Content
Citation(s) & Footnote(s)
Book(s)Full supplied-source audit — September 27, 2026
Ferreira, Andre Dominghetti; Abrahão, Juliana Costa de Rezende; Carvalho, Gladyston Rodrigues; Carvalho, Alex Mendonça de; Andrade, Vinicius Teixeira; Gonçalves, Flavia Maria Avelar; and Malta, Marcelo Ribeiro (2021). Chemical and sensory characteristics in the selection of bourbon genotypes. Bioscience Journal 37, e37072. DOI 10.14393/BJ-v37n0a2021-54155. Published December 29, 2021 as printed on p.8.
All eight pages fully read, including author contributions, conflict/ethics/funding statements, 32 references and final open-access notice. Three tables, one 36-pair correlation matrix and the mixed-model equation visually inspected on pages 2–5. No separate supplement or appendix identified. This replaces reliance on the earlier extraction/screening statement with recorded full reading.
Bourbon refers to coffee; exclusion from direct whiskey evidence remains appropriate. Original source identity preserved.
Design and proposed contribution — pages 1–4
Twenty coffee genotypes, including 17 Bourbon selections and three standards, across five Minas Gerais municipalities, using three blocks and ten plants per plot (six central plants evaluated). First three harvests were averaged; chemical assays run in triplicate. Trained judges evaluated five cups per sample, eight attributes scored 0–8 plus a 36-point base. The study uses REML/BLUP and a sum-of-ranks selection index.
It seeks chemical proxies and breeding selections for cup quality, not whiskey sensory mechanisms. Coffee cherry processing, drying and bean chemistry are not transferable as bourbon-whiskey results. Technical assay replicates and cups are not independent genotypes or sites.
Findings — pages 4–7
Authors nominate BA6, BA2, BA5, BV14 (called BA14 in results/table) and BA12, predicting 1.65% sensory-score gain. Reported genotype-mean heritability is 0.92 and selection accuracy 0.96; neither means 92% of an individual's taste is genetically fixed nor 96% correct predictions in an external test. These are model-derived quantities conditional on design/assumptions. Predicted selection gain is not realized offspring improvement.
The article itself concludes its chemical evaluations did not efficiently detect the sought genetic/phenotypic chemistry–sensory relationships. Positive acidity–score association cannot establish a causal improvement from adding acid. Grain damage and conductivity mechanisms in discussion are interpretations, not directly manipulated causal tests.
Visual/numerical audit
- Table 3 (p.5) repeats the ANR numbers verbatim in the LK column for all 20 genotypes, while its LK mean is 26.91 and displayed values are approximately 72–79. The numerical column is internally incompatible with its mean and reported negative gain; do not reconstruct missing LK values.
- Page 4 swaps interaction coefficients for ATT and SS: narrative 0.09/0.29, Table 2 0.29/0.09.
- Figure 1 shows LK–CPT =0.18, while the results paragraph includes that pair among moderate/high positive correlations of 0.60–0.89. CE–CPT is 0.60, a possible intended pair, but not silently substituted.
- Figure 1 is captioned Pearson correlation, while discussion calls these genotypic correlations. The target quantity/procedure needs clarification before reuse.
- Methods list BV13/BV14, whereas Table 3 calls both BA13/BA14. Preserve the genotype-name ambiguity.
- Table 1 uses N for two apparent longitude entries; this is not a usable longitude hemisphere label. Do not map without reconciliation.
- The printed model does not explicitly display the genotype-by-location term whose estimates appear later; kinship construction for undocumented farm-origin selections is not explained.
- General statement that all chemical data use g/100g conflicts with the distinct titratable-acidity unit and does not specify conductivity units. AOAC1970 cited in methods is absent from the bibliography. RA in results appears to be an inconsistent variable label.
- The rank-index values are not transparently reproducible from the printed columns. Missing/corrupt LK values preclude a trustworthy recalculation of the selection rule.
- Sensory-gain check: mean of the five displayed selected Fsc values is 84.774; relative to 83.39 this is about 1.66%, broadly consistent with reported 1.65% after rounding. That limited consistency does not validate the corrupted other columns.
No corrected dataset or publication erratum was supplied. These warnings concern the visible source, without inventing corrected estimates.
Academy relevance
Pair only as a methods comparison with Figueiredo2013 No access: both concern coffee genotype/environment and show why a chemical measurement is not automatically a proxy for sensory quality. They do not prove whiskey terroir or provide a whiskey tasting scale. A proposed research-literacy exercise can distinguish predicted gain from achieved outcomes and spot duplicated table columns.
No new whiskey Zettel warranted. Preserve as excluded coffee source with a complete critical Literature Note, not a factual basis for whiskey lessons. No public-course edits. Proton hash and browser rendering remain separate pending checks.
