The evolution of morphological predictability in Romance paradigms

Most languages make use of different inflected forms of a word (e.g. Eng. go, goes, gone, going, went) to convey grammatical meanings. In many languages, these sets of word forms, known as 'paradigms', can show very substantial complexity. Given the large number of words in a language, it is impossible to evaluate whole inflectional systems by hand. In recent years, however, we have witnessed advances in computing power and the development of tools and resources that make it possible to do this automatically (see Ackerman & Malouf 2013, Stump & Finkel 2013, Beniamine 2018, Kirov et al. 2018). We now have precise assessments of different systems, where we can look at multiple measures and facets of morphological complexity: the number of classes, the number of principal parts of different kinds, the uncertainty (i.e. conditional entropy) involved in predicting one form from another, etc.

While the analysis of paradigms has changed radically over the last decade, this has impacted primarily paradigm synchrony. The novel quantitative assessments of the Paradigm Cell Filling Problem (Ackerman et al. 2009), i.e. the challenge that speakers of highly-inflecting languages with inflection classes face to produce all forms of any lexeme, have barely extended to the analysis of paradigm diachrony or evolution (see e.g. Maiden 2018). Given that paradigms have been widely considered to be among the most stable and resilient structures in language change (e.g. Meillet 1958, Nichols 1986), adapting these approaches to the study of morphological change and analogy would open up heretofore unexplored domains of enquiry, and the possibility to answer questions such as: Can quantitative predictive structure in paradigms be used to infer cognacy relations between related languages and to reconstruct ancestral paradigms? Is it possible to distinguish and quantify the role and weight of analogical vs sound changes in paradigmatic structural change?

This project will attempt to answer these and related questions by focusing on the Romance language family, specifically on their verbal inflection. Romance provides the ideal setting for this endeavour because i) adequate resources exist for a number languages (i.e. large and reliable inflected lexicons [Bonami et al. 2014, Pellegrini & Cignarella 2020, Beniamine et al. 2021, Herce 2023, Herce & Pricop 2023] and compilations of ordered and automatically-applicable historical sound changes [Marr & Mortensen 2022]), and ii) there is a well-documented direct ancestor to the whole family, i.e. Latin. The project will attempt to answer its questions from two angles:

• Through comparative work, i.e. starting from synchronic inflectional systems. I will survey every facet of verbal inflectional complexity in Spanish, Portuguese, French, Italian, and Romanian, assessing variation within the family compared to unrelated inflectional systems, and to artificial randomly-generated systems. The goal is to assess what aspects of morphological complexity are more stable and hence more useful for potential use in the reconstruction of ancestral paradigms, and to distinguish general human cognitive and language-related properties of inflectional systems from purely mathematical dependencies between the various complexity metrics that have been proposed in the literature.
• Through computerized forward reconstruction (see Sims-Williams 2018, Marr & Mortensen 2022), i.e. starting from the Latin verb. I will compile all sound changes attested in the evolution from Latin to the major Romance languages, and apply them to a Classical Latin verbal inflected lexicon (see Pellegrini & Passarotti 2018), to derive *expected* modern paradigms. Differences between the expected and actual modern inflectional forms will be generally indicative of morphological change. This approach will allow us to quantify the irregularizing/complexifying effect of regular sound change vs the simplifying/regularizing role of analogy in paradigmatic change (see Surtevant 1947).

Projektseite SNF