I don't think it has much to do with unionization at all so much as that there is generally an overabundance of students applying to graduate school. Combined with the fact that the NIH/NSF more or less set the wages anyway, there really isn't anywhere for a union to lean. After all, unless you could organize a strike of a significant portion of the graduate student populace in the US, I doubt Congress or the NIH/NSF will take any notice.
Overabundance is definitely an issue, and one can indeed argue that in the biomedical sciences wages are set or guided independent of institution, and therefore similar work could expected from trainees. However, in the humanities (and whenever there's a teaching load involved), anecdotally it seems more ripe for abuse--the work you do in TAing a class may not at all benefit your future career but is just required of you.
That being said, the overabundance itself can also be a need for unionization after a fashion: can organized labor influence the number of new students to reverse the trend? Not to mention the general lack of increases in quality of life: wage levels of graduate students have been flat for decades (don't have a graph to point to, but it's in http://www.amazon.com/Economics-Shapes-Science-Paula-Stephan...).
