Characterising mouth-feel properties of red wines

Characterising mouth-feel properties of red wines Two studies are outlined in this paper illustrating the importance of applying sensory analytical methods to quantify specific perceptions when a red wine is tasted. A formal sensory descriptive analysis approach was carried out to obtain detailed mouth-feel sensory data of 11 red wines made in an identical fashion from grapes sourced from four viticultural regions, assessed after one and three years of cellar storage. Ratings of attributes such as,,, and discriminated among the wines, and those wines that could be considered of higher quality had more complex mouth-feel attributes. The effect of cellar storage was illustrated for two of the wines from this study, with substantially lower ratings for all attributes, but with differentiation on the basis of the attribute. A separate study investigating the effect of different added grape and apple tannin fractions on mouth-feel of a model wine showed that ratings for the attributes, coarse grain and medium grain differed greatly among the fractions. An increasing degree of polymerisation (DP) related to increased scores for each of these attributes, while seed tannin of DP was rated more coarse and than a similar sized skin tannin fraction. These examples illustrate the benefits of using a trained panel to rate specific, defined mouth-feel terms to advance our knowledge of influences on wine sensory qualities. I.L. Francis 1,2, R. Gawel 1,3, P.G. Iland 2,4, S. Vidal 1,2, V. Cheynier 5, S. Guyot 6, M.J. Kwiatkowski 1,2, and E.J. Waters 1,2. 1 The Australian Wine Research Institute; 2 Cooperative Research Centre for Viticulture; 3 Roseworthy Wine Tasting Programs Pty Ltd; 4 Department of Horticulture, Viticulture and Oenology, Adelaide University; 5 Unité Mixte de Recherche Sciences pour Oenologie pellier cédex, France; 6 Unité de Recherche Cidricole-Biotransformation des Fruits et Légumes, France. Corresponding author s email: Leigh.Francis@awri.com.au T HE APPLICATION of formal sensory analysis plays an important role in modern wine research studies. To enable adequate understanding of the relative importance of particular compounds to wine aroma or flavour characteristics, or to establish the link between viticulture or winemaking practices and wine quality attributes, it is essential to have quantitative data of specific sensory properties of the samples studied. It can be argued that the sensations that are perceived when a wine is tasted, and not a wine s aroma (i.e. sensations obtained by nosing a glass) or appearance, are of uppermost importance to the consumer and consequently to the winemaker who produces the wine for sale. The ability to obtain appropriate, reliable information regarding these sensations is thus of considerable interest for those involved in wine production. The perceptions obtained when sipping a red wine include the true tastes, of which acidity and bitterness are probably the most important for red wines, but saltiness, sweetness and umami (savoury taste, see Day 2 for an overview) may also be perceived. A red wine can also give a degree of hotness or irritation, and there can be the sensations of fullness, weight and vis- cosity. In addition, of probably greater significance to general wine sensory impression is the retronasal perception, by the olfactory bulb, of volatile flavour compounds released by the wine held in the mouth and after the wine is swallowed or expectorated. The amount of volatile flavour overall, the degree of flavour complexity, and the types of flavours perceived will all be of considerable importance. When thinking about perceptions from a red wine when tasted, astringency is an outstanding aspect that certainly is one of the factors which differentiates the taste of white wines from red, and which can vary substantially among different red wines. Astringency is a touch sensation and is manifested by sensations of dryness, roughness or puckering (Lawless et al. 1994), and there is strong evidence that it is caused by the interaction of salivary proteins with so-called tannins, resulting in precipitation or complexation of these proteins, leading to a loss of the lubricating properties of the saliva (see the review by Gawel 1998). A number of astringent sensations can be apparent in red wines, and in common with other sensory perceptions, will be described in many different ways even by experienced tasters. In an effort to improve the clear communication of mouth-feel perceptions, a terminology list in the form of a wheel has been developed (Gawel et al. ). The mouthfeel wheel terminology was partly intended to be a useful starting point for sensory panels to allow the rating of the intensity of defined attributes of a set of samples, to provide a profile of the mouth-feel characteristics. Each of the terms on the wheel is defined so that each taster is aware of what is meant by the term, either by written definitions or 18 MAY JUNE 2 > VOL 17 NO 3> WINE INDUSTRY JOURNAL

by use of finger touch standards. Using the terminology in practice over a number of tasting studies, it has been found that terms from several sections of the wheel were particularly useful to the judges on sensory panels for discriminating wines. The degree of smoothness/roughness (for example, silk,,, emery, furry ), perception of a particulate sensation in terms of fineness versus coarseness (e.g. talc, chalky,, grainy ), the presence of a puckering sensation (e.g. adhesive, chewy,, pucker ), together with the term, have all been found to be of value in assisting profiling wine s sensory characteristics. The application of the terminology for research studies involves a number of steps that are commonly used for sensory descriptive analysis protocols (see Noble 1988). Following panel recruitment, using judges who are selected on the basis of prior wine tasting experience, interest and availability, several tasting sessions are held with a small number of wines, selected from the larger set to be examined, tasted and then discussed in a round table setting. This allows a selection of appropriate attributes/terms that are required to distinguish among the samples to be tested. A training process is carried out to further familiarise the panellists with the terms and wines, and to practice scoring the wines for the intensity of the attributes, typically using a line scale. Following this process, a final consensus attribute choice by the panel will have been made, and further practice rating sessions carried out under controlled conditions in tasting booths. Finally, formal rating sessions are carried out to generate the quantitative data, with the panellists rating the attributes with the wine held in the mouth and also after spitting. Normally only a maximum of three to four wines per session are assessed, with a random presentation order for each judge, replicate assessments and with colour masking lights and black glasses being employed to avoid any biases due to colour of the samples. Barossa 1 1 2 3 Riverland 4 PC2 (6%) Eden Valley 1 Barossa 3 Coonawarra 2 PC1 (89%) Figure 1. Biplot of principal components 1 and 2 for mean sensory scores of in mouth mouth-feel descriptive analysis data for 11 Shiraz wines, made under identical small lot winemaking conditions, and assessed after one year of cellar storage post bottling. Wines are labelled by region of grape origin. Note that the samples should not be considered as representative of the particular regions. ASSESSING THE INFLUENCE OF THE VINEYARD ON WINE MOUTH-FEEL One of the first questions to be answered by using formal sensory descriptive analysis methods applied to detailed mouth-feel attributes was how widely a set of wines might vary in mouth-feel properties as a result of viticultural differences only, where winemaking was kept constant. A set of Shiraz wines was made under identical conditions from grapes sourced from 11 different vineyards in the 1997 vintage, from four South Australian viticultural regions (Riverland, Barossa Valley, Eden Valley and Coonawarra). The vineyards were selected on the basis of giving fruit with differing composition, and the grapes were grown under a range of management practices (pruning, crop thinning, irrigation and trellising). Note that the vineyards are not likely to be in any way representative of these regions. The total soluble solids (TSS) levels of the grapes at harvest were between 22.5 and 24.9ºBrix and the wines were made using standard small lot replicated conditions described by Ewart and Sitters (1991) modified to accommodate the specific requirements of the study. These modifications were made following input from The Australian Wine Research Institute/Adelaide University Tannin Project researchers and from winemakers from the project s Industry Reference Group. Notably, no additions of commercial tannin preparations or any oak treatments were applied. The wines were assessed as one year old wines by a sensory panel consisting of eight judges, most of whom had been involved in the development of the mouth-feel terminology wheel, and some of the data from this sensory analysis has been reported elsewhere (Gawel et al. 1). The 11 wines were rated for a number of mouth-feel attributes, both with the wine held in mouth and after expectoration. The wines were assessed in duplicate under colour masking lights in isolated booths, with two or three samples presented per session. In addition, a subset of the 11 wines was assessed after three years of cellar storage using a closely similar tasting protocol. Figure 1 shows the overall results for the mean data of the 11 wines assessed at one year presented in the form of a principal component analysis (PCA) map. The in mouth ratings are shown, and this representation gives a good overview of the major differences among the 11 wines. The wines were separated largely along the first principal component, which was responsible for close to 9% of the variation among the samples. Thus, those samples plotted on the left of the diagram were rated highest in and lowest in the other attributes, in particular,, and. Those wines plotted towards the top of the diagram were rated relatively highly in while those to the bottom half of the figure were higher in. All of the Riverland wines were clustered on the left of the diagram, being among the highest in, together with two wines from Barossa vineyards. In contrast, a separate Barossa Valley wine ( Barossa 3 ) was rated substantially higher in,,, fine emery and. This wine was made from fruit sourced from vines which received less irrigation and which were lower yielding compared to the two other Barossa wines. The Eden Valley and Coonawarra wines were rated highest as a group in,,, and WINE INDUSTRY JOURNAL > VOL 17 NO 3 > MAY JUNE 2 21

(a) In mouth scores Riverland 2 6 5 4 3 (LSD=13) (b) After spitting scores adhesive (LSD=12) ns 7 6 5 4 3 (LSD=25) Riverland 2 (LSD=8) (LSD=8) (LSD=22) (LSD=19) (LSD=6), with one of the Coonawarra wines being rated particularly highly in the mouth-feel attribute. This data shows that with identical winemaking and with fruit of closely similar TSS, wines can be characterised and differentiated on the basis of quite subtle mouth-feel attributes. The fact that many of the attributes were correlated, as indicated by the narrow angles among the attribute vectors in Figure 1, is not surprising, as this indicates that the attributes reflect the overall astringency ( tannin ) perceived in the wines, with the wines with higher astringency also having more complex mouth-feel properties. These wines were also scored, using a separate panel of experienced tasters, for overall quality using the standard Australian point score system, and it was found that the scores for the samples on the left of Figure 1 ranged from.3 to 15.3, while the wines on the right of the diagram were scored from 15.7 to 16.3 (n=9 judges 2 replicates). Thus, the quality of the wines appears to be in part related to their mouth-feel properties, although other factors were likely to have influenced the judges overall impression of a wine s quality. To better assess more detailed differences among the wines, it is useful to look at actual mean values of a smaller number of wines in the form of a radar or cobweb plot. Figure 2 shows mean data for the in mouth and after spitting ratings of four of the wines from the study. As was seen from the PCA plot in Figure 1, the Riverland 2 wine was rated highest in and low in all the other attributes. The sample was rated quite similarly to the Riverland wine, but differed in having higher scores for. The other two wines were perceived by the panel as substantially and statistically significantly higher than the Riverland and Barossa samples in most attributes. These two wines were discriminated by the panel on several attributes:, and (in mouth) and, and adhesive (after spitting). Note that the after spitting scores for were larger than the in mouth scores for each of the four wines, probably partly reflecting the common experience of a building of astringency over time, or perhaps that differing physiological (LSD=) (LSD=27) Figure 2a and 2b. Sensory profile plot of mean intensity ratings and least significant differences (LSD, P<.5), for attributes rated a) with the wine held in mouth and b) after expectoration, for four of the 11 Shiraz wines studied, assessed after one year of cellar storage post bottling. Note that the samples should not be considered as representative of the particular regions. pucker emery talc 6 5 4 3 chalky Figure 3. Sensory profile plot of mean intensity ratings, for attributes rated with the wine held in mouth (attributes underlined) and after expectoration, for two of the 11 Shiraz wines studied, assessed after three years of cellar storage post bottling. Note that the samples should not be considered as representative of the particular regions. mechanisms may have been at play (Gawel 1998). These four wines were assessed by a separate sensory panel, made up predominantly of judges who had performed the earlier assessments, following three years of cellar storage. Figure 3 provides data for the and wines assessed at the later time. The scale on the Figure 3 is comparable to the scale used on Figure 2, and it is striking that these wines were rated substantially lower in all attributes than when the wines were assessed at one year of age. Although the influence of differing panel performance cannot be ruled out, these differences are likely to be due to changes in the composition of the wine, as other young red wine samples rated by this panel at close to the time of this later tasting had comparable scores to the data shown in Figure 2. The point line scale that was used by the panel to rate the wines for mouth-feel, a labelled magnitude scale (Green 1993, Lawless and Heymann 1998), has written anchors marked along the line, with a score of 6 units indicating weak, 17 units indicating moderate, 34 strong, 53 very strong and being strongest imaginable. 22 MAY JUNE 2 > VOL 17 NO 3> WINE INDUSTRY JOURNAL

Thus the Coonawarra and Barossa wines had average scores between strong to very strong as young wines, while after maturation they were perceived by the panel as at the low side of the moderate range for this attribute. Notwithstanding the lower scores, the panel could differentiate the two wines, with the Coonawarra sample having a higher mean score for chalky (in mouth) and (after spitting). Other attributes were not rated differently. RELATING GRAPE COMPOSITION AND WINE MOUTH-FEEL Having developed and used formal sensory procedures for quantifying mouth-feel characteristics in red wines, it was considered of great interest to attempt to relate differences in tannin chemical structure to wine mouth-feel properties. There is surprisingly little knowledge available regarding the effect of tannin (more correctly proanthocyanidin ) chemical make-up on sensory properties, although a good picture of the types of tannins that can be extracted from the grape berry during winemaking has been developed over recent years (see discussions in Souquet et al., Cheynier, and Brossaud et al. 1). Grape seeds contain tannins made up of epicatechin, catechin, and epicatechin gallate units. There can be varying sized tannins, from two, three and more units, and from the combinations of the basic units a wide range of different tannins is possible. Skin tannins are similarly composed of a number of subunits, but they will only have a small amount of epicatechin gallate, and instead consist predominantly of epicatechin, catechin, and epigallocatechin, as well as a number of other, less prevalent units. The relative lack of epicatechin gallate groups within the structure in skin tannins (together with the presence of epigallocatechin) has been speculated to confer a softening mouth-feel effect to wines, but little sensory data has yet been obtained to allow this hypothesis to be tested. The size of skin tannins has also been found to be greater than seed tannins (Souquet et al 1996, Prieur et al 1994), with seed tannins having an average units size i.e. so-called degree of polymerisation, DP, with a maximum DP up to about 3 units. Skin tannins have a mean DP of 3 units, and some tannins can be up to 8 units (Souquet et al. 1996). An investigation was carried out to attempt to explore the question as to what mouth-feel characteristics different tannins might confer on wine, and full details will be published separately (Vidal et al. 2). Initial work in this study involved isolation of tannins from Shiraz grapes. Because the investigation involved tasting by a trained sensory panel, a fairly large quantity of tannins was required, and a 5 kg sample of grapes was deseeded manually to allow extraction of tannins from the seeds separately from the skins, using a suite of chemical solvents. Following the isolation procedure, tannins of differing size were prepared from the extracts, with the aim of having small tannins, of three units average size i.e. mdp3 (mean degree of polymerisation 3), as well as tannins of intermediate and large size. Tannins were also isolated from apples, to provide a further comparison, as it is known that apple tannins are chemically simpler than grape tannins, being composed exclusively of epicatechin units. Two fractions were obtained from Shiraz grape seeds: DP3 and DP, while from grape skin DP3, DP12 and DP fractions were produced. From the apple source, DP3, DP and DP7 samples were investigated. In addition, a further sample of seed tannins was obtained, which was the DP fraction treated enzymatically to remove gallate esters from the tannin structure ( degalloylated ), thus providing material comparable to the apple and skin tannins of similar size, neither of which contain these gallate derivatives. The fractions were carefully characterised chemically and then presented for sensory analysis. It was decided to present the fractions to the panel at.5 g/l in a base model wine medium (13% v/v aqueous ethanol/tartrate ph 3.7 buffer). An experienced panel of 16 judges was convened, of whom all except one had participated in previous mouthfeel sensory studies. Following attribute selection and training/practice rating sessions, the panel rated the mouth-feel attributes of the nine fractions, as well as the model wine with no addition, in triplicate, with only two or three samples presented per session. The attributes rated were fine grain, medium grain, coarse grain,, chalky, pucker, adhesive, overall astringency and bitterness. There were statistically significant differences among the samples for each of the attributes, with the exception of bitterness. This lack of difference in bitterness among the fractions was a somewhat surprising result, as previous studies (for example, Lea 199) had indicated that smaller tannins should be more bitter than larger tannins. Why the small tannins were not bitter in our study is not certain, but this result may be due to the degree of care taken to ensure there were no monomeric compounds in the fractions obtained, as it is well known that catechin and epicatechin are predominantly bitter (Kielhorn and Thorngate 1999). The panel in our study had demonstrated the ability to perceive and rate bitterness in training sessions and in other sample sets. Figure 4 shows the mean data for the nine fractions plus the model wine, for three of the attributes scored by the panel. Many of the other attributes were quite closely correlated with these three, and these attributes are presented as they provide a good indication of the major differences among the samples. As might have been expected, the model wine was rated lowest in these three terms. The three DP3 samples were grouped, indicating that they were scored similarly by the panel, with the apple DP3 fraction being rated somewhat lower in than the grape fractions. The apple DP fraction was rated as more coarse grained and than the DP3 counterpart, with the apple DP7 fraction being rated higher again in these attributes, showing the major impact of the size of the tannin on mouth-feel. The grape skin tannin fractions showed a similar pattern, with the skin DP12 sample being higher in medium grained rating than 24 MAY JUNE 2 > VOL 17 NO 3> WINE INDUSTRY JOURNAL

22 18 16 12 8 6 4 2 MW Skin DP3 Apple DP3 2 4 6 8 12 medium grain the apple fraction of comparable size, but more coarse and than the DP3 skin sample, and less coarse grain and than the DP sample. The seed DP sample is of particular interest. Although it is of only moderate size, it was rated similarly to the skin DP, being substantially higher in coarse grain and than the skin DP12 fraction. The enzyme treated seed DP fraction, labelled seed DP degall on Figure 4, was, however, scored very similarly to the Skin DP12 fraction, providing good evidence that the gallic acid groups on the skeleton of seed tannin affect the sensory properties, conferring a distinct coarseness and higher level of than the skin tannins. In considering these results, it must be borne in mind that the fractions were extracted from seeds and skins under fairly severe conditions, and the tannins isolated would probably not be identical to those extracted from grapes under normal winemaking conditions. During winemaking, a number of chemical and microbiological processes will take place, to alter the phenolic compounds extracted from grape material into more complex species. The next stage in this research is to investigate the mouth-feel properties of grape tannins in the presence of other wine components, and also attempt to isolate more complex tannin fractions from a red wine. Carrying out detailed sensory studies will give new insights into tannin structural features of importance to wine quality, but the chemical aspects remain a challenge, in particular isolating sufficient quantities of these compounds to enable sensory work to be carried out. CONCLUSION Formal sensory descriptive analysis methods are powerful tools to enable detailed characterisation of a set of samples. Using a trained panel to quantify mouth-feel attributes of red wines, important information on aspects of sensory Skin DP Apple DP Skin DP12 Seed DP3 2 4 6 Apple DP7 Seed DP Seed DP degall 8 12 coarse grain Figure 4. Mean values for three mouth-feel attributes rated for nine tannin fractions added to a model wine, together with the values for the model wine (MW), all rated by panellists while held in the mouth. Tannins were extracted from Shiraz grape seeds ( Seed ), Shiraz grape skins ( Skin ) and apple fruit ( Apple ), and were of varying degree of polymerisation (DP). The Seed DP degall fraction was the Seed DP sample enzyme treated to remove gallic acid groups from the tannin. See text for more details. properties of red wines when tasted can be obtained. The effect of vineyard on wine mouth-feel was shown to be substantial, with wines of similar level of overall astringency able to be differentiated in specific characteristics, and the effect of cellar storage on wine astringency could be described and quantified. The application of descriptive analysis to characterise grape berry polyphenolic fractions obtained through laboratory techniques was also informative. It was clear that larger tannins are more astringent, and coarse than smaller ones, and that the larger seed tannin fraction studied was more coarse grained than the comparable skin tannin fraction, which was likely due to esterified gallic acid derivatives. None of the tannins contributed bitterness. The improved understanding of how grape derived tannins may impact on wine mouth-feel obtained from this study, will provide a foundation for further work to establish influence of management practices on red wine quality. ACKNOWLEDGMENTS We wish to thank the sensory panels for their substantial investment of time and effort. The project was supported by Australia s grapegrowers and winemakers through their investment body the Grape and Wine Research and Development Corporation, with matching funds from the Federal Government. Some of the work described here was supported by the Commonwealth Cooperative Research Centres Program. We wish to thank C.A. Henschke & Co, Southcorp Wines, Majella Wines, the INRA experimental station of Pech-Rouge (France), and Nuriootpa Viticulture Station (SARDI) for providing grapes, as well as Renata Ristic and Dr Robert Asenstorfer for assistance with the winemaking. REFERENCES Brossaud, F., Cheynier, V., Noble, A.C. (1) Bitterness and astringency of grape and wine polyphenols. Aust. J. Grape Wine Res. 7: 33-39. Cheynier, V. () Tannins in grape and grape products. Brooker, J. D. (ed.). Tannins in livestock and human nutrition: proceedings of an international workshop. ACIAR Proceedings No. 92. Australian Centre for International Agricultural Research: Canberra, ACT: 9-94. Ewart, A.J.W., Sitters, J.H. (1991) Small scale winemaking as a research tool: influence of fermenter size and juice clarification on resultant wine quality. Aust. N.Z. Wine Ind. J. 6: 128-132. Gawel, R. (1998) Red wine astringency: a review. Aust. J. Grape Wine Res. 4: 74-95. Gawel, R., Iland, P.G., Francis, I.L. (1) Characterising the astringency of red wine: a case study. Food Quality and Preference, 12: 83-94. Gawel, R., Oberholster, A., Francis, I.L. () A mouth-feel wheel : terminology for communicating the mouth-feel characteristics of red wine. Aust. J. Grape Wine Res., 6: 3-7. Green, B.G. (1993) Oral astringency: a tactile component of flavor. Acta Psychol. 84: 119-125. Kielhorn, S., Thorngate, J.H. (1999) Oral sensations associated with the flavan-3-ols (+)-catechin and (-)-epicatechin. Food Qual. Pref. : 9-116. Lawless, H.T., Corrigan, C.J., Lee, C.B. (1994) Interaction of astringent substances. Chem. Senses 19: 1-154. Lawless, H.T.; Heymann, H. (1998) Sensory Evaluation of Food: Principles and Practices. New York: Chapman and Hall. Lea, A.G.H. (199) Bitterness and astringency: the procyanidins of fermented apple ciders. Rouseff, R.L. (ed.) Dev. Food Sci. 25. Elsevier: Amsterdam; 123-3. Noble, A.C. (1988) Analysis of Wine Sensory Properties. Linskens, H.F.; Jackson, J.F. (eds.) Wine Analysis. Mod. Methods Plant Anal. New Ser. 6: Springer-Verlag: Berlin Heidelberg; 9-28. Prieur, C., Rigaud, J., Cheynier, V., Moutounet, M. (1994) Oligomeric and polymeric procyanidins from grape seeds (Vitis vinifera). Phytochem. 36: 781-784. Souquet, J.M., Cheynier, V., Brossaud, F., Moutounet, M. (1996) Polymeric proanthocyanidins from grape skins. Phytochem. 43: 59-512. Souquet, J.M., Cheynier, V., Moutounet, M. () The proanthocyanidins of grape. Bull. OIV 73: 61-69. Vidal, S., Francis, I.L., Guyot, S., Kwiatkowski, M.J., Gawel, R., Cheynier, V., Waters, E. (2) The mouth-feel properties of grape and apple proanthocyanidins in a wine- like medium. J. Sci. Food Agric. (submitted). WINE INDUSTRY JOURNAL > VOL 17 NO 2 > MAY JUNE 2 25