False Memories as the Error Term of Learnable Construction
Type: intersection (second-order) Slug: intersection—false-memories-error-term-learnable-construction Parents: intersection—false-memories-construction, intersection—learnable-nature-hippocampal-construction Last updated: 2026-05-14 Epistemic status: Extrapolative
The convergence
Construction is learnable structure prediction (intersection Learnable×Construction). False memories are over-construction — the system constructs scenes that are too semantically coherent (intersection F). Combined: false memories are the error term of the learnable construction optimisation — they reveal where the learned “laws of experiential coherence” break down.
Why this isn’t obvious from either parent
Learnable×Construction (intersection 5) frames construction as discovering laws from data but doesn’t consider failure modes. False Memories×Construction (F) frames false memories as over-construction but doesn’t connect to learnability. Together: false memories aren’t just errors — they’re informative errors. They reveal the specific locations in the learned construction grammar where the prior is too strong relative to evidence.
Generative prediction
If the temporal pole’s similarity code is a learned prior (not innate), then false memory rates should change systematically with domain expertise. Experts in a domain should have different false memory patterns than novices — not fewer false memories overall, but false memories at different locations in the similarity space, reflecting their restructured prior. A sommelier’s false memories about wine should differ from a non-expert’s not in magnitude but in direction — they false-recognise wines that are similar in the expert similarity space, not the novice one.
What makes this non-trivial
Standard false memory theory predicts more expertise = fewer false memories (better discrimination). This predicts restructured false memories — same error rate, different error pattern. This is testable and would distinguish the learnable construction account from all prior theories.
Falsification: If false memory patterns are not informative about construction grammar structure (random errors, not systematic), the error-term claim is false.