Title Pending 18053

This is an accepted article with a DOI pre-assigned that is not yet published.This paper introduces a computational model of a novel hierarchical approach to parameter setting. We argue that parameters, whether innate or derived from innate properties or another source, remain necessary within the Minimalist approach. We distinguish and compare two methods of parameter setting in the computational modeling of language acquisition: the direct parameter setting approach and the grammar selection approach. The merits and limitations of each approach are discussed. We observe that the strengths of one address the weaknesses of the other, and vice versa. Consequently, we suggest a hybrid method termed the Clustering Approach, according to which the parameters are hierarchically organized based on their occurrences in the input. The Clustering Approach retains the benefits of both grammar selection and direct parameter setting approaches while circumventing their shortcomings. We compare the Clustering Approach to a previous hybrid method, illustrating how it resolves a significant issue found in the latter. The Clustering Approach assumes that parameter setting is strictly data-driven due to its accurate update mechanism during learning. Our simulation results demonstrate that the Clustering Approach offers a unique framework for modeling the acquisition of parametric variation, especially in scenarios involving a larger set of parameters.