Regular and polyregular theories of reduplication

Abstract

We explore the generative capacity of morphological theories of reduplication. We computationally classify theories of reduplication using a hierarchy of string-to-string function classes. Reduplication as a process requires only the regular class of functions. We show that various morphological theories necessarily treat it as a more expressive polyregular function, while others maintain regularity. We discuss the significance of this formal result for reduplicative functions and recognition.

On regular copying languages

“…The kind of recursive copying exhibited by w 2 n means that this language does not have the constant growth property that Joshi (1985) identified as a criterion for mild context-sensitivity. Excluding this recursive copying from phonology seems relatively well-justified, on the grounds that triplication is attested (Zimmermann 2019;Rawski et al 2023). But the situation may be different for syntax, where Kobele (2006), for example, has argued for recursive copying of the w 2 n sort on the basis of Yoruba relativized predicates.…”

“…For recent phonological analyses, see Zimmermann (2021a) and Zimmermann (2021b). For a more detailed discussion on the string-to-string function version of this problem, seeRawski et al (2023). [ 42 ]…”

Is grammatical tone item-based or process-based?

“…Canonical arguments against item-and-process approaches have to do with worries that they are too powerful; however, as Hockett (1954) & Anderson (1992) point out, these fears may have more to do with the fact that a specific formalism of process-based accounts is (or was, at the time) lacking. If one limits the types of processes possible in an item-and-process account, such an approach can be as restrictive as an item-and-arrangement account (see also computational work showing that item-based and process-based accounts are computationally equivalent (Karttunen 2003; Roark & Sproat 2007), or in fact that some process-based approaches are computationally simpler than item-based approaches (Dolatian et al 2021; Rawski et al 2023)).…”