Evaluation of rapid descriptive sensory methods with different panels in the characteristics variations of beers packaged in distinct materials

Two new rapid descriptive sensory evaluation methods have been gaining ground in the field sensory evaluation. The Projective Mapping method uses similarities and dissimilarities as a criterion, while Pivot Profile, uses reference criteria. This research aimed to assess panels with 12 and 24 judges, comparing its reproducibility, and evaluate if a non trained panel with a smaller numbers of judges is sufficient for results reliability. Samples of Pilsen beers in different packages were distributed, as well as a reference sample, with different sensorial characteristics. It was possible to observe a slight discrepancy between the results obtained in each of the applied tests. We observed a need for short-term training before the application of the test, aiming for better use of the descriptive terms by the judges. Also, the number of judges influenced the obtained results, being the panels of 24, in both tests, the ones that best described the indicated characteristics.

the Sensory Analysis and Consumer Studies Laboratory (Lasec) of the Operations, Processes and Technology Sector (SOPT) of the Food Department (ALM) from the Faculty of Pharmacy (FAFAR) at the Federal University of Minas Gerais (UFMG), where they were conditioned at 5 °C until analysis.

Judges and application of tests
For the study, 72 judges were recruited. Recruitment took place through flyers at the Faculty of Pharmacy (FAFAR) at the Federal University of Minas Gerais (UFMG) and through digital media on social networks, informing different days and times for application of the test.
The target audience was consumers and beer lovers, of both genders and aged between 18 and 60 years. People who showed some aversion to the product as a restriction to gluten or some ingredient present in the formulation, as well as people with some health problem and with continuous use of medications were excluded from the test. In addition, availability, motivation and interest in participating were considered.
All judges were informed about the product, its ingredients and the test procedures, and expressed their agreement to participate through the Free and Informed Consent Form. This research project was approved by the Ethics Committee of the Federal University of Minas Gerais (registry number 96268418.7.0000.5149).
The tests were applied in two distinct moments, variating the number of judges. In the first moment, 12 judges composed the panel for the Projective Mapping test and 24 judges for the Pivot Profile test. In the second phase (after 48 hours), the panel included 24 judges for the Projective Mapping test and 12 judges for the Pivot Profile test, totaling 72 judges in 4 sessions.
The number of judges in each test was chosen considering that data in the literature are still inconclusive regarding the best scenario and once these are rapid tests, we aimed to perform two viable realities for running the test in a short period of time.
Samples were presented to the judges in a coded form with three-digit numbers, randomly and balanced, and served in glass bowls. A glass of mineral water for cleansing the palate was also available between the tasting of the samples.
For the Projective Mapping test, the judges received five samples of Pilsen-type beer: B1 (brand A, glass packaging); B2 (brand A, aluminum packaging); B3 (brand B, glass packaging), B4 (brand B, aluminum packaging) and B5 (reference brand, glass packaging).
They were asked to prove each of the samples and to place each of them, in a white sheet (65 x 45 cm), according to their similar characteristics or differences. It was explained to Research, Society andDevelopment, v. 9, n. 9, e08996137, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i9.6137 6 the panel that they should group them according to their own criteria, with no right or wrong answers. For each test, the X and Y coordinates of each sample were determined, considering the lower-left corner as the origin of the coordinate system (PAGÈS, 2005).
For the Pivot Profile test, judges were asked to prove five pairs of samples (one pair at a time), consisting of the pivot sample (reference) with a coded sample in four pairs and two pivot samples in one pair. Afterwards, they were asked to write which attributes of the coded sample was greater than the pivot sample. This description was not mandatory, so the panel was free to describe the pertinent characteristics of each one, being able to use one or more attributes (Esmerino et al., 2017).
The number of times which each attribute was quoted as "less than the pivot" (negative frequency) and "more than the pivot" (positive frequency) was automatically computed and summed. Subsequently, the negative frequency was subtracted from the positive one for each attribute. The resulting score was then translated by adding the absolute value of the minimum score to all the scores. Thus, the minimum score takes the value of zero and all other scores are positive, yielding a translated frequency table (Fonseca et al., 2016;Lelièvre-Desmas et al., 2017;Thuillier et al., 2015).

Statistical procedures
The data were evaluated through Correspondence Analysis (CA) for the Pivot Profile test and Multiple Factor Analysis (AFM) for the Projective Mapping test. For both tests, the hierarchical cluster analysis was performed in order to group the samples. All analyzes were performed using the programming language R (R development core team, 2007) and the FactoMineR package (Lê, Josse & Husson, 2007).

Results
The Pivot Profile analysis was performed with panels of 12 ( Figure 1a) and 24 ( Figure   1b) judges. Figure 1 indicates the sensory map containing frequencies of the main attributes cited in 12 and 24 members panel for Pivot Profile analysis. Circles represent the samples, while triangles illustrate the most significative attributes, only those showing cos 2 above 0.7 were selected. Research, Society and Development, v. 9, n. 9, e08996137, 2020 (CC BY 4. As presented in Figure 1(a) and Figure 1(b), the two dimensions of the map sum up to 88.15% and 83.72% of data variance, respectively.
It was observed in Fig. 1(a) that samples B1, B2 and B3 were similarly distributed in both dimensions, with attributes spread in the quadrants, being both described as "fruity, sweet and foamy". It is noticeable that the panel members were able to distinguish the control sample (B5) from the others, highlighting its pleasant aroma and smooth taste.
Furthermore, the analysis showed that sample B2 and B4 (packed in an aluminum can) was characterized as "bitter" and "gassy and dark", respectively. In addition, B4 was placed distant from the other samples.
When increasing the number of judges (Figure 1-b), four new groups appeared: B1, featured by strong aroma; B2, featured by full-bodied flavor and fruity aroma; B3 and B4 featured by the dark color, viscous appearance and pleasant aroma and B5, featured by watery flavor, salty, gassy and acid.
In addition, the 12-members panel grouped B1, B2 and B3 samples, separating it from B4, differently from the results of the Pivot Profile analysis with 24 judges.
Dendrograms obtained by cluster analysis from Pivot Profiling for both 12 and 24 judges corroborated with data assessed in correspondence analysis, in which, for 12 judges, we verified Research, Society andDevelopment, v. 9, n. 9, e08996137, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i9.6137 8 the prevalence of three groups. When applying the dendrograms for the 24 judges in Pivot Profile, four distinct groups are seen, as presented in Figure 2. For both tests, the removal/unification of used terms was necessary in order to refine the analysis' quality. Figure 3 (a and b) indicates the configuration obtained in the multiple factor analysis test, applied to 12 and 24 judges, respectively. The dimensions of this analysis sum up to 67,94% and 58,95% of the data variance. Research, Society and Development, v. 9, n. 9, e08996137, 2020 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v9i9.6137 The sample B1 represents brand A, glass packaging; B2 brand A, aluminum packaging; B3 brand B, glass packaging, B4 brand B, aluminum packaging and B5 reference brand, glass packaging. As indicated (Fig. 3-a), samples B2, B3 and B4 were grouped similarly, while samples B1 and B5 were placed in different quadrants.
Was observed that three sample clusters were formed for both Projective Mapping configurations, as illustrated in Figure 4.

Discussion
Regarding the sensory characteristics of the samples, in Figure 1 was observed that the panel members were able to distinguish the control sample (B5) from the others, highlighting its pleasant aroma and smooth taste. A possible explanation for that is the extraction technique called "dry-hopping", which is performed in B5 sample and could possibly contribute with the observed sensory perceptions. This technique allows differentiating aromatic attributes of hops from different geographic origins (Barry, Muggah, McSweeney & Walker, 2017). It is noticeable that in both 12 and 24-members panels, this control sample could be distinguished from others, corroborating with the hypothesis that the hop extraction method is crucial for the sensory characteristics of this beer, preserving its aromatic compounds.
In addition, sample B4 was placed distant from the other samples. The beer packaging in materials such as aluminum can give the product a metallic effect and bitter taste (Ivušić, Gjeldum, Nemet, Gracin & Marić, 2006). Research, Society and Development, v. 9, n. 9, e08996137, 2020 (CC BY 4.