\newtheorem{prequestion}{Question} \newtheorem{preanswer}{Answer} \renewcommand{\thepreanswer}{} \newenvironment{question}{\begin{prequestion}\rm}{\end{prequestion}} \newenvironment{answer}{\begin{preanswer}\rm}{\end{preanswer}} \newcommand{\latexword}[1]{{\rm\tt #1}} \newcommand{\lamport}{{\it \LaTeX: A Document Preparation System\/} by Leslie Lamport} \newcommand{\shortlamp}{{\it The Manual}} \author[R.~A.~Bailey]{R.~A.~Bailey\\Goldsmiths' College, University of London} \title{Topical Tips --- side by side figures in \LaTeX} \begin{article} \setlength{\unitlength}{0.1cm} \newcommand{\blob}{\circle*{3}} \begin{question} How do I set two figures side by side in \LaTeX? \end{question} \begin{answer} I assume that you mean how do you set a pair of figures something like Figures~\ref{anova} and~\ref{hasse} below. \begin{figure}[htbp] \noindent \makebox[\columnwidth]{% \begin{minipage}[b]{0.5\columnwidth} \[ \begin{array}{lc} \mbox{Source} & \mbox{df}\\ \hline \mbox{blocks} & b-1\\ \mbox{varieties} & k-1\\ \mbox{residual} & (b-1)(k-1)\\ \hline \mbox{total} & bk-1 \end{array} \] \caption{The analysis-of-variance table\strut} \label{anova} \end{minipage}% \hfill \begin{minipage}[b]{0.4\columnwidth} \begin{center} \begin{picture}(20,25)(-10,-2) \put(0,0){\line(0,1){20}} \put(0,0){\blob} \put(0,10){\blob} \put(0,20){\blob} \put(0,-4){\makebox(0,0)[t]{$bk$ plots}} \put(0,24){\makebox(0,0)[b]{$1$ universe}} \put(2,10){\makebox(0,0)[l]{$b$ blocks}} \end{picture} \end{center} \caption{The Hasse diagram\strut} \label{hasse} \end{minipage}% } \end{figure} That pair of figures consists of two objects in a single \latexword{figure} environment (more on that later). Here is how I did them. {\fontseries{n}\selectfont\begin{verbatim} \begin{figure} \noindent \makebox[\textwidth]{% \begin{minipage}[b]{0.5\textwidth} \end{minipage}% \hfill \begin{minipage}[b]{0.4\textwidth} \end{minipage}% } \end{figure} \end{verbatim} } \end{answer} (Well, what I've just said isn't totally true, but I'll come back to that.) The first object is an \latexword{array} inside displayed maths, while the second object is a \latexword{picture}. The objects can be anything you like, including tables and straight text. Figures~\ref{box} and~\ref{alice} are done similarly. In both cases I've put each object in a \latexword{minipage}: see pages 98--99 of \lamport, hereafter called \shortlamp. The \latexword{minipage} takes an argument specifying its width. Rather than defining the widths of the objects absolutely, I prefer to define them in terms of \verb+\+\latexword{textwidth}, which is described on page~94 of \shortlamp. I've chosen two fractions which seem to me in the right proportion and which add up to less than~1: in Figures~\ref{box} and~\ref{alice} the proportions are $0.3$ and $0.6$. Normally, \latexword{minipage}s are horizontally aligned around a central horizontal axis. I prefer to have the two \latexword{caption}s lined up nicely, so I have given both the \latexword{minipage}s the optional argument \latexword{b}, which aligns them at the bottom. The two \latexword{minipage}s are put inside a box which is just as wide as the text. This is made with \verb+\+\latexword{makebox} (see page~97 of \shortlamp\/), whose optional argument specifies the width of the box. There is a \verb+\+\latexword{noindent} before it, to make sure that there is no space on the left before the box starts. The \latexword{\%} sign immediately after the start of the box lets me start a new line in the input file without introducing a space before the first \latexword{minipage}: in other words, the first \latexword{minipage} is flush left to the edge of the text. In the same way, the \latexword{\%} sign after each \verb+\+\latexword{end}\verb+{+\latexword{minipage}\verb+}+ prevents a little inter-word space creeping in after each \latexword{minipage}. This fussiness is not evident in Figures~\ref{anova} and~\ref{hasse}, but does show up in Figures~\ref{box} and~\ref{alice}, which have visible outer edges flush with the edges of the surrounding text. Finally, \verb+\+\latexword{hfill} is put in between the two \latexword{minipage}s, forcing them as far apart as they will go: see page~96 of \shortlamp. \begin{figure}[hbp] \noindent \makebox[\columnwidth]{% \begin{minipage}[b]{0.3\columnwidth} \begin{picture}(30,30) \put(0,0){\framebox(20,10){E}} \put(0,10){\framebox(10,10){C}} \put(10,10){\framebox(10,10){D}} \put(0,20){\framebox(10,10){A}} \put(10,20){\framebox(10,10){B}} \end{picture} \caption{Five objects in one figure} \label{box} \end{minipage}% \hfill \begin{minipage}[b]{0.6\columnwidth} \begin{flushright} Tomorrow, and tomorrow, and tomorrow, creeps in this petty pace from day to day, to the last syllable of recorded time; and all our yesterdays have lighted fools the way to dusty death. \end{flushright} \caption{Philosophy} \label{alice} \end{minipage}% } \end{figure} I expect that you asked this question because of a misunderstanding about what a \LaTeX\ \latexword{figure} is. It is not really a figure at all: it is what \shortlamp\/ calls a {\em float}. That is, it is a chunk of text---or maths, or pictures, or whatever---that is treated as a whole and is printed in the first convenient place where there is room for it. It may contain any number of \latexword{caption}s (even none), each of which produces output beginning ``Figure \ldots''. I have used the word ``object'' for what you probably think of as a figure. You have to visually place the objects and captions within the \latexword{figure} so that the captions refer to the correct objects. If you do not want to force the two objects to the edges of the text, you can omit the \latexword{makebox} and put explicit space between the two \latexword{minipage}s, something like \begin{verbatim} \begin{figure} \centering \begin{minipage}[b]{0.5\textwidth} \end{minipage}% \hspace{1cm} \begin{minipage}[b]{0.4\textwidth} \end{minipage}% \end{figure} \end{verbatim} You can also stack two \latexword{minipage}s on top of each other in a single \latexword{figure}. If their total width is more than \verb+\+\latexword{textwidth} you do not even have to force them onto two separate lines. In this way you can glue five objects together in one figure in the arrangement shown in Figure~\ref{box}. Now to the lies. Since \BV\ is set in two columns, all the figures in this article have been done using \verb+\+\latexword{columnwidth} instead of \verb+\+\latexword{textwidth}. If you do the captions as I have suggested, then that for Figure~\ref{anova} appears to come out a little lower than that for Figure~\ref{hasse}. This is because the second \latexword{caption} contains the descending letter `g' whereas `variance table' contains no descenders. So the bottom of the `g' is lined up with the baseline of the first caption. To cure this, we can put in the useful command \verb+\+\latexword{strut}, which isn't in \shortlamp. It produces no visible output, and has height but no width. The principle is explained on page~100 of \shortlamp. The command \verb+\+\latexword{strut} produces a strut whose size depends on the current font size. It is big enough to cover the descenders and ascenders on ordinary letters. Two lines of ordinary text that both contain \verb+\+\latexword{strut} both take up the same amount of vertical space. So we finish the first and second captions with \begin{verbatim} table\strut} \end{verbatim} and \begin{verbatim} diagram\strut} \end{verbatim} being careful not to introduce an extra inter-word space before the strut. Any statistician reading this will probably object that Figure~\ref{anova} should be called a {\em table}. Here's the rub. I can put any number of \latexword{caption}s inside a \latexword{figure} and they will all come out ``Figure \ldots''. Similarly, I can put several \latexword{caption}s inside a \latexword{table} and they will all come out ``Table \ldots''. What I cannot do is mix ``Table''s and ``Figure''s in the same \LaTeX\ float. In particular, I cannot put a ``Table'' beside a ``Figure''. Nor can the average \LaTeX\ user. Let us hope that the \LaTeX3 team will give us this flexibility. \end{article}