I know that there are many opinions around about colours in scientific graphs. I have followed the recommendations (and colours) by Paul Tol, finding the argument about readability and attention to colour-blind readers convincing. Indeed, I’ve commented on these things to colleagues, always with an open mind as to the exact aesthetics. I know, for example, that many swear by ColorBrewer.
Recently, however, I got a comment by a reviewer who disliked my choice of colours strongly enough to comment on. We can discuss the quality of the figure, but it was particularly the colours the reviewer picked on and the difficulty of distinguishing them (something Paul Tol optimizes). While the recommendation of the lovely viridis palette was probably misguided (we’re talking about categorical data, not continuous data), it made me wonder whether we’re entering an age where attention to colours (in principle a good thing) has become so strong and opinionated that we start arguing about which colours we should use in plots — not what we communicate…
Didier Ruedin 