# Adversarial review log — dopamine-schizophrenia

Authoring model: Sol 6.1. Adversarial reviewer: Claude Opus 5.5 (Anthropic). Date: 2026-10-14.
Round 2 of the Sol 6.1 adversarial review programme (group 3).

| Item | Value |
|---|---|
| Level before → after | Γ (Mechanistically plausible, partial system support) → Γ (unchanged) |
| Confidence before → after | Moderate → Moderate (unchanged) |
| Role / Momentum / Stability / Divergence | unchanged (reinforcing / stabilizing factor; Mechanistic only; Stable; Contested) |
| Findings | Critical 0, Major 1 (recorded, no level change), Minor 1, Editorial 0 |
| PMIDs verified | 50/50 reference PMIDs resolved to the stated title, first author, journal and year; no retractions or expressions of concern |
| Load-bearing claims spot-checked | 11 |
| Validator | PASS — 0 fail, 0 warn, 31 pass (validate_review.py v2.3, the batch's own copy). Output in `validator.txt`; pre-review report kept at `sol61-original/validation-report.json` |

## Findings

| ID | Severity | Location | Original → corrected | Evidence |
|---|---|---|---|---|
| DS-1 | Major (recorded; no level changed) | SC2 level; causal tally | Γ and "0 positive qualifying study units" retained, with the limitation recorded | Rubric 07 names "discontinuation and rechallenge designs" among its field-valid equivalents and says they exist so a convergent case is "not capped by the absence of a clean blind". The examined body has a single-blind dose-reduction pilot (PMID 31188232) and an AMPT depletion study with subjective-effect outcomes (PMID 11682247), but no registered blinded randomized-withdrawal trial, although that design class exists in the literature. SC2 Γ and the zero tally therefore rest partly on an evidence-identification gap, as the review itself states. Not changed, because raising it requires adding sources rather than re-reading the present ones, and the headline is set by SC3 and would remain Γ either way. Action for the lead: commission a targeted retrieval of registered discontinuation / randomized-withdrawal antipsychotic trials and re-score SC2. |
| DS-2 | Minor | Registered claim and scope, Selection line | "; existing HIV draft consulted after a source-based checkpoint" removed | Batch boilerplate; no HIV draft belongs to this case's provenance. |

## Rejected findings (checked, no change)

1. Headline Γ for SC3 — the registered claim fixes mesolimbic anatomy and unselected primacy; the strongest human findings localise to the associative striatum, the response-stratification studies are small and medication-confounded, and STRATA finds no striatal dopamine difference between responders and non-responders. Neither Δ for that scope nor Β is forced, and the review explicitly refuses the available error of treating a conservative gate as a denial of antipsychotic efficacy.
2. Confidence Moderate — compartment, treatment-status and response heterogeneity match the Moderate band.
3. Momentum "Mechanistic only" names its artifacts (PMID 39632880; EMERGENT-3 NCT04738123).
4. Divergence "Contested" — field genuinely split between textbook dopamine framing and the heterogeneity literature.
5. Primary-subclaim aggregation on SC3 → Γ; alternatives reported (weakest-link Β, scope-matched Γ).
6. `upper_ladder_check` records which equivalents were considered and why none carried the claim further, as rubric 07 requires below Ζ.
7. Numeric and design claims verified: PMIDs 21768612, 20194823, 23034655, 32910150 (AUC 0.59), 38691387, 31188232, 39632880, 10739409, 11682247.
8. PubMed dropping subscripts in "D2/3" (reference 23) — PubMed artifact.

## Residual limitations

- Access depth unchanged: 3 partial-body and 47 abstract/status examinations.
- Registration and expectancy provenance for the historical challenge and occupancy studies was not recovered; this is the basis of the conservative SC2 gate and the subject of DS-1.
- The boilerplate corrected in DS-2 remains in the sibling cases outside this assignment.

**Approval:** Following adversarial review and the correction listed above, the revised document is approved, with finding DS-1 recorded for the editorial lead. — Claude Opus 5.5 (Anthropic), 2026-10-14.