Sensory Evaluation at the Winery

Heather McMahon

Winemakers spend extensive amounts of time using their expertise to evaluate their products. Wine is the most thoroughly scrutinized food; numerous resources are allocated for chemical and microbiological analysis, but the most important tools remain under utilized. The human senses are the most thorough analytical devices available. In many wineries sensory evaluation is performed improperly and by a single person, usually the winemaker. Idiosyncrasies in individual preferences can lead to disastrous economic effects for the winery as a result of biases. In addition, an unavoidable leave of absence, an allergy, or the common cold can temporarily cripple even the most reliable palate. Establishing an in-house sensory panel can reduce these effects; drawing from the advice of trained sensory judges reduces the illogical dependency on a single or few opinions and provides an invaluable resource.

One of the most important aspects of sensory evaluation is creating a standard, defined procedure. A winery should attempt to design a plan which minimizes variables associated with product evaluation, and execute that plan consistently. This formalized program should meet specific in-house objectives, such as assessment of the bottom line, maintenance or improvement of product quality, style definition, identification of problems, definition of differences among vineyard sources, evaluation of experimental wines, etc.

One sensory technique for the detailed analysis of a product is known as Quantitative Descriptive Analysis (QDA) (Meilgaard et al., 1991). This procedure involves the selection and subsequent training of individuals for the comprehensive evaluation and description of specific product attributes. Use of this method to carefully analyze and quantify a wine=s characteristics (both positive and negative), can enable a winemaker to make more objective decisions concerning viticultural and processing techniques.

There are several considerations in choosing panel members: 1) Interest, 2) Availability, 3) Promptness, 4) Health (especially lack of allergies), 5) Articulateness, and 6) Attitude about the product (ASTM, 1981). However, the single most important characteristic for a panel member is motivation. The most discriminating palate is not useful if the individual cannot commit time and concentration, but a motivated judge can be given extra training sessions to make up for a lack of natural discrimination ability. The pool of individuals who can participate in the panel should be as large as possible; judges may become less discriminating, loose impartiality, or have conflicting commitments. Every winery has a number of devoted customers who are knowledgeable and would be eager to participate in a wine sensory evaluation program. Optimally, 10-14 judges should be chosen for a single sensory session. (If this number is unreasonable, as few as six panelists may be used, but dissenting opinions may be more difficult to resolve and will have greater statistical consequences.) Some training is important to the success of a QDA program. Alterations to a base wine are an effective way of illustrating the presence and intensity of an attribute. Sensory standards are used to establish a common vocabulary and can be easily prepared from grocery store items as illustrated by Zoecklein et al. (1996). The winemaker should not impose definitions, but the panel should reach a consensus and be uniform in their comprehension and use of terminology. These reference standards may be recreated at any time to Aretune@ the panel, or aid a particular judge who may have difficulty with a specific characteristic. Simple statistical analysis of the results can be used to evaluate panel and judge performance, both for individual characteristics and session to session variability.

Due to alterations in aroma and flavor detection, standardization of wine and environmental temperature is vital. Room temperature is usually appropriate, but if over time the testing area is prone to large fluctuations, temperature modification of the product may be necessary. No more than six wines should be evaluated in a session in order to avoid fatigue. A standard glass type should be used, always filled to the same level (usually 30-40 mL of wine is sufficient for proper evaluation). Wine glasses should be covered with a watch glass to avoid blending aromas or causing olfactory adaptation. Optimally, samples should be labeled with a three digit random code and be presented in a balanced, random order to prevent bias. If samples are labeled by letter AA@ or the number A1@ for example, a panelist may believe this sample is the control or of the highest quality, and rate accordingly. A bias known as order error occurs if the entire panel evaluates the samples in the same order. The first product tasted always is rated more favorably than subsequent wines. If each panelist has a unique presentation order, this error is eliminated. Many wineries will evaluate a control followed by a test wine, with the judge or judges cognizant of this order. Prior knowledge of the sample identity leads to an inherent bias. For example, if a Brettanomyces contamination is suspected, a possibly imagined >mousy= taint in the test wine may be reported.

QDA uses an unstructured line scale (usually 10 - 15 cm) for each characteristic, anchored with antonym descriptors (commonly Aweak@ and Astrong@). Panelists make a perpendicular hatch mark on each line which is later converted to centimeters. Simple statistical analysis may be performed (one- or two-way analysis of variance (ANOVA)). Significant differences among means or standard deviations (and therefore the wines) are usually expressed in relation to a p-value or level of significance. For example, significance at p _ 0.05 indicates the probability of the conclusions being incorrect is 5%. In other words, the researcher is 95% confident about the reported results. Graphical presentation of the data is usually accomplished with a spider plot (examples follow). All characteristics are represented by different branches from a central point; the mean or standard deviation for that characteristic for each wine is placed on that line. Analysis of variance can also provide information about the accuracy and consistency of the judges, highlighting areas which need more extensive training.

Three sensory sessions were performed at the regional meeting of the American Society of Enology and Viticulture B Eastern Section, a workshop on Wine and Juice Production and Practical Monitoring given June 7-9, 1997 in Dulles, VA. Four different Rieslings were blindly and randomly presented to approximately 90 individuals associated with the wine industry. The wines were from three distinct macroclimate regions; wines A and C are from climates which are generally cooler than those for wines B and D. There was no panel training, but each judge was assumed to be familiar with evaluating wine. QDA standards were available as references. The scorecards were successively more detailed in each session.

Wine A: Chateau Ste. Michelle =96

Wine B: Rockbridge =96 White Riesling

Wine C: Schmitt-Sohne =95 Riesling

Wine D: Rapidan River =96 Riesling

Session II consisted of evaluating the same wines on more specific characteristics: sourness, fruitiness, and sweetness. A QDA system utilized a 10 cm line scale for each characteristic, anchors being Aweak@ and Astrong@. Perpendicular hatch marks made by the panelists on each line were later converted to centimeters by measurement from the left anchor.

Specific components of balance were evaluated in session II. Balance is the result of the relative intensities in the perceptions: sweetness _ acidity + phenols. In white wines, the phenolic content is usually low, and the relationship is frequently simplified to the contributions of Asweetness@ and Asourness@. Wines A and B were shown to be more sweet than sour, while the relationship was reversed for wines C and D. The strength of acidity in wine D amplified the perception of reduced sweetness in this wine. Acidity may be brought into balance with age, and while the older wine (C) in the trial was still more acidic than sweet, it is impossible to compare to previous levels without sensory records. Many vintners set a five year goal for achieving the supreme harmony between acid and sweetness, so criticisms concerning balance in any of these young wines could be premature. Rieslings gained popularity and distinction for their capability to withstand sudden cold winter freezes for prolonged periods of time. It is believed that heat during fruit maturation can reduce the acid and fruit characteristics, and thus Riesling is not suited to warm climates. However, the sample with the highest fruitiness (wine B) is from our own relatively warm climate. Another observed variable is that the differences in soil composition and plant environment (terrior) influence sensory attributes.

QDA analyses can be used to help define and describe sensory differences resulting from fruit maturity, terrior, and processing. Most of the differences in the categories are statistically significant, particularly for fruitiness and balance. Young Rieslings often have an apple >tone= in the finish, as both of the Virginia wines (B & D) possessed. A full, honeyed fragrance is a prevalent characteristic of mature Rieslings, and is illustrated by the California wine (A) and the Virginia wine B. California Rieslings are gaining recognition for tropical flower aromas. The only floral aroma evaluated was rose, which the California wine did exhibit, but the wine with the highest rose character was Virginia wine D. The slightly negative attributes of grass and cold tea were not perceived appreciably in any of the samples. It is interesting to note that although wine B had the highest fruitiness (as it did in Session II), the ratings for the individual attributes (citrus, peach, apple, apricots) would lead to the assumption that wine A is the most fruity. This is not due to panel error; there were fruity characteristics which were not specifically identified on the scorecard due to time limitations. Simply comparing fruitiness from sessions II and III illustrates an increase in the range from 1.72 (session II) to 2.45 (session III). An increase in mean range facilitates wine differentiation, and could promote more focused production decisions.