A brain figure has to do two things at once: show a spatial pattern honestly, and survive being reduced to a two-column width in print. Most published brain figures fail at one of them, usually by rendering a continuous statistical map with a rainbow colormap that invents boundaries the data does not contain.
The choices that matter are mostly about colour and thresholding. Perceptually uniform sequential maps for magnitudes, diverging maps centred at zero for contrasts, and never a rainbow for continuous data — the perceived rate of change in a jet colormap has nothing to do with the numeric rate of change, which means the reader sees structure that is an artefact of the palette. Thresholding decisions deserve the same scrutiny: a map shown at an arbitrary uncorrected threshold, with the threshold unstated, is a figure that cannot be assessed.
Surface rendering versus volume slices is a genuine choice rather than a matter of taste. Surfaces show cortical patterns without the foreshortening that makes gyral effects look patchy on slices; slices remain the honest option for subcortical structures, which have no surface.
Related¶
Nilearn — the plotting workhorse
Multiple comparisons — what the threshold means
Reporting guidelines — what a figure must state
Brainstorm — electrophysiology visualization