R03 Citation Verification — Batch C (Gemini R03 citations)
CCKE PROTOCOL v1.3 | Verifier: Claude (Fable 5) | Date: 2026-09-01 Method: Crossref API identity checks (ctx_execute JS fetch), doi.org handle-registry resolution for 404s, web search for claim-support and unsourced magnitudes.
Summary table
| # | Citation as given | Identity | Claim support | Verdict |
|---|---|---|---|---|
| 1 | Bodwitch et al. 2019, Cal Ag, 10.3733/ca.2019a0018 | Exact match (title, authors, journal, year) | n/a (identity check only) | VERIFIED |
| 2 | Polson & Petersen-Rockney 2019, Cal Ag, 10.3733/ca.2019a0017 | Exact match | n/a | VERIFIED |
| 3 | Valdes-Donoso et al. 2019, Cal Ag, 10.3733/ca.2019a0013 | WRONG OBJECT: a0013 = Crowder 2019, “Science and law enforcement teaming up to help ‘critters’” (Cal Ag) | $179/lb: NO at that exact figure. Correct paper: Valdes-Donoso, Sumner & Goldstein 2019, “Costs of mandatory cannabis testing in California,” Cal Ag, 10.3733/ca.2019a0014 (companion: PLOS ONE 2020, 10.1371/journal.pone.0232041). Best estimate is $136/lb; $89.58–$186.01/lb range depending on rejection rate (8-lb batch); up to ~$791/lb for 1-lb batches. $179 falls inside the range but is not the paper’s headline number. | FAILED (DOI) / PARTIAL (claim) |
| 4 | Gabriel et al. 2012, PLoS ONE, 10.1371/journal.pone.0040163 | Exact match: “Anticoagulant Rodenticides on our Public and Community Lands: Spatial Distribution of Exposure and Poisoning of a Rare Forest Carnivore” | PARTIAL: the 2012 paper reports 79% (46/58) fishers AR-exposed (96% of those to second-generation ARs). The 85% (86/101) figure is from the follow-up Gabriel et al. 2015 (PLoS ONE 0140640). “79% to 85%” conflates two papers; magnitudes real but attribution spans both. | PARTIAL |
| 5 | Gabriel et al. 2018, 10.1186/s13104-018-3196-0 (owls) | DOI DOES NOT EXIST: Crossref 404 AND doi.org handle registry responseCode 100 (handle not found). Not a second real paper — a fabricated BMC Research Notes-style DOI. The real owl paper is the one a prior cycle verified: Gabriel et al. 2018, Avian Conservation & Ecology 13(1), 10.5751/ACE-01134-130102 (confirmed live in Crossref this cycle). | NO (nonexistent DOI; underlying finding real via ACE paper) | FAILED |
| 6 | Bauer et al. 2015, PLoS ONE, 10.1371/journal.pone.0120016 | Exact match: “Impacts of Surface Water Diversions for Marijuana Cultivation on Aquatic Habitat in Four Northwestern California Watersheds” | n/a | VERIFIED |
| 7 | Butsic & Brenner 2016, ERL 11(4):044023, 10.1088/1748-9326/11/4/044023 | Exact match: “Cannabis agriculture and the environment: a systematic, spatially-explicit survey and potential impacts” | Topic matches (spatially explicit survey of CA cannabis agriculture) | VERIFIED |
| 8 | Ridgeway & Kilmer 2016, DAD, 10.1016/j.drugalcdep.2016.06.004 | Exact match: “Bayesian inference for the distribution of grams of marijuana in a joint” | YES — consumption-estimation methods paper | VERIFIED |
| 9 | Davenport & Caulkins 2016, IJDP, 10.1016/j.drugpo.2016.03.002 | WRONG OBJECT: resolves to Aaraj & Jreij Abou Chrouch 2016, “Drug policy and harm reduction in the Middle East and North Africa” (IJDP) | The real Davenport & Caulkins 2016 is in Journal of Drug Issues 46(4): “Evolution of the United States Marijuana Market in the Decade of Liberalization Before Full Legalization,” 10.1177/0022042616659759 — that one does concern market/consumption estimation. Wrong journal AND wrong DOI as cited. | FAILED (correct tuple supplied) |
| 10 | Queirolo et al. 2019, IJDP, 10.1016/j.drugpo.2019.03.011 | WRONG OBJECT: resolves to Kowalski, Hooker & Barratt 2019, “Should we smoke it for you as well? An ethnographic analysis of a drug cryptomarket” (IJDP) | No Queirolo 2019 IJDP article found in Crossref. Closest real Queirolo works: Queirolo et al. 2016, “Cannabis clubs in Uruguay,” IJDP, 10.1016/j.drugpo.2016.05.015; Queirolo et al. 2025, “Intended and unintended effects of cannabis regulation in Uruguay,” Policy & Politics. The claim (consumers bypass pharmacy channel, partly over potency caps) is directionally consistent with the Uruguay literature (pharmacy THC capped; legal market reaches <50–55% of users) but the cited source does not exist as cited. | FAILED |
| 11 | Caulkins & Reuter 2010, 10.1080/09595231003650619 | DOI DOES NOT EXIST (Crossref 404; doi.org handle not found). The real Caulkins & Reuter 2010 is “How Drug Enforcement Affects Drug Prices,” Crime and Justice 39(1), 10.1086/652386 (Univ. Chicago Press — not a 10.1080 Taylor & Francis DOI). | That paper argues enforcement intensity has weak effects on price/supply — consistent in direction with “eradication produces no measurable long-term supply suppression,” but it is about enforcement/price generally, not cannabis eradication specifically; the strong specific claim is an overreach of this source. | FAILED (fabricated DOI; real tuple supplied; claim PARTIAL) |
| 12 | Craven et al. 2019, 10.1080/17524032.2019.1601000 | DOI DOES NOT EXIST (Crossref 404; doi.org handle not found). Prefix pattern mimics Environmental Communication but resolves to nothing. Wrong-topic anyway: an env-comm journal would not establish combustion chemistry. | Myclobutanil→HCN on combustion is a real, documented concern but from other sources: Health Canada’s 2017 clarification on myclobutanil in cannabis (canada.ca); myclobutanil thermal decomposition >~205°C/401°F yielding HCN (SDS/regulatory data); Sullivan, Elzinga & Raber 2013 (J. Toxicology, 10.1155/2013/378168) for pesticide transfer into cannabis smoke; Steep Hill Labs 2016 contamination reporting. | FAILED (fabricated citation; underlying chemistry claim UNVERIFIABLE via cited source, supported elsewhere) |
| 13 | Schwabe et al. 2023, PLoS ONE, 10.1371/journal.pone.0282396 | Exact match: “Uncomfortably high: Testing reveals inflated THC potency on retail Cannabis labels” | PARTIAL: demonstrates potency inflation directly (observed mean 14.98% vs labels; ~70% of 23 Colorado samples >15% below label). It only mentions lab-shopping reports as circulating background — it does not demonstrate systematic lab shopping. “Widespread… systematic lab shopping” overstates the paper. | VERIFIED (identity) / PARTIAL (claim) |
| 14 | Polson et al. 2023, Land Use Policy, 10.1016/j.landusepol.2022.106351 | WRONG OBJECT: resolves to Grigorescu et al., “The resilience of sub-urban small farming in Bucharest Metropolitan Area… COVID-19” | Almost certainly a one-digit transposition of the real paper: Polson, Bodwitch, Biber et al. 2023, “After legalization: Cannabis, environmental compliance, and agricultural futures,” Land Use Policy, 10.1016/j.landusepol.2022.106531 (106351 vs 106531). Topic of real paper matches the claim area. | FAILED (DOI) / correct tuple supplied |
| 15 | Health Canada: >1.68M kg unsold cannabis destroyed 2018–2023 | Reconstructable: MJBizDaily analysis of Health Canada data (mjbizdaily.com, 2023) — >1.7 billion grams (≈1.7M kg / 3.7M lb) of unsold unpackaged dried flower destroyed 2018–mid-2023 (155M g 2019; 279M g 2020; 425M g 2021; 611.7M g 2022; 265M g H1-2023), plus ~24M packaged units. 1.68M kg is consistent (slightly conservative). Source is Health Canada compliance data as reported by MJBizDaily, not a Health Canada standalone publication. | YES | VERIFIED (with source = MJBizDaily/Health Canada data) |
Item 16 — unsourced magnitudes (provenance-or-withdraw)
| Item | Finding |
|---|---|
| (a) Compliance $150–300+/lb smallholder vs <$15/lb industrial | NO single reconstructable source for these exact numbers. Components exist: testing $136/lb (Valdes-Donoso 2020); CA cultivation tax reached ~$148/lb; UC Berkeley CRC 2020 compliance brief documents fixed-cost burden on small farms; <$15/lb industrial figure echoes Caulkins RAND WR-764 (2010) fully-legal production-cost estimates, not compliance cost. As stated: UNVERIFIABLE — recommend withdraw or restate from components. |
| (b) “>60% of CA cities/counties banned commercial retail” | YES, reconstructable and canonical: DCC data (Feb 2022): 62% of CA jurisdictions prohibit any cannabis retail licensing; 56% prohibit all cannabis business types. DCC itself runs “Retail sales remain illegal in a majority of California cities and counties” (real.cannabis.ca.gov); also MJBizDaily, Cannabis Business Times. Well supported. |
| (c) Legal consumer markup 35–50%+ from taxes/fees | Substantially supported: cumulative CA retail tax burden commonly reported at 28–45%+ (15% excise + 7.25–10.25% sales + local business taxes up to ~15%; worked examples ~43%). “35–50%+” is at the high edge but within reported ranges. PARTIAL — needs a named source (e.g., CDTFA rates + Tax Foundation/CalMatters analyses). |
| (d) Legal wholesale <$500/lb CA/OR vs prohibition-state $1,200–2,000/lb | First half: YES — CA outdoor flower ~$300/lb, OR as low as $100–300/lb widely reported (MJBizDaily, Cannabis Benchmarks-derived reporting). Prohibition-state $1,200–2,000/lb: plausible per illicit/limited-license market reporting but no canonical source found — PARTIAL; cite Cannabis Benchmarks spot index or withdraw the second half. |
| (e) Oregon collapse <$300/lb | YES — Oregon outdoor wholesale fell to record lows, reported at $100–300/lb (OLCC supply-demand reports; MJBizDaily; businessofcannabis.com). Reconstructable. |
| (f) Uruguay ~50% capture | YES, approximately: IRCCA-based estimates ~55% of users reached (with sharing assumptions); recent academic reviews put legal market at “<50% of cannabis used.” “~50%” is a fair round figure; cite IRCCA / Queirolo et al. 2025 Policy & Politics or the JIED market-revenues article. |
| (g) DEA DCE/SP “1 lb/plant” yield assumption | YES, documented: the ~1 lb/plant convention entered DEA/sentencing equivalency practice and is discussed critically in Caulkins, RAND WR-764 (“Estimated Cost of Production for Legalized Cannabis,” 2010), noting it is ~an order of magnitude above indoor-lit yields (e.g., Toonen et al. ~1.2 oz/plant) and includes leaf+bud. Also see USSC sentencing-guideline history (1 plant = 100 g, formerly 1 kg for 50+ plants). Reconstructable with proper citation. |
Verdict counts (items 1–15)
- VERIFIED: 6 (items 1, 2, 6, 7, 8, 15) — plus 13 identity-verified
- PARTIAL: 2 (items 4, 13)
- FAILED: 7 (items 3, 5, 9, 10, 11, 12, 14)
Item 16: supported/reconstructable — b, e, f, g; partially — c, d; unverifiable as stated — a.
Pattern note
Gemini’s wrong-DOI failure mode recurs: 3 fully fabricated DOIs (items 5, 11, 12 — nonexistent in both Crossref and the doi.org handle registry), 4 wrong-object DOIs (items 3, 9, 10, 14 — real DOIs pointing at unrelated papers, two of them near-transpositions: a0013 vs a0014; 106351 vs 106531). Item 12 is the worst class: fabricated citation in a topically irrelevant journal for a chemistry claim.