How to Write a Research Proposal
So you have to do a research survey or a research proposal:
what are you going to do? The following is intended to give
suitable guidance as to what is expected. It is not, however,
exhaustive, so feel free to ask questions. Just read this first,
and then ask questions.
First note that a research proposal is a survey together with
the identification of an open research issue that needs to be
addressed, and some indication of how it can be addressed. If a
research survey is done well, it is often the case that one or
more open research problems become almost self-evident. However,
the simple statement of an open research problem does not turn a
survey into a proposal. For example, Fermat's Last Theorem was an
open research problem for nearly 350 years. Surveying the
foundational mathematics and stating Fermat's Last Theorem does
not turn the survey into a proposal. Some notion of how the
theorem could be proved or disproved would also be required.
The degree of detail in the description of how the open
problem can be addressed is one measure of the quality of the
proposal. At the very least, a proposed method for addressing the
problem should be provided. For example, packet-switched networks
cannot easily provide quality of service to their traffic flows.
A method for addressing this problem might involve a half-baked
idea (e.g. We plan to solve the problem first in a network
where everyone is benign, no nodes crash, no routes change, and
the only thing that can be reserved is bandwidth. We believe that
we can then aggregate the bandwidth flows into a single number
(current reserved bandwidth.) together with a suggestion for how
that idea might be proved/disprove (e.g. We will simulate
our half-baked algorithm using the Frobniac benchmark and the
WizzleWare network simulator.)
Now, about the survey portion: writing a survey is not
an easy thing. It requires you to have studied a substantial
portion of the literature in your chosen topic area, understood it
well enough to figure out the "landscape," and have the maturity
to explain the landscape to others. For good examples, go and
read some articles in the ACM Computing Surveys. Not all of them
are good, but most are. (As a UW student, they are accessible to
General pointers on writing a survey:
- Do not give a listing of papers with short summaries of
each. A document that says "paper 1 has done this, paper 2 has done
the other" makes for a very poor survey. What you should do is
understand the field and figure out what the alternative techniques,
algorithms, etc. are and organize your survey accordingly. A
very useful thing to develop when writing a survey is a taxonomy of
the field, with suitable orthogonal axes. A good taxonomy is hard to
develop, but well worth the effort in the understanding that it brings
to the field. Having done that, individual papers simply serve as
examples that clarify what you are saying, and fit into appropriate
places in the taxonomy.
- Do NOT directly copy material from papers without
attribution. This is called plagiarism. It is both an academic
offense and amounts to fraud. If you paraphrase something that is in
a paper, put a reference to that paper. If you take a passage
verbatim, put it in quotes ("...") and put a reference. Violations of
this rule are easy to detect because of the change in writing style.
Further, with Web search engines available it is easy to find the
source within a very short time. A strategy that may help you avoid
this danger is to never read and write at the same time. Do not have
the paper in front of you when you are summarizing the ideas it
- Pay close attention to your writing.
- USE A SPELL CHECKER. Far too many documents have spelling
errors in them. There is absolutely NO excuse for spelling
errors giving the wide availability of tools for checking spelling.
- Make sure the grammar is correct. Get a copy of "Strunk and
White: The Elements of Style." In a very thin volume it will tell you
all you need to know to write correct English.
- Make sure that you are communicating your ideas clearly. Just
because you understand what you have written does not mean that
someone else will understand. Just because your document is
grammatically correct does not mean that the ideas are presented in a
clear manner. Identify your audience and write accordingly. Finish
your survey a few days ahead of time, get some sleep, and the re-read
it carefully. Better yet, get someone else to read the document and
- Pay attention to your phraseology. A technical paper is a formal,
scientific document. There is no place in such documents for phrases
like "this is crap," "the idea is not worth anything," "deserves
applause," "way too much," etc. (you get the idea) are not to
be used. You will have read enough papers by the time you start
writing the survey to get an idea. Similarly, contractions
(e.g. don't, won't, etc.) have no place in formal
writing. A technical report should not read like a conversation.
- Take your WYSIWYG wordprocessor, throw it in the garbage, and get
a copy of LaTeX. WYSIWYG systems
cause you to focus on the wrong thing, viz layout. Layout is
the least important aspect of a technical document. Structure is the
most important aspect. Pleasant layout can never compensate for poor
structure. Systems such as LaTeX cause you to focus on structure, and
leave the layout to the typesetting system.
- Read (and follow) this
advice in research and writing especially the piece about
how to organize your thesis which can (with little modification)
apply to writing any good piece of research.
- Content: A typical research survey/proposal/paper contains the following:
- Title page, containing, at a minimum:
The "title page" may or may not be on a separate page, depending on
the nature of the research work. If it is a technical report, it will
be on a separate page (and will usually contain the technical report
number, the date of the report, and the name and address of the
organization for which this is a technical report). If it is a
conference or journal article, it will not be on a separate page.
- Affiliation (i.e. where the authors are from
- Abstract. A abstract is a 200--300 word summary of the paper. It
should state the problem that you are addressing and the major
results. Note that there are abstracting services that make these
abstracts available for others and people have to be able to read your
abstract and understand what the paper is. While you should have some
notion of where you are headed before you begin, you probably do not
know enough to write the abstract until the rest of the document is
- Introduction. This section is a crucial part of your
paper. Whether the paper is intended for submission to a conference or
journal, a proposal to a funding body, or simply a term paper in a
course, this is the section in which you have to get the reader's
attention. If this section does not get the reader interested in the
work, s/he will be reading the rest of the paper with a negative view,
seeking out every little mistake. On the other hand, if the problem
being addressed by the paper is clearly spelled out, and the motive
for addressing the problem is explained plainly, the reader is likely
to read the rest of the paper in a positive light, overlooking simple
errors, and focusing instead on the main results. Note that while you
may not like this manner in which reviewers comment on your work, you
should simply treat it as a fact of life, as sure as the law of
With these points in mind, this section should be written very
carefully. It must address motivation and contribution. More simply,
it must answer the questions:
Rather than attempting to create interesting literature, it is
probably best to state these points very simply and clearly. Thus,
there should be a sentence like "In this paper we
prove/solve/address/etc. the Frobniac problem ...." The rest
of that paragraph should expand on this statement. The context of the
research should be clearly identified. For example "The Frobniac
Hypothesis is of critical importance to Applied Philological Semiotics
forming the basis of many Semiotic Philological Security Systems in
use today." This would likewise be expanded upon in the remainder of
the paragraph. Finally, the contribution should be clearly stated. A
sentence such as "This document provides the following significant
contribution(s) to scientific and engineering knowledge" followed by a
simple enumeration of the contribution(s). It may also be helpful to
elaborate on these contributions. Note that even a humble survey
should have some contribution. This will typically be the insight
that your organization of the literature brings to a particular area
(e.g. if you have developed a taxonomy of the area, this is a
contribution). Note that if there is already a published survey of an
area, perhaps you should consider picking a different area.
- What problem does this paper address? (Problem)
- Why is this problem important? (Context)
- What is the significant contribution of this paper? (Contribution)
The critical point is that the three questions above are clearly
answered, and not lost in an overly elaborate description. This point
can be thought of as "Baroque is broke."
- Literature survey. There is little that can be generalized here,
unfortunately. The key points have already be mentioned above:
If you were writing an actual research paper, rather than a proposal
or a survey, then this section would typically be labeled background
or related work. (Note that some people will separate these two
concepts, providing only background material as required to understand
the research at this point, and giving a related work section
separately (typically near the end of the document). Either approach
is generally acceptable, though check the typical pattern for the
conference or journal to which you are submitting.)
- Do not give a sequential summary of papers.
- A taxonomy may be a very useful way of organizing a survey.
- Body of research. This is only applicable if this is a research
paper, and not a survey or proposal. Put in as many sections as
necessary to describe what you have done. Think about the organization
of both the sections and the material in each section. You are
explaining your work to someone who is not as familiar as you are with
the material. Think carefully as to how you want to describe the work
to them. For example, if there are a number of cases, it may make
sense to start with the simplest case and build on top of it. Also,
the sections should be linked together so that the reader can go from
one to the other without being surprised.
- Conclusion. This section is only applicable to a research paper
or a survey. It should summarize the main results, emphasizing what
the paper has achieved. While it is useful, perhaps, to start with a
cut and paste of the contribution paragraph(s) from the introduction,
do not leave it at that. No reviewer wants to read the same thing
twice, particularly in such a short time interval. Make sure you
reword, at the very least.
- Future Work/Proposal. Usually small in a research paper, and
often absent in a survey, this section is the key aspect of a research
proposal. At a minimum it indicates how the work can be extended in
the future. In the case of a proposal, it should identify how you can
take what you have done and turn it into a real research project
(either for publication or to solve some corporate problem) if you
were given the chance. As already stated above, a simple statement of
open problems is insufficient. Some idea of how those problems can be
tackeled is required.
- References. While this may be considered trivial, it is
frequently poorly done. Therefore, note carefully: each reference
If you are using LaTeX to write your surveys, BibTeX styles should
tell you what you need to include. (If you are not using LaTeX, see
the earlier point on the subject!) If you are still not sure, look at
ACM Computing Surveys papers to figure out what you need to do.
- Full set of author names (first name as well as last)
- Full (and correct!) title of the publication
- Where the publication has appeared
- Page numbers
- Volume and issue numbers (for journals)
- Year of publication
Consistency is very important. Do not use different names for the
same conference or journal. Do not use different names for the same
author (e.g. one reference uses middle initials and a different
reference does not).
It is often the case that you have found material on the web. Do not
simply use the BibTeX miscellaneous or unpublished formats and
reference the URL. Rather, find the conference or journal in which
the item was published. Note the assumption here that the item was
published in a conference or journal. If it was not, then it has
probably not been peer reviewed. While some such material is
acceptable, it should not represent a significant fraction of the
material that you use for a research document.
How long is a research proposal? The correct answer to this
question is "As long as it needs to be and no longer." Unfortunately,
this causes some students to produce great tomes of dense material and
others to produce fluffy dossiers devoid of substance. Neither are
appealing to read, though at least the fluffy ones have the virtue of
being quick and easy to read.
The only way to avoid these extremes, it appears, is to provide
specific guidance regarding the number of pages the document should
contain, together with font size and margins, etc. as this is
(regrettably) also necessary. The following seem to be about the
right amount, for either a survey of a proposal.
- 25 to 30 pages. Absolutely no longer than 30 pages without prior
approval. This includes everything (references, appendices, etc)
- 11 pt if you are typesetting in LaTeX and 12 pt if
you are typesetting in Word.
- 1.5 spacing
- 1 inch margins on all sides
Questions Pertaining to this Document
As a result of having this document on the web, I have started to
receive various questions. Since one person asked the question,
it is possible that others would also like to know the answer.
- Does it really have to be 25 pages?
Well, let's think about this for a bit: if I write a document that
is 25 pages minimum in LaTeX how long is that in practice?
Since this includes everything, it is really only 20 pages,
as there will be a title page, an abstract, possibly a table
of contents and a list of figures, and a page for references
(possibly two). If you use report style, and their are five
chapters, the average chapter will be four pages, with the
first page 3/4 full and the last page on average half full.
So that reduces 20 to 16.25 pages. At 1.5 spacing, that is
about 27 lines. At 11 point with 1 inch margins, depending
on the font you use, there will be around 15 words per
line. So 27x15x16.25 is more or less 6500 words. Is it
that hard to write 6500 words on a technical topic?
- Is there any minimum number of references we must use?
You should have all references necessary to cover the area you are
investigating. If that is a very large number, then you are
probably not focused enough. If that is a very small
number, then it may not be a topic of much interest to
people (and thus does not satisfy the significance
requirement for good research).
A Final Note
This is a work in progress. It started with an e-mail from Tamer Ozsu
to graduate students who had just submitted some (presumably poor
quality) surveys. Having faced the problem of graduate students
asking what was expected in a research survey or proposal, and
explaining the same thing many times, I was quite happy to get a copy
of this e-mail from Tamer, and have used it as a base with which to
create this document. I expect the document will change somewhat over
time, and would encourage those of you who have graduate students who
face this same problem to link to this page, or to e-mail me your own
suggestions that I may add them to this document.
If you still have questions
drop me a line.
Last modified: Mon May 2 14:08:12 EDT 2005