* the explanation of 8.6 is wrong - it's cbind that's coercing all columns to a common type, not data.frame
* 8.7 is a bad example of good R code - use column names not indices!
* in 9 (and elsewhere) it's not necessary to continually strips names off vectors or use return in functions
* 11.2.1 to count TRUEs in logical vector, sum it.
She is going to start using it in January when she gets some new corpus material.
She also said that they are going to use Tinn-R.
I have to add, though, that I (as a non-programmer and aspiring social scientist) liked this tutorial much better than many other R tutorials out there. In know a little bit about programming but that was enough to pretty much understand the whole tutorial. (I have the sneaking suspicion that I'm an outlier so don't expect many other social scientists to know about basic CS concepts.)
(And boy am I happy that I now can feel confident ditching SPSS in favor of R. That gruesome SPSS!)
2) The graphics in R are pretty good. I've used matlab and matplotlib and gnuplot. I'd say comparable functionality to matplotlib in 2D, with better 3D graphics via rgl (though I guess in that case you'd use a python OpenGL library).
1 week to get used to the syntax and do basic data manipulation such as contructing, editing data frames, reading and writing files etc. You can also get going with the simple graphing functions quite quickly.
1 week to learn the statistical stuff - scatter plotting, correlations, linear regression modelling. Of course, I'm talking about just learning the basic syntax. Doing regression properly could take a long time to learn.
The R Book by Crawley is quite a good introduction, especially if you want to brush up on some basic statistics. It's not very in-depth though, so you might have to move on to more advanced texts after a month or so.
Downsides include poor programming tools/debugger and pass-by-value. Still, it's one of my favorite languages.