This paper re-evaluates the influential "literary fiction conjecture," which posits a causal link between reading literary fiction and enhanced Theory of Mind (ToM). This claim, first empirically demonstrated by Kidd and Castano (2013), is now mired in a significant replication crisis, with high-powered studies finding no such effect. This impasse, it is argued, stems from the conceptual weakness of "literary fiction" as a monolithic experimental variable. This paper synthesizes cognitive psychology with the literary-linguistic field of cognitive poetics to propose a new, mechanistic framework. It argues that the observed effect is not a product of genre, but of specific textual features (e.g., Free Indirect Discourse, high "levels of intentionality" ) that function as a cognitive demand on the reader's ToM faculties. This process, however, is critically moderated by the reader's psychological state of Narrative Transportation. The paper presents a three-phase Literary-Cognitive (L-C) model that traces the pathway from (1) Textual Triggers, through (2) Narrative Transportation and Social Simulation (engaging the dorsomedial prefrontal cortex (dmPFC) network ), to (3) Cognitive Exercise and temporary ToM enhancement. This model resolves the apparent contradiction between the original finding and its subsequent failures, providing a more precise, testable, and durable account of narrative's influence on social cognition.