From Biomarkers to Behavioral Intelligence: Why Cortisol Alone Fails — and How Multimodal Longitudinal Modeling Unlocks Canine Cognition

Current approaches to canine cognition and welfare rely heavily on population-level aggregation and isolated physiological biomarkers such as cortisol. Drawing on empirical evidence of salivary cortisol's failure in real-world pet dogs (Ferrans et al., 2025) and the structural limits of the breed-average GWAS paradigm (Gnanadesikan et al., 2020), this paper proposes individual longitudinal behavioral intelligence as a new paradigm. It introduces the Cross-Modal Cognitive Index (CMCI) and the DogGraph architecture, formalizing canine cognition as a temporal modeling problem: Cognition(t) = f(Behavior(t), Physiology(t), Context(t), History). Cortisol is repositioned from primary signal to one noisy channel among many, weighted by its historical reliability for the specific individual. Barkley White Paper | Precision Behavioral Intelligence Series | No. 08.