Abstract
Astringency is one of the most important sensory attributes in red wine and is attributed to the presence of phenolic compounds. The objective of this study was to determine the relationship between tannin, anthocyanin, and small (SPP) and large (LPP) polymeric pigment concentrations in numerous Washington State red wines and to relate these values to perceived astringency. Three groups of wines were formed from a larger collection of wines based on tannin concentration: low (<400 mg/L CE), medium (400–800 mg/L CE), and high (>800 mg/L CE). Cabernet Sauvignon wines representing low, medium, and high tannin concentrations were selected and evaluated by 18 untrained panelists. Sensory results indicated that panelists gave significantly higher astringency ratings to wines high in tannin concentration compared with wines low in tannin concentration (p ≤ 0.05). In a second study, a trained panel evaluated Cabernet Sauvignon and Merlot wines representing combinations of low, medium, and high tannin, anthocyanin, SPP, and LPP concentrations. Results indicated that perceived astringency was significantly correlated with tannin, SPP, and LPP levels, while bitterness was significantly correlated with SPP, LPP, and tannin levels (p ≤ 0.05). These results indicated a relationship between polyphenolic compounds (tannins, SPP, and LPP) and the sensory attributes of astringency and bitterness.
Consumers differ in their preference for wine attributes, notably for astringency, which has been described as one of the most important characteristics of wine (Guinard et al. 1986). Astringency is not a taste but a mouthfeel (Singleton and Esau 1969, ASTM 1989) and is the feeling of dryness or roughness that results from increased friction between the tongue and the surfaces inside the mouth (Lea and Arnold 1978). The friction arises from the precipitation of polyphenolic compounds with oral salivary proteins, resulting in the dry or rough sensation of astringency (Gawel 1998).
Tannins are the main class of polyphenolic compounds thought to contribute to astringency (Singleton and Noble 1976) and are found in the seeds and skins of winegrapes (Cortell et al. 2005, Cheynier et al. 1998). Tannins are water soluble phenolic compounds with molecular weights ranging between 500 and 3000 daltons (Bate-Smith and Swain 1962). The relationship between polyphenolic compounds in wine and the sensory perception of astringency has been explored in a number of studies. In one study, model wine solutions similar in composition to dry white table wine with three levels of phenolic concentration were evaluated for perceived astringency (Arnold and Noble 1978). Results from the trained panel analysis showed that an increase in the polyphenolic concentration corresponded to an increase in perceived astringency, indicating the influence of polyphenolic compounds on perceived astringency.
In order to relate astringency to chemical analysis, it is necessary to determine the polyphenolic content in wine. A survey of California red wines (Adams et al. 2004) used a protein precipitation assay (Harbertson et al. 2003) that provided chemical determinations for tannin, anthocyanins, large polymeric pigment (LPP), and small polymeric pigment (SPP) concentrations. SPPs do not precipitate with protein, while LPPs precipitate with protein (Harbertson et al. 2003). Reported tannin concentration ranges for red wines were 160 to 1563 mg/L catechin equivalents (CE) (Cabernet Sauvignon) and 132 to 792 mg/L CE (Syrah) (Adams et al. 2004). The ranges of anthocyanin, SPP, and LPP concentrations and the relationships between tannin, anthocyanin, SPP, and LPP concentrations on the sensory perception of astringency were not presented. However, another study reported a strong correlation (r2 = 0.82) between the Harbertson polymeric pigment assay for tannin concentration and perceived astringency using an experienced panel (Kennedy et al. 2006). Both tannin and LPP concentrations have also been positively correlated to perceived astringency (Boselli et al. 2004).
The objective of the present study was to explore the relationship between perceived astringency and tannin, anthocyanin, SPP, and LPP concentrations in Washington State red wines using sensory and chemical methodologies.
Materials and Methods
Chemical assay.
Bovine serum albumin (BSA; fraction V powder), sodium dodecyl sulfate (SDS; lauryl sulfate, sodium salt), triethanolamine (TEA), ferric chloride hexahydrate, potassium metabisulfite, and (+)-catechin were purchased from Sigma (St. Louis, MO). The materials used for preparing buffers (hydrochloric acid, malic acid, sodium chloride, sodium hydroxide, 100% ethanol, quinine sulfate, and glacial acetic acid) were purchased from J.T. Baker (Phillipsburg, NJ). A Genesys 10 UV scanning spectrophotometer (Thermo Electron Corporation, Madison, WI) was used for the chemical assay, together with a Vortex-Genie 2 (Scientific Industries, Bohemia, NY) and Galaxy 140 centrifuge (VWR, Bridgeport, NJ).
Wines were selected from a collection of wines previously analyzed for tannin, anthocyanins, SPP, and LPP using the Hagerman and Butler (1978) method for tannin analysis, as modified by Harbertson et al. (2003). Based on tannin concentrations determined by the chemical assay, the wines were divided into three groups; high (>800 mg/L CE), medium (400–800 mg/L CE), and low (<400 mg/L CE). For the untrained and trained sensory panel evaluations, wines from each grouping (low, medium, and high) were selected.
Astringency assessment by untrained panel.
The untrained sensory panel was composed of 18 volunteer participants, 9 male and 9 female, all between the ages of 21 and 65. The panelists were recruited from Washington State University (WSU) and consumed red wine at least twice a month. All participants signed an informed consent form and the project was approved for human subject participation by the WSU Institutional Review Board. Panelists were awarded a small nonmonetary incentive for participation. Prior to evaluation of the wine, subjects familiarized themselves with astringency standards. The standards consisted of three levels of alum (McCormick, Baltimore, MD): 5 g/L, 10 g/L, and 15 g/L (Lee and Lawless 1991). The alum was dissolved in a base red wine (2006, Franzia, Ripon, CA).
Three Cabernet Sauvignon wines were selected from a larger collection of wines previously analyzed for tannin, anthocyanin, SPP, and LPP. Wines representing the low (<400 mg/L CE), medium (400–800 mg/L CE), and high (>801 mg/L CE) tannin concentration wines were selected. The high tannin wine (1071 mg/L CE) was from Columbia Valley (2001), the medium (631 mg/L CE) from Yakima Valley (2002), and the low (250 mg/L CE) from Columbia Valley (2001). These wines were part of a larger study that classified wines into three groups based on the tannin concentration determined using the assay developed by Harbertson et al. (2003). Prior to the sensory panel evaluation, wines were maintained at room temperature for at least 24 hours.
A randomized balanced block design was used for sample presentation. Specifically, the wines were presented in a random serving order as two replicate flights of three samples (low, medium, and high tannin concentration) for a total of six samples. Each wine was coded with a three-digit code and red lights were used during evaluation to mask any color differences between wines (Meilgaard et al. 1999). Each sample consisted of 10 mL wine, poured into an ISO (International Standards Organization) tasting glass, and covered with a plastic putridity. The evaluation of the wine samples was conducted over one panel session. For the interstimulus protocol during wine evaluations, panelists were instructed to chew one cracker, rinse with deionized water, and wait at least 2 min between samples. The panelists rested at least 5 min between flights. Panelists rated the astringency of each sample using a 15-cm unstructured line scale, anchored with “not astringent” at 1 cm and “extremely astringent” at 14 cm.
Astringency intensities were expressed as distance along the 15-cm unstructured line scale. Two-way analysis of variance (ANOVA) with replication and Tukey’s multiple comparisons were used to analyze the results for significant differences (XLSTAT, Addinsoft, Paris,France). The ANOVA performed used a mixed effects model, with the panelists held as a random effect and the wine as a fixed effect. Significance was defined as p ≤ 0.05.
Astringency assessment by trained panel.
Ten volunteers were recruited from the WSU student and staff populations and were selected to participate in the trained sensory panel based on their availability. The panel consisted of 3 males and 7 females between the ages of 22 and 58. Panelists received a small nonmonetary compensation for their participation.
Panelists were trained over six sessions to identify and rate astringency through the presentation and discussion of wines of varying astringency levels. Syrah, Cabernet Sauvignon, and Merlot were used during training to limit potential panelist bias toward different wine varieties during the evaluation sessions. It was thought that if trained with a variety of wines, then panelists would be able to focus on the single attribute of astringency without allowing the difference in flavor characteristics to bias their evaluation. For astringency training, astringency standards of three levels of alum were used (5 g/L, 10 g/L, and 15 g/L prepared in a base red wine). Panelists were also trained to recognize bitterness using a bitterness standard of 0.1 g/L quinine sulfate (J.T. Baker, Phillipsburg, NJ) prepared in base red wine (Jackson 2002). Following initial training with standards, panelists were trained using red wines of different anthocyanin, tannin, SPP, and LPP concentrations determined using the assay described by Harbertson et al. (2003). Panelists were trained so that low perceived astringency ranged from 0 to 5 cm on an unstructured line scale, medium from 5.1 to 10 cm, and high from 10.1 to 15 cm.
Four formal evaluation sessions were conducted over two weeks. For all evaluations, wines were presented using a randomized balanced block design. Wines were coded with a three-digit random number and evaluated under red lights to mask any color differences. During evaluations, panelists were instructed to chew one cracker, rinse with deionized water, and wait at least 2 min between samples. Instructions and scales were presented to the panelists using a computerized sensory software program (Compusensefive, release 4.6, Compusense Inc., Guelph, ON) and data were collected. Panelists recorded their response on two 15-cm unstructured line scales, anchored with “not astringent” at 1 cm and “extremely astringent” at 14 cm or “not bitter” at 1 cm and “extremely bitter” at 14 cm.
All data were analyzed using two-way ANOVA with replication and Tukey’s multiple comparisons (XLSTAT). The ANOVA performed used a mixed effects model, holding the panelists as a random effect and the wine as a fixed effect. For trained panel analysis, principal component analysis (PCA) was conducted using the Pearson correlation matrix with no rotation. For these analyses, the means of each quadruplicate wine evaluation were used. The significance value for all analyses was established as p ≤ 0.05.
Results and Discussion
Results of the untrained panel analysis of variance showed that wine had a significant effect on astringency ratings, indicating that panelists could differentiate wines based on perceived astringency (Table 1⇓). The panelist effect was also significant, suggesting that the evaluation of astringency depended upon the individual’s perception and use of scale (p ≤ 0.05).
The average ratings for astringency by the untrained panel were significantly different between the high and low tannin wines (Table 2⇓). These results showed that the wine higher in tannin concentration was perceived as significantly more astringent than the wine with the low tannin concentration, results consistent with previous reports (Kennedy et al. 2006, Arnold and Noble 1978). However, in this study, no significant differences in astringency between the medium and high tannin wine were observed.
The wines evaluated by the trained panel were selected primarily based on their tannin concentration, with further selection based on SPP and LPP concentrations (Table 3⇓). In order to achieve different levels of wine components, two varieties of wine were studied, Cabernet Sauvignon and Merlot. These wines were selected primarily because of their difference in tannin concentration, but also because of their differences in anthocyanin, SPP, and LPP concentrations.
In order to simplify presentation of the results, each red wine was designated a letter based on tannin, anthocyanins, SPP, and LPP concentrations (Table 3⇑). Wine A showed low concentrations of all chemical components while wine C had high concentrations of the components. Bitterness ratings were significantly different between wines A and C with no differences observed between wines A and B or B and C.
Astringency ratings were significantly different between wines A and B and wines A and C (p ≤ 0.05). Wine C, which had the highest tannin concentration, was not rated as significantly higher in astringency than wine B, which had a medium tannin concentration. The trained panel may not have been able to determine differences in perceived astringency between the medium and high tannin wines because of masking by other aspects of the wine, including fruit character, pH (Peleg et al. 1998), alcohol (Serafini et al. 1997), and viscosity (Peleg and Noble 1999). Overall, wine A, which had the lowest tannin, was rated as significantly lower in astringency and bitterness compared with the higher tannin wines. The trained panelists gave significantly different astringency ratings to the two Merlot wines evaluated, which differed from the results found elsewhere (Cliff et al. 2007).
Similar results between the untrained and trained panel were found in that significant differences between low and high tannin wines were observed with no significant differences in perceived astringency between medium and high and low and medium tannin wines. In both the trained and untrained panel, the difficulty in distinguishing between wines of medium and high tannin concentration was attributed to a carryover effect of astringency (Arnold 1983), which occurs when a sensation lingers following the evaluation of a sample, affecting the evaluation of the next sample (O’Mahony 1986). During a time-intensity study of astringency in wine, some residual astringency could be detected 100 seconds after tasting the wine (Valentova et al. 2002). If insufficient time was allowed between samples, then panelists may confuse a lower astringency wine with a higher astringency wine because of the lingering sensation from the previous sample. Unsalted saltine crackers were found to be more effective at reducing astringency carryover than pectin rinses over repeated sips of red wine (Ross et al. 2007); however, even with the use of crackers to cleanse the palate, astringency gradually built up over time. In the current study, panelists used unsalted saltine crackers between wine samples.
When Pearson correlations were performed (Table 4⇓), astringency had a significantly positive correlation with SPP (r2 = 0.758, df = 10), LPP (r2 = 0.653, df = 10), and tannin (r2 = 0.506, df = 10) concentration. Bitterness was also positively correlated to SPP (r2 = 0.596, df = 10), LPP (r2 = 0.656, df = 10), and tannin (r2 = 0.658, df = 10). Bitterness, astringency, and LPP were not correlated with anthocyanin concentration. SPP had a strong correlation with LPP (r2 = 0.953, df = 10) and tannin content (r2 = 0.826, df = 10). A strong correlation was observed between LPP and tannin (r2 = 0.956, df = 10).
In the current study, astringency showed a strong positive relationship with tannin, SPP, and LPP, the latter being a pigmented polymer consisting of four or more monomeric units. In agreement, earlier studies have also suggested a positive relationship between astringency and tannin concentration and astringency and polymer concentration (Vidal et al. 2003, Arnold and Noble 1978). However, a positive relationship was also noted between SPP and perceived astringency. In one study (Vidal et al. 2003), trained panelists evaluated the model wine samples using the mouthfeel wheel and specific mouthfeel descriptors with astringency subqualities (Gawel et al. 2000). While the present study defined astringency as drying or puckering in the mouth without attention to subqualities, it may have resulted in the panelists not being able to specifically express the mouthfeel quality perceived, instead classifying their perceptions more generally as “astringent.” The relationship between SPP and astringency should be further explored with a panel trained to define more specific subqualities of astringency, such as fine, grainy, dry, and chalky. Moreover, interpretation of the results may have been influenced by potential limitations of some tannin-protein precipitation assays based on the relative amounts of tannin (Silber and Fellman 2006, Hagerman and Robbins 1987, Martin and Martin 1982).
As observed with astringency, bitterness was positively correlated to SPP, LPP, and tannin. The relationship between bitterness and SPP was supported by past research in which a positive association between monomer concentration and bitterness was reported (Arnold and Noble 1978). However, another study reported that perceived bitterness was not affected by degree of polymerization (Vidal et al. 2003).
The panelists in the current study rated actual wine samples. Panelists in the previous studies rated model wine solutions, as the panelists agreed that model wine solutions were more likely to be distinguished than those prepared in a red wine base (Vidal et al 2003).
Anthocyanin concentration did not have a strong relationship with any of the other factors examined. In a previous study in model wine, anthocyanin fractions did not influence either perceived astringency or bitterness (Vidal et al. 2004).
Results from the trained panel analysis of variance showed that wine had a significant impact on astringency and bitterness ratings (p ≤ 0.05) and that panelists were able to differentiate wines based on perceived astringency and bitterness (Table 5⇓). A significant panelist effect was observed for both perceived astringency and bitterness (p ≤ 0.05), suggesting that the panelists may have used different parts of the scale. The panelist-by-wine interaction effect was not significant for bitterness but was significant for astringency ratings, indicating that the panelists used the scale more similarly when evaluating bitterness compared to astringency, which may have been due to the simpler definition associated with the term bitterness. Astringency may be divided into numerous subqualities such as drying, puckering, chalky or fine, medium or coarse. Even with training, the panelists may have defined astringency in a slightly different way, resulting in the observed panelist effect.
Principal component analysis (PCA) was conducted to examine interrelationships between the different variables and allow the separation of the three wines (Figure 1⇓). PCA of the sensory and analytical data explained 92.8% of the variation in the data, with 72.1% and 20.7% explained by the first (PC1) and second principal components (PC2), respectively. PC1 was primarily defined by LPP, SPP, tannin, astringency, and bitterness, while PC2 was a function of anthocyanin concentration. PC1 contrasted the medium tannin wines (wines B), which were high in SPP, LPP, tannin, bitterness, and astringency against the low tannin wines (wines A), which were low in those parameters but high in anthocyanin. The high tannin wines (wines C) were high in all of these attributes.
Factor scores indicate that wine A (Merlot) had low SPP, LPP, and tannin concentration, together with low ratings for astringency and bitterness (Figure 1⇑). Wine B (Cabernet Sauvignon) had medium tannin, SPP, and LPP levels and high perceived astringency but was lower in anthocyanin concentration. Wine C (Merlot) was high in tannin, LPP, SPP, anthocyanin, SPP, and LPP with high perceived astringency.
Conclusions
The hypothesis that tannin concentration had a strong positive relationship with astringency perception in Washington State red wines was supported through the determination of correlations between tannin concentration and astringency perception. This research showed that tannin concentration may not be an independent factor influencing perceived astringency and the interaction between tannin and polymeric pigment concentration impacts perceived astringency.
Footnotes
Acknowledgments: The authors gratefully acknowledge the financial support provided by the Washington State Wine Commission.
- Received April 2007.
- Revision received September 2007.
- Revision received November 2007.
- Copyright © 2008 by the American Society for Enology and Viticulture