Resonance
In substrate dynamics, resonance is a measure on the substrate itself, not on any reader: how strongly a set of frames belongs together. It has two ingredients — the coupling already built across them (Hebbian, accumulated by co-activation over time) and the semantic coherence of their content. Coupling carries the history of use; coherence carries the fit of meaning. Resonance is the two together, which is why frames that have been worked together and are about the same thing resonate most.
What resonance does is set what gets retrieved together. A retrieval gathers the units whose resonance clears a threshold — resonant units come back as a group. Held high, the threshold returns only the most resonant: the coarse summary at the top of a subgraph. Lowered, retrieval reaches deeper and more of the detail below clears the bar, down to fine grain. So resonance, read through a threshold, is the knob behind granularity — how coarse or fine the retrieved context is (see fractal-composition). Concretely, the threshold is a setting in the re-ranker.
Most retrieval systems treat semantic similarity as the cardinal signal for what belongs together. We read the second-order effects on top of it: co-retrieval and action — what gets used and worked on together — alongside structural composition. Belonging is set by use and structure as much as by meaning. This is the quantity a membrane’s channels gate: the membrane is the boundary, the threshold on resonance is what opens a channel, and how far the threshold drops is how deep retrieval reaches. Permeability, depth, and granularity are the same gate read at different grains.
Open problems
- The function that combines coupling and coherence into a single resonance value — product, weighted sum, something with a threshold of its own — is unsettled.
- Whether resonance is best read per-edge, per-frame, or per-subgraph, and how those aggregate.
- How resonance relates to the variety a retrieval admits: depth and breadth may trade off, or move together.