Sorry to be late to the game and thanks for the interesting conversation (I love reading these things after the fact – everyone has said everything there is to say, and I get to soak it all in). I would add only a couple of things. First, whatever courts purport to do analytically, it strikes me that they have very different predispositions for previously-distinctive words than for marks that began as generic/descriptive terms. As long as there’s no evidence that the TM holder itself participated in genericide (cf. Bayer), and as long as there’s a satisfactory alternative (“Kleenex brand tissues” “Xerox brand copier” vs “Thermos”), courts seem loathe to abrogate previously-existing rights. I’ve never studied genericide so this may be off-base but it’s an impression that I have..and as Mike points out, even when they find a mark generic, they will sometimes find a way to provide limited protection anyway. On the flip-side, they seem equally loathe to remove a word from the vernacular if it’s one of a handful of terms that logically encapsulate a product. (It’s almost like a gut-level, implicit merger doctrine, I think – doesn’t that feel like what was going on in Pretzel Crisp?). Second, I just want to note one point in Kevin’s last comment – the idea that names are different from descriptive terms. If something is perceived as a surname, then it’s essentially treated the same as a descriptive term – i.e., it requires secondary meaning for protection. I think there’s an argument that “Pilates” would not be perceived as a surname, so your point above (that the situation in Pilates was inherently distinctive vs generic, rather than descriptive vs. generic) strikes me as right; but I just wanted to make the TM-geek observation that surnames are generally not treated as inherently distinctive.
Cheers and hope you are all well – Stacey