The ABA’s amicus in Lee v. Tam in favor of neither party is largely strong (though I don’t agree with everything therein) but does contain one argument that is unworthy of the ABA’s trademark experts:
The theory that otherwise valid marks become unprotectable because they are unregistrable also cannot be reconciled with Section 43(c) of Act, 15 U.S.C. § 1125(c) (2012), which protects the distinctiveness of famous marks against likely dilution. Under that statute, whether or not a mark is registered is merely one of four nonexclusive factors used to determine whether a mark is sufficiently famous to qualify for protection. See 15 U.S.C. § 1125(c)(2). The nonexistence of a registration therefore does not preclude actions brought under that statute.
This is a classic fallacy of the excluded middle. It is completely consistent with §43(c) that dilution protection could be available for a registrable but unregistered mark but not for an unregistrable mark. The ABA takes the position that unregistrable only sometimes means invalid and unprotectable; I strongly doubt that it wants invalid marks to be given dilution protection—but on their theory that registration isn’t dispositive to dilution, such marks should be eligible. (Actually, the ABA's position that invalid marks ought to get some "protection" under the common law and §43(a) implies that maybe the ABA does want dilution to be potentially available. However, that "protection" has historically been not a prohibition on use but an obligation on the second-comer to sufficiently distinguish itself from another's generic, functional, or abandoned use, which is a confusion-based remedy that has no obvious purchase on dilution. Even the remedy suggested by the court of appeals in Belmora to implement Bayer's §43(a) right based on its Mexican rights was supposed to respect the defendant's rights to use its registered mark in the US.)
The difficulty with the ABA’s theory overall, though it is not an insurmountable one, is that there is no statutory guide for which unregistrable marks ought to be considered “invalid” and thus entirely unprotectable versus which should be merely unregistrable and protectable under §43(a). History may provide some guidance, but it’s actually pretty wobbly on a lot of the categories of unregistrable marks. This is one of the points I make in my forthcoming article on registration.