Hierarchical processing of continuous natural language: From monologue to conversation

Converging evidence from cognitive neuroscience has shown that linguistic information in continuous natural language is processed and integrated across multiple timescales and represented along cortical gradients corresponding to temporal hierarchies. However, most existing studies have focused on monologic language, leaving open the question of whether hierarchical processing also extends to naturalistic conversation. This review synthesizes theoretical frameworks and empirical evidence on the hierarchical processing of continuous natural language. By integrating findings from linguistics, cognitive science, and neuroimaging, we argue that both monologic and conversational language exhibit structured organization across linguistic levels such as words and sentences, while conversational language additionally involves interaction-dependent units such as turns and topics. Conversational processing also depends on interpersonal mechanisms such as alignment, grounding, and prediction between interlocutors. Recent hyperscanning studies provide emerging evidence that, for conversational units such as turns and topics, level-specific neural signatures of conversational organization appear more evident at the dyadic than at the individual level. Together, the available evidence supports the view that hierarchical organization is a general property of continuous natural language, while highlighting important differences in how such organization is neurally expressed in monologic and conversational settings.