Paradigmatic predictability metrics as signals of phylogenetic relatedness: a proof of concept in Romance and Pamean diachrony

Morphological complexity metrics like entropy, and notions like the Paradigm Cell-Filling Problem, have recently (re)gained popularity for the synchronic analysis of inflectional systems. The potential of these quantitative approaches, however, remains largely untapped with respect to diachronic research. This paper constitutes a first exploration of whether/how these methods can be used profitably in this domain. It consists of a proof of concept that explores quantitatively the diachronic stability of different aspects/metrics of an inflection class system's predictability relations (in Romance), as well as the possibility to use these diagnostically to identify cognate tenses and inflection classes in related languages (in Pame) even after most of the morphological material has been replaced or altered beyondrecognition. The results suggest that quantitative predictability metrics can preserve a strong phylogenetic signal over extended periods of time and fruitfully complement traditional qualitative approaches to explore diachronic relations.