(November 2002)
Below we offer numerous guidlines, hints, suggestions and strong(!) opinions on how to effectively communicate the results of your research. Communicating your results effectively is an invaluable part of doing science, and one that requires considerable effort and experience. Of course, communication is ultimately a very personal matter; accordingly, personal styles differ widely and you may disagree with some points below. But even if you do, the guidelines will at least encourage you to think of a good reason why you disagree, thereby serving their original purpose yet again, namely to
encourage you to devote a lot of thought to communicating science effectively.
Implementing some of these suggestions may require huge amounts of time and effort. Nevertheless, don't shirk these! Just as in a market economy,
only polished products sell well !
The reader of your paper or the audience of your talk (=customer, buyer) expects and deserves to see only the final, optimized product (the history of its development seldom interests them). If your product is too far from perfect, they'll rapidly stop paying attention.
a) Do not use the same symbol for different purposes in nearby places. If this is unavoidable, only do so in different sections or chapters, and alert the reader to the fact that you are changing notation.
b) It is advisable to spend a lot of time at the beginning of your project (but after you have understood what you are supposed to calculate) in inventing the most convenient notation for the purpose at hand. If, during the course of your calculations, you come across a more convenient notation or convention, adopt it henceforth. This will save yourself a lot of time in the long run. Moreover, you cannot expect the reader to put up with some inconvenient notation just because for historic reasons you happened to start with a bad choice, or did not have time to improve it (this is a very bad excuse) -- typically, the reader will become annoyed, or worse, stop being interested altogether. Only polished products sell well.
c) Compact notation is advisable when explaining or summarizing the general strategy of your calculation, where the reader is just supposed to understand what your starting point is, what type of manipulations you do, and what type of result emerges, but where details would be distracting rather than helpful. Nevertheless, the compact formulas should be explicit enough that a sufficiently expert reader would be able to reproduce your results when using them as starting point.
d) Explicit (long) formulas are advisable when you want to explain details of the calculation that you think the not-so-expert but interested reader (e.g.your roommate) will benefit from seeing in detail. Typically, this will happen when you want to draw attention to a particular crucial step, trick, or manipulation, that is absolutely crucial for success, or that is rather nonobvious and took you a long time to figure out and that you are very proud of, or that will save whoever repeats the calculation a lot of time, or that manages to simplify calculations that other experts before you have done in more cumbersome ways, etc.
e) Epxlicit (long) formulas are also advisable if they represent key results. Typically, these would be convenient intermediate results that a lot of further results depend on, or the ultimate, final result of your calculation that you hope you will become famous for, or a formula for a curve that is plotted in a figure (in which case the figure caption should refer to that formula), etc.
f) Explicit (long) formulas are not advisable when they just represent a typed-up version of your personal detailed notes, which any reader with enough time and diligence will be able to straightforwardly reproduce by just writing out explicitly your general, compact formulas. (Readers will usually trust you to have done this correctly!)
a) Figures are what most readers remember best about your paper, talk or poster. Therefore, it pays to optimize them.
b) Usually, the key figures in your paper or thesis will also be the key figures in your talks. Therefore, design them in such a way that it is possible to adopt them for a talk without further changes. (Your collaborators will appreciate this, too!) In particular, make sure that the figures in your paper are beamer-compatible if used in a laptop presentation, because typically you will prepare your talk by cutting-and-pasting figures from your published paper into your slides.
c) Beamer-compatibility implies:
- Curves should not be too thin: ideally, use line thickness 2 for all curves in your figure, or, if that makes the figure too busy (because it has insets, etc.), use line thickness 2 at least for the curves representing your main results.
- Colors should not be too faint: i.e. no light yellow, light green, light grey, etc.; colors that work well are black, blue, red, dark orange, magenta, violet, dark brown. At the same time, the colors you choose should produce line styles that can be properly distinguished when printed out greyscale with a non-color printer, e.g. black, blue, dark orange, or solid, dashed, dash-dotted lines (or both color AND different line styles).
TEST the beamer-compatibility of your figures with an actual beamer (even at the stage of writing your paper, long before preparing your talk) - nothing is more embarrassing than having to apologize to an audience that the main result of your talk (which you copy-pasted from your published paper) is unfortunately not visible on the screen because the color is too faint. Also TEST the non-colorprinter-compatability of your figures by printing out a test copy - nothing is more annoying than being told by the journal's production team that you unfortunately have to redo the figures because they don't come out well in black and white.
- If you plan to publish the figure in black-and-white, the figure caption should start with "(Color online)". Moreover, the caption and text should AVOID refering to colors, as in "the red line shows ..."; instead, refer to line styles, as in "the dashed line shows".
d) Make sure the size of the legends (numbers on axes, etc) is not very much smaller (at most a factor of 2/3) than the printed text around the figure. Remember, you might want to use the same figure in a talk, too, and unreadably small legends on figures really annoy the audience.
e) Put as many of the parameters required to make the plots in the figure into the figure itself, e.g. in the from of little boxes (e.g. T=0, V=0, V_g/Gamma = 1, etc.) Not only does this save space in the figure caption, it also increases the figures suitability for talks.
f) In theory papers, use dimensionless quantities for making plots as much as possible (e.g. G/G_0, T/Gamma, V_g/Gamma, etc.) Why? Firstly, it increases the applicability of the results (they hold not only for one value of Gamma, but for any one). Secondly, deciding which combinations to use is usually a good excercise for understanding what scales are important in your problem. When there are several possibilities (usually there are), think hard about which one is best (usually there is an optimal one, which makes the main message of the plot maximally clear). If you find a better one only after the plot is finished, redo the plot, even if it costs a lot of time -- only polished products sell well !
g) Exceptions to rule-of-thumb 2.d) are required, of course, when comparing theory to actual experimental data, in which case dimensionful quantities are usually needed.
h) NEVER, EVER, plot anything in arbitrary units!!! (people will be reminded of Jan Hendrik Schön, who did just that!)
i) If possible, discuss and explain all important, novel or striking features of your figures in the text. For example, if there is a striking, unexpected dip or peak in your figure, the reader expects an explanation, ideally via a simple back-of-the-envelope argument. If you do not understand these features yourself, maybe you should not put this figure in your paper. Or, at the very least, you should add a comment that you do not understand them but find them so interesting that you want to advertize them for further study in the future...
The typical target audience for papers are interested experts, who have little time to read cumbersome epics and to which you don't have to explain all the basics. Hence, brevity is advisable, even in long papers. However, the urge to find perfect, maximally brief formulations at the outset can lead to "writers block", i.e. slow you down terribly. To overcome "writers block" when starting to write a paper, here are some suggestions:
(i) Begin by deciding which figures convey your main message best.
(ii) Make a telegramm-style outline of the organization of the paper and write down the key equations. Show this to your advisor and force him/her to read it!!
(iii) Write a telegram-style introduction (polishing the introduction is best left to the very end); the introduction should not only explain why the project you worked on is exciting, but also state in not-too-technical terms what the main results are (and, if the paper is long, where to find them; e.g. name the key figure or equations). This will help the reader to decide whether it is worth reading further, or, if they only have time to skim, where to skim.
(iv) Start writing section 2 carefully (formulating complete sentences cannot be postponed forever!) Usually, writers block has disolved by this time...! Once you have written the first complete section, show it to your advisor or collaborator to get feedback on your writing style, notation, etc. It really pays off to do this at an early stage!
(v) Write the abstract and elaborate the introduction only when the bulk of the paper has been written and is approaching its final form.
(vi) Write the conclusions. They often restate what is in the introduction, but in more technical language, since the reader can be expected to have read the contents in between.
(vii) Go through the paper and make sure you made clear what your main results are. Every paper should have at least one sentence of the type: "This figure/equation/conclusion is our (first, second, third...) main result."
(viii) Iterate the main body of the paper X times (where X is a large number...).
(ix) The last step: polish abstract, introduction and conclusion one more time: 95 % of all readers will read only these!
(x) Get outside opinions: Ask a friend, office mate or group member to read the paper. Then send it to an expert you know and politely ask for comments (if your paper is interesting, usually they'll be happy to give you feedback. Telling them by what time you plan to submit usually speeds up getting a response). Sometimes, it is even advisable to send it to competitors, as a gesture of good will and to give them a chance to complain about delicate phrasings, left-out references, etc. Take all comments very seriously -- if your friend finds something unclear, the chances are that the referee will, too.
a) For papers, the typical target audience are interested experts, who have little time to read cumbersome epics and to which you don't have to explain all the basics. Here, brevity is advisable, even in long papers.
b) The style of theses is different than for papers -- your typical target audience are your fellow diplom- or PhD-students (and Professors from a different field that have no clue about yours!). They should be able to learn, if they so wish, from your thesis precisely what you did and how you did it.
c) Often, your thesis will be the starting point for the next person joining the group to "learn the basics" and the "tricks of the trade". Therefore, somewhat more details than in a paper, and an attempt at pedagogical presentation, are appropriate.
d) This does not mean that you have to reproduce long passages from the textbooks or the literature. But it does mean that if you rely on some "standard" methods, you should give precise references for where they are best explained (preferably including Eq. and page numbers; just citing a 700 page book by name is not necessarily helpful!)
e) Even if you use, without derivation, some standard results or methods, it never hurts to try to summarize, in a few sentences, the key idea behind their derivation or what the main goal or strategy is of the approach (even if "well-known") that you are using. This really helps the unexperienced reader, and to the experienced one it demonstrates that you really understand what you are doing.
f) If you are lucky enough to already have published or submitted papers by the time you start your thesis, you will want to use as much as possible of this material, with as few changes as possible (because it has already been optimized and iterated X times). However, since this will have been written for experts, your target audience (the next group member) will have a tough time to understand some of it. The cleanest and perhaps most efficient way to bridge this gap is to add introductory sections to scetch the background, and technical appendices, in which you explain the details which the experts take for granted to the next group member (and to yourself ten years from now, by which time you usually will have forgotton all details!)
a) The purpose of a poster is like that of a blackboard: it is meant to help you explain to a colleague in a conversation what you were doing. It should therefore contain the things (equations, figures, key assumptions, key results) that you would usually write on a blackboard if you were explaining your project to a colleague or visitor at home in your office.
b) The only part of a poster that people are likely to take note of without you being there, are the the title, perhaps the abstract, and the (very important!) pretty pictures. For everything else, they'll ask you to explain it to them. They will NOT "read" your poster in detail like they read your paper.
c) Therefore, everything on the poster should be:
BIG, SIMPLE, ONLY KEY WORDS, NO LONG PARAGRAPHS.
d) Posters should contain references, because interested colleagues will be happy to write them down.
e) Pin up a few A4-copies of your poster next to the actual poster, and likewise a few preprints, if relevant. Interested colleagues will be happy to take some, in particular if you are not there when they are looking for you.
Every talk you give is a potential job talk -- your potential future colleague or boss may be in the audience, and in the distant future when hiring decisions are made, he or she may remember the excellent or terrible impression you made in the one talk he ever heard you give. Therefore: do all you can to give the absolutely best talk possible, every time -- your future may depend on it!
a) Your first slide should have enough material on it to be able to give a five(!)-minute introduction before changing to the second one. First slides which have only the title of the talk and the name of the speaker on them are, though regrettably common, annoying, because the audience typically gets to see them only for 10 seconds.
b) The introduction should, much like that of a paper, tell the audience why the field you work in is interesting, what are the big questions, why you find them exciting, what are the issues you have decided to study. Advertize your field, not just your own work. Don't rush here -- a good introduction is essential in persuading the audience to actually listen to you instead of letting their thoughts wander. Usually, one can tell within the first two minutes whether a talk will be really good or not.
c) Give the outline of what will come only after you gave the introduction (if you do it in the opposite order, quite often the audience will not be able to appreciate the outline.) Don't use more than four or five points in the outline -- more will be confusing, and will be forgotten.
d) Summarizing, the first slide should ideally contain:
(i) Title of talk, speaker name (preferably underlined), coauthors, collaborators
(ii) One or two beautiful figures serving as motivation, and one or two more that summarize your main results. These figures can be small -- they just serve as a table of contents at this point, you'll discuss them again in detail later.
(iii) A one-line sentence that summarizes the main question that you plan to address.
(iv) A brief outline of at most four or five points. (If you run out of place, you can put this on slide two, but it usually "sticks" better in your audience's minds if they can stare at it for 5 minutes while listening to slide one.)
(v) A one-line sentence that summarizes the main result.
e) Try to dedicate each subsequent slide to a single "key idea or concept", which should be mentioned in the title of the slide.
f) Summarize the take-home message of each slide by a one-line sentence at the bottom, highlighted in an attention-grabbing way (e.g. red with yellow background).
g) As a rule of thumb, plan to talk at least 2 to 3 minutes per slide (even better, 3 to 5). This gives the audience sufficient time to absorb what is on the slide. Figures that are just flashed for 10 seconds and disappear again are completely useless -- very few people are able to grasp the contents so quickly. Also remember that some people actually try to make notes while you talk -- they are really grateful if they have sufficient time to copy a key figure or equation, and really annoyed if they don't. Also, rushing through a talk is generally counterproductive -- the audience first gets nervous, then annoyed, and once you've lost them by rushing too much, they stop listening altogether.
g) Rule-of-thumbs (e) and (g) imply that you will sometimes have to combine two or three of your favorite figures onto one slide, plus a few words of explanation, an equation, a reference, and a take-home message. At the same time, you will have to judiciously fiddle with placing, spacing and sizes of all components to avoid the slide looking too "busy". (Building up the slide component by component in power point talks helps to ameliorate the latter problem.) Preparing such a slide takes much more time than putting each figure on a separate page. However, the benefit of being able to talk about a single concept in a coherent way, and give the audience enough time for this to sink in, outweighs the extra time spent in preparing the slide. -- Only polished products sell well!
h) Every slide should have at least one figure (with very rare exceptions!). If you don't have a results-figure, at least make up an explanatory cartoon-figure that sheds light on the equations you are talking about.
i) No slide should have more than three equations (with very rare exceptions!)
This rule-of-thumb encourages you to resist the urge of "deriving results" in front to the audience's eyes. Of course, presenting derivations is sometimes useful and meaningful, in particular for back-of-the-envelope-arguments. But usually, 99 % of the audience is lost within 10 seconds after you plunge into a detailed derivation, and the 1 % that is not would have asked you afterward if they were really interested in the details.
Remember, a talk is not a lecture: your job in a talk is to get across ideas, intuition (in the form of revealing cartoons) and results (in the form of easy-to-remember figures and plots), not detailed derivations!
j) Put the key references for each slide inconspicuously (i.e. very small type -- place is scarce!) but readably at the top or bottom. Make the reference complete enough that an interested reader will be able to track down the paper, e.g. "first author et al., PRL, 10, 193 (2000)" for the important references, or "first author (2000)" for the not-so-important ones. No harm is done in putting a reference to your own paper on several different slides. REMEMBER to also include references mentioning the work on which your's builds, or that of your competitors. Nothing upsets people more than not being cited properly.
It is not always necessary to actually mention the references while talking. But at least they should be visible, to keep those that do care about them happy.
k) The summary slide should, ideally, have mini-versions of the figures illustrating your main results. This helps to reinforce the "take-home" pictures that the audience should try to remember.
l) The summary slide should contain a comment or two about open questions, future prospects, etc. (This will help the chairman to ask at least one meaningful question in case the audience has none!)
m) If you give a power point talk, resist the urge to be cute by using too many silly animation features. Formulas and figures that swing or swirl in from left to right, or fade away, or blink before settling down, are terribly distracting. Keep it simple: when a new component of a figure is to appear, let it simply pop up at the right place without further fuss.
o) Stick to one color scheme throughout the talk: Color 1 for titles of slides, color 2 for references, color three for formulas and explanations, color four for main results,etc. Similarly, choose a certain size lettering for each type of item, and use it consistently througout.
p) If you are not a very experienced speaker, type up word for word what you plan to say for the first few minutes in the introduction, and memorize this. Usually, the words you will actually end up saying will be different; nevertheless, the process of formulating an introduction word for word will force you to organize your thoughts properly and will help you avoid getting tangled up right at the beginning. By the time you have finished the typed-up part, your nerves will usually have calmed down quite considerably. -- Formulating passages word for word while preparing may also be helpful for other delicate moments of the talk, e.g. when you have to explain politely why your competitor is wrong, or why experiment so and so is inconclusive, etc.
q) Schedule a practice talk with your friends and collaborators at least one week before the actual talk. Usually, you will start preparing the talk too late to have it in perfect shape by the time the first deadline arrives, and you will be grateful for the extra week until the actual talk.
r) If it is an important talk (invited talk at major conference, job talk), schedule a second practice talk, and even a third one, if you are a nervous type of person. After all, your future may depend on it!
