The article "Empirical Tests of the Gradual Learning Algorithm" by Paul Boersma and Bruce Hayes evaluates the capabilities of the Gradual Learning Algorithm, a constraint-ranking algorithm for learning optimality-theoretic grammars. The authors compare it with the Constraint Demotion algorithm, which initiated the learnability research program for Optimality Theory. The Gradual Learning Algorithm is argued to have several advantages, including the ability to handle free variation, deal effectively with noisy data, and account for gradient well-formedness judgments. The study examines Ilokano reduplication and metathesis, Finnish genitive plurals, and English light and dark /l/ as case studies. The algorithm is shown to successfully learn these phenomena, demonstrating its robustness and adaptability. The article also discusses the continuous ranking scale and stochastic candidate evaluation, which are key components of the Gradual Learning Algorithm. The authors conclude that the algorithm's ability to handle free variation is crucial for modeling human language learning capabilities.The article "Empirical Tests of the Gradual Learning Algorithm" by Paul Boersma and Bruce Hayes evaluates the capabilities of the Gradual Learning Algorithm, a constraint-ranking algorithm for learning optimality-theoretic grammars. The authors compare it with the Constraint Demotion algorithm, which initiated the learnability research program for Optimality Theory. The Gradual Learning Algorithm is argued to have several advantages, including the ability to handle free variation, deal effectively with noisy data, and account for gradient well-formedness judgments. The study examines Ilokano reduplication and metathesis, Finnish genitive plurals, and English light and dark /l/ as case studies. The algorithm is shown to successfully learn these phenomena, demonstrating its robustness and adaptability. The article also discusses the continuous ranking scale and stochastic candidate evaluation, which are key components of the Gradual Learning Algorithm. The authors conclude that the algorithm's ability to handle free variation is crucial for modeling human language learning capabilities.