Trained-panel descriptive analysis of nine rums separates sensory modalities and shows why premium labels, nominal ABV and flavor wheels cannot substitute for observed profiles.
Citation And Scope
Ickes, C. M., and Cadwallader, K. R. (2017). Journal of Food Science, DOI 10.1111/1750-3841.13936. Supplied early-view PDF has eleven sheets and Vol00/Nr0 footers; final pagination is not inferred. All sheets reviewed; Tables1–3 on PDF3,5,6 visually checked.
Design And Findings
PDF1–6: eight trained panelists evaluated nine selected rums, seven premium and two mixing, all labeled 40% ABV. Twenty-three panel days included orientation, term/reference development, scaling, practice and testing. Samples were coded, randomized, served in covered black glasses under red light at22°C. This is a trained descriptive panel, not a consumer liking test or representative market sample. Fifty-two initial modality-specific terms became sixty-nine after additional sessions using a wheel; the final list had thirty-eight attribute/modality entries and twenty-seven unique terms. Training time and wheel exposure are confounded, so the study does not isolate a causal effect of the wheel. PDF5–10: differences were reported in alcohol aroma, warming, sweet-associated aromas and other attributes despite the same nominal label ABV. This supports measuring perceived warming rather than deriving it solely from label ABV; the experiment does not isolate its chemical causes. Premium and mixing products did not form a clean descriptive division. PCA emphasizes variation in the selected attributes, not quality, value or liking. The first component explains88.4% and the second3.6%; explained variance alone is not a significance test.
Quality Checks And Interpretation Limits
Table1 visibly specifies “10 oz.” in a “2 oz.” cup for its alcohol aroma reference—an unresolved preparation error. Do not copy it into a home exercise. Table2/Table3 inclusion is inconsistent: brown-spice aftertaste and coconut aroma-by-mouth appear in Table3 despite unstarred adjusted F values1.60 and3.05 in Table2; alcohol aroma-by-mouth has a starred raw effect but is absent from Table3. Caramel aroma's rum-by-replication interaction is starred, contrary to the blanket prose claim of none. Without raw data, do not invent corrected p values or treat the reported twenty significant attributes as independently audited. Small nonrandom product selection, possible sensory fatigue, training/assessor effects and brand/age/process confounding limit category-wide claims. Age trends are associations, not controlled maturation experiments. Sweet-associated words do not establish measured sugar content or a particular molecule.
Blending Lab Application
Collect unaided notes first, then optional vocabulary-assisted notes, retaining which was which. Provide clear modality labels and reference definitions without claiming that a wheel calibrates a person. Preserve warming as an observation, separately from ABV. Compare personal preferences blindly rather than treating premium positioning as sensory superiority. Adapt everyday food references only after checking suitability and instructions; the paper's laboratory references are not turnkey consumer recipes.
Research Connections
Shared sensory vocabulary does not establish assessor calibration
A premium market position is a different claim from sensory superiority
No access