The Morphological Basis of Paradigm Leveling

This chapter argues that the Latin honor change was caused by more than simply a sporadic pressure for paradigm uniformity or Uniform Exponence constraints to assert themselves over IO-Faithfulness constraints. It shows that the spread of [r] to nominative forms did more than just create uniform paradigms; it also extended a pattern of non-alternation that was already dominant in the lexicon. Details of the change, such as its restriction to polysyllabic nouns and non-neuters reflect the fact that these were especially strong contexts for [r] stems. Furthermore, the ‘backwards’ direction of the leveling, with oblique forms influencing the nominative singular, can be explained by a particular model of Latin noun paradigms, in which an oblique form served as the base, and nominative forms were derived from oblique forms by rules operating on surface forms. More generally, this result provides evidence for a model of paradigm learning in which learners choose the base form that is the most informative — i.e. that preserves the most distinctions between classes of words, and allows the remainder of the paradigm to be predicted with the greatest accuracy and confidence.