Does the mind selectively activate frames from only some words?

A small exchange exposes the structure. Adrian said "bullseye" as a compliment. Claude said "so to speak," reading it as metaphor. Adrian then named the wordplay: the design context had a dot, a cover, and an arrow through it — bullseye + dot + arrow form an archery frame. Three words activate the frame; "cover" sits adjacent linguistically (it's in the same description) but does not enter the frame. The human mind does this automatically, holding three words tight in resonance and letting the fourth alone.

The cognitive operation here is not similarity computation. Cosine-style measures over word embeddings would treat all four words by their pairwise relations and would not produce the selective-frame effect. The selectivity is doing work computational similarity does not capture: it suppresses some adjacencies even when they are linguistically present, in order to surface the frame that holds three of the words together. The frame is detected by what it activates, not by what it computes.

This bears on the chronic difficulty AI has with wordplay and joke-detection. The standard diagnosis is that AI lacks world knowledge or fails at pragmatics. The diagnosis available from the resonance frame is more specific: AI processes the words serially or in parallel without the selective-suppression that frame-activation requires. It treats all available linguistic relations as candidate signal, which is the opposite of what frame-activation does. Why do AI systems miss jokes and wordplay so consistently? is the AI-side companion claim.

The strongest counterargument: this is just contextual pragmatics, well-studied in linguistics and learnable from data. True for many cases, but the selectivity in the bullseye example is not driven by context-as-conventional-co-occurrence — "cover" co-occurs in the same context but is suppressed; "arrow" is more distant in the surface text but is amplified. The selectivity tracks frame-coherence, not co-occurrence frequency.

This connects to longstanding work on frame semantics (Fillmore) and on the non-compositional nature of much human meaning-making. What the bullseye example adds is the specific cognitive operation: not just that frames matter, but that the mind selects from available linguistic material to activate a frame, and the selection is the meaning-making move.

Source: Making Sense - brief for co-authored essay on language