Why is grammaticalization irreversible?

Grammaticalization, i.e. the change by which lexical categories become functional categories, is overwhelmingly irreversible. Prototypical functional categories never become prototypical lexical categories, and less radical changes against the general directionality of grammaticalization are extremely rare. Although the pervasiveness of grammaticalization has long been known, the question of why this change is irreversible has not been asked until fairly recently. However, no satisfactory explanation has been proposed so far. Irreversibility cannot be attributed to the lack of predictability, to the interplay of the motivating factors of economy and clarity, or to a preference for simple structures in language acquisition.
    I propose an explanation that follows the general structure of Keller's (1994) invisible-hand theory: Language change is shown to result from the cumulation of countless individual actions of speakers, which are not intended to change language, but whose side effect is change in a particular direction. Grammaticalization is a side effect of the maxim of Extravagance, i.e. speakers' use of unusually explicit formulations in order to attract attention. As these are adopted more widely in the speech community, they become more frequent and are reduced phonologically. I propose that degrammaticalization is by and large impossible because there is no counteracting maxim of "Anti-extravagance", and because speakers have no conscious access to grammaticalized expressions and thus cannot use them in place of less grammaticalized ones. This is thus a usage-based explanation, in which the notion of imperfect language acquisition as the locus of change plays no role.