You are on page 1of 9

38 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 42, NO.

1, MARCH 1999

Writing Research Article


Introductions in Software
Engineering:
How Accurate is a Standard Model?
Abstract—In this paper, a standard model for describing the structure
of research article introductions, the CARS model, is evaluated in terms
of how well it can be applied to 12 articles which have received “Best
Paper” awards in the field of software engineering. Results indicate
that although the model adequately describes the main framework of
the introductions, a number of important features are not accounted
for, in particular, an extensive review of background literature, the
inclusion of many definitions and examples, and an evaluation of the
—LAURENCE ANTHONY research in terms of application or novelty of the results.

Index Terms—Engineering, genre, pedagogy, software.

Manuscript received March 30, 1998; [11], authorial comment [12], and
revised August 7, 1998.
This work is a substantially revised T HE research article (RA) or
paper is one of the most impor-
voice [13], and an analysis of RAs
at the discourse level has led to
version of a paper published in the
conference proceedings of tant genres that both scientists general models for the structure of
The Japan Conference on English and engineers will write [1], [2]. the various sections of the RA [5],
for Specific Purposes, 1997. [14]–[16].
The author is with the Successful publication of research
Department of Information findings can lead not only to an
and Computer Engineering, enhanced reputation, but also
Faculty of Engineering,
peer acceptance, professional ad- One important question that has
Okayama University of Science, yet to be answered, however, is to
1-1 Ridai-cho, Okayama 700-0005, vancement, research grants, and
Japan so on [3]–[5]. It is not surpris- what extent such generalizations
(e-mail: anthony@ice.ous.ac.jp). ing therefore, that journals such can be made about the writing of
IEEE PII S 0361-1434(99)01808-1. research articles. In other words,
as the IEEE TRANSACTIONS ON
PROFESSIONAL COMMUNICATION fea- how well do the general descrip-
ture work covering all aspects tions of RAs accurately account for
of the RA, from approaches to the writing in a specific discipline?
writing and techniques to improve Harmon, for example, concedes
clarity, to key elements in the dif- that “for just about any character-
ferent RA sections and strategies istic one might ascribe to the struc-
to improve chances of publication ture in scientific and engineering
[3], [6]–[9]. In the field of genre papers, numerous counterexam-
analysis, a subdiscipline of English ples can be cited” [8, p. 132]. In
Language Teaching (ELT) and ap- this paper, an attempt is made
plied linguistics, a separate body to answer the above question by
of work has also focused on the assessing how accurately Swales’
RA. Here, corpus-based studies Create a Research Space (CARS)
have revealed preferred ways to model [5, p. 141] for the structure
use tense [10], citation patterns of RA introductions accounts for
the features of writing in a corpus
0361–1434/99$10.00  1999 IEEE
ANTHONY: WRITING RESEARCH ARTICLE INTRODUCTIONS IN SOFTWARE ENGINEERING 39

of 12 RAs in software engineering not only limits the time that can be details of the corpus articles are
(see Fig. 1). given to individual texts, but also summarized in Table I.
tends to render the corpus size
The CARS model was chosen for for a particular discipline rather Analysis and Validation of Re-
a number of reasons. First, the small. For example, in Swales’ sults First, using the descriptions
model has been through several 1981 study, a corpus of 48 articles offered by Swales [5], the corpus
revisions since its conception in from 14 different fields was used, articles were analyzed, and indi-
1981 [17], incorporating the find- an average of only 3.4 articles vidual sentences were classified
ings of Cooper [18], who applied per discipline; and in Crookes’ into an appropriate step of the
it in electrical and electronic en- 1986 study, 96 articles from 12 CARS model. Sentences that could
gineering, and Crookes [19], who different disciplines were selected, be classified into more than one
applied it in both the “hard” and an average of still only 8 per dis- step, those that were unclear, and
“social” sciences. As such, the cipline [17], [19]. In view of this, those that did not fit the model
model can be considered one of here a more detailed study of 12 were noted. Of course, one of the
the stronger descriptions of text articles from a single, high-profile major problems in doing an analy-
structure to date. Second, the journal in software engineering sis of this type is the subjectivity of
model is widely referred to in the was undertaken. The journal cho- the interpretations made. Indeed,
field of ELT, and its acceptance sen was the IEEE TRANSACTIONS Crookes [19] is highly critical of
is reflected in the number of text- ON SOFTWARE ENGINEERING, one Swales and many others’ works
books on writing which directly of the most influential journals in this regard, and Swales himself
quote it or whose accounts have in software engineering, with a describes the labeling of an earlier
been strongly influenced by it, monthly distribution of over 11 version of the CARS model as
e.g., [14], [20]. Finally, as one thousand copies to institutions “unsubstantiated and ill-defined”
of the major pioneers of genre and individual subscribers [21]. [17, p. 14].
analysis, Swales’ research into
the structure of RA introductions To ensure the validity of the in-
Although all articles appearing in terpretations made here, after
illustrates many of the concepts
such a prestigious journal are of the initial analysis was complete,
and techniques used in the field
very high quality, a large num- the results were discussed with
as a whole.
ber of submissions are made by four specialists in software en-
nonnative speakers of English gineering, a technique used by
and so the possibility of errors Selinker [23], Tarone et al. [13],
RESEARCH DESIGN in English remains [18]. Distin- and others [2], [18]. The insights of
Selection of Corpus In previ- guishing between the writing of the informants were gained during
ous genre studies of this type, native and nonnative speakers, a series of recorded interviews
a general picture of text struc- however, can be difficult even with lasting from two to four hours
ture has been obtained using a the journal supplying biographic and later communication via e-
large corpus of articles from a data on the authors. Here, to mail. In addition to covering the
wide range of disciplines. Unfor- reduce the potential problems problem sentences noted above,
tunately, this selection procedure of analyzing nonnative speaker the specialists were also asked
writing, 12 full paper articles from about their own views on writing
a total of 22 which had received RA introductions and the nature
Fig. 1. Swales’ CARS model.
“Best Paper” awards in 1996 were
selected from the journal. (Due to
TABLE I
time restrictions ten “Best Papers”
AUTHORSHIP DETAILS
in the area of computer security
were excluded from the study). Al-
though the criteria for the awards
are not stated explicitly in the
journal, the editors indicate that
the articles were evaluated more
in terms of their significance than
writing style. Nevertheless, all the
papers had to go through two
rigorous review processes where
errors in English would be checked
by up to six reviewers and editors,
and in at least one case reviewers
suggested changes to the struc-
ture of the paper [22]. Authorship
40 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 42, NO. 1, MARCH 1999

of writing in software engineer- A software requirements specifi- value of the research and how it
ing in general. One cautionary cation should be a comprehen- extends previous results. Finally,
note should be made however; as sive statement of a software sys- the original Step 3-3 was renamed
Swales points out, tem’s intended behavior. (Intro. Step 3-4 and includes only state-
1.1, line 1) ments about the remaining RA
Before developers of certifiably structure.
Although Selinker (personal
safe software can take advan-
communication) is clearly
tage of the concurrent and real- The modified version of the CARS
right to point out that without
time constructs of Ada, rigorous model used in the analysis here is
specialist informants genre
analysis techniques to analyze presented in Fig. 2. It should be
analysts are in danger of
their timing properties must be noted that apart from the addi-
‘not knowing what they don’t
developed. (Intro. 2.2, line 14) tional evaluation step, it is almost
know,’ over-reliance on spe-
identical to the original model. The
cialist informants may invite
To distinguish between these steps, only other change made was the
the opposite danger of ana-
it would seem necessary to also addition of the “and” condition
lysts believing all that they
consider the location of the state- in Move 2 and 3 Step 1s, as it
hear [5, p. 129].
ments in the text with those ap- was anticipated that more than
pearing prior to a review of liter- one type of step could be used
To avoid this, four specialists rather ature (Step 1-3) classified as Step at a particular place. The issue
than one were interviewed, giving 1-2, and those appearing after may be raised that presenting a
the researcher a broader scope the review classified as Step 2- modified version is unnecessary
on which to make judgements, 1D [27]. Similarly, it was difficult and that the evaluation step can
and they were not informed of to distinguish between Step 3- be incorporated into Step 3-2.
the others’ responses. This was to 1B, which is said to describe “the Indeed, this is possible, but I hope
reduce the so-called “halo effect” main features of the research,” and the analysis here will show that
[24], when informant responses the Step 3-2 option “announcing in software engineering at least,
are biased by what they think principle findings” [5, p. 159]. It it is beneficial to consider it as a
the interviewer wants to hear or was decided, therefore, to classify separate step.
what they think the correct answer only general statements about how
should be. For further discussion or what was done into Step 3- Move Structure and Occurrence
on the use of the interview tech- 1B, and specific statements about of Steps Based on the Modified
nique see [25], [26]. research method, descriptions of CARS model, the corpus article
tools or techniques developed, introductions were analyzed and
and/or specific results into Step shown to have the move structure
3-2. given in Table II, and occurrence of
RESULTS steps shown in Table III. As can be
Clarification of Step Descrip- Other problems were found in the
tions in the CARS Model During description of Move 3 steps. Swales Fig. 2. The “Modified” CARS model.
the first stage of analysis, it was describes Step 3-3, for example, as
found that many sentences could including “in varying degrees of de-
be classified into more that one tail the structure—and occasion-
step of the CARS model. The prob- ally the content—of the remainder
lem was mainly due to the am- of the RA” [5, p. 161]. To detail the
biguous descriptions of the steps content of the RA, however, would
offered by Swales [5]. For example, naturally include details of results,
Step 1-2 in the model is said to thus merging with the purpose
include statements about “the cur- of Step 3-2. The description is
rent state of the art,” “phenomena” complicated further if Swales’ 1994
in the field, and “current require- account is considered [28]. In this
ments for further progress” [5, p. study, two further possibilities are
146]. The last type here, however, mentioned, i.e., a statement about
clearly overlaps with statements secondary findings and statements
in Step 2-1D, which are said to about the value of the research
establish a niche by expressing [28, p. 192]. To deal with these
the “needs/desires/interests” of problems, statements about sec-
the field [5, p. 156]. Based on ondary findings were included in
these definitions, the following Step 3-2, and a new “Evaluation of
statements from the corpus could Research” step was created, which
be classified into either step. included statements about the
ANTHONY: WRITING RESEARCH ARTICLE INTRODUCTIONS IN SOFTWARE ENGINEERING 41

seen from the tables, each corpus the introductions is less consis- this view but suggest that vari-
introduction exhibits all the moves tent with Swales’ [5] model. For ations in step occurrence among
described in the proposed model, example, although steps 1-3, 3- different disciplines may be greater
supporting the general framework 1B, and 3-3 are obligatory, steps than originally thought. They also
offered by Swales [5]. In the inter- 2-1A and 2-1C are redundant, suggest that even within a single
views, the software engineers were and steps 1-1, 2-1D, and 3-1A field such as engineering, there
also asked to comment on what occur in only half or less of the may be considerable variations
they would include in an article introductions. Swales comments between its associated subdisci-
introduction. Their responses are that the occurrence of steps will plines. Cooper [18], for example,
summarized in Table IV. Again, be discipline-dependent, stating found that electrical and elec-
there is close agreement with both that the hard sciences will show tronic engineering articles showed
the corpus introductions and the a preference for using steps 2- few examples of Step 1-3, with
Modified CARS model. 1B and 2-1D over steps 2-1A and the number of citations in the
2-1C, which are more common introductions and articles as a
From Table III, however, we see in education, management, and whole averaging only 7.6 and 17.6,
that the occurrence of steps in linguistics. Results here support respectively. Here, the occurrence

TABLE II
MOVE STRUCTURE IN CORPUS ARTICLE INTRODUCTIONS

TABLE III
STEP OCCURRENCE IN CORPUS ARTICLE INTRODUCTIONS
42 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 42, NO. 1, MARCH 1999

of Step 1-3 statements is far more the only sections common to all an average of over 1000 words.
extensive, with the average num- the articles are the introduction They are also significantly longer
ber of citations almost doubling and the conclusion. However, to than those of other engineering
to 14.6 and 42.9, respectively. For investigate this further, the corpus disciplines [18]. In addition, the
example, article authors were contacted corpus introductions show a large
directly via e-mail and asked why amount of cycling. For example, in
Jaffe, in his dissertation, defined they included the step. From five Intros. 2.1 and 2.2 there are four
a rigorous basis for [21]. (Intro. that responded, two authors gave cycles of Move 1 and 2 steps before
1.1, line 7) reasons related to helping the the aims of the research are finally
In work related to ours, Gerhart reader follow the structure of the stated, and in Intro. 1.1, the total
et al. [8] have put forth a set of paper. On the other hand, three number of cycles increases to 12.
criteria for evaluating (Intro. said they include it almost as For example, the following are the
1.2, line 18) a convention in all their article opening sentences to Intro. 1.1.
introductions.
A software requirements specifica-
The results are also in contrast
tion should be (Move 1)
with other views on engineering Cyclicity, Length of Moves, and
writing. For example, although the “Preaching to Cannibals” Unfortunately, requirements spec-
Pierson et al. [6] recommend that Phenomenon Although Swales ifications are often incomplete
writers do not include a summary accepts the possibility of cycling (Move 2)
of the remaining RA structure between different moves in the We know that many serious con-
(Step 3-4), in the introductions CARS model, he goes on to de- ceptual errors are (Move 1)
here almost all of the authors scribe introductions in engineering Therefore, it is important to
chose this option. Interestingly, as showing “brevity and linearity” (Move 2)
Cooper [18] also found a high [5, p. 159]. Other authors too have
Previously, we defined formal cri-
occurrence of Step 3-4 statements recommended that introductions
teria for (Move 1)
in electrical and electronic engi- should be brief [29]. Looking at
neering RAs. She explains that Table II, however, we can see
this is because they do not follow that in software engineering at Comparing the length of individual
the standard IMRD (Introduction– least, the precise opposite is found. moves and steps, on the whole
Method–Results–Conclusion) pat- The corpus introductions here Move 3 is the longest followed
tern or any other discernible pat- are surprisingly long, varying in by Move 1 and then Move 2 (see
tern. The same appears true here: length from 591 to 1479 words, Table V). From the table, we can
TABLE IV
SPECIALISTS’ VIEWS ON THECONTENT OF ARTICLE INTRODUCTIONS
ANTHONY: WRITING RESEARCH ARTICLE INTRODUCTIONS IN SOFTWARE ENGINEERING 43

see that in Move 1 a great deal bachelors degree can almost get gresses, more specific statements
of time is spent making general- some idea of what I’m trying to do are made. For example, the fol-
izations about the field (Step 1- in the paper.” lowing Move 3 cycles appeared in
2) and/or summarizing previous Intro 3.2, lines 16–45:
research (Step 1-3) (see above). “When you write a journal paper
Again, this contradicts the views you have to be very careful that However, in this paper, we focus
of many authors who state that even a naı̈ve person, that means on SGSPNs (Move 3—Move 1/2
there should only be a brief review who is not very familiar with your steps)
of literature in the introduction work, will be able to read that
journal paper and understand We consider analysis techniques
[8], [30], [31]. When the special- for a given SGSPN (Move 3—
ists were asked why so much what [you’re] trying to say.”
Move 1 steps)
background and previous liter- “I regard [article writing] as a
kind of ‘preaching to the canni- In this paper, we give a repre-
ature were covered in software
bals’.” sentation of Q, but (Move 3—
engineering article introductions,
Move 1/2 steps)
they explained that many readers Clearly, this “preaching to the
would be unfamiliar with much of cannibals” phenomenon is an im- This problem is attacked and solved
the terminology and background portant consideration for the soft- in this paper (Move 3)
necessary to understand the re- ware engineer when writing RA One more technique used by au-
search. This is because not only introductions. We see this in the thors is to include definitions of
software engineers, but engineers way the author gives detailed back- important terms and examples
from a wide range of disciplines ground information in the form to illustrate difficult concepts.
will subscribe to the journal in of topic generalizations and sum- Although there is no step in the
order to acquire results which can maries of previous research piece CARS model under which to clas-
be used to solve their own par- by piece, commenting on its prob- sify these statements, they were
ticular problems. This profoundly lems or gaps, and relating it to found extensively after Move 1
affects the writing of the articles. the aims of the present research. steps, and two introductions in-
For example, when the specialists Hence, there is extensive use of cluded long examples to clarify the
were asked to describe their own Move 1–Move 2 cycles, and even research problem and approach
writing, the following comments cycles involving a Move 3, where being used. For example,
were made: the aims of the research are first
presented in very general terms, internally complete, i.e., closed
“I try to explain things in a simple
but then as the introduction pro- with respect to statements (In-
way so that anyone who has a
TABLE V
LENGTH OF STEPS IN THECORPUS ARTICLE INTRODUCTIONS
44 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 42, NO. 1, MARCH 1999

tro. 1.1, line 7) introduction where an appeal to used software modules without
By module, we mean a single the audience is made is in Step incurring the expense of completely
procedure (Intro. 2.1, line 12) 3-3, a possibility only hinted at by reanalyzing each calling program.
Swales [27]. (Intro. 2.1, line 22)
scheduling algorithms/tech-
niques (e.g., rate monotonic sched- Most designs are flawed, and Nit-
From Table III, it can be seen pick finds a counterexample—even
uling) (Intro. 2.2, line 11)
that Step 3-3 appears in all the in a huge space—surprisingly fast.
corpus introductions, justifying (Intro. 2.3, line 26)
Note, there were occasions when the creation of a new step in the
citations appeared to be used On-the-fly elimination can save a
Modified CARS model. It is also
for the purpose of illustrating or significant amount of memory in
one of the longest steps in Move
exemplifying a point in the text in the generation process (Intro.
3, accounting for almost one third
addition to their role of signaling a 3.1, line 39)
of the move and 14% of the intro-
Step 1-3, as the following example duction as a whole (see Table V).
shows, In this step, the present research The second option, although less
is evaluated, almost always pos- popular, appears in seven of the
The Arch style is more recent, and itively, with respect to either or 12 introductions, and accounts
has an associated metamodel [38]. both of the following criteria: 1) for 24% of Step 3-3 as a whole.
(Intro. 1.3, line 5) the applicability of the research Here, the “novelty” is signaled in
and 2) the novelty of the research. terms of differences, uniqueness,
The first criterion is central in en- and “extension,” as the following
THE IMPORTANCE OF STEP 3-3 gineering as one of the specialists examples show.
(EVALUATION OF RESEARCH) describes,
“[Engineers] want to know did you Our work differs from that above in
Swales chose the labeling of moves that (Intro. 1.2, line 19)
in the CARS model to reflect the build it, does it work, how long,
how fast is it, because they want Our algorithm is unique in that it
nature of introductions, that of
to use it they want to see some rests on a model of concurrency
“persuading” the reader to accept
proof of concept.” that (Intro. 2.4, line 18)
the research being presented [5],
[32]. As mentioned above, the per- Thus we see its appearance in all This work extends that of [9] and
suasion is achieved to some extent but one of the corpus introduc- [11] in several ways (Intro. 2.2,
by cycling between statements tions, accounting for 58% of the line 38)
which establish the field (Move 1) total length of Step 3-3 sentences.
and those which indicate a gap or For example, One further possibility, accounting
extend previous findings (Move 2). for the remaining instances of Step
Results here, however, show that One advantage of our algorithms is 3-3, is for the author to discuss
one of the major sections of the that they let us analyze frequently the limitations of the research.

TABLE VI
OCCURRENCE OF STEP 3-3 EVALUATION TYPES
ANTHONY: WRITING RESEARCH ARTICLE INTRODUCTIONS IN SOFTWARE ENGINEERING 45

Although this is rare in the cor- however, when a more detailed literature background are asked to
pus, we do find three instances. description is required. The model teach technical writing courses to
Interestingly, these are always is first hampered by weak defi- scientists and engineers. A related
followed by a contrastive statement nitions of individual steps, and problem is how both teachers and
signaling a more positive aspect of because it has been developed students deal with texts that do
the research (see Table VI). For based on a wide variety of disci- not fit the model. In current text-
example, plines, many steps are redundant books that use the CARS model,
or only rarely used, namely, steps there is rarely an opportunity to
1-1, 2-1A, 2-1C, 2-1D, and 3-1A. deal with these problem cases, so
We have not yet demonstrated that A more serious problem is the if at some point they are encoun-
all these goals are achievable absence of a separate “evaluation tered, they are likely to be treated
by this style. We have examined of research” step in the original simply as exceptions to the rule. Of
several key properties and built model, which is shown here to be course, as the results here show,
several diverse experimental sys- not only obligatory, but a crucial the many exceptions to the rule
tems, however. (Intro. 1.3, lines element in achieving the aims of may in fact be standard practice
15–16) the introduction. in a certain discipline.

CONCLUSION Ultimately, the CARS model is


The aim of this paper was to an- intended to be used as a pedagogic
swer the question: How well does tool in the classroom. If the limita- In order to effectively interpret the
the CARS model accurately ac- tions of the model are understood, current descriptions of RA intro-
count for the writing of intro- then I feel it can be used effectively. ductions, and indeed, all aspects of
ductions in software engineering? The danger, of course, is that many the research article, it is important
From the results above, we can teachers of technical writing, com- that their range of applicability and
see that in terms of describing ing from backgrounds unrelated to limitations are understood. This
the overall framework, the model the target discipline, will be unable can only be achieved by concen-
is very successful; only the clas- to correctly interpret the model trating more on determining the
sification of definitions and ex- and inevitably use it “as is.” This unique features of writing that are
amples into an appropriate step is common in Japan, for example, found in different scientific and
was missing. Problems emerge, where teachers with an English engineering disciplines.

REFERENCES
[1] T. Dudley-Evans, “Genre analysis: An investigation of the introduction
and discussion sections of M.Sc. dissertations,” in Talking about Text:
Studies Presented to David Brazil on His Retirement (Discourse Analysis
Monograph no. 13), M. Coulthard, Ed. Birmingham, UK: English
Language Res., Univ. Birmingham, 1986.
[2] , “Genre analysis: An approach to text analysis for ESP. Advances
in Written Text Analysis,” in Advances in Written Text Analysis, M.
Coulthard, Ed. London and New York: Routledge, 1994.
[3] H. A. Estrin, “How to write for scientific and technical journals,” IEEE
Trans. Prof. Commun., vol. PC-25, pp. 32–33, 1982.
[4] S. Kushner, “Tackling the needs of foreign academic writers: A case
study,” IEEE Trans. Prof. Commun., vol. 40, pp. 20–25, 1997.
[5] J. M. Swales, Genre Analysis: English in Academic and Research
Settings. Cambridge, U.K.: Cambridge Univ. Press, 1990, p. 95.
[6] M. M. Pierson and B. L. Pierson, “Beginnings and endings: Keys to
better engineering technical writing,” IEEE Trans. Prof. Commun., vol.
40, pp. 299–304, 1997.
[7] H. F. Lippincott, “Some tips for clear writing,” J. E. Harmon, “The
structure of scientific and engineering papers: A historical perspective,”
IEEE Trans. Prof. Commun., vol. PC-26, pp. 11–12, 1983.
[8] J. E. Harmon, “The structure of scientific and engineering papers: A
historical perspective,” IEEE Trans. Prof. Commun., vol. 32, pp. 132–137,
1989.
[9] M. J. Charney and J. H. Williams, “From start to finish: Approaches
to introductions and conclusions in technical writing textbooks,” IEEE
Trans. Prof. Commun., vol. 33, pp. 220–225, 1990.
46 IEEE TRANSACTIONS ON PROFESSIONAL COMMUNICATION, VOL. 42, NO. 1, MARCH 1999

[10] L. Malcolm, “What rules govern tense usage in scientific articles,”


English for Specific Purposes, vol. 6, pp. 31–44, 1987.
[11] S. Jacoby, “References to other researchers in literary research articles,”
ELR J., vol. 1, pp. 33–78, 1987.
[12] D. E. Adams-Smith, “Medical discourse: Aspects of author’s comment,”
The ESP J., vol. 3, pp. 25–36, 1984.
[13] E. Tarone, S. Dwyer, S. Gillete, and V. Icke, “On the use of the passive
in two astrophysics journal papers,” The ESP J., vol. 1, pp. 123–140,
1981.
[14] R. Weissberg and S. Buker, Writing Up Research: Experimental Research
Report Writing for Students of English. Englewood Cliffs, N.J.: Prentice-
Hall, 1990.
[15] S. Wood, “An examination of the rhetorical structures of authentic
chemistry texts,” Appl. Linguistics, vol. 3, pp. 121–143, 1982.
[16] A. Hopkins and T. Dudley-Evans, “A genre-based investigation of the
discussion sections in articles and dissertations,” English for Specific
Purposes, vol. 7, pp. 113–122, 1988.
[17] J. Swales, Aspects of Article Introductions. Birmingham, U.K.: Univ.
Aston, Language Studies Unit, 1981.
[18] C. Cooper, Aspects of Article Introductions in IEEE Publications, unpub-
lished M.Sc. dissertation, University of Aston, U.K., 1985.
[19] G. Crookes, “Toward a validated analysis of scientific text structure,”
Appl. Linguistics, vol. 7, pp. 57–70, 1986.
[20] T. N. Huckin and L. A. Olsen, Technical Writing and Professional
Communication for Nonnative Speakers of English. Singapore: McGraw-
Hill, 1991.
[21] Information for year end 1996, supplied by the Membership and
Circulation Marketing Manager, IEEE Trans. Software Eng.
[22] Personal communication with original authors of corpus research
articles.
[23] L. Selinker, “On the use of informants in discourse analysis and
‘language for specialized purposes,” Int. Rev. Appl. Linguistics, vol.
17, pp. 189–215, 1979.
[24] J. C. Richards, J. Platt, and H. Platt, Dictionary of Language Teaching
and Applied Linguistics. Harlow, U.K.: Longman, 1992.
[25] D. Nunan, Research Methods in Language Learning. Cambridge, U.K.:
Cambridge Univ. Press, 1992.
[26] M. Brooks, “In-depth interviewing as qualitative investigation,” in
Classroom Teachers and Classroom Research, D. T. Griffee and D.
Nunan, Eds. Tokyo, Japan: Japan Assoc. Language Teaching.
[27] Such an interpretation has since been supported by Swales (personal
communication).
[28] J. M. Swales and C. B. Feak, Academic Writing for Graduate Students.
Ann Arbor, MI: Univ. Michigan Press, 1994.
[29] J. F. Brown, Engineering Report Writing. Solana Beach, CA: United
Western, 1985, p. 13.
[30] R. A. Day, “How to write a scientific paper,” IEEE Trans. Prof. Commun.,
vol. PC-20, pp. 32–37, 1977.
[31] T. Brusaw, G. J. Alred, and W. E. Oliu, Handbook of Technical Writing.
New York: St. Martin’s, 1993.
[32] J. W. Gilsdorf, “Writing to persuade,” IEEE Trans. Prof. Commun., vol.
PC-30, pp. 68–73, 1987.

Laurence Anthony is a Lecturer in the Department of Information and Computer


Engineering at Okayama University of Science, Japan. He received the M.A. degree
in TESL/TEFL from the University of Birmingham, U.K., and the B.S. degree in
mathematical physics from the University of Manchester Institute of Science and
Technology (UMIST), U.K. His primary research interests are in discourse analysis,
genre analysis, and corpus linguistics.

You might also like