(… at least for large documents)

Two week-ends ago, I spend a whole day trying to apply a consistent style to a thesis. I spent hours trying to be obeyed by a word processor because it would systematically change the style of some element, somewhere in the 100-or-so pages. Including figures was also a nightmare: we had to keep an eye on the (limited) memory of the computer (otherwise we got unexpected screen freeze, a lot of noise from the hard disk (paging), etc). Generating a bibliography was also another daunting task, even with the use of a dedicated reference manager …

Now I don’t know if I have to blame what we call “word processors” or human laziness …
First, these pieces of software are not really “processing words”. Grep, sed, vim, LaTeX, XSLT… actually are processing words, transforming them from a raw text format to something else, possibly transforming the whole text into something more readable on paper.
Second, word processors tend to let people write whatever they want, in whatever style (some word processors even retain the different style of things pasted into the text), in whatever order. In this aspect they resemble mind mapping software. But if a draft of experiment can be written like this, a good thesis needs a good planned underlying structure (imho). And current word processor software doesn’t push you to do this.

In this aspect, LaTeX may have a steep learning curve but it somehow force you to think about the structure before adding text and embroideries. Changing the style is only done in one place and is easy (once you know the command to type). I already used LaTeX for other reports, presentations and thesis and will certainly keep it for my Ph.D. thesis this year.