Authors should ensure that each figure contains the data needed only to answer a specific question. If a figure is meant to answer too many questions at once, it may appear cluttered and confusing to readers. Authors should consider breaking complex figures into multiple simpler figures, if possible.

In

Figure 2A, which has a scale that starts at 90, it appears as though substantial differences were found between treatment groups. In

Figure 2B, which has a scale that starts at 0, it is clear that the differences between groups were actually much smaller. Bar graphs should always have scales that start at 0. Smaller scales are more appropriate for other types of figures, such as line graphs and dot plots.

In addition, figures should draw readers' attention to the information in the figure, not to the figure itself.^{3} In statistical graphs, authors should minimize the number of nondata elements.^{3,6} For example, authors should include only necessary symbols and labels and consider labeling every other index line (eg, 20, 40, 60 instead of 10, 20, 30, 40, 50, 60) to keep the focus on the data.^{7} Authors should also avoid a 3-dimensional format for a figure that has 2 dimensions of data. The display may make it more difficult for readers to identify important comparisons or determine exact values.^{2}