33

I heard this suggestion from a mathematics professor:

In writing a mathematical Ph.D. thesis, it is far more tolerable to be tediously-lengthy than having a gap in the proofs.

I think what he means is that whenever in doubt, adding more details to make the argument clearer is always better, even if sometimes doing this may make the proof too wordy.

Now if I really follow his advice literally, it seems there are too many details for me to write. For example, I don't even feel safe to write "the case n=1 is trivial" when proving by induction, or "by a direct calculation we have the following result". Another example is, whenever writing a commutative diagram (some of my diagrams are 3-D), I doubt if I need to prove that every small triangle or rectangle is commutative. (Actually I have asked this question but have received no answer so far.)

Thus my question is: How detailed should be the proofs be in a mathematical Ph.D. thesis?

David Richerby
  • 33,823
  • 6
  • 74
  • 143
Zuriel
  • 3,477
  • 1
  • 26
  • 40
  • 30
    My thoughts, though without much to back them up, is that your thesis should be aimed roughly at you-when-you-started-your-Phd ie someone with the right background who hasn't studied your field yet, rather than a paper which is aimed at the experts in your field. – Jessica B Feb 03 '15 at 11:37
  • @JessicaB, thanks for the comment! Actually I am quite confused by the example of a commutative diagram: should I always check that every small rectangle is commutative? I seldom see it being done in papers; doing this will make my proofs really long and tedious. – Zuriel Feb 03 '15 at 11:40
  • 4
    I seldom see it being done in papers. Me too. However, have you checked other PhD theses? – Nobody Feb 03 '15 at 11:45
  • @scaaahu Not yet... Are you suggesting that I should check that every small rectangle is commutative in a Ph.D. thesis? – Zuriel Feb 03 '15 at 11:50
  • 1
    In my field (not Mathematics, though), that kind of things on a paper would go into supplementary materials or an appendix. – Davidmh Feb 03 '15 at 11:51
  • A lot will come down to what your examiners think. If your work is very strong, they will probably require less thoroughness than if it is not so strong. Part of the aim is to ensure you know what's going on, so in a way it's a little like undergraduate exams. – Jessica B Feb 03 '15 at 11:51
  • 9
    Yes, you should definitely check that every rectangle commutes, but no, you probably don't need to write out all the details. I can't say this for certain, though, because the commutative diagram in my thesis was infinite, but I was able to succinctly say in general why each rectangle commuted. – Tara B Feb 03 '15 at 12:22
  • 8
    One rule of thumb I applied when writing up my thesis is: only say something is clear or obvious if you're confident that you could flesh out the details on the spot if challenged in the viva (defence). – Shane O Rourke Feb 03 '15 at 19:03
  • 1
    "I don't even feel safe to write 'the case n=1 is trivial' when proving by induction, or 'by a direct calculation we have the following result.'" You could of course write those in the "Proof Overview" with the note that "proof is worked in detail in the appendix." – Joshua Taylor Feb 03 '15 at 20:30
  • 1
    @TaraB I once tried to see if there was a commutative diagram in my thesis. I concluded that the definition of a fibred link does contain the composition of two maps. – Jessica B Feb 03 '15 at 21:10
  • 2
    I edited the title so it matched the question. The original title presented a false dichotomy between gaps and verbosity. – David Richerby Feb 04 '15 at 12:01
  • I think @JessicaB's advice to address the thesis to new PhD students (say at the qualifier stage) is solid. Would checking the small rectangles commute have been trivial to you in your first or second year? In general I expect much more exposition in a thesis than in a paper. Bear in mind that for current research, theses are often beginning grad students' best resources - act as if you are writing to them. Because in a way you are. – Sasho Nikolov Feb 04 '15 at 18:42
  • 1
    Regards "obviously", "trivial" etc, see also http://academia.stackexchange.com/questions/10797/the-use-of-words-such-as-clearly-obviously-etc-in-a-technical-paper – Keith Feb 05 '15 at 02:54
  • @ShaneORourke One rule of thumb that I give my students is that if they find themselves writing that something is "clear" or "obvious", that claim is almost certainly false. And I would shorten your advice: Don't write anything in your thesis if you are confident that you could flesh out the details on the spot in your defense. – JeffE Jun 25 '15 at 13:35

10 Answers10

42

I usually give the following advice:

If something is clear, just write it down clearly.

Of course, there is a caveat: What is "clear" clearly depends on the reader, and also what is considered a "clear explanation" certainly depends.

But the message is: If you are tempted to write "The case $n=0$ is obvious." think a minute about the explanation and how to write it down. If you then think that writing down the explanation does not add anything, leave it out.

But in general, I would say, being more verbose is the better option.

Another test you could apply is: Write down "obviously…" but note down a longer explanation somewhere else. After a week or two reread that part and try to recover the result again. If you don't have any problem then you are probably fine without a longer explanation.

Some resource that may be helpful for your proof-writing skills is the talk given by Leslie Lamport on proof-writing last year at the Heidelberg Laureate Forum: Here is a blog post about it and the whole talk is here.

Anton Menshov
  • 6,030
  • 4
  • 32
  • 54
Dirk
  • 38,078
  • 9
  • 96
  • 154
  • 4
    +1. Re your last paragraph: the relevant literature is Lamport's two articles How to write a proof and How to write a 21st century proof. Now, I am not worthy to argue with Lamport, but on reading these two papers, my reaction was "this will make papers more correct and longer, both by a factor of five." I am not entirely convinced the gain in correctness is worth the loss in conciseness, but then, I'm not a mathematician any more. – Stephan Kolassa Feb 03 '15 at 13:29
  • 1
    No need to make them longer: just hide the lower levels, dynamically in electronic format, or by placing them in an appendix in printed form, and using a proof sketch of shallower levels only in the main text. – 0 _ Mar 15 '16 at 06:49
  • 1
    "The case $n=0$ is obvious." --- Often in cases like this just adding a handful of words can enormously increase the explanatory aspect, on the off-chance that the reader is having some kind of mental block, or simply to lessen the reader's reading load. Depending on the situation, it could be "The case n=0 is vaccously true" or "The case n=0 follows from a simple calculation"--maybe even give the calculation, as in "The case n=0 follows from 0+0 = 0". – Dave L Renfro Sep 04 '21 at 04:08
  • @DaveLRenfro Very good examples! – Dirk Sep 04 '21 at 13:43
20

Think about who is going to read your thesis. My thesis had four audiences: me, my advisors (2), and my mom. Unless you're Tate or Serre you probably have a pretty similar number.

I wanted all the details in it so that I wouldn't have to sweat over them again when I rewrote it into a journal article. I knew my advisors could skim past what they thought was obvious. I knew the whole thing was going to be incomprehensible to my mom and so it might as well be long to be impressive. So I wrote it long.

Matthew Leingang
  • 1,582
  • 10
  • 16
19

The useful distinction is not between "long-winded" and "concise", etc. Verbosity per se is not helpful, nor is succinctness bad (it's good).

Wordiness does not automatically prevent gaps in arguments. If anything, it may merely obscure them. Terseness in arguments is not the same thing as "gaps in reasoning".

Yes, discussions can be shortened by deliberately omitting the least-interesting fragments of proofs or explanations. For experienced people, who trust themselves or are trusted to be able to fill-in "standard" gaps, this doesn't matter. Perhaps a novice should doubt to some degree their own capacity to distinguish "standard" from "critical" issues, and this distinction is exactly what other people will wonder about. In fact, a typical thesis may exactly be a protracted exercise in making sure that one can carry out all those (often eventually boring, not too dramatic) routine arguments "once in one's life".

paul garrett
  • 88,477
  • 10
  • 180
  • 343
2

Detailed enough that you could have followed it when you were a third year grad student.

Noah Snyder
  • 32,772
  • 7
  • 71
  • 128
2

You must be wary of "COIK." Clear Only If Known. In a thesis (dissertation?), one is generally too close to the content to spot jumps. Too much is better than the committee sending it back for multiple edits, i.e., they get lost at point A, so they send it back, you edit, now they get lost at point B, you edit, etc.

If some of the material is too tedious for the main text, however, one option is to dump that stuff into appendices.

jakebeal
  • 187,714
  • 41
  • 655
  • 920
gumbo
  • 31
  • 1
0

Saying it is "far more tolerable" does not mean it is desirable. Proofs should not be tedeously lengthy if they don't need to be. You supervisor is telling you that there is more tolerance in the assessment of a dissertation than in the assessment of a paper submission, which is true. This certainly doesn't invert the normal principles for giving a clear parsimonious proof.

Ben
  • 68,453
  • 9
  • 142
  • 263
0

It should be stronger than the Italian geometers who made entire theories which were false.

FourierFlux
  • 1,600
  • 2
  • 11
  • 15
0

This, as everything else, is going to depend on context. What role is this result and proof going to play in your document? Have you already published the proof in question elsewhere? How cool is it really? These are some questions that can drive your approach.

If this is as yet unpublished material, you can't afford to leave the details out, so they're going to have to go somewhere. If your whole thesis is this one result, put your proof(s) in the main body. You can still break the more tedious parts off (say by impromptu lemmas) and put them in appendices, if that helps your flow, but all the details should appear. If the things you think are trivial really are explainable with a few words, use those words.

If the thesis is about more than the result (how the result applies to other areas, how it applies to solve some other class of problems, did you just create a new branch of mathematics), then you can try to describe the proof in a way that relies more on intuition (or give an overview), not just so that laypersons can get a better grasp of it, but so that everyone who wants/needs to read it can get a good feeling of why it must be true, and then it becomes easier to explain its importance. Then you can put the hard, dry version elsewhere, possibly in another chapter ("Chapter 9: The Nitty-Gritty") or an appendix, but the details still must appear somewhere. Contrary to Big T Larrity in Code Monkeys, this is one occasion where you never want to leave 'em guessing.

Ken
  • 460
  • 1
  • 5
  • 6
0

Whenever a proof, a theorem or any other such thing can help the reader to understand any single step of the procedure then do write them down fully and even verbosely, if necessary. On the other hand, whenever it is only about mere technicalities that do not add any deeper understanding or insight it is best practice not to be too pedantic, for the sake of readability.

"Things must be made as simple as possible, but not any simpler." (A. Einstein)

gented
  • 777
  • 4
  • 10
-1

In general there are two aspects to this:

Overkilling It: You do not have all the time in the world. Therefore, you can't spend years on the thesis. As Long as you backed up the assumptions during the introduction and background section of your thesis; and you are able to connect the dots, during your contribution section, you should be fine. Because you did provide enough information (e.g., background material, citations, etc.) and then pointed out your contribution, so the reader can follow and do more readings on this based on your direction points.

Fault Assumptions With no Backup: In this case, you just wrote something to fill the papers; with assumptions that you didn't thought it through while writing your thesis. The experience academics look for these faulty parts of the thesis and will "grill" you during your defense, because they know you didn't thought it through; regardless how good the other parts of the thesis are.

So the rule of thumb here is as long as you know how to defend these 'gaps'; you should be fine and don't overkilling it because the clock is ticking.

o-0
  • 7,610
  • 1
  • 18
  • 35
  • 4
    This may be good advice for some fields, but I think the standard you describe is too low for a thesis in mathematics. Remember that the student is writing a long mathematical proof; it's critical that every step, no matter how small, is correct and can be verified by a reader. – Nate Eldredge Feb 03 '15 at 15:08
  • 4
    "As Long as you backed up the assumptions during the introduction and background section of your thesis;" I don't understand what that means for a thesis in mathematics. Could you give an example? – Pete L. Clark Feb 04 '15 at 06:14