Tutorial 14: How to actually write a paper

July 6, 2011

Matt recently told us how to get ideas for papers, but if you’ve not previously published, you may be wondering how you get from idea to actual manuscript.  I’ve written about twenty palaeontology papers now, not counting trivial ones like encyclopaedia entries and corrections (plus a few in computer science).  So while there are plenty of people out there with much bigger CVs than mine, I’ve accumulated enough different experiences over the last six or seven years that hopefully I can shed a bit of light on the process.  DISCLAIMER: this means I am going to be citing myself like crazy, and will look like a complete egomaniac.  That really is not the point of this exercise.

Before I plough in, a digression: you may legitimately wonder why, if I’ve written 20 papers, my publications page lists only fourteen.  A couple are in press but not yet out: my work on those is done, I just have to wait for the wheels to grind exceeding fine.  A few more are in review.  Others, though once completed, are now in the process of being revised, either in response to reviewers’ comments or because they were rejected outright and need retooling for submission to another journal.  Maybe the most interesting category, though, is that I have two or three papers that I think are dead: they’ve been submitted and rejected, and I think I will probably never resubmit them.  In two cases — dinosaur diversity surveys — the manuscripts have aged badly, because the rapid rate of new dinosaurs being named is rendering them more and more obsolete.  To bring these up to publishable standard again would involve rebuilding the database and redoing the stats, and I just can’t summon up enthusiasm for that work when I have other projects going on that are so much more fun.

Anyway, we’re not here to talk about how to abandon finished manuscripts — we’re here to talk about how to get them finished in the first place.

In my projects, I have used three broad approaches.  Let’s look at them in turn.

Approach 1. Gather notes first

If you take this approach, you’ll begin by gathering all your thoughts on the subject of the manuscript-to-be into one place — these days, most likely a single file or folder on your computer, but in the old days it might well have been a physical notebook.  Don’t think about the structure of the manuscript, or about narrative flow, at this stage.  Don’t worry about what to include and what to exclude: just gather everything you can, pour it into a pot, and stir it.  You can think about the other stuff later.

The idea here is to separate “left-brain” and “right-brain” activity, so you can concentrate on one of them at a time.  During the gathering phase, you’re being creative, an artist playing with ideas.  When you’re done, you switch into engineer mode, and your task becomes to synthesise some or all those ideas into a coherent argument.  It’s easier to think about one of these things at a time than both at once, so the theory goes.

Handy household hint: you don’t have to put all your ideas into a single paper.  Find a set of thoughts that fit together into a narrative, and build the paper around that.  The other ideas will find homes in subsequent papers, they’re not lost.

The right-brain-then-left-brain approach sounds good; but in practice, I’ve found this doesn’t work well for me.  In fact, looking back over my submissions, it looks like I’ve only done it twice, and both times it’s resulted in a huge amount of work.  Those two papers are the Taylor et al. (2009) paper on habitual sauropod neck posture and Taylor et al. (2011) on sexual selection of sauropod necks.  These were three- and four-way collaborations between myself, Matt, Darren, and for the latter David Hone.  And for such short papers (eight and twelve pages respectively) they took an amazingly long time to put together.  They went through long sequence of revisions and rewrites before reaching their final forms, and lead authorship swapped hands many times along the way — we all held it at one time or another on both papers.

So what was the problem?  Only this: that “synthesise all those ideas into a coherent argument” sounds like a straightforward mechanical process, but it’s not.  It’s an art in itself — it requires taste, judgement, and most of all a lot of hard work.  (And of course, this is especially true when working with a team, when everyone has different ideas.)  When you start the composition process using a big ol’ bucket full of observations, it’s hard to tie them all together in a sequence that makes sense; and sure enough, we didn’t.  Early “complete” drafts of the neck-posture paper contained all the same information as the published version, but they were incoherent and repetitive.  One moment you’d be reading about some assertion that Stevens and Parrish had made about ONP being habitual, then next you’d be reading something about semi-circular canal orientation, then it would be be some observation on extant animal behaviour, then it would be back to DinoMorph, and so on.  Reading it felt like being batted around inside a pinball machine.

We probably could have submitted it in that form, and found a venue for it.  But we didn’t, because we wanted our paper not just to contain a bunch of relevant facts, but to lay out an argument, a connected sequence of observations and deductions, that would tell a story, make a compelling case.  We wanted our paper to convince.  And doing that is an art — hence the many, many, revisions and rewrites.  We got there in the end, and all three of us are happy with how the paper came out, but it was a real hack to get there, and it left me wondering whether we’d gone about it the wrong way.

On the other hand, what other way is there to write a genuinely three- or four-cornered collaborative paper?  Most of the other collaborative papers I’ve been involved with have had a very clear lead author who contributed the bulk of the prose, with the remaining authors contributing specific passages of text – and of course other input, just as important, such as the discussions that gave rise to the project in the first place.  The genesis of the neck-posture project was that we each contributed a stack of notes — some about what has previously been written on the subject, some about the flaws in those assertions, some about the behaviour of extant animals — and I just don’t see a significantly better way of melding all those into a coherent narrative than the multiple-pass approach that we adopted.

So anyway, what I’ve mostly done instead is:

Approach 2. Just write a manuscript

There is something enormously empowering about firing up OpenOffice (or MS-Word, if you must), choosing File → New, looking at that brand new white page, and typing a title.  Once you’ve done that, you’re up and running.  You’re really doing it.  It gets much harder to procrastinate.  Even if you end up changing that title half a dozen times, and rejigging the order of the manuscript, and rewriting the conclusions, and retaking the photos and doing the figures again, and reworking the statistical analysis because new data has come in, none of that changes the fact that once you’ve started a manuscript, you’ve started.  It doesn’t matter if it gets three new heads and five new shafts along the way, it’s still the axe the George Washington used to chop down the cherry tree.

So with this approach, the idea is that after accumulating information and internalising it, you just sit down and start writing — telling the story in an order that makes sense and draws the reader in.  The liberating thing is not trying to use any of the actual wording of your notes, not feeling obliged to work them all into the manuscript, just writing.

A technique that people often recommend at this stage — and one that in theory at least I endorse — is not to bother with your citations and references at this stage, or even with boring typographical details like italicising your genus and species names.  You don’t want to let yourself get sucked into any of that detailed clerical work — it will break the flow of your thoughts, and prevent you from getting them all down in a sequence that makes sense.  You want to be writing in the same spirit that you would explain the ideas to an intelligent friend in a pub, after maybe the second pint, waving your hands wildly to get you through the difficult bits, but not worrying about that because the point is to get your idea across — or rather, your sequence of ideas, that gets the listener from A to B.  You can go back in fill in the references later.

Like I said, in theory I endorse this technique.  In practice, I don’t seem to be able to do it: when I start to write, the citations just thrust themselves into my mind and I’m not able to write a perfectly simple sentence like “the humeri of brachiosaurids are the longest known in any sauropods, exceeding 2 m in length” without shoving in a “(Janensch 1961)”, and nine times out of ten going and re-checking that paper so I can specifically cite the table on page 187.  Whether this is a good or a bad thing, I couldn’t say — maybe if I could discipline myself not to do this, I’d save myself the pain later in the writing process of having to shuffle the text to get it into an order that tells the right story.

A digression on story-telling

I’ve used the metaphor of story-telling a couple of times, and I think it’s absolutely central here. You want to draw your reader through the paper.

Of course, what we mean by “story” is very different from one paper to another.  For example, in my short paper surveying dinosaur diversity (Taylor 2006), the story could hardly have been different from how it turns out: here’s where the data was from, here’s what I did with it, here are the results, and then end with some discussion.  By contrast, there were lots of different ways I could have structured my plea to the ICZN to recognise electronic publication (Taylor 2009b), but I went for an approach where the section headings outlined the core argument even if you didn’t read the actual text: 1. Background: the availability of the name Darwinius masillae; 2. The Code is in danger of becoming an irrelevance; 3. Paper journals are going away; 4. The time to act is now; 5. Electronic documents are different from electronic media; 6. We must come to terms with the ubiquity of PDF; 7. The current rules are too hard to get right (and finally a Conclusion).

For a fairly hardcore descriptive paper like the Xenoposeidon description (Taylor and Naish 2007, natch), you’re more limited in how much of a story you can tell, and pretty much constrained by the usual Introduction, Systematic Palaeontology, Description, Systematics, Discussion structure.  But even there, we laid out what I think is a fairly compelling story by splitting the “description” section into two parts: one that was purely descriptive, and a subsequent one containing all the comparisons.  Only after those two sections did we progress to the phylogenetic analysis that weakly corroborated our inferences.

For my Brachiosaurus/Giraffatitan paper (Taylor 2009a), though, I subverted the usual structure by postponing the Systematic Palaeontology until after I’d done all the necessary descriptive work to support the generic separation, rather than presenting the systematic conclusion up front and then going back and justifying it.  I also gave that paper a very, very short introduction (116 words incuding citations and taxonomic authorities), pushing the rest of what would normally be considered introductory material back into a separate Historical Background section.  Why?  Because that way I could put the end-of-introduction subsections on Anatomical Nomenclature, Anatomical Abbreviations and Institutional Abbreviations up front on the first page where they belong, rather than buried on the sixth page as they would otherwise have been.

Well, it seems that I have have drifted a bit from what I intended to talk about, and got onto the subject of how to structure a paper; but since that’s sort of relevant, I won’t let it spoil my day.

Anyway, the two approaches I’ve discussed so far really bracket the range of ways to put a manuscript together, and most projects will fall somewhere on the continuum between them.  But every now and then an opportunity comes up to use a third way:

Approach 3. Convert from another medium

I already mentioned my paper on electronic publication (Taylor 2009b), but long-time SV-POW! readers will remember that much of the material in this paper was cannibalised from a sequence of SV-POW! posts (notably Non-Open Academic Publishing Is Dead) as well as a few comments that I’d left on relevant posts on other people’s blogs.  On paper, you’d say this is a lot like approach 1, in that while I had much of material to hand, it needed sorting, integrating and rewriting.  In fact, it went much more smoothly than the neck-posture paper’s editing process, perhaps because all the source material was my own rather than having come from several different authors; and perhaps because the posts and comments were already in a chronological order which reflected the way my thoughts had arrived at the position where they eventually landed.

But a more interesting example of this route is the survey of the history of sauropod studies (Taylor 2010).  This started life as a slideshow, the accompaniment for my talk at the conference Dinosaurs: A Historical Perspective.  (Yes, the very same talk which Fiona fell asleep in, when I rehearsed it on her.)  To put together a talk, you already have to have your story together — the sequence in which things happen, the sections that you chop it all up into, the references forward to things you’re going to say, and back to things you’ve said.  So transcribing that all down into a manuscript is surprisingly straightforward — at least, it was for me, for this project.  It really was, almost literally, as case of taking each slide in turn, writing a little essay about what it depicted, and moving on.

So I kind of recommend that.  In fact, I’d go further: do not ever give a conference talk without immediately transcribing your slides into a manuscript.  If you do, you’re throwing away a super-easy publication: you’ve already done all the hard work.

(I didn’t know I was going to say that, just as I didn’t know I was going to digress onto story-telling earlier.  Turns out that this is an essay in the literal French sense of “an attempt”, something that you only figure out as you’re writing it.  Now that I’ve said you should always turn your slideshows into papers, I find myself wondering whether I’ve taken my own advice …  Hmm, quick check of the old publications list and I see that, hmm, I have roughly three sets of unconverted slides.  So that gives me something to do in the evenings, then.)

Actual writing

I only have two things to say about this, and they have both been said better by other people (computer scientists, as it happens):

Say what you mean, simply and directly.

— Brian W. Kernighan and P. J. Plauger


Present to inform, not to impress; if you inform, you will impress.
— Frederick P. Brooks

In short, do not write “the taxon under consideration exhibits a tendency towards velocitous aerial locomotion” when you could write “it flies fast”.

A final thought

All I’ve done here is list and discuss what’s worked for me (and some of the things that haven’t).  If these things don’t work for you, don’t do them.  If you find a way that works better, then by all means use that.

But if you’re struggling to find a way to get started, then follow Approach 2 above.  Just start writing, and keep going until you’re finished.

Hope that’s helpful.


[Once more, I’m sorry that the reference list is so me-centric, but I had to use my own papers as examples because I don’t know the genesis of anyone else’s.]

16 Responses to “Tutorial 14: How to actually write a paper”

  1. You left out “Delete Angry Birds and any other browser game apps.”

  2. Matt Wedel Says:

    About halfway between the “gush it out without stopping for refs” storytelling method and the “transcribe each slide into a mini-essay” method is something I’ve done more of over time: write the sections of the paper without knowing in advance how they’re going to fit together, and then organizing them into a coherent structure at the end. I did this for both the 2005 book chapter and the 2007 paper on prosauropod pneumaticity. In both cases, my original outlines included more sections than I actually ended up writing as separate things-in-themselves. I started with a list of n big ideas I wanted to cover, but found that there was enough overlap between them that by the time I’d written, say, (n-3) of them, I’d said everything I wanted to say. In both cases I did a fair amount of Tetris revising: sliding sentences around in paragraphs, paragraphs around in sections, and sections around in the manuscript, until everything made sense (at least to me).

    Even if you write your paper by some other method, it’s extremely useful to set it aside for a few days when you’re done and then read through it in one go with fresh eyes. A well-organized paper just feels smooth; in one that still needs work, the out-of-order bits are jarring, like speed bumps. But with a little perspective it’s easier to see how to rearrange them.

    IIRC, Mike did a LOT of this sort of reorganizing on both the 2009 neck posture paper and 2011 sexual selection paper. He might describe what he did differently than I have, but the effect of what he did felt, to me, almost identical to my Tetris revision mode.

  3. Mike Taylor Says:

    Tetris is an interesting metaphor for this process — it’s odd that I’ve never heard you mention it before. Of course it doesn’t really work, because the point of Tetris is not just making things fit, but continuing to absorb NEW things, on an unreasonable schedule, which you can only do by throwing out parts of old things. A better analogy might be the Rubik cube.

    You’re right about putting it aside and coming back to it a few days after it’s “finished”. But it’s so hard not to just submit! This is one way that co-authors help you: having to circulate a draft enforces a cooling-off period!

  4. Nathan Myers Says:

    In the old days, people typed paragraphs on paper, then cut them out and re-arranged them until they exhibited some sort of flow, and finally stitched them together with transitional sentences. I’m not sure the new way is better.

  5. Matt Wedel Says:

    Well, it’s faster, kills fewer trees, and pokes out fewer eyeballs.

  6. You’re revealing our institutional secrets! Pulling back the ivory curtain!!!

    Expect the men in black suits to show up at your doors any day now…

  7. […] científica Eu li hoje um post simplesmente excelente sobre caminhos para escrever um artigo científico. Os exemplos que o autor […]

  8. Bobby Says:

    When I’ve gotten reviews back that indicate certain passages of my manuscript are too convoluted and require restructuring, I’ve printed those pages off before so I can see every page, and literally cut and tape them back together where it flows better as Nathan described – it’s fairly effective, and at the time my computer on campus was just too shitty for there to be an effective means of accomplishing that electronically.

  9. […]  I could say more about the nuts and bolts of writing and submitting papers, and I will do so in Tutorial 14.  But for now, I am leaving this here.  Because the single, simple point that this article makes […]

  10. […] — why not?  I know I can put sentences together, I know I can plan out a narrative sequence (I have to do this for my palaeontology papers).  Looked at objectively, this doesn’t look like something that I’m fundamentally […]

  11. […] explained by Mike here, there are several approaches to this, and all are valid, and none works smoothly. However, for the […]

  12. […] institutional abbreviations at end of the Introduction section, several pages into the […]

  13. […] himself by re-reading old SV-POW! posts (yes, we do this). He was struck by my exhortation in Tutorial 14: “do not ever give a conference talk without immediately transcribing your slides into a […]

  14. […] coming, and there are things you can do to improve your chances. Be aggressively curious. Write. Publish. Give good talks (and give lots of talks so you can become good at it). Broaden your skill […]

  15. […] you are, and you’d like to get your name into published scientific work (whether you pursue writing and publishing yourself or not), get drawin’, and upload those babies using CC-BY. Make sure […]

  16. […] we’ve touched on a similar subject in a previous tutorial, but today I want to make a really important point about writing anything of substance, whether […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: