Beyond Havlik's Law: a dataset for quantitative investigation

This article explores a methodology for large-scale quantitative study of the jer change as it is reflected in Contemporary Standard Russian (CSR) as represented by Russian Wiktionary. The aim is to find out what proportion of the modern forms can be accounted for by a pure unmodified application of Havlik's Law, so that exceptions can be more systematically identified and investigated. We employ a computerised inflectional morphology of LCS to generate the ancestors of CSR forms, as well as a forward-reconstruction algorithm that attempts to convert these into CSR, with a strict implementation of Havlik's Law. Where the forward-reconstructed forms do not match the Wiktionary forms, we manually annotate them for the potential causes of the missed match, including the ones most frequently invoked in the literature: “difficult” consonant clusters, analogy and Church Slavonic influence. We find that especially consonant clusters and Church Slavonic influence are prominent potential causes. Overall we find that Havlik’s law holds in a large majority of cases. We also provide a small pilot study where we test some of the morphophonological rules proposed by Isačenko (1970) to demonstrate the methodology’s potential.