Why not discuss good parts of papers?
Ad Lagendijk
Monday, 26 May 2008 10:39 UTC
I am pretty new here. So my post could be totally displaced, but I will try anyway.
It seems to me that the aim of a forum on good papers is to stimulate that many more good papers will be written.
Diversity
The vehicle that is mostly used in this forum to achieve this goal is that of posters pointing out – what they consider to be – examples of good papers. I have no doubt that these papers are of indeed very high quality. The problem I have is with the diversity. Let us restrict ourselves for the time being to the Nature journals. Their scope varies wildly from medicine, biology to astronomy and climate science. The culture in these disciplines differs considerably. So my view is that pointing out an excellent paper on global warming is not going to help very much a physics postdoc who wants to write a good paper.
Dissect a paper
My suggestion for referencing a good paper is to do, what old-fashioned biologist know very well, dissect it.
A paper has a title, a list of authors, a list of affiliations, a corresponding author, an abstract, paragraphs, figures (graphs), figure captions, tables, table captions, conclusions, equations, acknowledgements, a list of references, footnotes, appendices, etc.
Example: Figures
It would be worth while to identify examples of good figures and captions. Criteria could include: use and abuse of arbitrary units, absence of spaghetti text in captions (caption saying: “see for explanation text”), good labeling, sensible color coding, self-explanatory caption.
Benefit
I could imagine that many more junior scientists from all science disciplines could learn and profit from such atomic examples.
But maybe this change of focus would be beyond the scope of this forum.
Ad Lagendijk
sciencesurvivalblog
-
Replies
Jump to resultsResults
-
Ad,
I think that the imperatives of structure, logic, flow, grammar are supra-disciplinary. Of course a physics student can take what is good from a ‘good’ paper on global warming.
Logic, grammar, sentence construction — and all the things you mention — are not restricted to any one discipline.
I, for one (but I can’t speak for the other moderators, obviously) would love to see what you call atomic examples. Feel free to post some :)
-
Thanks for your stimulating thoughts, Ad. As Richard writes, it would be nice if you could post examples of what you mean. For example, your section “dissect a paper” seems to me to be the same as a journal’s guide to authors/format requirements; your “figures” section of its figure advice, and so on. I think it is a good idea to have examples of “good” figures for others to follow, so please do post some in this forum. Nature uses this approach for some aspects of writing guidance: we provide an example “first paragraph” (abstract) and detailed statistical, protein-structure and chemistry guidelines.
Nature is an interdisciplinary journal, and one of our editorial criteria for publication is that an advance reported in a paper is of discernible interest to those in other fields. So we certainly hope that the papers we publish are comprehensible to all scientifically trained readers, at least “in the main” if not in every detail.
In my opinion, the principles of writing a clear scientific paper (or any document) apply to all fields. Yes, there are specific differences and conventions within fields, but these are not impediments to those from other fields understanding the paper, if it is well-written — which to start with means that the authors have a clear idea of what they wish to convey.
-
Okay, in the spirit of the suggestion (but it’s a pain to take figures and legends out of PDFs, upload them to another website then link), here is a good paper I read today in a field that I normally don’t have fun reading, biochemistry. It’s Kawaguchi et al., Biochemistry 2008, 47:5387-5395. The paper is here
I would say it is good for a number of reasons. One thing I loved was their clear, succinct explanation as they went along in the results of why they did certain controls and how it helped promote or eliminate certain hypotheses.
They also made some new, testable hypotheses in their discussion (future directions). I like that a lot.
The figure legends are pretty much self-sufficient. You don’t have to flip between methods, results and the figure itself to understand what it’s about.
I’ll try to insert an example here. Arrgh, I can’t easily. What a hassle. I’ll get back if I can.
-
Here’s a go:
The methods are also clear enough that one could redo the experiments oneself by only reading this article. My one very minor complaint is that I can’t find the number of times they needed to repeat the quantification of enzymatic activity to generate their error bars.
-
Atomic examples is a very good advice, something we had discussed before. We had discussions on good titles and good first sentences. We probably need more good figure and table samples.
It would be great if we could simply post the examples in this forum, but this will often be difficult for technical and copyright reasons.
-
Perhaps, Heather could post her example as a topic itself, we could all discuss the merits of various aspects of the paper.
-
Okay – will do.
-
Sorry, for the delay in answering some of your very useful comments.
Good paper is good paper, no matter what
I just do not agree. A good paper in mathematics could be a sequence of theorems and proofs. Without almost no additional text in between. So even within the natural sciences papers are written in a different way in different subfields (abstract versus verbose, for instance).Much of what I suggested is already in the author guidelines.
Yes and no. Many of the useful guidelines are not maintained by editors and reviewers. For instance I still see regularly captions to complicated figures that consist of two short lines of which the second line reads “see text for explanation”. In my second post I will go into more detail by showing examples.
I still see published figures that are too busy, with many insets. Curves that are too thin. Almost unreadable labeling.Copyright
I did not realize when I suggested to discuss details of papers, it could violate copyrights. Interestingly enough one can quote text without consent of publisher and authors, but one cannot reproduce figures without their content. -
As I explained before I think it might help to improve writing of scientific papers by analyzing small parts of (good) papers.
When you are trying to improve yourself, for instance in skiing or in playing golf, your instructor will try to correct your mistakes and bad habits. From that you can learn quickly. You could also study excellent examples, to try to prove yourself. But learning from mistakes is in the beginning more efficient.
So by the same token when promoting good scientific writing it is more illustrative to point out examples with mistakes than advertising good examples. However, this criticism can only be done with great care. Criticism can easily be counter productive, leading to a less congenial atmosphere. In addition anybody makes mistakes and we do not want a situation to arise where we continuously criticize in public each others writing.
I will nevertheless give a few illustrative examples with some criticism. But before I will do this, I want to make a very clear statement.
BEGIN OF QUALITY STATEMENT
The papers I discuss in the following are – as far as I can judge – all excellent papers. I wish I would have been a coauthor. I congratulate the authors with their very fine paper.
End QSI assume that readers have access to the web sites of Nature,
Science, and the American Physical Society. Probably papers could also be downloaded from authors’ web sites. In all cases I have notified corresponding authors of those papers that I discuss their papers here.Figure example 1
Read QS first.
P. Neumann et al. Science 320, 1326
There is no y-axis scale for Fig. 2B and 2C. Labeling of axes of Fig. 2D is too small.Figure example 2
Please read QS first.
I. Fushman, Science 320, 769
Fig 3 B. Excellent figure. Good labeling. Good colors. Y-axes has happily no arbitrary units. Calibration signal is also shown. Not busy.Figure example 3
Please read QS first.
“S. Bertaina et al., Nature 453, 203 (2008) ":http://www.nature.com/nature/journal/v453/n7192/full/nature06962.html
Inset is busy. Why all these arbitrary units? Are the two curves in the inset arbitrary with respect to each other? Labeling excellent.Caption Example 1
Please read QS first.
T. Biben and L. Joly
“Phys. Rev. Lett. 100, 186103 (2008) ":http://scitation.aip.org/getabs/servlet/GetabsServlet?prog=normal&id=PRLTAO000100000018186103000001&idtype=cvips&gifs=Yes
Fig 2 is a pretty busy figure. Its caption reads:
FIG. 2 (color online). Wetting properties of a crenelated surface
(see text for details).I would prefer as a reader not to be referred to the text for the understanding of a figure.
GOTO statements in text
Another interesting micro-property of a text is how many GOTO statements there are in the text. Like: “See previous section”. In software engineering GOTO statements are outlawed (“spaghetti code”). In a scientific text they make the text more difficult to read.
Please read QS first.
Song et al. Nature 453 629 (2008)
I count 18 explicit references to the supplementary information.Caption question/suggestion
For the purpose of writing this post I read a number of captions. I am in favor of a habit that would improve a caption a lot. Good captions are usually long (5+ lines). Why not add one or two lines that summarize the conclusion following from the figure? Mildly interested people that initially only have the time to glance at the figure learn something without having to read the main text, and might be more inclined to actually read the full paper. -
Thanks for these useful posts, Ad.
Maths papers (all equations and no text) are not published in Nature journals so I have no direct experience. We do publish equation-heavy papers, but they always have a summary paragraph and at least some explanatory text.
Figure legends. Nature journals’ figure legends always start with a “title” sentence to explain what the figure is about. (This is partly because of the online layout in the full-text version.) The legend then has to describe the parts (a, b, etc) and the symbols. Then more can follow, we allow the “see text” option if there is not much text (as is often the case in the physical sciences). I agree with you that complicated figure descriptions are easiest for the reader if they are encapsulated in the legend.
In some fields of science, authors want to have a lot of figures, and because of print space constraints, they create composites of multi-panelled figures, where each panel is of a very different type (of experiment, or medium for example). We ask that panels should be used within a figure only if the parts are logically connected. But there is certainly a tendency in some disciplines and by some authors to cram in as many panels as possible, without much logical connection between them, and this is very difficult for everyone (referees, editors and readers) to untangle. On the other hand, referees tend to ask for more data to be shown to demonstrate a point.This lack of focus can be even worse in the online-only supplementary information, which is not edited. Sometimes authors provide movies that contain massive redundancies (seconds or even minutes of empty panels, many repeats of one process, etc) which are quite boring, as a “reader”, to have to watch, once one has got the point (or while waiting to get the point).
Thank you for your examples. I know that it takes a lot of effort to post these given that Nature Network has not yet been upgraded to a more user-friendly platform (this should happen very soon now, I hope). I wonder what the other moderators think about moving these examples into separate comment threads for each, to facilitate discussion? I can do this if people think it is a good idea, as they are a bit buried in this thread.
Results
-
