Highly grammaticalized forms have primarily a structural function rather than a meaning.

For instance, in many languages, the paradigm of pronominal indices cross-referencing the subject is identical with the paradigm of indices cross-referencing the possessor. For instance, ‘I sleep’ has the same pronominal form as ‘my sleep’. This is generally amenable to a (synchronic or diachronic) structural account (viz. the verbal constituents whose subject is thus marked are or were nominalized). It would be wrong to interpret the situation in terms of a possessive meaning of the subject relation, so that the construction of ‘I sleep’ “really” means ‘my sleeping’ or even ‘I am the possessor of a process of sleeping’.¹

¹ This is done throughout in the Lacandon grammar of Bruce S. 1968.