An Experience-Dependent Relational Memory Architecture Based on a Numerical Relational Field and Recursive Relational Structures

This study presents and experimentally evaluates a non-address-based computational memory architecture in which prior experience persists not as stored event copies or fixed semantic labels, but as changes in a persistent relational composition topology. The architecture operates on anonymous numerical states and recursive relational structures. Its memory function is tested through repeated exposure, interference by intervening experiences, relational generalization and discrimination, and direct intervention on persistent state across 24 pre-specified stimulus families. Exact, near, and structurally isomorphic inputs reuse previously formed structure, whereas inputs containing the same numerical values with altered relational topology behave like fresh experiences. Direct topology intervention further shows that the persistent composition structure, rather than coordinate identity or transient runtime state, determines subsequent structural reuse. The results support persistent relational composition topology as a computational state variable for experience-dependent memory. The study is limited to the computational architecture and its experimentally verified memory behavior; it does not claim identity with biological neural mechanisms. This version revises the manuscript framing to present the work explicitly as an independent computational-memory architecture study and clarifies the experimental scope, controls, and interpretation. The underlying experimental results are unchanged.