximum length admitted in general in SE journal
possible of the relevant information. Some guidelines are instructions to authors spans from 2000 to 5000 words. The
available from the Inform Software Tech web site and pre- revision process is, as a general rule, as rigorous as for nor-
sented by Kitchenham (2004), who synthesizes the infor- mal contributions. The publication process may be shorter
mation reported in medical guidelines to perform than for standard articles, but, in many cases, they are pro-
systematic reviews. More recently, Budgen et al. (2006) cessed as the other categories of papers. There are also
have investigated the possibilities of aggregating evidence cases in which such articles may be reviewed only by the
in SE by looking at how this is done in other domains (like editor in charge of the corresponding section (J Funct Pro-
Nursing & Midwifery and Criminology). In doing so, they gram). Reviewers must make sure that the paper improves
encountered domain-specific obstacles to perform system- the state of the art ‘‘in some significant way” (ACM T
atic reviews. Potential authors of systematic reviews need Database Syst R).
to deal with problems such as not standardized key-words According to Tichy (2000), short papers would suit well
and abstracts of poor quality (Brereton et al., 2007). certain experimental studies, such as those reporting on
Zelkowitz and Wallace (1998) remind that a major prob- negative results or on repetitions with identical or very sim-
lem with this type of research is the fact that published lit- ilar results.
erature tends to favor positive results to the detriment of
negative results with the resulting bias. Even with these lim- 3.9. Papers oriented towards practice
itations, though, they consider surveys a validation method
that allows confirming existing hypotheses or enhancing 3.9.1. Educational papers
data collected in one project.
Box 14 Reflection box on: educational papers
3.8. Short papers – Educational papers appear more commonly in con-
ferences than in journals.
Box 13 Reflection box on: short papers

– Short papers deal usually with practical or technical Educational papers are mentioned both in conference
topics. guidelines and in journal instructions, but they are
– They usually undergo the same revision process as described mainly in the former case. At ICSE 2006, sound
full-length papers. research design and empirical analysis are strongly encour-
– Negative results or repetitions of previous results fit aged, whilst educational experience reports are accepted
well short papers. provided they justify and explain the conclusions using
careful analysis and providing a thoughtful retrospective”.
Author's personal copy

1708 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

Evaluation criteria are: ‘‘originality, importance of contri- In reviewing ‘‘how to” reports, reviewers are asked to
bution, soundness, empirical analysis, quality of presenta- check if examples are provided and if the conditions of
tion, and comparison to related works (as appropriate)”. applicability/inapplicability are clear. Similarly, tool
Typical good examples of educational journal papers are reports should explain how the tool can be used in common
published on the IEEE Transactions on Education, whereas practice and for which applications it is most effective.
conference papers are to be found in the Proceedings of the ‘‘How to” or tool report papers, like for instance
IEEE Conference on Software Engineering Education and (Arranga, 2000), are appearing less often in scientific pub-
Training. lications. Instead, they are being replaced by online publi-
cations and web sites or blogs especially dedicated to
3.9.2. ‘‘How to”/ tool report papers technologies and ‘‘how to” issues.

3.9.3. Application papers


Box 15 Reflection box on: ‘‘How to”/ tool report
papers
Box 16 Reflection box on: application papers
– ‘‘How to” and tool articles explain how to perform
specific development tasks or on how to use specific – Application papers are similar and related to expe-
tools. rience reports, tool reports and ‘‘how to” papers.
– The conditions of applicability/inapplicability of – They may focus on applications in different
the process or the tool at hand must be detailed. domains.
– They usually stress problems and obstacles to
overcome.
‘‘How To” articles in the IEEE Software Guide ‘‘provide
readers with step-by-step instructions on how to perform a As commented previously, different types of articles may
specific software development task”. They should discuss be expected under the name of ‘‘application papers”. In
the distinction between tool-related and process-related some cases, an application paper may explain how to
aspects, and in this sense they seem to be intended as com- implement a certain tool, method or technique. In some
plementary to ‘‘Tool reports”. In the IEEE Software Guide others, it may focus on the application of software in spe-
tool report articles ‘‘report the development of a new soft- cific environments and on solutions of specific problems.
ware tool and its application”. However, ‘‘how to” and Finally, it can be understood as an experience paper. We
tool report articles differ in the audience they address. don’t gain a deeper insight with the alternative name of
Whilst ‘‘how to” articles are intended for those ‘‘software ‘‘applied research paper” in use in the IEEE Software
developers and/or managers who have never performed Guide, whose purpose is to ‘‘Report to readers on the avail-
the specific task, or who have performed it previously with ability of some new research technology, technique, tool, or
unacceptable results”, tool reports address ‘‘Software language”. However, we can tentatively establish a differ-
developers, software managers, high-tech executives, entiation. Application or applied research papers are really
applied software researchers”, and assume that ‘‘readers similar to tool and ‘‘how to” reports, but more concen-
are not experts in the technology or class of tools being trated on applications in specific domains. In addition, they
addressed”. The word limit varies accordingly. ‘‘How to” are also related to experience papers, with the difference
reports (2300–4300 words) have a word limit slightly under that experience reports focus on lessons learned, while
the threshold fixed for major papers, such as tool reports application reports on obstacles and problems to over-
(4300–5400), among others. Structurally speaking, the come. As a matter of fact, in an applied research article,
two reports are also different, as reported in Table 4. according to the IEEE Software Guide, whilst the feasibility

Table 4
The structure of a ‘‘tool report” article and a ‘‘how to” paper according to the IEEE Software Guide
Introduction, explaining the problem the tool addresses; Introduction, explaining the relationship of the task to software
development, who performs the task, when and how often;
The tool section, indicating who the intended users are, how and when it How to section, providing step-by-step instructions and relative examples;
would be used, and how it solves the problem;
State-of-the-art section, discussing alternative solutions to the problem, Tailoring the procedure section, describing the conditions under which the
and comparing the tool with similar ones; above procedure is applicable, and how to tailor the procedure in
mismatched conditions
How was the tool built, a ‘‘short” section (maximum 1500 words)
explaining the development process, the architecture, and the
algorithms; and
Summary and conclusion. Summary and conclusion
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1709

of the research has already been proved, it is expected that whereas ‘‘position papers” at EWSA 2006 ‘‘present concise
more experimentation will confirm the effectiveness of the arguments about a topic of software architecture research
results. Expected readers are: ‘‘Software developers and/ or practice”.
or managers who wish to use new technology experimen- Some journals distinguish further opinion articles,
tally. Applied software researchers who want to learn what sometimes depending on whether they give a technical
others are doing”. Research results should be likely to be in opinion or not. The Commun ACM, for example, separates
widespread commercial use within the next 5 years. The ‘‘viewpoints” that are ‘‘carefully reasoned commentaries in
article should address the classes of applications ideally sui- which all positions are substantiated by facts or principled
ted to experiment, as well as further research endeavors. arguments”, from ‘‘technical opinions”, that are treated as
The problematic nature of application papers is also informal commentaries on technical matters. IEEE Comput
stressed by the ICSE 2006 guidelines, which publish appli- Graph publishes opinions both as viewpoint articles and as
cation papers within the achievement and challenges track. ‘‘Graphically Speaking” articles. In viewpoint articles, the
Here papers address the critical issues pending in order reader can find technical opinions, as well as the discussion
to permit major software engineering achievements, and of challenges and limitations in todays methods, and of
the technical challenges which remain in theory and prac- potential areas of new research. These papers are evaluated
tice. Achievements focus on SE contributions in different on the basis of the following criteria: uniqueness, brevity
application domains, whereas challenges are expected to (2–4 pages), high-quality figures, and the structured format
focus on major unsolved problems regarding currently introduction actual visualizations–results. Different crite-
implemented systems, with the aim of raising interesting ria are considered for opinion papers of the ‘‘Graphically
and fundamental research questions. However, despite Speaking” section. They must deal with some interesting
their clear-cut purpose, such contributions do not adopt aspect of interactive computer graphics, should not be
necessarily a specific methodology, and achievement and obvious, should be short but provide a clear basis for the
challenges contributions may appear as either case studies viewpoint presented. A proper IEEE Software Guide opin-
or problem statements. On the other hand, ‘‘Problem state- ion essay should present the position, explain how it relates
ments” as a type of paper appear in CAiSE 2006 too, with to overall software development, and should provide evi-
no specific reference to application, but rather to ‘‘usage dence and justification for the defended position, while
and development”. explaining opposite camps and common practices. IEEE
Software Guide potential reviewers should check if the posi-
3.10. Opinion papers tion is adequately supported with ‘‘conjecture, logic, intui-
tion, or experimental data”. The reporting of alternative
positions and ‘‘pointers to background information, posi-
Box 17 Reflection box on: opinion papers tion justification, and opposing points of view” are also
required for this paper. Finally, reviewers should check
– Opinion papers may cover general topics of interest whether a sufficient number of readers are likely to disagree
for the field, comment on previously published with the position taken.
papers, or explain a technical position. In our ‘opinion’ a good example of an opinion paper is
– Requirements for publication include, among the given by Booch (2007) as part of the series On Architecture.
others: convincing arguments, brevity, and
interestingness. 4. Discussion

What we learned from this study is that the guidance


Opinions may be given on a general topic of interest in currently available to write and review SE papers is frag-
the field, or on a previously published paper – in this case, mented and still insufficient to help authors and reviewers
often in the form of letters. In IEEE Comput Graph, an thoroughly in their tasks. In what concerns empirical and
opinion paper ‘‘[. . .] aims at describing and challenging experimental research, this is probably a consequence of
the status-quo of interactive computer graphics, and pre- the methodological problems discussed in much SE litera-
dicting its many opportunities by learning from the past, ture (Fenton, 1993; Tichy et al., 1995; Glass, 1995; Zelko-
questioning the present, and anticipating the future that witz and Wallace, 1998), and commented on in the
within a few decades evolved from the ideas and imagina- introduction to this study. The not yet clear indications
tions of researchers and scientists to become a commodity about what to include in certain types of papers reflect
in all areas of our lives”. In the IEEE Software Guide, they the absence of well-established procedures to carry out
should encourage readers to think ‘‘out of the box”, research. Certainly, the members of the community have
expand their perspectives, and raise controversy. Gener- implicit knowledge of what is not written on paper. For
ally, opinion papers can be ‘‘provocative”, or ‘‘serious, example, if certain requisites of good papers, such as
whimsical, or humorous” (Sigmod Rec). The general tone importance, relevance, or significance may sound too gen-
is similar in conferences. ‘‘Vision papers” presented at eral to a standard reader, they may have, on the contrary, a
RE 2006 should be ‘‘revealing and thought-provoking”, clear meaning for authors and reviewers, and especially for
Author's personal copy

1710 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

the latter who, as a general rule, are experienced in writing need to be based on the consensus of at least a great part
and judging papers. Still, it is true what stated by Wieringa of the community. Kitchenham et al. (2002) and
et al. (2005), that articles might be judged of a good or a Jedlitschka and Pfahl (2005), for example, claim that it is
bad quality for different reasons. How to properly write not yet clear which factors should be reported in the
or judge papers depends on what is considered to be a research article regarding, for example, participants, con-
proper research methodology or research procedure, and, text or threats to validity. Nonetheless, consensus seems
as we have seen, the debate on this topic is in part still necessary in other aspects as well. First of all, we did not
open. find a great agreement on the terminology used for each
Nevertheless, despite the problems just mentioned, this article genre. The confusion existing on terms such as
study has come up with a first classification of SE articles, ‘‘experiment” or ‘‘empirical research” is a good example
which is reported in Tables 5 and 6. It is undoubtedly an of such problem. In this article, however, we have also seen
improvable classification, but still it represents a necessary that it is not yet clear the difference between ‘‘case studies”
basis for further developments. The commentaries included and ‘‘industry reports”, or between ‘‘experience reports”
under each article genre describe all of them in terms of and ‘‘application or applied articles”, to cite but a couple
requisites for publication and communicative function, of examples. It is very difficult to establish what a good
sometimes pointing to unresolved issues such as an unclear paper is, if we do not agree first on what we are speaking
communicative purpose or possible terminological about.
problems. This terminological confusion is particularly undesirable
Several researchers, especially those studying empirical if it is definitely proved that certain article genres are more
and experimental research, point out that improvements appropriate than others to bridge the gap between theory

Table 5
Summary of the article genres described in the paper and their major aspects (part A)
Publishable papers: general
– Regular papers may be shaped as different genres, such as, for example, experimental papers, case studies, or surveys
– A typical SE paper has a three-section structure: Introduction – Main Body – Conclusion
– Appropriate style must be comprehensible
– Relevance for practice, novelty, and originality are the most important requirements for publication
Extended versions of conference papers
– Extended versions of conference papers are a typical genre of SE
– Extended versions have passed through a double round of reviews
– Many journals require that approximately one third of the content be new
– Publication on a journal allows discussing the research beyond the time and space limitations set by conferences
Empirical research reports
– Empirical research includes different research strategies such as case studies, experimental research, or qualitative research
– Clear and comprehensive guidelines on how to write a proper empirical research paper are not yet available
– The experimental context is considered important for all types of empirical research
– Empirical research is increasing in SE
Observational studies:
– Observational studies include all empirical research strategies that impose little control over the developmental process
– Contextual factors, such as the type of industry in which the products are used, or the experience and skills of software staff, affect the utility and
generality of the results
Case studies:
– Case studies’ most distinguishing feature is their descriptive character
– It is difficult to distinguish case studies from experience reports or first-hand experience reports
– The environment of the research is especially important to evaluate case studies
– They allow for different data collection procedures, including controlled experiments
– Their major advantage is that they reflect real world situations
– Some authors argue that, for this reason, they can communicate better to practitioners
Field studies:
– Neither the function of field studies nor their difference from case studies is clear in most sources
Experimental papers:
– The terms ‘‘experiment” and ‘‘experimentation” are often used improperly in SE
– Many researchers complain about the lack of standardization regarding collection measures for experimental research
– Some also argue that current SE experimental research is insufficiently representative of industrial settings
– It is particularly difficult to carry out true experimentation in SE real settings
Meta-analyses:
– Meta-analyses allow to combine results of various empirical studies
– The lack of standardization in reports of empirical research makes it difficult to apply meta-analyses in SE
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1711

Table 6
Summary of the article genres described in the paper and their major aspects (part B)
Experience papers
– Experience reports are called by different names (‘‘lessons learned reports”, ‘‘industry experience reports”, etc.), and it is not clear if different names
imply variations
– Experience reports appear to be connected with application and practice
– In some sources, practitioners are the expected authors of experience reports
– Soundness of the report and objectivity are important publication requirements
Theoretical papers
– Theoretical papers are often of a methodological or computational nature
– Relevance for practice is the most important requirement for publication
Tutorial articles
– Tutorials are mainly educational, and introduce the work done in a field in a highly comprehensible manner
– Two types of tutorial are considered in the IEEE Software Guide: one focusing on applications and the other on research activities
– Tutorials address audiences with different background and levels of experience
Surveys
– Surveys complement tutorials, but address a more experienced audience
– Surveys do not need to be novel, but should make a contribution to the understanding of an area
– They help researchers as well as practitioners to keep up to date
– They can have a role in bridging the gap between practice and research
Short papers
– Short papers deal usually with practical or technical topics
– They usually undergo the same revision process as full-length papers
– Negative results or repetitions of previous results fit well short papers
Papers oriented towards practice
Educational papers:
– Educational papers appear more commonly in conferences than in journals
‘‘How to”/ tool report papers:
– ‘‘How to” and tool articles explain how to perform specific development tasks or on how to use specific tools
– The conditions of applicability/inapplicability of the process or the tool at hand must be detailed
Application papers:
– Application papers are similar and related to experience reports, tool reports and ‘‘how to” papers
– They may focus on applications in different domains
– They usually stress problems and obstacles to overcome
Opinion papers
– Opinion papers may cover general topics of interest for the field, comment on previously published papers, or explain a technical position
– Requirements for publication include, among the others: convincing arguments, brevity, and interestingness

and practice. In this respect, distinguishing a preferential qualified as interdisciplinary as well. Glass et al. (2004) and
audience for each type of paper, the IEEE Software Guide Vessey et al. (2005) studied the relationship of SE with
confirms what we commented upon in the introduction, other disciplines in terms of ‘‘reference discipline”, i.e.
that different types of papers are better suited for certain ‘‘any discipline outside the SE field that SE researchers
readers. Further research should try to find out which types have relied upon for theory and/or concepts”. They con-
of papers are best received by practitioners, if interaction cluded that hardly ever does SE rely on a reference disci-
between research and practice remains a crucial problem pline, but this conclusion clashes with the way many SE
in the field. Practitioners would certainly benefit from journals present themselves. Most of them insist on their
major terminological consistence. Regarding the article interdisciplinary and sometimes multidisciplinary nature.
genres more likely to be useful for SE practitioners, Is the application type of paper the one more likely to deal
researchers point out two genres: case studies (Ardimento with interdisciplinary content? If so, how do writers handle
et al., 2003; Kitchenham et al., 1995; Segal, 2005) and sur- such content? No answer comes from the sources we have
veys, sometimes also called literature reviews (Dyba et al., studied.
2005). More vague indications exist on the utility for prac-
titioners of experience and tool reports, which however 5. Conclusion
have not yet been identified and described unequivocally,
or of tutorials. Evidence must be provided in all cases. In this article we have summarized the fragmented infor-
Finally, application papers are indeed a good example mation available on the different types of papers published
of papers without a clear communicative function. Consid- in SE journals and conferences. Our sources are the
ering that they often deal with specific applications of soft- instructions to authors of major publications in the field,
ware in different domains, we could wonder if they could be calls for papers of major conferences, as well as the relevant
Author's personal copy

1712 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

literature published on the topic. What have we achieved as IBM Syst J – IBM Systems Journal (1.255)
a result? IEE Proc-Softw – IEE Proceedings. Software (0.300)
Firstly, we can now count on a first classification of IEEE Comput Graph – IEEE Computer Graphics and
research articles in SE, as summarized in Tables 5 and 6. Applications (1.152)
Each type of paper has been described in terms of desirable IEEE Internet Comput – IEEE Internet Computing
requisites for publication and communicative function. (2.304)
Relationships between different article genres have been IEEE Micro – IEEE Micro (1.238)
commented on, and a special emphasis has been placed IEEE Software – IEEE Software (1.333)
on unresolved issues, such as the terminological problems. IEEE T Reliab – IEEE Trans. on Reliability (0.715)
Secondly, we have learned that, without solving the IEEE T Software Eng – IEEE Trans. on Software Engi-
methodological issues currently matter for debate in SE, neering (1.967)
the problem of properly writing and judging certain types IEEE T Vis Comput Gr – IEEE Trans. on Visualization
of SE papers remains open. Different authors have com- and Computer Graphics (1.457)
plained about the lack of agreed-on criteria to write and Inform Software Tech – Information and Software
judge papers, as explained in other parts of this study. Technology (0.435)
However, how a good paper should be depends on well- J ACM – Journal of the ACM (2.197)
established patterns of research and methodologies, which J Comput Sci Technol – Journal of Computer Science
in some cases are currently under discussion in SE. and Technology (0.353)
Finally, this study is fundamentally theoretical, as it J Funct Program – Journal of Functional Programming
reports on what certain actors in the community regard (1.400)
as desirable attributes for publication. Future research J Math Imaging Vis – Journal of Mathematical Imaging
can further refine article genres characteristics by studying and Vision (2.197)
published articles in the framework of our classification. J Res Pract Inf Tech – Journal of Research and Practice
in Information Technology (0.205)
Acknowledgements J Softw Maint Evol – Journal of Software Maintenance
and Evolution (0.457)
The authors would like to thank Rafael Capilla and J Syst Software – Journal of Systems and Software
Hans van Vliet for their comments and suggestions on ear- (0.744)
lier versions of this paper. This work was partially sup- J Visual Comp Animat – Journal of Visualization and
ported by a grant from the Spanish Ministry of Computer Animation (0.864)
Education and Culture. Knowl Eng R – The Knowledge Engineering Review
Presence-Teleop Virt – Presence: Teleoperators & Vir-
Appendix A. List of journals tual Environments (0.628)
Sci Comput Program – Science of Computer Program-
List of journal titles cited in the paper, in alphabetical ming (0.734)
order with their corresponding acronyms. When available, Sigmod Rec – Sigmod Record (1.155)
the acronym of the JCR has been used in the first place. Simul Model Pract Th – Simulation Modelling Practice
For the journals present in the JCR we also include the and Theory (0.361)
Impact Factor in round brackets. Simulat Pract Theory – Simulation Practice and Theory
Software Pract Exper – Software: Practice and Experi-
ACM Comput Surv – ACM Computing Surveys ence (0.406)
ACM T Database Syst – ACM Trans. on Database Sys- Software Process – Software Process: Improvement and
tems (1.833) Practice
ACM T Graphic – ACM Trans. of Graphics (3.652) SoSyM – Journal on Software and System Modeling
ACM T Math Software – ACM Trans. on Mathemati- Theor Pract Log Prog – Theory and Practice of Logic
cal Software (1.463) Programming (1.372).
ACM T Softw Eng Meth ACM Trans. on Software
Engineering and Methodology (3.000) Appendix B. List of conferences
Algorithmica – Algorithmica, Springer (1.017)
Commun ACM – Communications of the ACM (1.797) List of conferences (in alphabetical order with their cor-
Comput Aided Design – Computer-aided Design (1.173) responding acronyms) whose calls for papers have been
Comput Graph Forum – Computer Graphics Forum consulted.
(0.972)
Computer – IEEE Computer (1.282) ASE 2006 – IEEE/ACM Int. Conf. on Automated Soft-
Dr Dobbs J – Dr. Dobbs (0.050) ware Engineering
Empirical Softw Eng – Empirical Software Engineering ASWEC 2007 – Australian Conference on Software
(0.966) Engineering
Author's personal copy

M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714 1713

CAiSE 2006 – Conference on Advanced Information the 29th International Conference on Software engineering (ICSE), pp.
Systems Engineering 581–590.
Booch, G., 2007. The irrelevance of architecture. IEEE Softw. 24 (3), 10–
CSMR 2006 – European Conference on Software Main- 11.
tenance and Reengineering Brereton, P., Kitchenham, B.A., Budgen, D., Turner, M., Khalil, M.,
EASE 2006 – Evaluation and Assessment in Software 2007. Lessons from applying the systematic literature review process
Engineering within the software engineering domain. J. Syst. Softw. 80 (4), 571–
ESEC/FSE 2007 – European Software Engineering 583.
Brooks, A., 1997. Meta analysis – a silver bullet-for meta-analysts.
Conference/ACM SIGSOFT Symposium on the Foun- Empirical Softw. Eng. 2 (4), 333–338.
dations of Software Engineering Budgen, D., Brereton, P., 2006. Performing systematic literature reviews in
ESEM 2007 (METRICS + ISESE) – Empirical Soft- software engineering. In: ICSE, pp. 1051–1052.
ware Engineering and Measurement Budgen, D., Kitchenham, B., 2005. Realising evidence-based software
EWSA 2006 – European Workshop on Software engineering a report from the workshop held at ICSE 2005. SIGSOFT
Softw. Eng. Notes 30 (5), 1–5.
Architecture Budgen, D., Charters, S., Turner, M., Brereton, P., Kitchenham, B.,
ICSE 2006 – Int. Conf. on Software Engineering Linkman, S., 2006. Investigating the applicability of the evidence-
ICSM 2006 – Int. Conf. on Software Maintenance based paradigm to software engineering. In: WISER’06: Proceedings
ICSR 2006 – Int. Conf. on Software Reuse of the 2006 International Workshop on Interdisciplinary Software
ISESE 2006 – International Symposium on Empirical Engineering Research. ACM Press, New York, NY, USA, pp. 7–14.
Djokic, S., Succi, G., Pedrycz, W., Mintchev, M.P., 2001. Meta analysis –
Software Engineering a method of combining empirical results and its application in object-
METRICS 2005 – IEEE International Software Metrics oriented software systems. In: Proceedings of the Seventh International
Symposium Conference on Object-Oriented Information Systems, pp. 103–112.
MoDELS 2006 – Int. Conf. on Model-Driven Engineer- Dyba, T., Kitchenham, B.A., Jorgensen, M., 2005. Evidence-based
ing Languages and Systems software engineering for practitioners. IEEE Softw. 22 (1), 58–65.
Fenton, N., 1993. How effective are software engineering methods? J. Syst.
OOPSLA 2006 – Int. Conf. on Object-Oriented Pro- Softw. 22 (2), 141–146.
gramming, Systems, Languages and Applications Fletcher, P., 1995. The role of experiments in computer science. J. Syst.
QoSA 2006 – Int. Conf. on the Quality of Software Softw. 30 (1–2), 161–163.
Architectures Glass, R.L., 1995. Structure-based critique of contemporary computing
RE 2006 – IEEE International Requirements Engineer- research. J. Syst. Softw. 28 (1), 3–7.
Glass, R.L., Vessey, I., Ramesh, V., 2002. Research in software engineer-
ing Conference ing: an analysis of the literature. Information Softw. Technol. 44 (8),
SEKE 2007 – Int. Conf. on Software Engineering and 491–506.
Knowledge Engineering Glass, R.L., Ramesh, V., Vessey, I., 2004. An analysis of research in
WICSA 2007 – Working IEEE/IFIP Conference on computing disciplines. Commun. ACM 47 (6), 89–94.
Software Architecture Goldberg, A., 1982. Editorial policy. ACM Comput. Surveys 14 (2), 151–
157.
Haddow, G., Klobas, J., 2004. Communication of research to practice in
library and information science: closing the gap. Library Information
References Sci. Res. 26 (1), 29–43.
Hjørland, B., 1998. Information retrieval, text composition, and seman-
Abran, A., Moore, J.W., Bourque, P., Dupuis, R., Tripp, L.L., 2004. tics. Knowledge Organization 25 (1/2), 16–31.
Guide to the Software Engineering Body of Knowledge (SWEBOK). Hjørland, B., 2002. Domain analysis in information science. Eleven
IEEE. approaches – traditional as well as innovative. J. Document. 58 (4),
Anthony, L., 1999. Writing research article introductions in software 422–462.
engineering: how accurate is a standard model? IEEE Trans. Profess. IEEE, 2002. IEEE keyword taxonomy. available at: http://www.com-
Commun. 42 (1), 36–46. puter.Org/mc/keywords/software.htm.
Ardimento, P., Baldassarre, M., Bianchi, A., Visaggio, G., 2003. Empirical IEEE, 2006. Genre submission guide. IEEE Software, available at http://
studies as a means for technology transfer. In: Proceedings of the www.computer.org/portal/site/software.
Second Workshop Series on Empirical Engineering, p. 8. Jedlitschka, A., Pfahl, D., 2005. Reporting guidelines for controlled
Arranga, E.C., 2000. Cobol tools: overview and taxonomy. IEEE Softw. experiments in software engineering. In: Proceedings of the Fourth
17 (2), 59–69. International Symposium on Empirical Software Engineering (ISESE),
Babar, M.A., Kitchenham, B., Maheshwari, P., 2006. The value of pp. 95–104.
architecturally significant information extracted from patterns for Journal Citation Reports, 2006. Journal citation reports. Available
architecture evaluation: a controlled experiment. In: Proceedings of the at http://portal.isiknowledge.com/portal.cgi?DestApp=JCR&Func=
Australian Software Engineering Conference (ASWEC). IEEE Com- Frame.
puter Society, Washington, DC, USA, pp. 379–390. Kajiya, J., 1993. How to get your siggraph paper rejected. http://
Basili, V.R., 1996. The role of experimentation in software engineering: old.siggraph.org/publications/instructions/rejected.html.
past, current, and future. In: Proceedings of the 18th International Kitchenham, B., 2004. Procedures for performing systematic reviews.
Conference on Software Engineering (ICSE). IEEE Computer Society, Tech. Rep. TR/SE-0401, Keele University, available at: http://
Washington, DC, USA, pp. 442–449. www.elsevier.com/framework_products/promis_misc/inf-systrev.pdf
Basili, V.R., Shull, F., Lanubile, F., 1999. Building knowledge through (July).
families of experiments. IEEE Trans. Softw. Eng. 25 (4), Kitchenham, B., Pickard, L., Pfleeger, S.L., 1995. Case studies for method
456–473. and tool evaluation. IEEE Softw. 12 (4), 52–62.
Best, B., Jürjens, J., Nuseibeh, B., 2007. Model-based security engineering Kitchenham, B.A., Hughes, R.T., Linkman, S.G., 2001. Modeling
of distributed information systems using UMLsec. In: Proceedings of software measurement data. IEEE Trans. Softw. Eng. 27 (9), 788–804.
Author's personal copy

1714 M. Montesi, P. Lago / The Journal of Systems and Software 81 (2008) 1694–1714

Kitchenham, B.A., Pfleeger, S.L., Pickard, L.M., Jones, P.W., Hoaglin, Succi, G., Pedrycz, W., Djokic, S., Zuliani, P., Russo, B., 2005. An
D.C., Emam, K.E., Rosenberg, J., 2002. Preliminary guidelines for Empirical Exploration of the Distributions of the Chidamber and
empirical research in software engineering. IEEE Trans. Softw. Eng. Kemerer Object-Oriented Metrics Suite. Empirical Softw. Eng. 10 (1),
28 (8), 721–734. 81–104.
Kwasnik, B.H., Crowston, K., 2005. Introduction to the special issue: Tichy, W.F., 1998. Should computer scientists experiment more? IEEE
Genres of digital documents. Information Technol. People 18 (2), 76– Computer 31 (5), 32–40.
88. Tichy, W.F., 2000. Hints for reviewing empirical work in software
Laitenberger, O., Rombach, D., 2003. (Quasi-)experimental studies in engineering. Empirical Softw. Eng. 5 (4), 309–312.
industrial settings. Lecture Notes Empirical Softw. Eng., 167–227. Tichy, W.F., Lukowicz, P., Prechelt, L., Heinz, E.A., 1995. Experimental
Marcos, E., 2005. Software engineering research versus software develop- evaluation in computer science: a quantitative study. J. Syst. Softw. 28
ment. SIGSOFT Softw. Eng. Notes 30 (4), 1–7. (1), 9–18.
Miller, J., 1999. Can results from software engineering experiments be van Vliet, H., 2005. Some myths of software engineering education. In:
safely combined? In: Proceedings of the Sixth International Sympo- ICSE’05: Proceedings of the 27th International Conference on
sium on Software Metrics (METRICS). IEEE Computer Society, Software Engineering. ACM, New York, NY, USA, pp. 621–622.
Washington, DC, USA, p. 152. van Vliet, H., 2006. Reflections on software engineering education. IEEE
Miller, J., 2004. Statistical significance testing: a panacea for software Softw. 23 (3), 55–61.
technology experiments? J. Syst. Softw. 73 (2), 183–192. Vessey, I., Ramesh, V., Glass, R., 2005. A unified classification system for
Moed, H., Visser, M., 2007. Developing bibliometric indicators of research in the computing disciplines. Information and Softw. Tech-
research performance in computer science: an exploratory study. nol. 47 (4), 245–255.
Tech. Rep. CWTS Report 2001-01, Research Report to the Council for Wasserman, A.I., 1984. Editorial policy. ACM Comput. Surveys 16 (2),
Physical Sciences of the Netherlands Organization for Scientific 103–110.
Research (NWO). Wieringa, R., Maiden, N., Mead, N., Rolland, C., 2005. Requirements
Montesi, M., Mackenzie-Owen, J., 2008. Research journal articles as engineering paper classification and evaluation criteria: a proposal and
document genres: exploring their role in knowledge organization. a discussion. Requir. Eng. 11 (1), 102–107.
Journal of Documentation 64 (1). Wohlin, C., 2005. An analysis of the most cited articles in software
Montesi, M., Mackenzie-Owen, J., accepted for publication. From engineering journals – 1999. Information Softw. Technol. 47 (15), 957–
conference to journal publication. How conference papers in software 964.
engineering are extended for publication in journals. Journal of the Wohlin, C., Runeson, P., Höst, M., Ohlsson, M.C., Regnell, B., Wesslén,
American Society for Information Science and Technology. A., 2000. Experimentation in Software Engineering: An Introduction.
Perry, D.E., Sim, S.E., Easterbrook, S.M., 2004. Case studies for software Kluwer Academic Publishers, Norwell, MA, USA.
engineers. In: ICSE’04: Proceedings of the 26th International Confer- Wu, J., Graham, T., Smith, P.W., 2003. A study of collaboration in
ence on Software Engineering. IEEE Computer Society, Washington, software design. ISESE 00, 304.
DC, USA, pp. 736–738. Zelkowitz, M.V., Wallace, D.R., 1998. Experimental models for validating
Pickard, L.M., Kitchenham, B., 1998. Combining empirical results in technology. Computer 31 (5), 23–31.
software engineering. Information Softw. Technol. 40 (14), 811–821. Zelkowitz, M., Wallace, D., Binkley, D., 2003. Evaluation of new software
Posteguillo, S., 1999. The schematic structure of computer science research engineering technologies. In: Juristo, N. and Moreno, A.M., (Ed.),
articles. English Specific Purposes 18 (2). Lecture Notes on Empirical Software Engineering, pp. 229–263.
Punter, T., 2003. What information do software engineering practitioners Zendler, A., 2001. A preliminary software engineering theory as investi-
need? In: Proceedings of Second Workshop on Empirical Studies in gated by published experiments. Empirical Softw. Eng. 6 (2), 161–180.
Software Engineering. Rome, Italy. Zhang, X., Windsor, J., Pavur, R., 2003. Determinants of software
Segal, J., 2005. When software engineers met research scientists: A case volatility: a field study. J. Softw. Maintenance 15 (3), 191–204.
study. Empirical Softw. Eng. 10 (4), 517–536.
Segal, J., Grinyer, A., Sharp, H., 2005. The type of evidence produced by Michela Montesi’s research area is scientific communication and the role
empirical software engineers. In: Proceedings of the 2005 Workshop on of information scientists in mediating communication between research-
Realising Evidence-based Software Engineering (REBSE). ACM ers, and research and practice. She obtained a PhD in Information Science
Press, New York, NY, USA, pp. 1–4. from the Complutense University of Madrid in 2004, and has spent the
Shaw, M., 2002. What makes good research in software engineering. Int. last two years as a guest researcher at the University of Amsterdam.
J. Softw. Tools Technol. Transfer 4 (1), 1–7.
Shaw, M., 2003. Writing good software engineering research papers: Patricia Lago is Associate Professor in Software Engineering at the Vrije
minitutorial. In: Proceedings of the 25th International Conference on Universiteit (VU University) in Amsterdam. Her research interests focus
Software Engineering (ICSE). IEEE Computer Society, Washington, on software and service architectures, software architecture knowledge
DC, USA, pp. 726–736. management and modeling. She co-authored various publications in
Sidiropoulos, A., Manolopoulos, Y., 2005. A new perspective to international journals, books and conference proceedings. She is co-
automatically rank scientific conferences using digital libraries. Infor- organizer of the International Workshop on SHAring and Reusing
mation Process. Manage. 41 (2), 289–312. architectural Knowledge (SHARK). She also acts as reviewer in several
Sjøberg, D.I.K., Hannay, J.E., Hansen, O., Kampenes, V.B., Karahasa- international conferences and journals in the field. A full publication list is
novic, A., Liborg, N.-K., Rekdal, A.C., 2005. A survey of controlled available at www.cs.vu.nl/~patric.
experiments in software engineering. IEEE Trans. Softw. Eng. 31 (9),
733–753.

View publication stats

You might also like