Table 2.
Classification of sensory discriminative tests.
| Test | Type of Evaluation | Statistical Analysis | Advantages | Limitations | Variations | Ref. |
|---|---|---|---|---|---|---|
| Triangle test | Identification of a different sample from a set of three samples. | Mixed model logistic analysis; mixed ANOVA 1; Tukey’s test |
Does not require specification of the nature of the difference |
Lack of accuracy; ineffectiveness and sensory fatigue; requires large sample sizes to be effective | Tetrad test; duo–trio test |
[85,87] |
| Tetrad test | Group similar samples from a set of four samples. |
Hypothesis testing | Fewer assessors can be used to recover the same confidence in the result | Sensory fatigue | [89,98] | |
| Duo–trio test | Three samples are displayed; one of them is the reference. Identification of the most similar sample regarding the reference. | Hypothesis testing | Easier performance in complex or hard-to-evaluate products; the ability to evaluate how significant sensory differences are between samples |
Sensory fatigue; large assessor groups need to be used to increase confidence in the data; low statistical power |
CRM 2; BRM 3; A-Not AR 4; 2-AFCR 5; different positions of references; ABX | [90,91,92,99] |
| ABX test | Two control samples and a treated sample are presented to assessors, and they are asked to match the “X” sample to one of the references. | Hypothesis testing | Participants do not need anyprior knowledge of the samples; assessment of fewer products |
No guidance over an attribute to focus on; less sensitive test; relies on the assessors’ memory |
[100,101] | |
| A Not-A test | Reference A and other samples are presented to assessors, and they must decide whether the other samples assessed are similar to the A sample. | Chi-squared test; Thurstonian distance |
Single presentation test; usable with high carryover effect samples |
Less recommended when assessors are untrained and/or with no experience with the products |
[91,102] | |
| Paired Comparison | Compares two samples without concerning the intensity of perception. | PCA 6; Friedman test; Bradley–Terry model |
Simple and intuitive task; sensitiveness to differences between stimuli |
Time-consuming. Low statistical power |
Simple difference tests or directional paired comparison tests (or 2-alternative forced-choice tests); multiple paired comparison test; FC 7 | [91,95,103] |
| FC 7 | Assessors must choose one of the two samples. | ANOVA 1 | Simple task | A tendency for “noise” in the datasets | Triangle test; AFC 8; can be based on the triangle test becoming 3-AFC or paired comparison test becoming 2-AFC; 4I2AFC 9 | [95,98] |
Legend: 1. Analysis of variance; 2. constant-reference mode; 3. balanced-reference mode; 4. A-Not A with a reminder; 5. 2-AFC with a reminder; 6. principal component analysis; 7. forced-choice; 8. duo–trio test alternate forced-choice; 9. four-interval two-alternative forced-choice.