The companion paper, The Relational Decoder (Zenodo, April 2026), proposed a theoretical framework in which social relationships operate through a predictive processing system constrained by social pain: the decoder builds a forward-flow model of another person, runs that model continuously, and generates aversive output when predictions are violated by relational asymmetry, withdrawal, or loss. That paper was a psychological account with biological gestures. This paper provides the biological substrate it assumed. A single computational principle recurs across four biological systems that predate social cognition: (1) visual filling-in at the physiological blind spot, in which primary visual cortex runs a generative model on surrounding context when retinal input is absent; (2) phantom limb pain, in which the somatosensory system generates aversive output for a missing limb whose body schema prior persists after amputation; (3) the auditory absence prediction error, in which the brain generates a dedicated neural signal — measurable subcortically, documented at single-neuron resolution in rats — when an expected stimulus fails to arrive; and (4) pain as hierarchical predictive processing, which establishes that the aversive output mechanism itself operates by the same Bayesian inference architecture as visual and auditory perception. Across all four systems, aversive response to signal absence scales with prior depth: the deeper the prior, the harder the decoder runs when the source goes silent. We argue by exaptation that social cognition did not independently evolve this architecture. It inherited it. The fill-the-gap-with-a-prior mechanism is present in early vertebrates. Its opioid pharmacology is conserved across mammals, from canine separation distress vocalizations to mu-opioid receptor knockout mice. The "social" in social cognition is the input domain, not the architecture. Evolution pointed an existing decoder at other minds. The 2025 identification of a dedicated absence prediction error signal in the auditory thalamus, combined with Willems et al.'s (2025) demonstration that omissions of expected aversive stimuli trigger reward-like prediction error signals in dopaminergic midbrain structures, provides strong evidence for the biological signature of the decoder running on a missing source. From this we derive the grief-as-omission-PE hypothesis: grief intensity should scale with relational prior depth — relationship duration, contact frequency, and forward-flow integration — independently of attachment style and secondary stressor load. This prediction distinguishes the decoder account from attachment theory and stress-and-coping frameworks and is testable at multiple levels of a five-level research hierarchy, from rat psychophysics to human bereavement neuroimaging. The two papers together constitute a unified account from neuron to relationship. The biological architecture is the substrate. The psychology is the deployment. Testing the substrate first, where measurement is precise, is the appropriate scientific sequence.
Lloyd Taylor (Wed,) studied this question.