Normative latent cognitive structure metrics delineate cognitive heterogeneity across schizophrenia liability spectrum in healthy controls, siblings, and patients.

Cognitive heterogeneity is a fundamental aspect of schizophrenia (SCZ), but conventional domain-specific score-based approaches reduce cognition to single values, overlooking its structure and primarily reflecting the severity of cognitive impairment rather than qualitative differences. To address this gap, this study derived and validated a normative latent cognitive structure (N-LCS) serve as a reference to evaluate whether individual cognition aligns with the norm, providing a basis for characterizing cognitive heterogeneity in SCZ. Three deviation metrics represent the deviation magnitude (CDM), the angular deviation (CDA), and the domain driving the deviation (CDD). CDM showed a linear increase from healthy controls (HC) to unaffected siblings (UC) and SCZ patients. CDA differed between HC and the other two groups, but not between UC and SCZ. Reasoning remained the most frequent CDD, whereas processing speed declined across the SCZ liability spectrum. These findings imply that heterogeneous changes in cognitive structure already emerge in the UC stage, preceding both clinical onset and measurable score-level differences. Cognitive heterogeneity in SCZ can be delineated by three deviation dimensions: 53.1% of patients had reasoning as the CDD, with moderate CDM and CDA, suggesting that their cognitive structure is preserved relative to the N-LCS, whereas 31.2% had processing speed as the CDD, with larger CDM and intermediate CDA, and higher HAMA scores. The N-LCS captures signal beyond conventional score-based measures, supporting its utility and informing individualized treatment.