Deletion or epenthesis? On the falsifiability of phonological universals

• Ambiguity in analysis can compromise typological evidence.
• Use of a statistical-learning based procedure for analysis avoids circularity.
• Re-examination of the epenthesis typology data calls into question previous universalist claims.

This paper presents a revised typology of consonant epenthesis and explores the theoretical implications of such a typology. Through careful re-analysis, the basis for a proposed universal of coronal preference and dorsal avoidance is shown to be lacking in evidential support. In fact, epenthesis as a verifiable phenomenon – and not just a theory-internal label – is called into question once careful attention is paid to the issue of choosing between epenthesis and deletion as competing analyses of the same data. The ambiguity between multiple possible analyses, and the lack of formal transformations (from ‘data’ to ‘evidence’) are shown to be general problems within phonological theory. Phonological ‘universals’ can be invoked to arbitrate between competing analyses, but when the typological evidence for those ‘universals’ is derived from the same data, a problem of circularity arises. In order to break this closed loop, a quantitative evaluation metric is proposed that is theory-independent with regards to substantive universals. This metric is essentially a statistical threshold for learnability (itself empirically testable) that allows for independent testing of certain theoretical claims.