When two reviewers from different institutions sign opposing findings under the same scope, Vela does not pick a winner. It stores both as first-class objects, links them with a typed Contradiction edge, and carries the disagreement forward until new evidence lets the field choose. Companion to /proof/pericyte-correction: that page shows scope narrowing; this one shows scope holding.
Lennon 2024 lands first.
A neuroscience reviewer at MIT compiles Lennon and colleagues' postmortem cohort showing astrocyte reactivity is a primary driver of cognitive decline in early AD. Confidence 0.66, scope human · sporadic AD · prodromal stage.
Astrocyte reactivity is a primary causal driver of cognitive decline in early Alzheimer's disease.
- scope
- human · sporadic AD · prodromal stage
- evidence
- 1 EvidenceObject · postmortem cohort · n=42
- cite
- Lennon et al., Cell Reports 2024
- signed by
- reviewer:demo-curator-A → vrev_4d2c… (illustrative; no real PI signed this)
Park 2025 disagrees, signed-publishes anyway.
Three months later a reviewer at Seoul National University compiles Park and colleagues' longitudinal MRI study from the same cohort era. Park finds astrocyte reactivity tracks decline but does not lead it; microglial inflammation does. Confidence 0.71, identical scope.
Microglial inflammation, not astrocyte reactivity, is the primary causal driver of cognitive decline in early Alzheimer's disease.
- scope
- human · sporadic AD · prodromal stage
- evidence
- 1 EvidenceObject · longitudinal MRI · n=128
- cite
- Park et al., Nature Neuroscience 2025
- signed by
- reviewer:demo-curator-B → vrev_a3c8… (illustrative)
A Contradiction object spawns automatically.
The substrate detects two findings under the same scope with opposing causal claims and emits a signed Contradiction event linking them. Neither finding is invalidated. Neither reviewer is overruled. The edge is the data: a downstream reader sees both claims, both sets of evidence, both signatures, and the explicit acknowledgement that the field has not chosen.
- finding_a
- vf_8e22ab… · "astrocyte reactivity is causal"
- finding_b
- vf_b71f09… · "microglial inflammation is causal"
- scope (shared)
- human · sporadic AD · prodromal stage
- nature
- causal_attribution_disagreement
- requires
- further evidence under matched scope
- signed by
- reviewer:bot · automatic detection at apply-time
On the constellation, both nodes light at full opacity with a winter-coloured edge between them. Hover either; the linked node and edge highlight together.
Downstream objects see both lines.
Three teams that previously cited "astrocyte reactivity is causal" had pinned
to vf_8e22ab. The dependency graph notifies them that a Contradiction now
attaches to that finding. None of their work is invalidated. Their queues
surface the edge so the next reviewer pass acknowledges it.
-
+ 2 days · Phase Ib design · Cambridge MA A target hypothesis at one biotech that picked astrocyte reactivity as the primary mechanism for inclusion criteria adds a margin gloss to its planning doc: "primary causal driver under active dispute (Lennon 2024 vs Park 2025); consider stratification by GFAP+ subgroup." Trial design proceeds; the next DSMB review surfaces the contradiction inline.
-
+ 4 days · Review article · Boston A neurodegenerative-disease review in galley acknowledges the disagreement in the discussion section, citing both
vf_8e22abandvf_b71f09, with a single line on the open question. Reviewer signs the change as a caveat event on the review's own claim. -
+ 11 days · Funder allocation · NYC A foundation officer reviewing a neuroinflammation grant cycle pulls the constellation, sees the Contradiction as a first-class object, and prioritises funding for studies that would distinguish the two mechanisms. The disagreement becomes a roadmap, not a footnote.
Contradiction is not a bug. It is the field telling itself what it does not yet know.
The substrate stores both. The reviewer who picks one without acknowledging the other is the one being unscientific.
A Contradiction edge resolves only when new evidence under matched scope lets the field choose — typically a Correction event narrowing one of the two findings, or a replication study breaking the tie. Until then, both sit on the constellation as signed claims with explicit attribution. The substrate enforces the discipline of not pretending the disagreement does not exist.