The number of phenomena which are gathered together under the term ‘grammaticalization’ is quite large and in some ways quite diverse. For the different types of grammaticalization similar motivating factors have been suggested, similar principles, clines and hierarchies. Some of Lehmann’s (1982[1995], 1985) parameters, which have long been considered to characterize processes of grammaticalization, are now under attack from various quarters, and indeed the phenomenon of grammaticalization itself has been questioned as an independent mechanism in language change. This paper addresses a number of problems connected with the ‘apparatus’ used in grammaticalization and with the various types of grammaticalization currently distinguished. It will be argued that we get a better grip on what happens in processes of grammaticalization and lexicalization if the process is viewed in terms of an analogical, usage-based grammar, in which a distinction is made between processes taking place on a token-level and those taking place on a type-level. The model involves taking more notice of the form of linguistic signs and of the synchronic grammar system at each stage of the grammaticalization process.