AbstractIt has long been recognized that many instances of change that have been discussed within the framework of grammaticalization studies notoriously defy categorization, for instance because they share properties of grammaticalization and lexicalization (Brinton & Traugott 2005), or because they share some properties of grammaticalization, but not all of them, as in the case of discourse markers (e.g. Ocampo 2006). In order to avoid these classification issues, we will argue that it is more useful to reduce grammaticalization and related changes to their “main mechanisms” (formal reanalysis and semantic reinterpretation), “primitive changes” (micro-changes on the levels of phonology, morphology, syntax, semantics and/or discourse), and “side effects” (e.g. obligatorification or layering). In grammaticalization and related changes, formal reanalysis and semantic reinterpretation tend to coincide with different sets of primitive changes. Primitive changes will be defined as ternary parameters with the values reduction, expansion, or zero, and it will be seen that they tend to cluster in different ways. Some of these clusters may coincide with changes traditionally labeled “grammaticalization”, “degrammaticalization”, or “lexicalization”, but changes may also cluster in alternative ways. This novel approach to composite changes we term the “clustering approach”, and we aim to show that this model of analysis allows for a more fine-grained account of composite changes than definition-based taxonomies.
Reconciling optimization with secure compilation