I say the same thing to all my students, and to everyone who comes to me to be on a committee. I could have sworn I’ve written it down before, but here it goes again.
Note that some of what’s here is just general advice on how to write any paper / proceed on any project. Its just that since a dissertation is bigger, it becomes more important to do these things.
Also: this is my opinion. My opinion might matter to you (if you want me to sign off). But you also need to make the other members of your committee happy.
Update 8/8/2008: Added “Expected Contributions” Discussion
1. Terminology Thesis vs. Dissertation
While the terms “thesis” and “dissertation” are used interchangeably they really differ in meaning in an important ways. Over time, colloquial usage is making common usage equivalent, but I think the difference is important, so even if the words aren’t used this way nowadays, the concept is good to get into your mind:
- A dissertation is a long document discussing a topic. To quote a dictionary: a treatise advancing a new point of view resulting from research; usually a requirement for an advanced academic degree
- A thesis is a statement to be defended (see What is a Thesis? below). To quote a dictionary: an unproved statement put forward as a premise in an argument
To get a Ph.D. you need to write a dissertation (a long document). That document should (or must, to get me to sign) present a thesis sufficiently well that your committee decides you deserve a Ph. D.
Here at Wisconsin, there is a milestone we call the prelim (short for preliminary exam). At CMU we called this a “thesis proposal.” I like the latter terms as it really explains what is going on. I use the terms somewhat interchangeably.
2. Now on to the real advice
First, read Olin Shivers' advice:
It is my opinion that in order to get a Ph. D. you need to write a dissertation that presents a thesis. My opinion might matter to you: I won’t sign off as a committee member unless I am convinced you have done this.
To me, the reason you do a Ph. D. dissertation is to prove to your committee that you are able to do a coherent, significant, and independent piece of research work, and therefore are worthy of the “credential” of a Ph. D.
Presenting your work as a thesis is an effective way to present it in order to make that proof. Specifically, I don’t think your Ph. D. dissertation should be “I did a lot of hard work for 5 years, let me out.”
3. What is a Thesis
A thesis is a statement to be defended. One you can state your thesis, then it is clear what you need to do: show that it is true or false. And show that it is a “good enough” thesis to warrant giving you a Ph. D.
Your thesis should be stateable in a sentence that begins “It is my thesis that” or “In this dissertation I will show that.”
In proper scientific writing, you should say it:
It is the thesis of this dissertation that X.
I need to be able to find that sentence in your introduction (in both your prelim and your dissertation).
A thesis must be a statement that can be true or false, and must be shown to be one or the other. How much work you did is not an issue. So this is a thesis:
It is my thesis that the sun will rise tomorrow morning.
This is not a thesis:
In this document, I will explore planetary motion and describe a set of experiments that illustrate important properties such as sunrise.
4. What makes it a Ph.D. Thesis
In order to make a dissertation, you need to have a thesis that is “good enough” so that if you defend it, you deserve to be called “Doctor.”
Here are the properties I think it must have:
- It must be a thesis - it might be true or false.
- It must be novel. Whether its true or false is/was not known to be true before your work. Demonstrating the thesis adds to the world’s knowledge. That is, it can’t be obviously true or false, and no one could have shown it before. (this latter reason is why you need to do a related work section)
- It must be important - the world will be a better place if we know that your thesis is true of false. (for this reason, you will need to have a motivation section in your document)
- It must be provable/testable. When you’re done, you’ll need a way to prove that you’ve actually shown the thesis to be true/false.
- It must take “significant work” in order to prove it. We don’t want to give a Ph. D. to anyone :-)
A good Ph. D. Thesis has those 5 properties (thesis, novel, important, provable, significant).
Your dissertation is a document that does 3 things:
- It states a thesis
- It shows that its a good thesis
- It shows that the thesis is true or false. Generally, this is the majority of the thesis.
It might do other things, like give background so that the reader can understand #3.
Having a thesis is critical because it defines (operationally) when you are done: if you can prove (or disprove) your thesis, then you’ve done enough work.
Your thesis must make a contribution (or multiple contributions) to the field of computer science. You should be explicit about what these contributions are (there should be a list summary in the intro, and they should be reviewed at the end). In the prelim, you should be explicit about what you expect the contributions to be.
Not everything you do will be a contribution. In fact, it is unlikely that the artifacts that you produce (e.g. the software you write) will be a contribution. Its work you do to support/document/prove the contribution, but usually not a contribution unto itself.
5. The Prelim
The “Preliminary Exam” is a right of passage. At most departments there is something like this (at CMU we called it a “thesis proposal”).
Basically, the idea of a prelim is that you will explain what you will do for a dissertation, so the committee can decide whether this will earn you a Ph. D. or not. Its about setting expectations.
You can view the prelim as kindof a contract: If you do the work you propose and write it up correctly in a dissertation, then the committee will sign off. They can’t (or shouldn’t) to say “sorry that’s not enough.“The prelim sets up sufficient conditions for you being finished, not necessary ones. If you find out that things change and you do something less/different, the committee may still sign off.
So, for example, a prelim is a great place to say something like:
Then, if the committee signs off on the prelim, then they can’t reject your dissertation if you only provide simulations. Of course, if you do built the real thing, then its probably OK.
An important aspect of the prelim is that it sets the expectation for what will be “enough” to be “done”. You might do this by explicitly listing the work - for example “my dissertation will have the following 4 experiments…”. I prefer to define “done” by the thesis.
For me, the main thing a prelim document does is shown that your thesis is a good thesis (or will be once you do it and write it up in a dissertation).
So your prelim document/presentation must show that you have the 5 things a good thesis must have:
- State the thesis that you will work on. I must be able to identify the sentence.
- Provide a thorough enough discussion of the related work to both show that your thesis is novel (i.e. not known to be true/false already). (see related work below).
- Provide enough motivation that the committee knows why the thesis is important.
- Show that the thesis is provable by explaining how you will prove/test your thesis to show that you are done. This is a big piece of expectation setting.
- Give a sense of why the thesis is significant by explaining why the problem (demonstrating the thesis) is hard, and what you intend to do about it. You may not know exactly what you are going to do, but you should give indication that you have a plausible plan, and that its feasible to succeed in a reasonable amount of time.
With the prelim #5 has an additional aspect: you need to show that your thesis isn’t too hard. You need to convince the committee that you have a high likelihood of being able to prove your thesis in a reasonable amount of time.
In my mind, everything else in a prelim is optional. Note: your other committee members might have other ideas of things that have to go into the prelim. For example:
- You need to demonstrate that you have fluency with the literature in the field (by providing a big enough survey of it).
- That you have a mastery of the required techniques (by writing a tutorial or survey)
- That you can write a big document (by having a length requirement)
I neither require, or even recommend these things unless some other committee member insists.
In your prelim document and talk, you should be explicit about what you expect your contributions to be. I recommend a slide in the talk, and a numbered list of expected contributions. They should be explained in the document, but its important to have a brief summary explicit and easy to find.
6. The Related Work Section
In my mind, there are two seperate issues in the thesis with regard to related work (in the prelim and the dissertation):
- You need to explain that your thesis is novel.
- You need to give enough background that your readers can understand your document. (by telling them where they can read more if they need to). Alternatively, you can view this as acknowledging the giants whose shoulders you stand on.
It is my preference that these two roles be distinct. And #1 is really whats essential.
Other people might have other expectations. At least one person I know wants a big literature survey in the prelim so you can demonstrate that you really know the literature in your field.
A dissertation can be a nice place for a good literature survey - when else do you get to take all the space you need to discuss the work done before. This is optional (in my mind). And should be done in a way that makes it clear what’s essential for the two things above.
7. Some History
Various forms of this document have been written by me over my career as a Professor. This version started around 2004. It got to its current form around 2008.
While I believe what is written here, I am much less picky about it than I was when I wrote it. It quite simply isn’t worth the energy to write a “great” thesis - your goal should be good enough. No one really reads theses.
My opinions on this were shaped when I was a graduate student at CMU. Some of this by my committee (Brad Myers and Bob Sproull really forced me to think about what a thesis is, and Andy Witkin and Paul Heckbert helped me make sure I actually wrote one). A lot of it also came from other students. Olin Shivers, Todd Rockoff (who told me “a thesis is a statement to be defended”), Scott Nettles, Greg Morrissett, …