When to Write What?

When you write your thesis, your of level of knowledge of its chapters changes. It increases as you read, implement and experiment. It also decreases, however, as you move on to new parts and your brain throws out old stuff to make room for new stuff.

If you write a part too early, you have to rework later to accommodate for the new knowledge; or delete pages, as you learn that the topic is far less important that you thought two months ago. If you write too late, on the other hand, you need effort to rediscover the forgotten information.

Both rework and rediscovery waste time. Choosing when to write in which form is key to minimize waste.

I differentiate between two forms of writing: prose is the form of text of the final document. It adheres to rules of grammar, spelling and style. It is aimed at the thesis reader. Notes are its predecessor. They do not adhere to grammar, spelling or style. They are aimed at you, the thesis author, to simplify the creation of prose later on.

To avoid waste, write notes early and prose late during a thesis project.

Take notes on anything important you are likely to forget:

1. Summaries and noteworthy points from research papers: you read related work early in your thesis project. You know most about a paper right after you read it. As soon as you move on, it will fade from your mind. Write notes to help you quickly find a paper and remember its gist when you write the related work section’s prose. (Personally, I hand-write notes on the paper printout.)

2. Useful definitions you come across are often hard to find later, when you want to reference them. Collect definitions that appear useful in a single document, together with a reference to their source. Add any that might be useful. You can always delete them later on.

3. Insights, decisions and problems: during the evaluation or case study, you typically make numerous decisions and solve many problems along the way. Note them down so that you remember them when you write prose. Again, you won’t need them all, so don’t polish.

Take these notes in a lightweight document you feed comfortable with. It should be as easy as possible to add them and shuffle them around. Style, layout, spelling, grammar, etc. do not matter. At all. For me, text files or Markdown work well. Word too. Latex does not.

So how can you tell when it’s time to start writing the prose of a chapter? There are two indicators that it’s too early:

1. You lack central information on a chapter. Until you have not finished your measurements or developed your opinion, you cannot write about them. This holds for chapters such as approach, case study or evaluation. In other words, the core contribution.

2. You don’t know the exact role a piece of content plays in your thesis. This holds for the supporting chapters of a thesis–introduction, related work, future work–until the core contribution is done. This point is more subtle: since you feel that you know the content of the introduction already, why not write it now? If you do, its size and connection to the rest of the prose is based on speculation. Speculation equals throwing away text later.

One of these indicators applies to each chapter in a thesis, until the main body of work (implementation, experiments, evaluation, …) is finished. It thus applies to all of the prose in a thesis.

To me, writing prose for any of the chapters early–before the core contribution is finished–is an anti-pattern I call premature writing. It is related to the programming anti-pattern premature optimization, where you invest time into optimizing code early, when it’s unknown if it is really important for system performance. It wastes time you lack later to optimize the system when it’s finished (and you can tell where optimization is really necessary). Premature writing affects your document similarly, since it sinks your writing time in sections that you probably don’t need, or don’t need to this extent.