Top critical review
15 people found this helpful
Buy this book, but do not read it
on 7 December 2013
This book has good and bad aspects. First let me look at the bad part: the text Tufte has written. The problem is, most of it is founded on a single principle: maximise the information-to-ink ratio. Now as a scientist myself, this sounded like a great idea. It would not just improve your graphs (and similar diagrams), but do it in a fairly systematic way. Why add more stuff to your graph if it doesn't add any more information? It will only confuse the reader for no benefit. It seems so simple. So obvious. And yet it is 100% wrong.
The problem with this idea is that it is based on a fundamentally flawed view of human perception. We don't just see individual blobs of inks (or darkened pixels), so that adding more makes it harder. Instead we see the various shapes they form (search the web for "Kanizsa's triangle"!), and we should be aiming to reduce the complexity of this. For example, if you have several graphs next to each other, then putting a box around each will keep them visually distinct, so you can focus on one at a time with no conscious effort. If you leave out the bounding boxes then they become a jumble of tiny objects that take some effort to group visually. A very small amount of effort, admittedly, but you've made it harder for no reason. But Tufte HATES putting boxes around things! After all, you've certainly added more ink, and added no more information, so by his flawed rule you have made things unambiguously worse.
So now for the good. Why should you buy this book if not to read it? Because it is filled with pictures representing data from a myriad of sources. Some of them are effective, some are not. Some are beautiful, some are ugly. All are worth reflecting on.
In conclusion: Ok, I exaggerate when say "do not read it". Go ahead and read what has Tufte has to say, since it too is worth reflecting on. But make sure you bear in mind that everything he says is based on a flawed principle. So when he tells you that you should get rid of gridlines on a graph or in a table, consider whether it's really a good idea. Perhaps you shouldn't, or perhaps you should just make them lighter. Perhaps even, after all, you really should remove them. Just don't take his word for it that it's a guaranteed win.