Comparing MaxEnt and Noisy Harmonic Grammar

MaxEnt grammar is a probabilistic version of Harmonic Grammar in which the harmony scores candidates are mapped onto probabilities. It has become tool choice for analyzing phonological phenomena involving variation or gradient acceptability, but there competing proposal making probabilistic, Noisy Grammar, derived by adding random ‘noise’ to constraint weights. In this paper these frameworks, and variants them, analyzed reformulating them all format where noise added candidate harmonies, differences between frameworks lie distribution noise. This analysis reveals basic difference models: relative probabilities two depend only on their scores, whereas it also depends violations incurred candidates. leads testable predictions evaluated against data variable realization schwa French (Smith & Pater 2020). The results support over Grammar.