Islands of Inference: A Neurodevelopmental Architecture for Fixed Delusions

Delusions can be strikingly selective: individuals may hold highly elaborated, unshakeable beliefs in one domain while remaining otherwise cognitively intact. Models emphasising dopaminergic salience or belief-updating failures account for the intensity and persistence of delusions but struggle to explain their circumscribed nature.

This paper develops a conceptual framework that complements dominant functional accounts. Subtle neurodevelopmental variation in cortical micro-architecture, as commonly occurs, may give rise to focal cortical regions ("islands") with greater representational depth than their surrounding networks. When such a region supports high-order social inference - particularly advanced theory of mind - it may generate internally coherent narratives that exceed the integrative capacity of neighbouring circuits. These inferences cannot be adequately contextualised, simulated, or revised by the broader network and therefore become fixed. The "Islands of Inference" model accounts for several core features of schizophrenia, including the social content of delusions, their typical onset in late adolescence, their resistance to counter-evidence, and their progression over time. It also clarifies the distinct roles of dopamine and cannabis: neuromodulators amplify the influence of the island's output without determining its content. Finally, the framework clarifies implications for empirical investigation. During delusional thought, focal regions within the mentalising network should support deeper internal representations than their surrounding circuits, even when those neighbouring areas are active and coherent.

This framework reframes delusions as reflecting a failure of architectural constraint rather than global cognitive dysfunction and suggests new directions for empirical investigation and treatment.