How clitic flags become affix flags: The role of word order

In this presentation, I discuss the well-known diachronic path from adposition (or clitic flag) to case affix (or affix flag), and I suggest that word order plays a bigger role in it than has generally been thought. There is a widespread view according to which clitics are intermediate between free words and affixes, both diachronically and synchronically. However, this view is not based on clear notions of “affix” and “clitic”, and if these are defined rigorously (Haspelmath 2023), the main distinction between them is that affixes are selective (occurring always on the same word class), while clitics are non-selective. For example, Turkish -e and English to are both dative flags, and they are both phonologically dependent and variable, but Turkish -e occurs only on nouns (misafir-e ‘to the guest’), while English to also occurs on possessive pronouns (to my guest) and other prenominal modifiers. It is for this word-order-related reason that we say that English to has not become a case prefix. It is not clear that it is helpful to say that English to is “less grammaticalized” than Turkish -e, or “not yet advanced to affix status”. It has sometimes been observed that there are fewer case prefixes than case suffixes (Reh 1986), and while there could be discourse reasons for a general suffixing preference (Himmelmann 2014), it seems quite likely that word order differences play an important role in determining the affix or clitic status of function morphs.