Sensory Characterization of Young South American Red Wines Classified by Varietal and Origin Laura Garcia Llobodanin, Lucia Pereira Barroso, and Inar Alves Castro
Typicality is the set of sensory characteristics that identify a distinctive type of wine. Thus, the aim of this research was to identify the sensory characteristics that contribute to define typicality of young South American red wines based on their varietal and origin, and to evaluate the effect of the vintage on this identification. To achieve this objective, visual appearance, odor, and taste of 138 wines from 2 vintages were submitted to a sensory evaluation using a descriptive analysis complemented with the frequency of citation method, performed by wine experts. The intensity of 17 odor and taste attributes was evaluated using a 5 points rating structured scale. The panel performance evaluation demonstrated its high level of expertise and reproducibility. The wines were separated into 3 clusters by multivariate analyses. Cluster 1 was primarily composed of Carm´en`ere, Malbec, and Syrah wines from Chile. Cluster 2 was predominantly composed of Tannat wines from Uruguay and Brazil, while Cluster 3 contained a higher proportion of Malbec and Merlot wines from Argentina and Brazil. Cabernet Sauvignon was equally distributed into all clusters. Wine experts were able to identify the wines according to their varietal and origin, suggesting that there is typicality in young South American red wines. The combination of descriptive analysis with the frequency of citation was useful in characterizing most of the wines, but the typicality perceived by the panelists was not achieved by multivariate analysis. Vintage did not alter the sensory characterization of the wines, and this result could be due the new viticulture or oenological practices used by the winemakers to compensate the environmental variation.
Abstract:
Keywords: cluster, sensory, typicality, vintage, wine
Although typicality has been used during the purchasing of wine, there is no evidence that this concept is real for young South American red wines. This study contributed to identify sensory characteristics associated to typicality and showed that vintage did not alter the wines classification.
Introduction South America is a relatively new wine production region, which is known as “New World.” This region includes South Africa, Oceania, Asia, and California. Recently, some “New World” wines have gained international recognition (King and others 2014). However, unlike the “Old World” wines from Europe, where better wines are composed of a mixture of grape varieties produced in a very specific region, relying on a geographic classification system (Grand Cru, Premier Cru, Protected Designation of Origin, or Protected Geographical Indication) (Parr and others 2010), “New World” wines are still predominantly monovarietals. In the South American countries, novice consumers have difficulty of evaluating quality and use to base their wine purchase decisions on extrinsic cues such as price, brand, origin, and varietal (Mueller and others 2010; D’Alessandro and Pecotich 2013). This information creates a sensory expectation for the wine that may or may not be achieved. According to Mueller and others (2010), product expectations at the initial purchase and intrinsic MS 20140470 Submitted 3/21/2014, Accepted 5/13/2014. Authors Llobodanin and Castro are with LADAF, Dept. of Food and Experimental Nutrition, Faculty of Pharmaceutical Sciences, Univ. of S˜ao Paulo, Ave. Lineu Prestes, 580, B14, 05508-900, S˜ao Paulo, Brazil. Author Barroso are with Dept. of Statistics, Inst. of Mathematics and Statistics, Rua do Mat˜ao, 1010, 05508-090, S˜ao Paulo, Brazil. Direct inquiries to author Castro (E-mail: [email protected]; www.ladaf.com.br).
R C 2014 Institute of Food Technologists
doi: 10.1111/1750-3841.12535 Further reproduction without permission is prohibited
sensory attributes during product consumption influence the repurchase decision. Thus, when consumers choose a wine bottle based on its varietal and country of origin or terroir, it is because they believe that there is a combination of sensory attributes that differentiates this varietal or region from the others. This concept is known as “typicality.” A wine is typical when some of its characteristics, which reflect both its origin and varietal purity, can be identified and make it recognizable as belonging to a distinctive type (Charters and Pettigrew 2007; Maitre and others 2010). Therefore, typicality includes sensory, technical, and environmental dimensions (Cadot and others 2010b). However, although typicality has been used during the purchasing of wine, there is no evidence that this concept is real for young South American red wines. Wine sensory characteristics are divided into 3 major aspects: visual appearance, odor, and taste perception. One class of compounds that contributes to these aspects is the phenolics (Fanzone and others 2012), especially the anthocyanins and polymeric proanthocyanidins (Kallithraka and others 2001). These polyphenols are biosynthesized in response to injuries caused by weather conditions and microorganisms, and during the wine making process, including aging and storage (Vitrac and others 2002). Thus, if phenolic substances are one of the main factors that determine the sensory characteristics of the wine and if they can change due to environmental conditions, the typicality, once identified, could also change according to the vintage. Vol. 79, Nr. 8, 2014 r Journal of Food Science S1595
S: Sensory & Food Quality
Practical Application:
Typicality of South American red wines . . . Table 1–Sampling of the wines by region of production and to 10 panelists participated in each session. The bottles were store grape variety. at 17 °C, opened immediately before the analysis, and panelists Country
Vintage 2009
Vintage 2010
Total
Argentinaa
2 Cabernet Sauvignon 5 Malbec 2 Merlot 5 Syrah 2 Tannat 5 Cabernet Sauvignon 1 Carm´en`ere 3 Merlot 1 Syrah 5 Tannat 9 Cabernet Sauvignon 9 Carm´en`ere 2 Malbec 3 Merlot 4 Syrah 3 Cabernet Sauvignon 2 Merlot 6 Tannat 69
3 Cabernet Sauvignon 6 Malbec 2 Merlot 7 Syrah 1 Tannat 4 Cabernet Sauvignon – 4 Merlot 1 Syrah 4 Tannat 10 Cabernet Sauvignon 9 Carm´en`ere 1 Malbec 3 Merlot 4 Syrah 2 Cabernet Sauvignon 3 Merlot 5 Tannat 69
35
Brazilb
Chilec
Uruguayd Total
28
54
21 138
a
Subregions: Valle Calchaqu´ıes, Valles de la Provincia de Catamarca, La Rioja, San Juan, Mendoza, Valle del Uco, Tunuyan, Valle del R´ıo Negro, Valle del Tullum, and Tupungato. b Subregions: Serra Ga´ucha, Planalto Catarinense, Serra do Sudeste, Bento Gonc¸alves, Vale de S˜ao Francisco, and Campanha. c Subregions: Valle de Elqui, Limar´ı, Choapa, Valle Central, Colchagua, Valle Aconcagua, Valle de Rapel, Maipo, Valle de Curic´o, Maule, and Cachapoal. d Subregions: Sur, Cerro Chapeu, Noreste, and Suroeste.
In this study, it was hypothesized that young wines could be sensorially characterized by varietal and country of origin, but this classification could change according to the vintage. To test this hypothesis, 138 young monovarietal red wines from Vitis vinifera grapes produced in 2009 and 2010 vintages were submitted to a sensory evaluation by a panel of wine experts.
S: Sensory & Food Quality
Material and Methods Wine samples A group of 138 red monovarietal wines from Vitis vinifera (Cabernet Sauvignon, Carm´en`ere, Malbec, Merlot, Tannat, and Syrah) produced during 2009 (69 wines) and 2010 (69 wines) in Chile, Argentina, Uruguay, and Brazil with retail prices ranging from U.S.$5.24 to 45.71 per bottle (750 mL) were chosen for the present study (Table 1). Production statistics regarding grape variety, production region, and price were taken into account to make the selection as representative as possible of the wines produced in these countries. When possible, the same brand was chosen for the 2 vintages. When the same wine was not available, a similar one was purchased, maintaining the same grape variety, production region, and price range. Although young, some of the wines were oaked. In addition to these 138 wines (Table 1), 13 bottles were randomly repeated during the sensory analysis, forming 13 pairs. These pairs were used to evaluate panel reproducibility. Sensory analysis The sensory analysis of the wines was performed by 10 professional sommeliers (25 to 60 y old), being 6 females and 4 males. All tasters had worked to wine distributors for more than 5 y. They had extensive experience in wine tasting, especially with regard to South American wines. The samples were assessed in groups of 8 to 10 wines per session. A volume of 30 mL of each sample was poured in a standardized crystal red wine tasting vessel (Open Up—Universal, 40cl, 13 ½ oz; ARC Intl. Group, France), and 7 S1596 Journal of Food Science r Vol. 79, Nr. 8, 2014
were prevented from seeing their label or shape. Sensory analyses were split into 10 sessions (2/week) for each vintage and were held for approximately 2 mo in 2010 and more 2 mo in 2011. Wines were presented individually, with a 10-min enforced break after every 2 to 3 samples. The panelists were required to expectorate the wines into a spittoon. They were informed that they were evaluating South American red wines from 6 different grape varieties (Cabernet Sauvignon, Carm´en`ere, Malbec, Merlot, Tannat, and Syrah) produced in 4 countries (Chile, Brazil, Argentina, and Uruguay). However, for each wine sample, no specific information was provided concerning the grape variety, country of origin, or price. Analyses were individually carried out, samples were coded with 3 randomized numbers, bottles were randomized between and inside the sessions, and they were monadically presented to the panel.
Descriptive analysis and frequency of citation methods The tasting sheet included visual appearance, odor, and taste perceptions. To perform the descriptive analysis, attributes were chosen from those defined by the school for wine professionals level 3 (WSET 2009), where all of the panelists had been trained. Thus, judges had a complete understanding of all the attributes present in the tasting sheet. All judges had been before trained by WSET system (from level 1 to level 3), and for this reason, no additional training was performed in this study. About 110 wine samples that did not take part of our sampling were selected as endpoint of the attributes scale during the tasters training. After, all tasters were submitted to a practical exam before of becoming certified (WSET 2009). Testing was performed at an open table, and a verbal feedback was given to the panel after finishing all steps of their reports. Initially, appearance was assessed by clarity (hazy, dull, clear, or bright), color (purple, ruby, garnet, and tawny), and color intensity (pale, medium(−), medium, medium(+), and deep). Then, the panelists were asked to swirl and smell each sample for approximately 15 s and evaluate the orthonasal attributes, rating their intensity on a 1 to 5 point structured scale. At this step, they classified the samples according to the level of development (youthful, developing, fully developed, and past its best), odor intensity, and odor characteristics (fruit, vegetal, and spice). Next, the frequency of citation methodology was used to complement the descriptive analysis. For this technique, the tasters select the most pertinent attributes from a list containing a relatively high number of terms (Campo and others 2008). In our study, a list containing 53 sensory attributes was given to the panelists, who indicated the presence or not of each descriptor in the sample. The expert panel was familiarized with the terms that composed the frequency of citation list. Afterward, the retronasal and taste attributes were evaluated after swirling the sample for approximately 15 s inside the mouth. The panelists classified the wines according to acidity, tannin, alcohol level, body, persistence, flavor intensity, and flavor characteristics (fruit, vegetal, and spice) using a 1 to 5 point structured scale. The same list of 53 attributes used to characterize the olfactory aspects was used for the taste characterization. The attributes assigned in the frequency of citation were converted into scores (“1” if a descriptor was marked and “0” if not), and the mean values of the panel were applied to the wine analyses. Finally, panelists were invited to suggest a varietal, country, and a retail price for the each evaluated sample.
Typicality of South American red wines . . .
Results The panel used in this study was evaluated for discriminatory capacity and reproducibility. Figure 1 shows the percentage of correct categorization obtained when the panelists were asked to identify the varietal, country, and price of the wines. In general, the panelists were able to recognize wines by varietal (Figure 1A), mainly Cabernet Sauvignon, Malbec, and Carm´en`ere.
30
** 25 20
** ns ns
ns 15 10 5 0
B
**
Merlot
Cabernet Tannat
Malbec Carménère Shiraz
50 ** 40
** **
30 ns
S: Sensory & Food Quality
ProporƟon (%) of correct categorizaƟon
A
ProporƟon (%) of correct categorizaƟon
Statistical analysis Initially, the performance of the panelists was assessed in terms of discrimination and reproducibility. After tasting each wine, the panelist was invited to suggest the varietal, country, and commercial price, from low (5.0 to 15.0 U.S.$), medium (15.1 to 25.0 U.S.$), or high (25.1 to 50.0 U.S.$). To determine whether the percentage of correct categorization was obtained by chance or not, the normal approximation of binomial statistics was applied, adopting a value of 0.05 to alpha risk. In the sensory analysis, 13 wine bottles were randomly duplicated in the sessions, totaling 26 samples, where each pair consisted of the same wine. The results from these 26 samples were submitted to a principal component analysis (PCA), where a 2-dimensional graph was generated by visual appearance and odor and flavor characteristics separately. The means of the attributes were used as active variables, while the wine samples were plotted as cases. Moreover, the mean or median values of the sensory attributes obtained in the 2 replicates were compared by Wilcoxon test. The intraclass correlation coefficient (ICC) was also calculated. After checking the panelist´s performance, a pattern of attributes able to define any possibility of typicality was searched. The data were submitted to multivariate correspondence analysis (data not shown), but no pattern was evident. Thus, the panelists scores obtained from the intensity scales of 17 attributes were converted to continuous variables, and hierarchical cluster analysis was applied to separate the wines according to visual appearance, olfactory, and taste characteristics. The wines, coded according to their respective clusters, were plotted in a factor-plane (1 × 2) generated by PCA using the same 17 sensory attributes. Fifty-three descriptors were recorded by the panelists as present (+1) or absent (0) for each wine sample. The mean value of each descriptor was compared among the 3 clusters by one-way ANOVA. To check the vintage effect in the clusters characterization, the mean values observed in 2009 were compared with the mean values observed in 2010, using a dependent T-test. All calculations were performed using the software STATISTICA version 9.0 (StatSoft, Inc., Tulsa, Okla., U.S.A.) and the SAS (Statistical Analysis Software) version 9.2 (SAS Inst. Inc., Cary, N.C., U.S.A.).
Similarly, they had significant success classifying the wines/country (Figure 1B), except for the Uruguayan ones. Panelists were also asked to suggest a high, medium, or low price range for the wines. They had a significant rate of success, mainly for lowand medium-priced wines (Figure 1C). Taking, for example, Figure 1(B), from the total responses obtained to Chile (389), 178
20
10
0
Brazil
Uruguay
Chile
Argentina
C 100 ProporƟon (%) of correct categorizaƟon
Panelist’s performance The panelist’s reproducibility was evaluated by comparing the score attributed to the same samples when evaluated in 2 different sessions. The following attributes were evaluated to check the panelist´s reproducibility: color intensity, development, odor intensity, odor fruit, odor vegetal, odor spice, acidity, tannins, alcohol, body, persistence, flavor intensity, flavor fruit, flavor vegetal, and flavor spice. Discriminatory capacity was measured by comparing the number of correct responses when the panelists were asked to identify the varietal, country, and price range against the number of correct responses obtained by chance. To evaluate the panelists’ performance, an average reproducibility index (Ri ) was also calculated as described by Campo and others (2008). To be able to calculate the Ri , about 10% of the wines (13 out of the total 138) were analyzed in duplicate, making a total of 151 samples.
80 ** 60 ** ns
40
20
0
LOW
MED
HIGH
Figure 1–Proportion (%) of correct categorization when tasters were asked to identify the varietal (A), country (B), and price (C), ∗∗ P < 0.01. It represents the % of the total responses obtained for each category that were correctly answered by the tasters. Vol. 79, Nr. 8, 2014 r Journal of Food Science S1597
Typicality of South American red wines . . . were correct answers (46%). This number (178) was higher than the number expected if the correct answer had been obtained just by chance (112). Thus, even in a blind test, the correct categorization of the wines by varietal, country, and price did not occur by chance. This observation suggests that the wines presented sensory characteristics that were used by the panel to identify their origin. In other words, the wines presented sensory typicality. At the next step of the panelist´s performance evaluation, PCA was used to compare how the panelists positioned repeated samples (n = 26 bottles, forming 13 pairs) according to visual appearance, olfactory (Figure 2A), and taste attributes (Figure 2B). The same wines occupied close positions on the plot, indicating similarities between the pairs according to the sensory attributes. A high level
A
of agreement between the wine pairs was evident. Considering that the samples were randomly distributed between 20 sessions over approximately 4 mo, these results suggest a high level of reproducibility. The Ri index evaluates the individual reproducibility of the panelists. It varies from 0 to 1 and a value of 0.2 or higher is considered acceptable (Campo and others 2008). The values for the panelists ranged from 0.22 to 0.49. Therefore, all the panelists responses were included in the present study. Table 2 presents the means and probability values obtained for each sensory attribute that was analyzed twice for the same sample. From all 15 attributes associated with visual appearance, odor, and flavor, only 2 were significantly different (odor intensity, P = 0.036 and fruit odor, P = 0.017). The ICC values showed an excellent or satisfactory
Figure 2–Representation of the replicate wines on the 2-dimensional plane generated by the 1st and 2nd PCs according to the “visual appearance and olfactory” characteristics (A) and taste perception (B); where winei and winei´ represent 2 replicates of a single wine (n = 13 pairs).
2.5 88
13´ 13´ 77
1.5
12 12
PC 2: 18.32%
9´ 9´
8´ 8´ 7´ 7´ 13 13
4´ 4´
0.5
44 12´ 12´ 10 10
11 11 99 11´ 11´
10´ 10´
55
3´ 3´
1´ 1´
66
-0.5 6´ 6´
5´ 5´ 22
-1.5 33
S: Sensory & Food Quality
2´ 2´
11
-2.5 -4
-3
-2
-1
0
1
2
3
4
5
PC 1: 54.65%
B
3
12 12
2
88 13´ 13´ 44
1
7´ 7´
PC 2: 12.32%
2´ 2´ 99
0
8´ 8´
77 12´ 12´
10´ 10´
1´ 1´ 22
5´ 5´ 10 10 11 11 5 5
3´ 3´ 33 4´ 4´ 13 13
9´ 9´
66
-1
11 -2
6´ 6´ 11´ 11´
-3 -4
-2
0
2 PC 1: 58.59%
S1598 Journal of Food Science r Vol. 79, Nr. 8, 2014
4
6
8
Typicality of South American red wines . . . Table 2–Visual appearance, olfactory, and gustatory scores of rietal or region of their production and to determine whether the the sensory characteristics attributed by the panelists to the vintage could change this characterization. To achieve this aim, wines tasted twice, including the “intraclass correlation coefcluster analysis was applied to all attributes measured by an inficient” (ICC). Sensory evaluation Attributea
1st analysis
2nd analysis
Pb
ICC
Color intensity Development Odor intensity Odor fruit Odor vegetal Odor spice Acidity Tannins Alcohol Body Persistence Flavor intensity Flavor fruit Flavor vegetal Flavor spice
3.71 ± 0.81 1.86 ± 0.52 3.13 ± 0.38 3.03 ± 0.27 2.38 ± 0.37 2.60 ± 0.31 3.08 ± 0.20 3.16 ± 0.21 3.06 ± 0.37 3.15 ± 0.32 2.85 ± 0.34 2.90 ± 0.41 2.93 ± 0.30 2.20 ± 0.30 2.38 ± 0.31
3.72 ± 0.65 1.85 ± 0.47 3.39 ± 0.33 3.27 ± 0.30 2.34 ± 0.40 2.54 ± 0.38 3.07 ± 0.32 3.07 ± 0.36 3.14 ± 0.42 3.13 ± 0.46 2.95 ± 0.38 3.03 ± 0.37 3.10 ± 0.50 2.28 ± 0.39 2.49 ± 0.44
0.969 0.894 0.036 0.017 0.906 0.838 >0.999 0.343 0.780 >0.999 0.410 0.197 0.147 0.450 0.266
0.873 0.825 0.254 0.257 0.450 0.556 0.298 0.419 0.325 0.494 0.357 0.668 0.096 0.050 0.253
a Values are expressed as mean ± SD (n = 13 wines). Attributes and their levels were based on Level 3 WSET tasting sheet. Two original attributes (clarity and color) were excluded from this analysis because they are categorical variables. b Wilcoxon test.
S: Sensory & Food Quality
correlation to 10 of 15 attributes, suggesting a high level of agreement when the same wine was analyzed during 2 different sessions. The results indicate the high capacity of the panel to carry out wine sensory analysis, giving reliability to the data obtained from each sample. The objective of this study was to identify a group of sensory characteristics that could discriminate wines according to the va-
tensity scale. Three clusters were selected. Then, the 138 samples were coded according to their respective cluster and plotted in the graphs generated by PCA (Figure 3). In this analysis, 61% of the total variability was explained by the first 2 dimensions. The contribution of each variable to the 1st, 2nd, and 3rd PCs and their respective eigenvalues are shown in Table 3. The first axis accounted for 50.6% of the variation and was explained by the intensity of olfactory and gustatory attributes, while the 2nd axis (10.3%) was strongly associated with visual appearance. The information relative to the 3 clusters obtained in our study is summarized in Table 4. Wines produced from the 6 varieties and 4 countries were spread across the 3 clusters. However, the highest proportion of Carm´en`ere (63%), Malbec (50%), and Syrah (45%) produced in Chile (63%) was present in Cluster 1, which is plotted in the left-hand quadrant (Figure 3). Cluster 2, which is plotted in the right-hand quadrant, contained a higher proportion of Tannat (48%) from Brazil (46%) and Uruguay (57%), while Cluster 3, which is plotted in the middle, primarily contained Malbec (43%) and Merlot (46%) from Argentina (46%) and Brazil (46%). The Cabernet Sauvignon wines were equally distributed among the 3 clusters. The mean prices of the wines in Clusters 1 and 3 were not different, but they were higher than those of the wines in Cluster 2. The wines included in Cluster 2 were clearly discriminated from the other wines due to the higher level of development and the lower levels of all other attributes. The wines included in Cluster 1 were perceived to display more intense levels of all attributes, except development and acidity, when compared with the wines from Cluster 3. Table 5 presents the descriptors
Figure 3–Projection of the sensory attributes and wines grouped according to the 3 clusters on the factor-plane (1 × 2). Wines were represented according to their respective clusters (1, 2, and 3). Both vintages (2009 and 2010) were included in this classification (n = 138). Variables designations: clarity, color intensity (ci), color, development, odor intensity (oi), odor fruit (of), odor vegetal (ov), odor spice (os), acidity, tannin, alcohol level, body, persistence, flavor intensity (fi), flavor fruit (ff), flavor vegetal (fv), and flavor spice (fs). Vol. 79, Nr. 8, 2014 r Journal of Food Science S1599
Typicality of South American red wines . . . Table 3–Contribution of each variable (attributes), eigenvalues, and % of total variation related to the 3 major principal components. Attributesa Eigenvalues % Total variation Clarity Color Intensity Color Development Odor intensity Odor fruit Odor vegetal Odor spice Acidity Tannin Alcohol level Body Persistence Flavor intensity Flavor fruit Flavor vegetal Flavor spice a
Range (low – high +)1
Factor 1 8.60 50.6
Factor 2 1.76 10.3
Factor 3 1.38 8.1
hazy dull clear bright pale medium(–) medium medium(+) deep purple ruby garnet tawny youthful developing fully developed past its best Light/low medium– ) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(−) medium medium(+) pronounced Light/low medium– ) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced
−0.26 −0.73 0.64 0.54 −0.69 −0.56 −0.49 −0.76 −0.70 −0.77 −0.72 −0.91 −0.87 −0.87 −0.77 −0.70 −0.83
−0.56 −0.35 0.51 0.58 0.32 0.16 0.36 0.23 −0.20 −0.26 −0.17 −0.03 0.18 0.28 0.15 0.34 0.16
−0.26 0.09 −0.07 −0.11 0.26 0.65 -0.50 −0.23 −0.16 −0.13 −0.10 −0.01 0.05 0.17 0.43 −0.43 −0.19
Attributes and their levels were based on Level 3 WSET tasting sheet.
by which the panelists perceived the wines inside each cluster. All other attributes not shown in Table 5 did not present significant difference among the clusters or were absent. Sensory attributes were observed at the highest frequency for the wines in Cluster 1 compared to Cluster 3, while almost no descriptors were identified in the wines in Cluster 2. Sensory attributes were compared between 2009 and 2010 in Cluster 1 (Figure 4A), Cluster 2 (Figure 4B) and Cluster 3 (Figure 4C). Few changes were observed between the years, suggesting that the sensory separation of the wines did not change according to these 2 vintages.
S: Sensory & Food Quality
Discussion Unlike in Europe, where the quality of a wine is associated with the origin (Charters and Pettigrew 2007), quality in South America is often predicted by the concept of typicality, which is based on a combination of varietal and origin. Typicality and quality used to be correlated and are based on intrinsic cues such as pleasure and appearance, gustatory, and potential characteristics (Charters and Pettigrew 2007). It does not mean that a typical wine is always good, but the concept reduces the risk of an incorrect purchase. Thus, after taking into account a range of price according to the consumption occasion, novice consumers base their purchase decision on extrinsic cues such as brand, packing design, critic´s rating scores, and typicality (Mueller and others 2010). If the expectation of the wine is achieved, the possibility of repurchase increases, conferring commercial value to the combination of “varietal and country.” For this reason, recognition of typicality has become an economic factor for “New World” winemakers. In our study, a sensory panel composed of wine experts was able to separate young South American red wines according to the varietal, origin, and range of prices. In this case, geographical origin involves vitiviniculture practices in addition to the physical and biological environment (Cadot and others 2012). However, using a combination of an intensity scale for visual, olfactory, and taste attributes with frequency of citation, and submitting the results to a multivariate statistical tool for grouping, just a trend for typification could be observed, because all countries and varieties were present within each cluster. In other words, although the panel successfully separated the samples according to varietal and S1600 Journal of Food Science r Vol. 79, Nr. 8, 2014
country, there was no agreement in terms of the attributes applied to this differentiation. Some studies have reported that wine experts learn to categorize wines according to the varietal and are able to recognize them blind, while the identification of origin is less successful (Ballester and others 2008; Maitre and others 2010). In our study, after 2 to 3 bottles were evaluated, the sensory evaluation worksheets were collected, analysis was systematically interrupted, and verbal feedback was given to the panel. We believe that this process improved the pattern of recognition for the following samples. After repeated tasting experiences, the taster builds a mental prototype of the sensory signature of the wines (Maitre and others 2010; Parr and others 2010). Moreover, the expert panel used in our study had a deep knowledge about South American wines, and sometimes during the sensory sessions, even the brand was recognized by some tasters. For this reason, in this study, the term “typicality” could also be taken as “familiarity.” The lack of a better typification of the wines observed in our study may be a consequence of several factors, including the methodology used to perform the sensory analysis. Different methods have been suggested to evaluate wine typicality, such as R analytical sensory description, free choice, Napping , quantitative descriptive analysis (QDA), just about right (JAR), and qualitative approach (Perrin and others 2008; Cadot and others 2010a; Maitre and others 2010). For example, Napping may be better correlated with the typical evaluation because the tasters arrange an “intuitive grouping” where 2 wines are positioned near to each other if they are perceived as identical. However, this method is limited to a smaller number of samples and does not characterize the product itself (Maitre and others 2010). Green and others (2011) identified sensory descriptors to be used for typicality, analyzing the same varietal produced in different regions. Part of their success could be attributed to the smaller number of samples and the use of another sensory method in which the participants were asked to write a few words to explain why they were classifying the wines in a specific group. According to Campo and others (2010), the frequency of citation method might represent a convenient alternative to conventional descriptive analysis. Thus, in our study, we opted to use both methodologies to characterize the wine samples.
Table 4–Proportion of grape variety and region of production Table 5–Sensory descriptors attributed to the wines according to (country), mean price, and sensory attributes of the wines that the clusters. were allocated into Cluster 1, Cluster 2, and Cluster 3. Clusters Clusters a Descriptors Cluster 1 Cluster 2 Cluster 3 Cluster 1 Cluster 2 Cluster 3 Pa Odor terms eucalyptus – – (n = 50) (n = 38) (n = 50) vanilla – – blackcurrant – blackcurrant Varietalb (% of total) CA: 63 CA: 05 CA: 32 coffee – – CS: 29 CS: 37 CS: 34 cedar – – MA: 50 MA: 07 MA: 43 cloves – – ME: 27 ME: 27 ME: 45 toast – toast SY: 45 SY: 23 SY: 32 blueberry – – TA: 17 TA: 48 TA: 35 mint – mint Countryc (% of total) AR: 34 AR: 20 AR: 46 blackpepper ( ) blackpepper( ) blackpepper CH: 63 CH: 11 CH: 26 prune – – BR: 07 BR: 46 BR: 46 chocolate – chocolate UR: 10 UR: 57 UR: 33 blackberry ( ) blackberry( ) blackberry Mean price 19.68 ± 8.01a 9.90 ± 4.01b 16.37± 8.77a
Typicality is the set of sensory characteristics that identify a distinctive type of wine. Thus, the aim of this research was to identify the sensory characteristics that contribute to define typicality of young South American red wines based on their varietal and origin, and to evaluate the effect of the vintage on this identification. To achieve this objective, visual appearance, odor, and taste of 138 wines from 2 vintages were submitted to a sensory evaluation using a descriptive analysis complemented with the frequency of citation method, performed by wine experts. The intensity of 17 odor and taste attributes was evaluated using a 5 points rating structured scale. The panel performance evaluation demonstrated its high level of expertise and reproducibility. The wines were separated into 3 clusters by multivariate analyses. Cluster 1 was primarily composed of Carm´en`ere, Malbec, and Syrah wines from Chile. Cluster 2 was predominantly composed of Tannat wines from Uruguay and Brazil, while Cluster 3 contained a higher proportion of Malbec and Merlot wines from Argentina and Brazil. Cabernet Sauvignon was equally distributed into all clusters. Wine experts were able to identify the wines according to their varietal and origin, suggesting that there is typicality in young South American red wines. The combination of descriptive analysis with the frequency of citation was useful in characterizing most of the wines, but the typicality perceived by the panelists was not achieved by multivariate analysis. Vintage did not alter the sensory characterization of the wines, and this result could be due the new viticulture or oenological practices used by the winemakers to compensate the environmental variation.
Abstract:
Keywords: cluster, sensory, typicality, vintage, wine
Although typicality has been used during the purchasing of wine, there is no evidence that this concept is real for young South American red wines. This study contributed to identify sensory characteristics associated to typicality and showed that vintage did not alter the wines classification.
Introduction South America is a relatively new wine production region, which is known as “New World.” This region includes South Africa, Oceania, Asia, and California. Recently, some “New World” wines have gained international recognition (King and others 2014). However, unlike the “Old World” wines from Europe, where better wines are composed of a mixture of grape varieties produced in a very specific region, relying on a geographic classification system (Grand Cru, Premier Cru, Protected Designation of Origin, or Protected Geographical Indication) (Parr and others 2010), “New World” wines are still predominantly monovarietals. In the South American countries, novice consumers have difficulty of evaluating quality and use to base their wine purchase decisions on extrinsic cues such as price, brand, origin, and varietal (Mueller and others 2010; D’Alessandro and Pecotich 2013). This information creates a sensory expectation for the wine that may or may not be achieved. According to Mueller and others (2010), product expectations at the initial purchase and intrinsic MS 20140470 Submitted 3/21/2014, Accepted 5/13/2014. Authors Llobodanin and Castro are with LADAF, Dept. of Food and Experimental Nutrition, Faculty of Pharmaceutical Sciences, Univ. of S˜ao Paulo, Ave. Lineu Prestes, 580, B14, 05508-900, S˜ao Paulo, Brazil. Author Barroso are with Dept. of Statistics, Inst. of Mathematics and Statistics, Rua do Mat˜ao, 1010, 05508-090, S˜ao Paulo, Brazil. Direct inquiries to author Castro (E-mail: [email protected]; www.ladaf.com.br).
R C 2014 Institute of Food Technologists
doi: 10.1111/1750-3841.12535 Further reproduction without permission is prohibited
sensory attributes during product consumption influence the repurchase decision. Thus, when consumers choose a wine bottle based on its varietal and country of origin or terroir, it is because they believe that there is a combination of sensory attributes that differentiates this varietal or region from the others. This concept is known as “typicality.” A wine is typical when some of its characteristics, which reflect both its origin and varietal purity, can be identified and make it recognizable as belonging to a distinctive type (Charters and Pettigrew 2007; Maitre and others 2010). Therefore, typicality includes sensory, technical, and environmental dimensions (Cadot and others 2010b). However, although typicality has been used during the purchasing of wine, there is no evidence that this concept is real for young South American red wines. Wine sensory characteristics are divided into 3 major aspects: visual appearance, odor, and taste perception. One class of compounds that contributes to these aspects is the phenolics (Fanzone and others 2012), especially the anthocyanins and polymeric proanthocyanidins (Kallithraka and others 2001). These polyphenols are biosynthesized in response to injuries caused by weather conditions and microorganisms, and during the wine making process, including aging and storage (Vitrac and others 2002). Thus, if phenolic substances are one of the main factors that determine the sensory characteristics of the wine and if they can change due to environmental conditions, the typicality, once identified, could also change according to the vintage. Vol. 79, Nr. 8, 2014 r Journal of Food Science S1595
S: Sensory & Food Quality
Practical Application:
Typicality of South American red wines . . . Table 1–Sampling of the wines by region of production and to 10 panelists participated in each session. The bottles were store grape variety. at 17 °C, opened immediately before the analysis, and panelists Country
Vintage 2009
Vintage 2010
Total
Argentinaa
2 Cabernet Sauvignon 5 Malbec 2 Merlot 5 Syrah 2 Tannat 5 Cabernet Sauvignon 1 Carm´en`ere 3 Merlot 1 Syrah 5 Tannat 9 Cabernet Sauvignon 9 Carm´en`ere 2 Malbec 3 Merlot 4 Syrah 3 Cabernet Sauvignon 2 Merlot 6 Tannat 69
3 Cabernet Sauvignon 6 Malbec 2 Merlot 7 Syrah 1 Tannat 4 Cabernet Sauvignon – 4 Merlot 1 Syrah 4 Tannat 10 Cabernet Sauvignon 9 Carm´en`ere 1 Malbec 3 Merlot 4 Syrah 2 Cabernet Sauvignon 3 Merlot 5 Tannat 69
35
Brazilb
Chilec
Uruguayd Total
28
54
21 138
a
Subregions: Valle Calchaqu´ıes, Valles de la Provincia de Catamarca, La Rioja, San Juan, Mendoza, Valle del Uco, Tunuyan, Valle del R´ıo Negro, Valle del Tullum, and Tupungato. b Subregions: Serra Ga´ucha, Planalto Catarinense, Serra do Sudeste, Bento Gonc¸alves, Vale de S˜ao Francisco, and Campanha. c Subregions: Valle de Elqui, Limar´ı, Choapa, Valle Central, Colchagua, Valle Aconcagua, Valle de Rapel, Maipo, Valle de Curic´o, Maule, and Cachapoal. d Subregions: Sur, Cerro Chapeu, Noreste, and Suroeste.
In this study, it was hypothesized that young wines could be sensorially characterized by varietal and country of origin, but this classification could change according to the vintage. To test this hypothesis, 138 young monovarietal red wines from Vitis vinifera grapes produced in 2009 and 2010 vintages were submitted to a sensory evaluation by a panel of wine experts.
S: Sensory & Food Quality
Material and Methods Wine samples A group of 138 red monovarietal wines from Vitis vinifera (Cabernet Sauvignon, Carm´en`ere, Malbec, Merlot, Tannat, and Syrah) produced during 2009 (69 wines) and 2010 (69 wines) in Chile, Argentina, Uruguay, and Brazil with retail prices ranging from U.S.$5.24 to 45.71 per bottle (750 mL) were chosen for the present study (Table 1). Production statistics regarding grape variety, production region, and price were taken into account to make the selection as representative as possible of the wines produced in these countries. When possible, the same brand was chosen for the 2 vintages. When the same wine was not available, a similar one was purchased, maintaining the same grape variety, production region, and price range. Although young, some of the wines were oaked. In addition to these 138 wines (Table 1), 13 bottles were randomly repeated during the sensory analysis, forming 13 pairs. These pairs were used to evaluate panel reproducibility. Sensory analysis The sensory analysis of the wines was performed by 10 professional sommeliers (25 to 60 y old), being 6 females and 4 males. All tasters had worked to wine distributors for more than 5 y. They had extensive experience in wine tasting, especially with regard to South American wines. The samples were assessed in groups of 8 to 10 wines per session. A volume of 30 mL of each sample was poured in a standardized crystal red wine tasting vessel (Open Up—Universal, 40cl, 13 ½ oz; ARC Intl. Group, France), and 7 S1596 Journal of Food Science r Vol. 79, Nr. 8, 2014
were prevented from seeing their label or shape. Sensory analyses were split into 10 sessions (2/week) for each vintage and were held for approximately 2 mo in 2010 and more 2 mo in 2011. Wines were presented individually, with a 10-min enforced break after every 2 to 3 samples. The panelists were required to expectorate the wines into a spittoon. They were informed that they were evaluating South American red wines from 6 different grape varieties (Cabernet Sauvignon, Carm´en`ere, Malbec, Merlot, Tannat, and Syrah) produced in 4 countries (Chile, Brazil, Argentina, and Uruguay). However, for each wine sample, no specific information was provided concerning the grape variety, country of origin, or price. Analyses were individually carried out, samples were coded with 3 randomized numbers, bottles were randomized between and inside the sessions, and they were monadically presented to the panel.
Descriptive analysis and frequency of citation methods The tasting sheet included visual appearance, odor, and taste perceptions. To perform the descriptive analysis, attributes were chosen from those defined by the school for wine professionals level 3 (WSET 2009), where all of the panelists had been trained. Thus, judges had a complete understanding of all the attributes present in the tasting sheet. All judges had been before trained by WSET system (from level 1 to level 3), and for this reason, no additional training was performed in this study. About 110 wine samples that did not take part of our sampling were selected as endpoint of the attributes scale during the tasters training. After, all tasters were submitted to a practical exam before of becoming certified (WSET 2009). Testing was performed at an open table, and a verbal feedback was given to the panel after finishing all steps of their reports. Initially, appearance was assessed by clarity (hazy, dull, clear, or bright), color (purple, ruby, garnet, and tawny), and color intensity (pale, medium(−), medium, medium(+), and deep). Then, the panelists were asked to swirl and smell each sample for approximately 15 s and evaluate the orthonasal attributes, rating their intensity on a 1 to 5 point structured scale. At this step, they classified the samples according to the level of development (youthful, developing, fully developed, and past its best), odor intensity, and odor characteristics (fruit, vegetal, and spice). Next, the frequency of citation methodology was used to complement the descriptive analysis. For this technique, the tasters select the most pertinent attributes from a list containing a relatively high number of terms (Campo and others 2008). In our study, a list containing 53 sensory attributes was given to the panelists, who indicated the presence or not of each descriptor in the sample. The expert panel was familiarized with the terms that composed the frequency of citation list. Afterward, the retronasal and taste attributes were evaluated after swirling the sample for approximately 15 s inside the mouth. The panelists classified the wines according to acidity, tannin, alcohol level, body, persistence, flavor intensity, and flavor characteristics (fruit, vegetal, and spice) using a 1 to 5 point structured scale. The same list of 53 attributes used to characterize the olfactory aspects was used for the taste characterization. The attributes assigned in the frequency of citation were converted into scores (“1” if a descriptor was marked and “0” if not), and the mean values of the panel were applied to the wine analyses. Finally, panelists were invited to suggest a varietal, country, and a retail price for the each evaluated sample.
Typicality of South American red wines . . .
Results The panel used in this study was evaluated for discriminatory capacity and reproducibility. Figure 1 shows the percentage of correct categorization obtained when the panelists were asked to identify the varietal, country, and price of the wines. In general, the panelists were able to recognize wines by varietal (Figure 1A), mainly Cabernet Sauvignon, Malbec, and Carm´en`ere.
30
** 25 20
** ns ns
ns 15 10 5 0
B
**
Merlot
Cabernet Tannat
Malbec Carménère Shiraz
50 ** 40
** **
30 ns
S: Sensory & Food Quality
ProporƟon (%) of correct categorizaƟon
A
ProporƟon (%) of correct categorizaƟon
Statistical analysis Initially, the performance of the panelists was assessed in terms of discrimination and reproducibility. After tasting each wine, the panelist was invited to suggest the varietal, country, and commercial price, from low (5.0 to 15.0 U.S.$), medium (15.1 to 25.0 U.S.$), or high (25.1 to 50.0 U.S.$). To determine whether the percentage of correct categorization was obtained by chance or not, the normal approximation of binomial statistics was applied, adopting a value of 0.05 to alpha risk. In the sensory analysis, 13 wine bottles were randomly duplicated in the sessions, totaling 26 samples, where each pair consisted of the same wine. The results from these 26 samples were submitted to a principal component analysis (PCA), where a 2-dimensional graph was generated by visual appearance and odor and flavor characteristics separately. The means of the attributes were used as active variables, while the wine samples were plotted as cases. Moreover, the mean or median values of the sensory attributes obtained in the 2 replicates were compared by Wilcoxon test. The intraclass correlation coefficient (ICC) was also calculated. After checking the panelist´s performance, a pattern of attributes able to define any possibility of typicality was searched. The data were submitted to multivariate correspondence analysis (data not shown), but no pattern was evident. Thus, the panelists scores obtained from the intensity scales of 17 attributes were converted to continuous variables, and hierarchical cluster analysis was applied to separate the wines according to visual appearance, olfactory, and taste characteristics. The wines, coded according to their respective clusters, were plotted in a factor-plane (1 × 2) generated by PCA using the same 17 sensory attributes. Fifty-three descriptors were recorded by the panelists as present (+1) or absent (0) for each wine sample. The mean value of each descriptor was compared among the 3 clusters by one-way ANOVA. To check the vintage effect in the clusters characterization, the mean values observed in 2009 were compared with the mean values observed in 2010, using a dependent T-test. All calculations were performed using the software STATISTICA version 9.0 (StatSoft, Inc., Tulsa, Okla., U.S.A.) and the SAS (Statistical Analysis Software) version 9.2 (SAS Inst. Inc., Cary, N.C., U.S.A.).
Similarly, they had significant success classifying the wines/country (Figure 1B), except for the Uruguayan ones. Panelists were also asked to suggest a high, medium, or low price range for the wines. They had a significant rate of success, mainly for lowand medium-priced wines (Figure 1C). Taking, for example, Figure 1(B), from the total responses obtained to Chile (389), 178
20
10
0
Brazil
Uruguay
Chile
Argentina
C 100 ProporƟon (%) of correct categorizaƟon
Panelist’s performance The panelist’s reproducibility was evaluated by comparing the score attributed to the same samples when evaluated in 2 different sessions. The following attributes were evaluated to check the panelist´s reproducibility: color intensity, development, odor intensity, odor fruit, odor vegetal, odor spice, acidity, tannins, alcohol, body, persistence, flavor intensity, flavor fruit, flavor vegetal, and flavor spice. Discriminatory capacity was measured by comparing the number of correct responses when the panelists were asked to identify the varietal, country, and price range against the number of correct responses obtained by chance. To evaluate the panelists’ performance, an average reproducibility index (Ri ) was also calculated as described by Campo and others (2008). To be able to calculate the Ri , about 10% of the wines (13 out of the total 138) were analyzed in duplicate, making a total of 151 samples.
80 ** 60 ** ns
40
20
0
LOW
MED
HIGH
Figure 1–Proportion (%) of correct categorization when tasters were asked to identify the varietal (A), country (B), and price (C), ∗∗ P < 0.01. It represents the % of the total responses obtained for each category that were correctly answered by the tasters. Vol. 79, Nr. 8, 2014 r Journal of Food Science S1597
Typicality of South American red wines . . . were correct answers (46%). This number (178) was higher than the number expected if the correct answer had been obtained just by chance (112). Thus, even in a blind test, the correct categorization of the wines by varietal, country, and price did not occur by chance. This observation suggests that the wines presented sensory characteristics that were used by the panel to identify their origin. In other words, the wines presented sensory typicality. At the next step of the panelist´s performance evaluation, PCA was used to compare how the panelists positioned repeated samples (n = 26 bottles, forming 13 pairs) according to visual appearance, olfactory (Figure 2A), and taste attributes (Figure 2B). The same wines occupied close positions on the plot, indicating similarities between the pairs according to the sensory attributes. A high level
A
of agreement between the wine pairs was evident. Considering that the samples were randomly distributed between 20 sessions over approximately 4 mo, these results suggest a high level of reproducibility. The Ri index evaluates the individual reproducibility of the panelists. It varies from 0 to 1 and a value of 0.2 or higher is considered acceptable (Campo and others 2008). The values for the panelists ranged from 0.22 to 0.49. Therefore, all the panelists responses were included in the present study. Table 2 presents the means and probability values obtained for each sensory attribute that was analyzed twice for the same sample. From all 15 attributes associated with visual appearance, odor, and flavor, only 2 were significantly different (odor intensity, P = 0.036 and fruit odor, P = 0.017). The ICC values showed an excellent or satisfactory
Figure 2–Representation of the replicate wines on the 2-dimensional plane generated by the 1st and 2nd PCs according to the “visual appearance and olfactory” characteristics (A) and taste perception (B); where winei and winei´ represent 2 replicates of a single wine (n = 13 pairs).
2.5 88
13´ 13´ 77
1.5
12 12
PC 2: 18.32%
9´ 9´
8´ 8´ 7´ 7´ 13 13
4´ 4´
0.5
44 12´ 12´ 10 10
11 11 99 11´ 11´
10´ 10´
55
3´ 3´
1´ 1´
66
-0.5 6´ 6´
5´ 5´ 22
-1.5 33
S: Sensory & Food Quality
2´ 2´
11
-2.5 -4
-3
-2
-1
0
1
2
3
4
5
PC 1: 54.65%
B
3
12 12
2
88 13´ 13´ 44
1
7´ 7´
PC 2: 12.32%
2´ 2´ 99
0
8´ 8´
77 12´ 12´
10´ 10´
1´ 1´ 22
5´ 5´ 10 10 11 11 5 5
3´ 3´ 33 4´ 4´ 13 13
9´ 9´
66
-1
11 -2
6´ 6´ 11´ 11´
-3 -4
-2
0
2 PC 1: 58.59%
S1598 Journal of Food Science r Vol. 79, Nr. 8, 2014
4
6
8
Typicality of South American red wines . . . Table 2–Visual appearance, olfactory, and gustatory scores of rietal or region of their production and to determine whether the the sensory characteristics attributed by the panelists to the vintage could change this characterization. To achieve this aim, wines tasted twice, including the “intraclass correlation coefcluster analysis was applied to all attributes measured by an inficient” (ICC). Sensory evaluation Attributea
1st analysis
2nd analysis
Pb
ICC
Color intensity Development Odor intensity Odor fruit Odor vegetal Odor spice Acidity Tannins Alcohol Body Persistence Flavor intensity Flavor fruit Flavor vegetal Flavor spice
3.71 ± 0.81 1.86 ± 0.52 3.13 ± 0.38 3.03 ± 0.27 2.38 ± 0.37 2.60 ± 0.31 3.08 ± 0.20 3.16 ± 0.21 3.06 ± 0.37 3.15 ± 0.32 2.85 ± 0.34 2.90 ± 0.41 2.93 ± 0.30 2.20 ± 0.30 2.38 ± 0.31
3.72 ± 0.65 1.85 ± 0.47 3.39 ± 0.33 3.27 ± 0.30 2.34 ± 0.40 2.54 ± 0.38 3.07 ± 0.32 3.07 ± 0.36 3.14 ± 0.42 3.13 ± 0.46 2.95 ± 0.38 3.03 ± 0.37 3.10 ± 0.50 2.28 ± 0.39 2.49 ± 0.44
0.969 0.894 0.036 0.017 0.906 0.838 >0.999 0.343 0.780 >0.999 0.410 0.197 0.147 0.450 0.266
0.873 0.825 0.254 0.257 0.450 0.556 0.298 0.419 0.325 0.494 0.357 0.668 0.096 0.050 0.253
a Values are expressed as mean ± SD (n = 13 wines). Attributes and their levels were based on Level 3 WSET tasting sheet. Two original attributes (clarity and color) were excluded from this analysis because they are categorical variables. b Wilcoxon test.
S: Sensory & Food Quality
correlation to 10 of 15 attributes, suggesting a high level of agreement when the same wine was analyzed during 2 different sessions. The results indicate the high capacity of the panel to carry out wine sensory analysis, giving reliability to the data obtained from each sample. The objective of this study was to identify a group of sensory characteristics that could discriminate wines according to the va-
tensity scale. Three clusters were selected. Then, the 138 samples were coded according to their respective cluster and plotted in the graphs generated by PCA (Figure 3). In this analysis, 61% of the total variability was explained by the first 2 dimensions. The contribution of each variable to the 1st, 2nd, and 3rd PCs and their respective eigenvalues are shown in Table 3. The first axis accounted for 50.6% of the variation and was explained by the intensity of olfactory and gustatory attributes, while the 2nd axis (10.3%) was strongly associated with visual appearance. The information relative to the 3 clusters obtained in our study is summarized in Table 4. Wines produced from the 6 varieties and 4 countries were spread across the 3 clusters. However, the highest proportion of Carm´en`ere (63%), Malbec (50%), and Syrah (45%) produced in Chile (63%) was present in Cluster 1, which is plotted in the left-hand quadrant (Figure 3). Cluster 2, which is plotted in the right-hand quadrant, contained a higher proportion of Tannat (48%) from Brazil (46%) and Uruguay (57%), while Cluster 3, which is plotted in the middle, primarily contained Malbec (43%) and Merlot (46%) from Argentina (46%) and Brazil (46%). The Cabernet Sauvignon wines were equally distributed among the 3 clusters. The mean prices of the wines in Clusters 1 and 3 were not different, but they were higher than those of the wines in Cluster 2. The wines included in Cluster 2 were clearly discriminated from the other wines due to the higher level of development and the lower levels of all other attributes. The wines included in Cluster 1 were perceived to display more intense levels of all attributes, except development and acidity, when compared with the wines from Cluster 3. Table 5 presents the descriptors
Figure 3–Projection of the sensory attributes and wines grouped according to the 3 clusters on the factor-plane (1 × 2). Wines were represented according to their respective clusters (1, 2, and 3). Both vintages (2009 and 2010) were included in this classification (n = 138). Variables designations: clarity, color intensity (ci), color, development, odor intensity (oi), odor fruit (of), odor vegetal (ov), odor spice (os), acidity, tannin, alcohol level, body, persistence, flavor intensity (fi), flavor fruit (ff), flavor vegetal (fv), and flavor spice (fs). Vol. 79, Nr. 8, 2014 r Journal of Food Science S1599
Typicality of South American red wines . . . Table 3–Contribution of each variable (attributes), eigenvalues, and % of total variation related to the 3 major principal components. Attributesa Eigenvalues % Total variation Clarity Color Intensity Color Development Odor intensity Odor fruit Odor vegetal Odor spice Acidity Tannin Alcohol level Body Persistence Flavor intensity Flavor fruit Flavor vegetal Flavor spice a
Range (low – high +)1
Factor 1 8.60 50.6
Factor 2 1.76 10.3
Factor 3 1.38 8.1
hazy dull clear bright pale medium(–) medium medium(+) deep purple ruby garnet tawny youthful developing fully developed past its best Light/low medium– ) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(−) medium medium(+) pronounced Light/low medium– ) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced Light/low medium(–) medium medium(+) pronounced
−0.26 −0.73 0.64 0.54 −0.69 −0.56 −0.49 −0.76 −0.70 −0.77 −0.72 −0.91 −0.87 −0.87 −0.77 −0.70 −0.83
−0.56 −0.35 0.51 0.58 0.32 0.16 0.36 0.23 −0.20 −0.26 −0.17 −0.03 0.18 0.28 0.15 0.34 0.16
−0.26 0.09 −0.07 −0.11 0.26 0.65 -0.50 −0.23 −0.16 −0.13 −0.10 −0.01 0.05 0.17 0.43 −0.43 −0.19
Attributes and their levels were based on Level 3 WSET tasting sheet.
by which the panelists perceived the wines inside each cluster. All other attributes not shown in Table 5 did not present significant difference among the clusters or were absent. Sensory attributes were observed at the highest frequency for the wines in Cluster 1 compared to Cluster 3, while almost no descriptors were identified in the wines in Cluster 2. Sensory attributes were compared between 2009 and 2010 in Cluster 1 (Figure 4A), Cluster 2 (Figure 4B) and Cluster 3 (Figure 4C). Few changes were observed between the years, suggesting that the sensory separation of the wines did not change according to these 2 vintages.
S: Sensory & Food Quality
Discussion Unlike in Europe, where the quality of a wine is associated with the origin (Charters and Pettigrew 2007), quality in South America is often predicted by the concept of typicality, which is based on a combination of varietal and origin. Typicality and quality used to be correlated and are based on intrinsic cues such as pleasure and appearance, gustatory, and potential characteristics (Charters and Pettigrew 2007). It does not mean that a typical wine is always good, but the concept reduces the risk of an incorrect purchase. Thus, after taking into account a range of price according to the consumption occasion, novice consumers base their purchase decision on extrinsic cues such as brand, packing design, critic´s rating scores, and typicality (Mueller and others 2010). If the expectation of the wine is achieved, the possibility of repurchase increases, conferring commercial value to the combination of “varietal and country.” For this reason, recognition of typicality has become an economic factor for “New World” winemakers. In our study, a sensory panel composed of wine experts was able to separate young South American red wines according to the varietal, origin, and range of prices. In this case, geographical origin involves vitiviniculture practices in addition to the physical and biological environment (Cadot and others 2012). However, using a combination of an intensity scale for visual, olfactory, and taste attributes with frequency of citation, and submitting the results to a multivariate statistical tool for grouping, just a trend for typification could be observed, because all countries and varieties were present within each cluster. In other words, although the panel successfully separated the samples according to varietal and S1600 Journal of Food Science r Vol. 79, Nr. 8, 2014
country, there was no agreement in terms of the attributes applied to this differentiation. Some studies have reported that wine experts learn to categorize wines according to the varietal and are able to recognize them blind, while the identification of origin is less successful (Ballester and others 2008; Maitre and others 2010). In our study, after 2 to 3 bottles were evaluated, the sensory evaluation worksheets were collected, analysis was systematically interrupted, and verbal feedback was given to the panel. We believe that this process improved the pattern of recognition for the following samples. After repeated tasting experiences, the taster builds a mental prototype of the sensory signature of the wines (Maitre and others 2010; Parr and others 2010). Moreover, the expert panel used in our study had a deep knowledge about South American wines, and sometimes during the sensory sessions, even the brand was recognized by some tasters. For this reason, in this study, the term “typicality” could also be taken as “familiarity.” The lack of a better typification of the wines observed in our study may be a consequence of several factors, including the methodology used to perform the sensory analysis. Different methods have been suggested to evaluate wine typicality, such as R analytical sensory description, free choice, Napping , quantitative descriptive analysis (QDA), just about right (JAR), and qualitative approach (Perrin and others 2008; Cadot and others 2010a; Maitre and others 2010). For example, Napping may be better correlated with the typical evaluation because the tasters arrange an “intuitive grouping” where 2 wines are positioned near to each other if they are perceived as identical. However, this method is limited to a smaller number of samples and does not characterize the product itself (Maitre and others 2010). Green and others (2011) identified sensory descriptors to be used for typicality, analyzing the same varietal produced in different regions. Part of their success could be attributed to the smaller number of samples and the use of another sensory method in which the participants were asked to write a few words to explain why they were classifying the wines in a specific group. According to Campo and others (2010), the frequency of citation method might represent a convenient alternative to conventional descriptive analysis. Thus, in our study, we opted to use both methodologies to characterize the wine samples.
Table 4–Proportion of grape variety and region of production Table 5–Sensory descriptors attributed to the wines according to (country), mean price, and sensory attributes of the wines that the clusters. were allocated into Cluster 1, Cluster 2, and Cluster 3. Clusters Clusters a Descriptors Cluster 1 Cluster 2 Cluster 3 Cluster 1 Cluster 2 Cluster 3 Pa Odor terms eucalyptus – – (n = 50) (n = 38) (n = 50) vanilla – – blackcurrant – blackcurrant Varietalb (% of total) CA: 63 CA: 05 CA: 32 coffee – – CS: 29 CS: 37 CS: 34 cedar – – MA: 50 MA: 07 MA: 43 cloves – – ME: 27 ME: 27 ME: 45 toast – toast SY: 45 SY: 23 SY: 32 blueberry – – TA: 17 TA: 48 TA: 35 mint – mint Countryc (% of total) AR: 34 AR: 20 AR: 46 blackpepper ( ) blackpepper( ) blackpepper CH: 63 CH: 11 CH: 26 prune – – BR: 07 BR: 46 BR: 46 chocolate – chocolate UR: 10 UR: 57 UR: 33 blackberry ( ) blackberry( ) blackberry Mean price 19.68 ± 8.01a 9.90 ± 4.01b 16.37± 8.77a
![Identification of sensory attributes that drive the likeability of Korean rice wines by American panelists. [Corrected].](/assets/img/hold.png)
