I posted this question on stats.SE, but the consensus was that it was better suited to tex.SE.


I've been thinking recently about conventions for figure width in data analysis reports that lead to PDFs (e.g., in A4 or Letter size). My usual context is R, Sweave, and LaTeX.

The default figure width in Sweave is 80% of the text-width (i.e., the width of a paragraph of text).



  • Is it reasonable for the width of figures to be wider than the text-width (i.e., to spill into the margins)?
  • Is the 80% rule of text-width a good one or would, for example 100%, or some other value be better?
  • Should the figure widths be consistent throughout a document (or perhaps with two sizes for small and large figures) or should the width be adapted completely to the content?

Any references or thoughts on best practice in this regards would be most welcome.

  • This doesn't seem like a question that has a good answer. See this.
    – TH.
    Commented Nov 14, 2010 at 9:59
  • 1
    @TH I agree that the question could be conceived in purely subjective terms. However, I also thought that there may be relevant typographic conventions with accompanying rationales. I think the latter is not subjective. Commented Nov 14, 2010 at 11:17
  • @TH I updated the question title to use the term "typographic conventions" rather than the word "aesthetic" in the hope that this reduces the perceived subjectivity. Commented Nov 14, 2010 at 11:19

3 Answers 3


I'm inclined to think that this is not a subjective question. Of course, it might be considered so if framed in terms of "aesthetics" or "art". However the "art of typography" often comes down to some fairly matter of fact questions and answers, of which I think your overall question and how it might be considered is one.

Staying away from the subjective (although not necessarily from the complex or hard), I would frame and therefore break down your design decisions with respect to the psychology of your target group of readers.

The two principles I live by in this regard are these:

  1. Principle of least surprise
  2. Productivity principle

The principle of least surprise leads to simple but effective guidelines such as these:

a) adopt the primary, i.e., most familiar style employed by your readership's main monographs or periodicals;

b) settle on a single style in any single work, avoiding inconsistencies unless you consistently provide contextual cues that make it easy for your readers to (possibly subliminally) recognise your pattern in this.

The productivity principle is a little harder. It's all about keeping the rate of your median reader's information assimilation for expended effort as high as possible. Of course, this means finding the psychological sweet spot over many factors (see another answer regarding this sort of consideration here). Some of these factors are psychophysical, e.g., relating the motor performance of eye saccades to the optimal size of lines on the page, or to considerations like: if your readers are around professor level or higher, they are likely to be 40+ years old, hence likely to be presbyopic, and therefore might be better served by a larger font in text and diagrams. Other factors are cultural and therefore expectations based, e.g., having much to do with how your readers learnt to digest textual and graphical information earlier in their lives. Some factors relate to their subject competency levels -- low competency levels often suggest introducing more whitespace on the page, higher subject competency readers would probably prefer the page to be more information dense. Then there are cognitive factors that fit somewhere between the psychophysical and higher order factors, e.g., the commonly applied rule of thirds, the golden ratio principle and so forth.

Nevertheless, as a useful rule of thumb, I often (but not always) set figurewidth/textwidth == textwidth/pagewidth since that maintains consistent proportions across the x-axis of the page.

Regarding the reasonableness of spilling a figure into the margin: under conventional circumstances, I would tend to think not since it would interrupt the steady flow of the eye as it scans across the body text elsewhere on the page. Nevertheless, while remaining true to principles 1 and 2 above, Tufte-style documents are a clear exception to the "don't spill figures into the margins" rule.

Summarising, while your question takes us to complex and hard places, "good" typography can find some serious psychological grounds. This, therefore ought legitimately be the stuff that TeX-typographers consider as part of their daily work. For me, unless the design is intended to push non-conventional boundaries, I always approach this matter by considering the psychological act of how the specifically-defined readership acquires textual and graphical information from the page. These considerations include weighing up psychophysical factors, and the target readership's hard and soft wired conventions, beliefs and expectations, etc. The beauty of this approach is that, apart from taking us from the realm of the subjective, it provides a framework to guide decision making about which rules to follow and how to break them when they don't apply.

  • 3
    impressive answer! Commented Nov 14, 2010 at 10:55
  • 2
    Geoffrey, excellent answer. I particular like the point, that typography to an extend is covered by conventions and rules, and this is what TeX was all about. Knuth was font of saying, that he distilled those rules to about a 100 and these have been used so successfully in the TeX engine. Having said that though, I have been using the Tufte-book class for some documents and so far they have been received quite well. However, a number of people were confused with Table captions being in the margin - which I partly solved by moving the Tables too!.
    – yannisl
    Commented Nov 14, 2010 at 11:43

Geoffrey gave some great tips, here are some images from a book I am busy with based on a hacked version of the tufte-book class. The examples are drawn specifically to illustrate the use of figures with LaTeX.

alt text

This one demonstrates the use of the `wrapfig' package, used in connection with margin figures. Margin figures work well with plenty of text as captions.

alt text

Some pottery with subcaptions.

alt text

This last one - although seems a bit unbalanced - was done on purpose to give the reader a feeling of the original size of the pottery (the last one is much larger than the first two). The margins need some more text though.


It doesn't sound like much of an answer, but what I have done with the same problem with business formats like letterheads is hire a graphic designer to help me figure out where the best places are to align vertically and horizontally, and which parts of the page either need whitespace or can be cramped. Off-the-shelf Latex packages don't do this for you.

If professional design help is not an option —nor spending 10-15 months studying graphic design— the next best thing is plagiarism: print out a pile of reports, decide which you find most pleasing, and measure these things with T-square and ruler. Write your package/module based on that.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .