I am finding it challenging to transform an early draft into a finished product. My biggest challenges are putting the prose in the proper order and using topic sentences and transitions so that the reader can easily follow my train of thought.

The most effective methods that I have found for revising are to either rewrite from a paper draft or to reverse engineer an outline and then re-organize and re-write. This feels like it is overkill, but sometimes it is the only way that I can get my head around what has become a sea of words.

I am not sure if the answers will be specific to a particular type of writing, but I am a scientist writing journal articles and proposals.


I would appreciate advice on effective methods that can be used for revising, perhaps an algorithm. Is there a set of steps that I can follow each time I need to revise?


I don't consider this a duplicate of a previous question about editing, because that question and its answers focus on grammar and copy-editing, which is not my concern here.

An answer to another question provides good advice on shortening a text, so I have excluded this aspect from my question.

  • 1
    When putting a bounty on a question, please clarify why do you think the question is not answered yet. For me, Lauren Ipsum's suggestion is the way to go. You do not even have the problem Kate mentions (starting too early with the draft), because you can iterate through your document again and again, rearranging the sections. What is missing for you? Commented Apr 1, 2011 at 8:29
  • @John Lauren's answer is good, but does not provide me with a viable solution. First, it starts from before the first draft has been written, and second, it requires a specific software.I content - in some ways dependent on - my writing software (latex). Commented Apr 1, 2011 at 14:49
  • 1
    @John Lauren's method applies at the level of sections and subsections, but some of the challenges that I face in occur with organization at the sentence and paragraph level. For example, the draft may cover the same topic in different locations. These need to be identified and consolidated. It may have sentences and paragraphs that need to move from one section or subsection to another or be fully deleted. Should they be deleted or placed in deep storage? By this point, I also have larger issues with making the paragraphs flow from one to the next. Maybe the answer is to get to work. Commented Apr 1, 2011 at 22:55
  • 1
    @David - I think maybe we're having trouble because to some extent it seems like the answer is in the question. I mean - if you're having trouble with using topic sentences - just start using topic sentences, you know? Same with transitions - if you've identified the problem to that degree of exactitude, the solution seems obvious, based on what you've given us - add transitions! So maybe we need to know more about why you are having trouble with these tools. You seem to know that you need them - are you unsure of where, or how, or what, or...?
    – Kate S.
    Commented Apr 2, 2011 at 19:55
  • 1
    @John Smithers: for the record, I don't work for Scrivener. Maybe I should take the hint and start asking for commissions, though! :D Commented Apr 4, 2011 at 17:31

9 Answers 9


Without going back to an outline, you could do the following:

  • Print your draft on paper.
  • Take some fluoresecent colored pens and mark all the words that are specific to your paper. Not words like "the" and "research", but words like "unification algorithm" or "eukaryotic cell". Use different colors for different topics in your paper.
  • If you see the same words/colors in many remote places, move those sentences closer together (you're obviously back behind your computer now ;-). Ideally, each section talks about a single topic, so it should have more or less only 1 color. Since your paper was a "sea of words" to begin with, you lose nothing by moving things around. Never mind the "flow" for now: just the topics.
  • Now that the topics are "sealed" into sections, remove all duplicate sentences. Keep only the ones that give new information.
  • Order the sentences so that new terminology/ideas are introduced in a logical order, i.e. new words/concepts are defined before they're used.
  • From each group of sentences, make a temporary title and put it in front of the group. It may not make it as a title into the final paper, but it brings an outline back into the paper for yourself.
  • Now, and only now, start looking at the grammar and the flow. Turn each group of sentences into a few paragraphs right there underneath their temporary title. Make sure that each group expresses exactly the idea in their temporary title.
  • For better "flow", start larger sections with a summary of what you explained in the previous section, and then show the connection to the new section. You can quickly see what you explained in the previous section by going over their temporary titles.
  • At the end, take all your temporary titles and turn them into a summary. You can use this as the introduction, the conclusion of the paper, or even both if you rewrite it slightly.
  • Then remove most of the temporary titles, keeping only the "big" ones to structure the paper for your readers.

Regardless of which technique(s) you use from the many excellent answers, let us all know how it worked out for you! Good luck.

  • Thanks @Koen There is a lot of excellent advice here, but your 'algorithm' will be the most straightforward to apply. Have you used it? I will let you know how it goes. Commented Apr 7, 2011 at 22:17
  • @David No I haven't used this method, which is why I'd be interested in hearing how it goes ;-) However, we did use the colored markers idea to pull a high-level software design from an existing requirements document once. It went pretty well given the context (tight deadline, management in panic mode, etc). Commented Apr 19, 2011 at 20:42

It's hard to be too specific without seeing your writing, and I certainly have nothing close to an algorithm for you, but I was caught by your idea of reverse-engineering an outline. Does that mean that you didn't write from an outline to begin with?

If so, and if you're having trouble with organization (as I assume you are, since you mention order, topic sentences and transitions), then I'd strongly recommend that you START with an outline, and THEN write your paper. If you already have all of your ideas, evidence, etc. set up in the outline, then you just need to base your work off that and your organization should almost take care of itself.

Maybe I've misinterpreted, and you outline, write, and then re-outline? I'm not sure, but I'd definitely recommend an outline as the first step for anyone struggling with organization.

  • thanks for the answer. I do start with an outline, but after I convert it into a first draft, I don't refer back to it. Subsequent revisions include moving text around and adding parts that were not in the original outline and eventually it can get unweildy. I guess that I could keep the outline format further through this process. Commented Mar 26, 2011 at 23:12
  • 1
    Is it possible that you're writing your first draft too early in the process? What is causing you to move the text around and add parts that weren't in the outline? If you're still doing research or gathering ideas, I think it's too early to write an outline OR a first draft.
    – Kate S.
    Commented Mar 26, 2011 at 23:16
  • your insight is helpful - I probably do jump from the outline too quickly. However, questions come up during the writing and revising that require additional analysis, which then needs to be integrated into different sections of the text. And some sections, like the intro and outline, can be written in advance of having results. Commented Mar 26, 2011 at 23:42

This sounds like a job for... Scrivener! :D

Building on Kate's question:

  • Create your outline.
  • Write each section of your outline as a separate Scrivener document. Label them appropriately. (II/A/3, III/B/4/i)
  • Once it's all on paper, you can rearrange the sections however you need.
  • If you find yourself writing extra analyses, create new individual documents, and then they can be moved around at the end. You can add them to the existing outline structure, or append them in the middle. There's nothing wrong with an outline which goes I, II, IIa, III, IIIa, IIIb, etc. if it helps you keep track.
  • Compile the document and read it in order several times as you go. This will help you to see if sections need to be reorganized.
  • your advice is good, except that it implies an unfortunate dependence on platform-specific software (i.e. I don't have a mac). Is there anything that Scrivener does that can't be done with sectioning? Commented Mar 27, 2011 at 23:59
  • Scrivener has a beta for Windows. :) Okay seriously, you could technically do exactly the above in one or in several text documents (Word or the like) and just label the separate documents. The Scrivener UI makes it simpler, but you could even do this on notecards if you really had to. Digital text is what allows this method to work; the program is just icing. Commented Mar 28, 2011 at 0:13
  • thanks for pointing that out - there is also a beta for Linux, which meets my needs; the beta doesn't support Latex so perhaps I'll start using it on my next project, and perhaps Latex support will be in the pipeline. Commented Mar 28, 2011 at 0:26

The technique that has been most useful to me is to begin each section, paragraph, and sentence with information that readers already know, and move new information to or toward the end.

I learned the technique from Joseph M. Williams's terrific book Style, which is chock full of ideas about how write with clarity and grace. Similar is Martha Kolnn's Rhetorical Grammar. Despite the scary title, it's not about grammar per se, but about how to use the grammatical structures you already know to create the effects you want to create in your readers.


Just speaking for myself, I have found it most helpful and useful to

  1. be merciless to eliminate all of the words that I can, including especially those pesky adjectival embellishments;

  2. locate and expunge all of the multisyllabic Latinate words, replacing them with their Anglo-Saxon equivalents.

To apply those rules to the lines above, I do a first pass to yield:

"I write better when I throw away words I don't need; and when I write plain English."

And a second pass might give:

"I use the fewest and the plainest words possible."

And I keep iterating on those principles

"And I go over it again and again until it's right."

This refining process clarifies my thought so that leading the reader as I wish becomes almost automatic.

I'm a software engineer and a lot of my writing is technical description or software proposals. The more I apply these rules, the clearer it becomes to the reader. And to me!


Caveat: I assume you have a backup of your work (using a version control software would be also beneficial here).

Go through every section/sentence. When you find stuff that could belong to a different chapter, name or number it and put this name into a reference list for this different chapter. E.g. the chapter is about cars from Porsche, then name the references POR01, POR02. The names must be easily retrievable with the find operation of your editor.

When you got through the whole text, start moving the sections where they belong to. Pick one at a time from your reference list, use Find and keep moving (and probably rewriting).

If you are in doubt what to do with a sentence, paragraph or section, then delete it. If you find out the next day, that you really need it, retrieve it from your backup or write it anew. But don‘t let it remain in a hideous spot in your text. Get rid of the crap.

These lists are like mini-outlines for each chapter or idea you have. This is not resorting your train of thought(s) yet. Only taking one thought at a time and look what belongs to it.

For putting together the train of thought correctly (and the overall outline), you should keep in mind, that you want to convince the reader. (I use convincing in a very broad sense here. If you write a manual for a video recorder, you try to convince the reader to use the recorder correctly.)

To convince him, the reader needs to know all necessary preliminaries before you can make your point. So what has the reader to learn upfront? Put it in the beginning of your text. What is the point you really want to make? Place it at the end.

Now the real challenge begins. How do you get from start to end? This is what Lauren's answer is about. At this point you should be able to follow that advice.

If you have problems, how to bridge the gap from one paragraph to the following, I suggest, that you post the problematic paragraphs (two or three of them) in a second question. With a concrete example the community can help you much better there. And you get ideas how to solve the other paragraphs. If you stuck again, post them.

@Kate The first time around, I have a nice outline with a logical flow, the second time around, the flow has been lost, and I am interested in knowing the steps and order of steps required to put Humpty back together again. – David 14 hours ago

For the future: Your problem is that you start editing and fixing your second round with the prose version instead of the outline. That's why you keep getting into trouble. You have an orderly first draft which works from an outline, with nice transitions, and then you start adding things pell-mell without regard for how they fit. Stop doing that.

If you are reading through your first draft and you think of things to add, add them to the outline instead. Keep adding everything to the outline. When you have no more to add, then you start rewriting your second draft.

For now: Yes, you are going to have to reverse-engineer some of your work. Here's how:

Take every sentence or thought — two or three sentences at the absolute most — and make it a bullet point. Do this for the entire piece.

When all your sentences are bullet points, start putting them back into the outline. Add your III/a, IV/2/b/47 extra sections as necessary. Your outline is going to look like this, with the bold section representing something new added in:

I. Gardening
A. Planting
1. Choose your plants
• Decide whether you want to plant flowers, ornamentals, vegetables, or fruit-bearing plants.
1a. Start your own from seeds, or buy seedlings?
a. Benefits of starting from seeds
1) Cheaper
2) Educational
b. Benefits of buying seedlings
1) Who has the time to raise plants from seeds? I work for a living.
2) Seedlings are generally sold when it's time to plant them, so you don't risk getting them into the ground before they're ready.
3) Supports local business

• Buy your seedlings.
2. Read the planting instructions.
3. Choose a location.
a. Considerations
• Sun vs. shade
• Acidity of soil
• Watering: Are you depending on rain, or will you water regularly?
b. Sunlight

• Sunny
• Shady

Do this as many rounds as you need to. Once you're satisfied that you have no more to add, THAT'S when you start smoothing out to prose and adding transitions and topic sentences.

  • Hey, you didn't mentioned, that he can do that with Scrivener ;) Commented Apr 5, 2011 at 8:31
  • that's because in my first answer he said my solution was "too Mac-dependent" and that he only uses Latex. I confess I have no idea how one writes on rubber gloves. But for the record... you can do this in Scrivener. ;) Commented Apr 5, 2011 at 11:30

Getting journal/conference articles ready for submission is a little different than for research proposals, but it is easier to talk just about articles because of their standard format.

A journal article will typically exist in four states of compression: the introduction will be a precis of the whole article, and the abstract will distill the original contribution laid out in the introduction. The title should at least be suggestive the contribution described in the abstract. The shortness of the these stages of compression is inversely proportional to how often they will be read: titles and abstracts are very widely read; introductions are read by those interested enough in the abstract to want to read more, but most people who read the introduction through will not care to get to grips with the body of the work. Each of these more compressed stages can be seen as an invitation to progress to the next stage.

Your problem sounds like you do not see how the content at the last stage relates to one of the earlier stages. There are several ways I can think of that may cause this:

  1. You have certain things you want to say in the body of your article that don't relate particularly well to your abstract. For instance, suppose you are writing about a drug trial of a therapy for some disorder and you happen to know a lot of interesting history of the treatment of that disorder. Your text might be full of discussion of changes in best practice that aren't obviously related to the drug trial that you set out to write up.
  2. You might set out to write up your demonstration establishing some fact, and then discover that your demonstration was not as watertight as you thought, or that the argument is immensely tricky, hard to follow, or tedious when written up in full. Then writing up starts to involve new research, and the text evolves in an unexpected way as part of a process that does not seem to end.

In each case, you can impose sanity by turning to the abstract. In the first example, you need to decide if your historical part is going to be part of the article at all: try writing two alternative abstracts, one including it and one that doesn't, and decide which will make for a better article; perhaps you will end up with abstracts for two articles. In the second case, you should be asking whether you can find an abstract that will give a worthwhile article without need for further research, and if not, it is time to stop writing up and start deciding what the research problem is that should be working on.

A couple of asides: If you are going to be breaking up text and putting sections of text on ice, distributed version control software, such as git can be useful, although the learning curve for this software is (still) steep. Then, while many guidelines to writing research papers suggest up to 300 words (more words than are normally found on an undecorated page) for an abstract, abstracts should very rarely be more than 150 words. The shorter the better, provided the constraint about all significant contributions being recorded is respected.

  • Thanks for the excellent answer. I currently use bzr for version control, and you are correct that it is an excellent tool. Commented Apr 8, 2011 at 3:34

All of the answers to date are good. You may also need to review and change your original outline and your train of thought. Its possible that you may simply be attempting to put a quantity of ideas and supporting material in one paper that deserves two or more papers.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.