Why it is a question
A19 §2: “Information can be stored perfectly well while the current cue fails to activate it.” That a cue exists that retrieves x does not imply the cue now given does. A19 §5 tabled what each intervention changes: a better cue changes access, a better decoder changes extraction, added context changes the information, more compute changes the resources, a better policy changes action. “Identical improvements in task accuracy could arise through any of these routes.”
A19 placed it in the pragmatic-transport line: “communication changes not only what information a receiver is given, but the addressing conditions under which existing information becomes recoverable.” CLAIM-298 states that as a thesis.
The record holds the mechanism A19 reached for and did not cite: LIT-269 (Bricken and Pehlevan, read), with SDM itself (LIT-921, read; LIT-928, Unavailable) and THEORY-261 on SDM recall. An answer is a design that holds the memory fixed, varies the cue, and measures retrieval and judgement separately, as A20 §5 proposed.