The Hierarchical Structure of Bad Writing

Over the past couple of months at Nature Network, there have been a variety of discussions about good writing (these discussions seems to have started shortly after I arrived. Are they trying to tell me something?). One of the early well written papers was commended for the quality of the language, which was simple and clear, and so enjoyable to read. Unfortunately, many papers have sentences that seem to be written with a different aim: to loose the reader in a whirl of syntactic complexities before finally beating them over the head with a dangling modifier, painfully.


This is not, alas, the only way a paper can be badly written. There is a whole hierarchy of levels of structure in a paper, and lack of attention to any of the levels can make it much more difficult for the reader (or referee!) to work out what’s going on. Not only can sentences be hopelessly mangled, but paragraphs and whole sections can be too. Indeed, with some care the whole paper can be given a structure that makes it difficult to follow. Journals like PNAS, who insist on their papers going straight from the introduction to the results without explain how the results were obtained, even enforce this.
The poorly-constructed paragraph can lurk anywhere, but is perhaps most likely to hide in the methods. It can express itself in a couple of ways, with opposite effects on the reader. In one phenotype, the paragraph meanders around the subject for a page and a half, showing all the necessary information, but only as an amorphous shoggoth which induces nightmares of you know where (the journal editors reading this have already had to make a -30 SAN check). I was deeply tempted to illustrate poor structure in writing, but I shall resist. Another way a paragraph can go wrong is when it can set off to explain a point, only to shift without warning onto a completely different subject. This causes the reader to jerk awake when they realise that the software manual they are reading has shifted to describing a Chinese train journey1. Evidently a couple of carriage returns went missing.
Disorganisation can strike at a higher level too. Sections, particularly introductions and discussions, can sometimes give the impression that they have been thrown together with the help of a d20 and a failed intelligence test. Like the shoggorific paragraph, an introduction can include all of the information needed to understand the paper, but presented in a totally arbitrary order. This leaves the poor reader confused as to the point of the paper: does it exist for any reason than because their boss wanted them to write it? Sectional schizophrenia is probably less common, perhaps because anyone capable of holding to one train of thought in a section is unlikely to want to hold to two. I suspect they can be found, and would predict they will mainly lurk in co-authored publications.
Why does this happen (a question Martin asked too)? I suspect part of the reason is simply that, as scientists, we are encouraged to write rather than to write well. Emphasis is often given more to putting the information down on paper. But on top of this, I think there is a problem that the organisation of a paper is often described in static terms: the introduction gives the back ground, the methods say what was done, the results say what were seen, and the discussion says why this is all so wonderful. The emphasis is in making sure everything is in the right place. Although this attention to content is desirable, emphasising this aspect of paper writing does, I think, make it more difficult to write good papers. I think using this description of a paper made my early writing difficult, so it me several years to work out how to write an introduction (and I’m still not entirely comfortable with discussions yet!). I slowly dropped this old model, for a different one.
The model I now use for writing papers is more dynamic: I think they should tell a story. The introduction is a way of walking the reader up to what was done, so that they can see why it was done. Take them on a short trip from where they are to where your work starts. All else is unnecessary detail. Then the methods and results are the meat of the action, before the discussion digests what has gone on, and the moral of the story is explained.
The same story-telling approach can also be applied at a different hierarchical scale. If we are telling stories, we should be aware that there are different ways of narrating2. Last week I was revising a manuscript which contained a verbal description of an algorithm for estimating a parameter. Our previous description set out the algorithm linearly, starting with the initialisation process and describing it one step at a time. This made it really difficult to follow what was being done (and why), so I split it into two stories: one small one (what happens each iteration), and one big one (how we go from one iteration to the next, including how we get to the first step). I think, well hope, it helped: reflecting the hierarchy in the algorithm made it easier to split the description into explanatory chunks. My point here is that it is a different way of describing the process, and when a description is difficult it is often worth trying a different was of telling the story. Sometimes it may take several attempts until it makes sense.
As I’ve already pointed out, the area still have trouble with is the discussion. There can be several points that need to be raised, and it can be difficult to do this and still keep a flow to the paper. It can be difficult to find a single logic that connects the threads. One trick is to use sub-headings: then the different sub-sections don’t need to flow together.
I also have difficulty working out how to finish articles.

1 Yep. It’s happened. See the VCE manual for details.

2 Like suddenly switching to an active voice. I dunno why it was done either.

About rpg

Scientist, poet, gadfly
This entry was posted in Uncategorized. Bookmark the permalink.

11 Responses to The Hierarchical Structure of Bad Writing

  1. Sabine Hossenfelder says:

    The problem is amplified by inappropriate advices. E.g. where I worked people used to tell each other, include bulleted or enumerated lists because it looks more structured (note, not because it is more structured). Result was, I’ve seen quite a number of papers that have completely unmotivated and meaningless lists in them, like the authors just took a paragraph and made it into a list. Other advices were of the kind, make plenty use of words like ‘Thus’, ‘Therefore’, and ‘Nevertheless’, whether appropriate or not, because it looks as if you’ve concluded something. Futher there is the bad habit of calling things ‘simple’, ‘trivial’ or ‘obvious’ to make the reader feel stupid for not getting it in the hope there won’t be any questions. It would have been more useful to just pass around a book on good writing, preferably one that doesn’t contain too many words. But maybe some enumerated lists.

  2. Henry Gee says:

    Bob – I take your point about the structure of the scientific paper being inimical to the telling of what should be a story. I’d never realized this before, but you’re quite write right. I’ve written innumerable articles on science and other things, and several books, but the most difficult thing I’ve ever had to write was my one and only scientific paper. I’ve since tried to write more, but I simply cannot fit the round thoughts in my head into the square holes that the format of the scientific paper demands. Apart from that first paper, I’ve never been able to get to the end of the introductory paragraph, let alone a first draft. There is something deeply off-putting about the way people are obliged to organize scientific papers. This seems like an appalling confession from a scientific editor, I know, but there it is.
    Why should scientific papers be so particularly difficult to organize? Am I making more of this than this problem merits? After all, organizing a book can be very hard and take a long time, and papers and books alike require a certain amount of discipline. It could be that since I left off being a researcher, I no longer care sufficiently about my (now very ancient) research work to put in the hours. And if you look at the citations of my one published paper, nobody else much cares, either.

  3. Maxine Clarke says:

    When I wrote scientific papers, back in the Devoninan (or was it the Jurassic?), I was taught to, and did, start with the figures. The number of figures you were allowed in the paper “told the story”. Then you added the results and methods, then the discussion, and then at the end, when it was all in a draft, write the introduction (including a summary). After that, revise, circulate, revise, circulate etc until nobody had any more criticisms.
    Then submit to a journal and start the “revise, circulate” process all over again but with a different set of people.
    So, has it all got more complicated since? Or am I oversimplifying it?
    I agree with Sabine that structure is often “bolted on”, but if structure is there because the paper has been built up from a set of “bullet points”, each one fleshed out, I think it works. (The “structure” in the final version can be a set of subheadings.)
    Final point: to what extent is(are) the author(s) writing the paper to fulfil some perceived obligation (a citation, a grant requirement) and to what extent because she/he/they want to communicate an exciting result? There is a differece in the readability of the end-result. I was very aware of this when I first joined Nature, as my job then was to edit all Articles (longer paeprs) and Reviews, whatever the discipline. Hence I edited a lot of articles by extremely prominent scientists, and I was quite shocked by the number of what I can only term “drafts” that were submitted, and had been accepted, for publication. This was a contrast with my own field, muscle crossbridge biophysics, a subdiscipline in which the process I’ve previously described was the cultural norm. I did observe that in general, papers in the physical sciences were more carefully constructed, and hence readable, than those in the biological sciences, but this is a very broad generalization, and there were plenty of exceptions in both directions.

  4. Cameron Neylon says:

    When I (try to) train students to write, I always ask them to start with a single sentence that explains the point. The sentence then gets expanded into a paragraph with a relatively small number of setences. Then each of those sentences gets expanded…repeat until appropriate scale/level of detail is reached. It is a way of imposing structure while at the same time saying ‘what does this paragraph mean’ and if it can’t be summarised in one sentence then it probably ought to be two paragraphs.
    Of course I don’t write that way at all…

  5. Brian Derby says:

    There is no correct way to write a paper because the context is important. Different sorts of results need different ways of presenting them and different journals have very different house styles. More scarily still, some Journals have page limits and that effectively limits them to a particular sort of paper. Finally and most importantly some people will only look at certain journals, because that is where they expect the good stuff to appear. Even with search engines, Scopus and WoK many people will not believe it is significant if it is in the wrong journal.
    In the physicy side of the physical sciences the structure is important. A classic style is to present the hypothesis (or theory/model) in a logical manner with all steps in the derivation present. This is then used to present results or predictions, which may or may not be tested against experiment. This tends to be rather long and so you won’t often see it in letter Journals where the tendency is to just to a bit – maybe a sexy result with a brief outline of the principles and then a challenge to prove/disprove it.
    The challenge is to make the argument understandable to a general audience yet rigorous enough to stand detailed criticism/inspection, especially if you are challenging an accepted view. I am working on a paper like that at the moment.
    In my experience the first draft often remains pretty much unchanged as the skeleton argument. What often happens is you anticipate rebuttals and tack answers to those on as you think of them. This can make the end product a little cumbersome unless you watch out.

  6. Cath Ennis says:

    Maxine, I’ve always started with the figures as a framework too. Then I’d go on to the methods (only those used to generate the figures) and results (only those in or directly related to the figures) before structuring the introduction and finally the discussion around those three sections.

  7. Bob O'Hara says:

    I also used to approach paper writing in the way Maxine describes, and I wonder if it’s part of the problem. It seems to presume that the purpose of the paper is to present the results of the research as a few new isolated facts, with the reasons why they are interesting coming later. The form is, again, rather static.
    Another way of looking at the purpose of a paper is that it exists to persuade the scientific community that some scientific statements (like “calcium is released from intracellular stores”) are more or less certain. So a paper should be an exercise in rhetoric, and the figures etc. exist solely to support the argument. The argument should then be built first, and the figures chosen to support the argument. I guess this means the introduction and some of the discussion should be written first!
    Hmm, more needs to be written about this, and I should also re-read Peter Medawar’s famous “Is the Scientific Paper a Fraud?” (pdf).
    Henry – had you had more experience of writing papers, you would have worked out how to create a narrative in a short section. Then you would have learned how to effectively write very short stories.

  8. Brian Derby says:

    As always, Peter Medawar has a way with words that explains with great clarity. I am not a biologist, so I cannot agree whether his analysis of the structure of a paper in the Life Sciences is still the case. In my style (noy in biology) results are discussed as they are presented. Is this normal style or is it to separate r & d?

  9. Bob O'Hara says:

    It’s usually a separate discussion. I’m in two minds about this – it’s useful to have the results as a distinct, compact section but it does break up the narrative. I guess what is best depends. As you wrote earlier:

    There is no correct way to write a paper because the context is important.

  10. Jan Wedekind says:

    I guess I was lucky for having a boss telling me early that “a good paper should tell a good story” – it’s definitely a good advice. After all, nobody would voluntarily read a bad story. Unfortunately, scientists more often than not force you to do just that because contrary to a novel you actually often need the information contained in the article.
    Still, I believe a good story isn’t necessarily hampered by the scientific structure, which in any case is just a loose frame and no absolute. My rules of thumb would like this:

    • start with the abstract (a bit what Cameron suggested above): it is the take-home-message of what you want to deliver in that paper.
    • write the conclusions.
    • write the introduction.

    It seems like a no-brainer, but quite obviously an abstract and the summary/conclusions should fit perfectly to each like two pieces of a puzzle. It’s amazing how often I find no trace of the abstract’s information in the conclusions.
    The introduction then basically is needed to set the stage for the story you want to sell tell. Anything irrelevant to a) the subject and b) abstract+conclusions should be cut out.
    I agree that it’s a good choice to then first do the figures (just like a comic storyboard) before writing.
    I find that when following this approach most things just fall into place by themselves and it is much easier to cut out paragraphs and pieces which are in fact superfluous or redundant.
    By the way, ever noticed how hard it is for some people to let go of anything they have already written? One should never forget that this is an iterative process. If you write your conclusions and notice there’s something missing in the abstract, then the obvious choice is to change it. Likewise, if there is something in the abstract you don’t need for conclusions, remove it. Unfortunately, I’ve met many people who believe their words in a computer are carved in stone and would rather stick to something irrelevant than making the paper better.

  11. Wayne Thogmartin says:

    d20 and failed intelligence tests – fond memories of middle school.