Hints and Tips on (Science and Engineering)
Bachelor’s and Master’s Thesis Writing
Peter McMahon1
May 2009
1 Introduction
As the title suggests, this document isn’t intended to be a comprehensive guide to
thesis writing. Instead it is a collection of random pieces of advice that I would have
liked to be told before I began writing my first thesis. I think most of the advice is
applicable to students writing theses in all fields of quantitative sciences and engineer-
ing. However, none of this has been rigorously tested — everything presented is my
uninformed opinion and I’m sure would have benefitted me, but perhaps may be less
helpful to others. Caveat emptor, and good luck!
2 Strategies for Thesis Writing
Students who begin writing their first thesis often make a critical mistake: they start
putting down words immediately, without first planning their thesis. You should plan
first. Make a high-level outline. Once this is done, go through each chapter and add
subsections. I recommend that you then go through each chapter and add your figures
or placeholders for figures. I used this technique when I wrote my Masters theses, and
I found that it made writing much easier: the bulk of the writing I needed to do was
just that for describing each figure. By deciding first on what figures you will include,
you will have compiled a list of figures that you still need to create. If you begin by
preparing most of your figures, you’ll reduce your chances of being surprised by missing
data later on.
Write your “background” chapter last. The reason for this is quite simple: it’s very
easy to spend an enormous amount of time writing this chapter, and leaving it until
you’re done with everything else makes sure that if you spend too much time on it, the
presentation of your actual work won’t suffer consequently. It’s very tempting to make
your review of background material a comprehensive tutorial of your field, because it’s
fun to write and you get to look up and learn new and interesting things as you go
1
UCT ECE [Link]. 2006; EE, CS [Link]. 2008. Please feel free to send comments and corrections to
pmcmahon@[Link].
along. In contrast, presenting your results is much more difficult: typically by the time
you begin writing up, you know so much more about your work than the average reader
of your thesis will, that it’s a daunting prospect considering how best to coherently
present everything you’ve done in the best light possible. Don’t use your background
chapter to delay writing the meat of your thesis!
Write! At some point, you have to stop doing experiments. It’s tempting to keep work-
ing on your experiments because there are always extra things you can try. However,
at some point (typically at least one month before your thesis handin deadline) you
have to transition from doing experiments and analysis to writing up. As you write,
there will be times when you need to “quickly” do some extra work when preparing a
figure, or a table of results, but you should resist the temptation to do anything more
that isn’t absolutely necessary. Your readers would much rather read a well-written
account of your N −4 weeks of research in your thesis than a rushed and sloppy account
of N − 2 weeks of work (where N is the total duration of your thesis in weeks). Your
thesis report is likely the only means your reader will have to evaluate your work, so
make sure your writeup reflects the same dedication to quality that you put into your
research!
Until you start writing up, you won’t know what data or knowledge you’re missing,
so starting your writeup early will help you avoid a rush before your deadline. It’s
very annoying discovering a week before you need to hand in that you forgot to run
an experiment with some set of parameters, and you don’t have enough time to do
it with just a few days left. At best, you may have to submit your thesis with some
graphs missing data points, and at worst the missing data points would have com-
pletely changed your conclusions. Either way, this is undesirable and can be avoided
if you find out well in advance of your handin date that you need more data.
3 The Structure and Style of Theses
The best way to learn about appropriate style for thesis writing is to read some theses.
(See the next section for suggested reading.) Academic style is quite different from all
other forms of writing, even report writing, that you’ve likely seen before starting your
research. It takes a while to get used to. During your thesis, you should have read
many academic papers from journals and conferences — these are typically written in
a style very similar to that which is appropriate for theses.
There is some room for creativity in structuring one’s thesis, but good theses usually
have the same basic divisions and flow:
• Introduction – ∼10 pages. Explain the purpose of the thesis.
2
• Background (Related Work) – ∼20 pages. Provide a summary of the subfield
you’re working in, to provide some context for the reader to put your work in.
Discuss related work done by other researchers, and explain how yours relates to
it.
• Experimental Design – 20-30 pages. Explain the specific problems you set out
to solve, and describe the design of your experiments.
• Results and Analysis – 20-30 pages. Present your results, and provide analysis
of them.
• Conclusion – ∼5 pages. Summarize what you did, what worked and what didn’t.
Discuss potential future work that may fill gaps in your work, or approaches that
seem promising to overcome problems you encountered but that you weren’t able
to tackle.
• Appendices – Include any material that would break the flow of the main body
of your thesis in appendices. For example, appendices often include presentation
of non-essential data; long proofs; elaborations of ideas for future work; code
listings or other “methods” information (circuit designs and suchlike), and other
documentation such as user guides.
The structure and style of a good thesis typically looks very different to what is taught
in “professional communication” classes in many universities. This is unsurprising: in
most cases, professional communication lecturers are neither scientists nor engineers,
and have never written a scientific document, let alone a science or engineering thesis.
No wonder their advice is ill-suited for science or engineering writing! Any grammar
lessons that one obtains in professional communications classes is advantageous, but I
recommend that you ignore everything else that they teach about report writing.
If your thesis is mathematical, then I can recommend a guide to mathematical writing
co-authored by Donald Knuth. Stanford offered a course in 1987, “Mathematical Writ-
ing”, that was co-taught by Prof. Knuth, and a transcript of the course was released
as a technical report that makes for very pleasant reading. There are many excellent
‘before’ and ‘after’ examples; the course is packed with practical advice to make your
writing easier to understand.
URL: [Link]
4 Examples of Outstanding Theses
• University of California, Berkeley David J. Sakrison Memorial Prize for a Ph.D.
Thesis (2005): “Limits on Efficient Computation in the Physical World” (Scott
Aaronson). This is the gold standard by which all theory theses should be mea-
sured — it is a very fun read, contains excellent overviews of background material,
3
and clearly explains all the results that were obtained.
URL: [Link]
• MIT Ph.D. Thesis (2001): “Photonic Crystals: From Theory to Practice” (Steven
G. Johnson2 ). This thesis is the gold standard for all physical sciences and
engineering theses. It is witty, but gives the reader a very serious and intuitive
introduction to the subject matter, placing the achieved results in context in a
very effective manner.
URL: [Link]
Meeting the high standards set by Aaronson’s and Johnson’s theses is not a realistic
goal for most of us (although it is good to know what superstars of thesis writing can
achieve). The following theses are also examples of great style and structure, and are
somewhat more conventional than the gold standards. I have tried to pick out theses
from several different areas.
• University of California, Berkeley M.S. Thesis (2008): “RDLC2: The RAMP
Model, Compiler & Description Language” (Greg Gibeling). This is an out-
standing Master’s thesis that provides a very clear and comprehensive coverage
of the material.
URL: [Link]
• Association for Computing Machinery Doctoral Dissertation Award Honorable
Mention (2008): “Seeing the World Behind the Image: Spatial Layout for 3D
Scene Understanding” (Derek Hoiem).
URL: [Link] [Link]?pub id=5825
• Association for Computing Machinery Doctoral Dissertation Award Honorable
Mention (2007): “Conditional Graphical Models for Protein Structure Predic-
tion” (Yan Liu).
URL: [Link]
• MIT David Adler Memorial Thesis Prize for Best Electrical Engineering [Link].
Thesis (2007): “Capacitive Sensing with a Fluorescent Lamp” (John J. Cooley).
URL: [Link]
• American Physical Society Atomic, Molecular or Optical Physics Dissertation
Award (2005): “Ultracold Bosonic Atoms in Optical Lattices” (Ana Maria Rey).
URL: [Link]
2
Besides Prof. Johnson’s work in computational methods for investigating photonic crystals, he is
also known for creating the Fastest Fourier Transform in the West (FFTW). He has co-authored an
outstanding textbook on photonic crystals that is available for free online as an e-book, and is another
superb example of technical writing. URL: [Link]
4
• MIT George M. Sprowls Award for Outstanding Ph.D. Thesis (1999): “Mostly-
Static Decentralized Information Flow Control” (Andrew Myers).
URL: [Link]
• MIT William A. Martin Memorial Thesis Award for an Outstanding Master’s
Thesis in Computer Science (2006): “Probabilistic framework for genome-wide
phylogeny and ortholog determination” (Matthew D. Rasmussen).
URL: [Link]
• MIT William A. Martin Memorial Thesis Award for an Outstanding Master’s
Thesis in Computer Science (2005): “End-user programming for the Web” (Michael
Bolin).
URL: [Link]
• MIT William A. Martin Memorial Thesis Award for an Outstanding Master’s
Thesis in Computer Science (2004): “Tree Pattern Inference and Matching for
Wrapper Induction on the World Wide Web” (Andrew W. Hogue).
URL: [Link]
• MIT William A. Martin Memorial Thesis Award for an Outstanding Master’s
Thesis in Computer Science (2001): “Adaptive Delivery of Real-Time Streaming
Video” (Nicholas Feamster).
URL: [Link]
• MIT M.S. Thesis (1993). “Robust, High-Speed Network Design for Large-Scale
Multiprocessing” (Andre DeHon).
URL: [Link]
• MIT Ph.D. Thesis (1996). “Reconfigurable Architectures for General-Purpose
Computing” (Andre DeHon).
URL: [Link]
• University of California, Berkeley Ph.D. Thesis (2001): “Decoherence, Control,
and Symmetry in Quantum Computers” (Dave M. Bacon). This is another fun
thesis that demonstrates the possibility of presenting difficult, interesting material
in an engaging and entertaining fashion.
URL: [Link]
Many professional societies have thesis awards, and these can be a useful source for
finding other theses that have been recognized as excellent. For example, the American
Physical Society has AMO awards, and the Association for Computing Machinery has
a Doctoral Dissertation Award. The ACM award is given out annually to what is
deemed the best Ph.D. thesis in the world in Computer Science submitted in the past
year3 .
3
Most often the winners are from MIT, UC Berkeley, Stanford, Carnegie Mellon, and the University
5
5 Technical Tools
Take the time to learn LATEX — you won’t regret it! If you’re not very technically in-
clined, then you might like a WYSIWYG LATEX editor, such as lyx. However, you get
significant speed advantages from writing directly in LATEX markup. Another major
advantage of writing raw LATEX is that you can focus on the content of your writing
and worry about the layout later. When you see how your material is being presented
in a WYSIWYG editor, it is very tempting to continually tweak the layout, which
distracts you. With LATEX, you can write your text and then only worry about making
the layout look just the way you want it to.
Cambridge University’s Engineering Department has a comprehensive collection of
LATEX resources. I recommend you use this as your starting point; it has links to
both Windows and Linux installation/usage instructions and both beginner and ad-
vanced LATEX tutorials. URL: [Link]
Good figures are critical to a thesis. If you use unwieldy tools, you’ll become frustrated
and undermine your work by producing subpar figures. The following software is useful
for producing good-looking diagrams for theses:
• OmniGraffle (Mac OS; commercial) — hands down the best diagramming soft-
ware available, but unfortunately restricted to Mac OS.
• Microsoft Office Visio (Windows; commercial) — the best option for Windows
users who have free access to it. An unconvincing choice if you have to pay the
retail price.
• Microsoft Office PowerPoint (Windows; commercial) — can be used to make
appealing block diagrams, but is found wanting for anything more complicated.
• InkScape (Windows/Linux; free) — not bad, but has many flaws. Quite possibly
the best free diagramming tool.
Make sure you use vector image formats for your diagrams and graphs. For figures
that are intrinsically in raster format (such as photographs of experimental setups, or
image outputs of computer simulations), use the highest resolution images possible —
few things create a worse first impression for a thesis reader than the appearance of
pixelated images every few pages.
of Washington, but occasionally theses from outside of the United States are recognized. The Technion,
and Weizmann Institute in Israel, and Ecole Polytechnique Federale de Lausanne in Switzerland have
produced winners or runners up in recent years.