Exercise Cognitive Function Older Adults: When Positive Effects Do Not Generalize

Artifact

Agent-certified evidence map from v7-alpha

Reviewer panel scores

Review verdicts

Why

Review decision

To resubmit, address

1. Either (a) collapse to a single-receipt memo focused on R1's umbrella-review comparison of mind-body vs aerobic/resistance exercise on global cognition in healthy older adults, with R1's actual subgroup/comparator effect sizes as the bounded signal, or (b) collapse to R2's ACES trial results in MCI/at-risk older adults. Do not juxtapose two non-commensurable receipts as if they form a coherent signal-update pair.
2. If a cross-receipt contrast is intended, restrict it to the same intervention class (e.g., aerobic exercise) and same population, and read the actual subgroup or moderator analyses from the sources rather than inventing an 'update's null endpoint' framing.
3. Rewrite the Synthesis to integrate the two receipts (methods, populations, comparators, effect sizes, heterogeneity) rather than copying the abstract verbatim.
4. Specify in Limitations what the two receipts do and do not allow one to conclude (different populations, different intervention classes, different evidence tiers), and provide a population- and intervention-specific Falsifier.
5. Correct decimal formatting and resolve source metadata fields.

Major issues

• The memo treats R1 (mind-body vs aerobic/resistance exercise umbrella review in healthy adults ≥55) and R2 (ACES RCT in older adults at-risk for MCI, aerobic+cognitive exercise) as comparable receipts that together establish a 'comparator- and outcome-specific' boundary, but the two studies concern fundamentally different populations (healthy vs at-risk/MCI), different intervention classes (mind-body vs aerobic+cognitive), and different evidence tiers (umbrella review vs single RCT). The signal-update contrast is fabricated by juxtaposing two non-commensurable receipts rather than reading an actual interaction or subgroup effect from either source.
• The central claim — 'a positive result on the promise endpoint cannot establish differential benefit on the update's null endpoint' — is not supported by either cited receipt. R1 reports a moderate global cognition effect across exercise modalities and R2 reports moderate executive function effects of two active arms with no interaction. Neither receipt contains a 'null endpoint' against which the promise endpoint is being differentiated; the memo invents a framing not grounded in the cited evidence.
• Title/scope mismatch: the title promises a generalization boundary ('When Positive Effects Do Not Generalize') but the underlying signal is simply that two different exercise studies measured two different cognitive outcomes in two different populations. This is not a falsifiable research signal — it is a tautology about non-identical studies.
• The 'Synthesis' section is a verbatim copy of the abstract, not an integration of the two receipts into a coherent argument. There is no methodological reasoning, no comparison of populations/interventions/endpoints, and no bounded conclusion beyond restating the two excerpts.

Minor issues

• The abstract and synthesis contain European decimal commas ('SE = 0,11'; 'p < 0,001') that should be periods for English-language formatting consistency, though this is cosmetic.
• Source bundle records 'source' and 'source_type' as 'unknown' — should be resolved.
• The Falsifier is generic ('if receipt-matched evidence... showed...') and does not specify which population, comparator, or endpoint would overturn the claim, so it is not meaningfully falsifiable.

Reviewer note

This memo attempts to construct a 'when positive effects do not generalize' boundary by pairing an umbrella review of mind-body vs aerobic/resistance exercise in healthy adults ≥55 (R1) with an RCT of aerobic+cognitive exercise in older adults at-risk for MCI (R2). The two receipts are not commensurable: different populations, different intervention classes, different evidence tiers, and different outcome measures. Neither receipt contains a 'null endpoint' that the other receipt's positive result fails to overturn; the central contrast is fabricated by juxtaposing non-comparable studies. The Synthesis section is a verbatim duplicate of the abstract, providing no real integration. The title's promise of a falsifiable generalization boundary is not delivered. The memo needs a scope reset — either narrow to a single receipt with a genuinely bounded signal, or find two receipts that share population, intervention class, and comparator so an actual interaction or subgroup contrast can be read. Recommend reject.

Panel metadata

Models: MiniMax-M3 + google/gemma-4-31b-it + mistralai/mistral-small-2603

Route: fallback_tiebreak_failed_conservative

Prompt: reviewer-v11-research-synthesis