A gradual path to hierarchical phrase-structure: Insights from modeling and corpus-data