Linking Synchronic and Diachronic Analysis of Paradigms

Speakers of even highly irregular inflecting languages can usually produce all forms of all lexemes (e.g. do, does, did). How they can do so on the evidence of skewed and partial input is known as the Paradigm Cell Filling Problem.

The increased availability of large inflected lexicons, and computing power and tools to assess cell-to-cell predictive uncertainty have enabled us in recent years to automatically measure multiple aspects of paradigm complexity and structural similarity. Unfortunately, this quantitative turn in morphology has been almost entirely limited to synchronic analyses.

This project seeks to harness the power of novel quantitative methods (e.g. Information and Set Theory, Bayesian models) for the study of paradigmatic change as well.

The focus will be on Romance verbs, as the most historically documented and best-understood inflectional system worldwide.Regarding synchronic variability, I will integrate the best and largest extant inflected lexicons in the family (5000+ complete verbs for the 5 largest Romance languages and Latin).

All forms will be annotated for lemma-level and cell-level cognacy to produce a reusable and expandable pan-Romance etymological inflected lexicon.

Turning to diachrony, one of the greatest obstacles to analyzing morphological change is distinguishing it from regular sound change. Given the large number of words and changes it is impossible to derive an optimal chronology of sound changes by hand.

I will therefore write software that applies sets of ordered sound changes automatically to large inflected lexicons (e.g. of Latin) to derive expected forms.

Comparing these to actual forms will reveal where morphological changes have occurred historically and enable us to explore what the best predictors are of analogical change: frequency, predictability, meaning, etc.

This will allow us to evaluate more tractably competing hypotheses of phonological and morphological change and trends of change.