Astronomical Techniques - Data Presentation and Standards

Now that you have obtained and optimally reduced your observations, and drawn stunningly important conclusions from them, there are still some hurdles to cross in contemporary scientific practice. You will need to present them in oral, written, and graphic form. Furthermore, in competition for resources (either grant support or time on facilities), you will have to write proposals for research. Some basic exposure to these arenas may help ease the starting pains.

The proposal

Many research projects begin with a proposal of some kind -- for travel funds, telescope time, access to a spacecraft, or construction of an instrument. In all these cases, you are trying to convince reviewers (individually or in a panel) that your project should be supported instead of somebody else's. That means that you must practice defensive proposing - leave no obvious excuse to exclude yours from consideration. Doing the advance homework is crucial - if you don't seem to know the relevant previous work or other relevant data, it will be easy to favor some other idea that looks like a safer bet. Show specifically how you will go from the proposed work to a scientifically useful result. Try to judge the right amount of background information for the most likely reviewing audience.

Most reviewers are overburdened in reading. Be as concise as the topic allows. Filling pages beyond the stated limit with tiny letters will not make a favorable impression. Everybody else also believes their research to hold the keys to creation. Finally, don't give up. Many pieces of excellent research (my latest successful HST proposal, for example) took several rounds to get past this first obstacle.

The talk

The first presentation of research results is often in an informal talk or seminar, in front of a mixed audience of specialists, people with more distant backgrounds, and students. Again, be attuned to the audience in gauging how much background and detail to give. There is a temptation to give too much detail on your own work, and leave out the reasons why someone should care, while the right mix is usually strong on background and weaker on the finer details of your recent additions. The following ideas are based on notes prepared by F. Israel at the Leiden Observatory for student's first seminar talks, and are useful here as well (after being purged of items that are linguistically superfluous for English speakers. Giving talks is an important part of scientific communication, not to mention getting jobs, and this is something you will need to learn along with the subject matter. (Note: I'm still thinking about which bits of this advice need to be modified in view of the triumph of PowerPoint and its kin).

(Almost) ideal colloquia are in fact given, but there is considerable art involved. There is no single recipe for a perfect colloquium; this is a matter of style and thus highly personal. This style is something you must learn for yourself - this can be difficult, but it's certainly possible and rewarding. Here as elsewhere, practice makes perfect. Once you have given a number of brief talks (for example, in the Friday afternoon get-togethers) and examined your own "performance" critically, you can learn much. Beware the urge to think "This will be easy". There are a few born geniuses who can give good colloquiua without practice or preparation, but for most people, the more the better. Consider that not only the speaker but the audience can be victims of a poor talk.

A good colloquium runs along smoothly, gives new insights, and is enjoyable, even for someone who doesn't already know the subject. A good speaker takes account of the specialized knowledge (or lack thereof) of the audience. Someone who already knows what you're saying is more likely to forgive a bit of review than someone outside your specialty is to forgive jumping straight into jargon. Don't assume that everyone understands why your problem is important until you tell them! Don't overestimate your audience. Review why hydrogen produces the 21-cm line even if its discoverers are in the audience.

Preparation: Prepare a talk shorter than the allotted time - prepare a 45-minute talk for an hour slot, for example. Don't be afraid you'll have a talk that's too short. They're usually too long, and going overtime usually irritates the audience. In practice, there are always interruptions and digressions. Don't get mad at interruptions - answer them, but briefly. Practice the beginning and end. A classic letdown is "Well, I think that's all I've got to say". Don't tell absolutely everything you know or did. Don't detail all the steps in data reduction, just the salient points. People frequently get so bogged down in trivia that there's no time for the results.

Try to talk only about matters that you understand well; appeal to authority is seldom convincing ("my advisor says that ...", "Einstein demonstrated that ..." - OK, so Frank picked a bad example there). Each speaker should strive for credibility. Try not to exceed the bounds of your expertise - build in some limits, and be sure to know the background material adequately so as to be able to deal with critical questions.

Execution: Begin with what you're talking about and why it's interesting (not more than about 1/5 of the time for this). There are projects of clearly widespread interest, but this isn't often so - even within the department. Once again, don't overestimate your audience, even among astronomers. Try to stick to one line of argument. Diversions or digressions may be necessary, but try to keep them limited, and be sure to recapitulate how they relate to your main theme. Finish with a clear conclusion. What did you say? What was new here? What further work is needed?

Movements and gestures: keep these under control. Watch closely (before it's your turn) various speakers and instructors and notice who walks around too much, who flaps like a bird, and who is too distracting a speaker. Establish your own guidelines from this. The same holds for use of the chalkboard, slides, and overhead projector - these are aids and should not dominate the presentation. Don't be a slave of your overhead transparencies. Here, too, observing the mistakes other people make can be very instructive.

Visual aids: Important: during a one-hour talk, use no more than 10-15 slides AND transparencies. If you overdo it, the talk will seem scattered, and you may confuse yourself. It is also boring if (for example near the end) the speaker flashes quickly through whole piles of slides or transparencies.

Slides should serve primarily as illustrations. A classic error is "as you see from this slide, A follows necessarily from B", instead of "In this slide you can see a good example of what I was just talking about". Be sure that the slides are convincing; another classic line is "Yes, but I have other slides where you really can see this clearly". See that the slides are legible and to the point. Try to project them before the talk. You can say only a little about each slide. If the slide contains a great deal of information of various kinds (not a good idea, but sometimes unavoidable) make the important point for your discussion clear; in graphical cases tell what the axes are and what values are covered (try sitting in the back of the lecture hall one day). Limit the number of slides. Ten in a colloquium is almost always too many. Thus, use only one or two slides to make a point (not 5 OH spectra that all look alike anyway). The transition between light and dark, usually demanded by slides, can be irritating to the audience if it's too frequent. Try grouping slides or overheads together to cut down on the number of transitions you have to make.

Overhead transparencies: the most common fault is writing too much. There are people who project their entire text - this is completely wrong unless there is a language barrier to worry about. If you want to rely on these (many people do), give the outline in large letters, short phrases or even via keywords. The same for tables: the shorter the better, get to the point. Limit equations to the quantities that you will really use. Don't underestimate the difficulty of finding some detail during a colloquium, and don't overestimate the audience's ability to read and understand quickly. Overhead striptease (covering part of a sheet on the projector) irritates many listeners. Use colors, but with restraint. They should be functional; some overhead sheets can be confusing for all but the color-blind.

Here's a real challenge - try a talk without overheads at all. One of the best colloquia at Leiden in the last five years was by Martin Schwarzschild on a very difficult topic - stellar dynamics. He gave this without overheads, without slides, and writing only 3 words on the chalkboard. In addition, he stood almost still, moving only his hands in a very expressive way.

Chalkboard: Make your talk livelier by sketching correlations, major principles, etc. on the board. Upcoming definitions, tables, etc. can be put on the board in advance. The same for co-authors, references, and so on.

Some random thoughts: Planned jokes usually misfire. This gets worse if you're outside your own environment - for an American speaker, half your jokes work in Britain, 15% in Holland, and almost none anywhere else. Look at the audience - don't talk to the screen or chalkboard, and don't stand in the way of slides or overhead projector. Don't read your talk from overhead sheets.

The research paper

This is the primary means of formal communication, though today it may be seen as often electronically as in print. It must be clearly written and complete enough for the reader to understand what you did and how, to the level that the measurements or derivation could be replicated. Most manuscripts could benefit from tight editing, improving the readability and expense (since most journals have page charges). There are basic instructions for manuscript preparation in each journal issue, and some have manuals of style available.

The best way to start writing a first paper is to find a similar one already published and copy its organization - a more individual style can evolve with practice and confidence. A traditional order of sections might encompass an introduction, details of the observations and processing, comparison with relevant theory, and finally discussion and summary. The introductory abstract is important; most readers may never get past this paragraph, and abstract searches will only key on the words that actually occur there.

Literature citations in most astronomical journals follow a style like Jones (1978) in the body of the text, with et al. for three or more authors. A streamlined citation in the reference list has been uniformly adopted by the major journals, with a form like

Jones, D.W. & Kincaid, Z.Z., 1981, ApJ, 321, 787

with abbreviations AJ, AA, MNRAS, etc. Preparing a paper can well begin with a literature search, either backwards from a recent reference list or forwards from a well-known paper using the Science Citation Index. In astronomy, always start with an ADS search. For individual objects, software tools such as the National Extragalactic Database or SIMBAD are important timesavers.

A common gaffe is to list numerical values to imaginary precision. Watch for significant figures and truncate or round accordingly.

Graphic results

The heart of many reports is in graphic presentations. There are some simple standards that can improve legibility. Graphs should be clear, uncluttered and informative - even when reduced for publication. Ideally, understanding should require minimal recourse to lengthy printed captions, with important information given in the figure (to a reasonable degree!).

Be creative and appropriate. Don't stay a slave to software defaults. Watch for labelling, especially when scaled units or logarithmic scales are involved. Lots of good (plus horrible, misleading, and stupid) examples appear in The Visual Display of Quantitative Information by Edward Tufte (Graphics Press, 1983). He sets forth some powerful and basic rules for presenting data in graphical forms. Among these are:

  • "Graphical excellence consists of complex ideas communicated with clarity, precision, and efficiency. It is that which gives to the viewer the greatest number of ideas in the shortest time with the least ink in the smallest space. (Tufte actually becomes a bit silly about mimimal ink). It is nearly always multivariate. And it requires telling the truth about the data."
  • He has a chapter on the Lie Factor in graphics, with these conclusions. "The representation of numbers, as physically measured on the surface of the graphic itself, should be directly propertional to the numerical quantitites represented. Labelling should be used to defeat graphical distortion and ambiguity. Show data variation, not design variation. Don't quote data out of context."

    I'll start a Hall of Fame for astronomical examples. First up, a very instructive figure from a paper by Nicole Vogt and collaborators (AJ 127, 3300, 2004). This lets you compare velocity and emission line structure with a galaxy's orientation and location within its cluster environment for a whole ensemble of galaxies. How long would you have to stare at 1D plots and tables to get this effect?


    Peer review is one of the most important and dreaded factors in scientific publication. Einstein didn't like it much the only time he ran into US-style reviewers (and as it turned out, the reviewer was right!).

    « Polarimetry | Data archives in astronomy »

    Course home page | Bill Keel's Home Page | Image Usage and Copyright Info | UA Astronomy
    2008	  © 2000-2008