21

I have a tendency to overcategorize things when writing a paper. Currently I’m working on my thesis and I’m concerned with my level of subchaptering. Right now I’m already down to x.x.x.x and I fear going down a level deeper would look bad. I think it adds to the overall clarity of the paper to categorize everything and would allow for more efficient lookup later on, but it’s not clear to me to what degree this should be done. Are there any general best practices? Are there any rules of thumb you use?

My faculty does not have any clear policy regarding this.

6
  • 2
    What did your supervisors say? What conventions have previous theses in your department used?
    – 410 gone
    Nov 11, 2014 at 12:43
  • 1
    I have sent an e-mail to ask about the conventions in past papers. I think, however, they will give an indication of their personal preference, not what a good paper should have. That's why I wanted to get the opinion of others.
    – esper
    Nov 11, 2014 at 12:46
  • possible duplicate of Can the word "subsubsection" be used in a thesis?
    – Moriarty
    Nov 11, 2014 at 12:53
  • 2
    ^ I'm not sure to what degree the above is a duplicate. Most writing can be restructured to fit within an "x.x.x" (3-level) scheme logically, maybe with a very occasional dip to 4 levels. I've never seen five levels of headings before, and I struggle to believe that such a level of compartmentalization is necessary or logical.
    – Moriarty
    Nov 11, 2014 at 13:01
  • 3
    @Moriarty: Leslie Lamport, in his two papers on how to write a mathematical proof, recommends quite a deep hierarchy, noting that this is easier using some kind of hypertext than dead trees. Sometimes, I'm tempted to follow his advice. Nov 11, 2014 at 14:09

3 Answers 3

24

Many, if not most, journals specify a maximum of three levels. In a book where the chapter is the top level, four may be ok considering each chapter can be as extensive as a paper. The problem of having too many levels is that the headings disrupt the reading. A good sign of this is when you end up having one heading per paragraph. If that is the case, you can probably remove the lowest level of headings and try to make the resulting segments of text flow by inserting bridges that makes paragraphs into a coherent text.

In my experience having read numerous student reports, theses and articles, three is a good goal. It is rare that a fourth level adds much in terms of structure apart from perhaps helping the writer. In fact, I often recommend students to keep a more detailed list of headings to enable them to see the structure of what they are writing but under the pretext that only a maximum of three should remain in their final version. I stand by that recommendation.

5
  • 1
    Thank you. Could you provide me with a link to this policy written out? I could propose my faculty to adopt the same policy.
    – esper
    Nov 11, 2014 at 13:27
  • "low level" heading can be put in the margin so they don't disrupt reading.
    – Ian
    Nov 11, 2014 at 13:42
  • @esper You could give your advisor the link to this answer :) Nov 11, 2014 at 16:30
  • @AustinHenley Thanks, that sounds like a good option!
    – esper
    Nov 11, 2014 at 16:31
  • Not in the case of the person answering, but, in my experience, often these kinds of rules of thumb are statements by supervisors who are not the most motivated and not deep in the subject matter, who just superficially want a thesis that is easy to understand. Most of the time that is also found in their research. It's like the discussion of the 130 kmh speed limit on German motorways: generally not wrong, but it also limits people who don't feel like dawdling on empty motorways at night at only 130 kmh instead of 250 kmh. Aug 10, 2022 at 16:31
5

I find that a good heuristic for structural depth is visual and conceptual navigation. If you think of a paper as a collection of (reasonable length) paragraphs, then for ease of navigation its structure should generally be a balanced tree with roughly 2-6 subunits at each level. Bigger than that, and it starts being hard to navigate, smaller and it starts feeling unorganized or pointlessly subdivided.

It is also important for each level of the structure of a paper to be a relatively even partition (possibly excepting the introduction and conclusion, which may be much smaller). If you find yourself with some sections much bigger or much smaller than others, then you may want to rethink your structure (e.g. should "Results" and "Discussion" be separate, or combined into "Results & Discussion").

Put the balancing and branch restriction heuristics together, and you've got a natural control on depth. One exception: some journals require a particular set of section headers, which may force the top layer of your tree to be unbalanced; you can still apply the heuristics for subsections and beyond, however.

2
  • 1
    Four is okay. Five is getting a bit dubious. But six!? Please, not ever.
    – Moriarty
    Nov 11, 2014 at 14:11
  • 1
    @Moriarty I'll argue it can be within the bounds of plausibility, particularly at the top level or in a very large manuscript. Three, of course, is a wonderfully resonant number...
    – jakebeal
    Nov 11, 2014 at 14:24
0

If you think that a deep structure adds to clarity, then it is ok to use it. I think, a thesis (PhD) should be fine with x.x.x.x.x. If your supervisor finds otherwise, he should tell you about this - or ask him.

A good supervisor will not just write: max structure is x.x.x.x. Instead, he would give you recommendations on where to change what and why - based on his opinion and maybe his experience (which is only reliable if he has written more than a master thesis a year before you).

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .