Abstract
Pinot noir is the most widely planted red grape variety in New Zealand and is considered a premium product based on the price per volume. To date no studies have attempted to characterize the different styles of the four main Pinot noir producing regions: Central Otago, Marlborough, Martinborough, and Waipara. The intensities of aroma, flavor, and mouthfeel attributes of commercial regional wines from two vintages were investigated. Descriptive analysis was carried out by a panel consisting of experienced but untrained wine professionals. Canonical variate analysis showed that the four wine regions were differentiated according to aroma (barnyard, black cherry, herbal, raspberry, red cherry, oak, spice and violet), in-mouth flavor (fruit density/concentration and red fruit), and mouthfeel (balance, body, and finish length) attributes. Results show that Pinot noir wines from the four regions of New Zealand are stylistically different and that experienced but untrained wine professionals produce reliable results for this type of sensory analysis.
The concepts of terrior and place of origin incorporate the notion that wines from specific geographic locations can be perceived as different. The significance of place is not unique to wine, but it is linked to quality more strongly than perhaps for any other agricultural product. In New Zealand, many wineries use place of origin for wine marketing and tourism, but little research had been conducted to determine if wines are different. The main objective of this work was to determine whether there are different sensory profiles for New Zealand Pinot noir wines from different wine-producing regions as described by wine professionals.
For place of origin or geographical branding to be successful, there must be a recognizable style (Maitre et al. 2010). However, the elaboration of those sensory characteristics that are crucial to stylistic differentiation is not straightforward. Wines that come from the same place may share many but do not need to possess all of the same sensory attributes (Rosch and Mervis 1975). A number of studies have characterized stylistic differences among wines from different regions using sensory analysis, including Pinot noir and Chardonnay wines from California regions (Guinard and Cliff 1987, McCloskey et al. 1996), ice wines from two Canadian provinces (Nurgel et al. 2004), and Cabernet franc and Bordeaux-style red wines from subregions of Niagara (Kontkanen et al. 2005, Hakimi Rezaei and Reynolds 2010).
Pinot noir wine production in New Zealand is recent in comparison to that from the vineyards of Burgundy, but there appears to be an emerging consensus of stylistic differences among the main regions (Central Otago, Marlborough, Martinborough, and Waipara). These differences have been noted in informal texts (Cooper 2008), but no scientifically robust information has yet been collected. In contrast, there have been a number of studies of New Zealand Sauvignon blanc, including experiments to determine aroma profiles of regional Sauvignon blanc and investigating the concept of typicality (Lund et al. 2009, Parr et al. 2007, 2010). Pinot noir is now the most widely planted red grape variety in New Zealand and wines are frequently compared to those from Burgundy.
Consensus training methods have generally been preferred for separating products when numerous attributes are used and when investigating wine specificities in which the intensities of attributes are evaluated (Cartier et al. 2006, Maitre et al. 2010, Parr et al. 2010). However, in one of the first works of this type using descriptive analysis, the data from seven of the 17 trained panelists had to be removed due to their unreliability (Noble et al. 1984). Some researchers have questioned the psychophysical assumptions inherent in consensus training methods when applied to complex stimuli as found in wine and have sought to use alternative intensity-based techniques such as citation frequency with trained panelists (Campo et al. 2010). Although descriptive methods are normally preferred with trained panelists (Hootman 1992, Murray et al. 2001, Wolters and Allchurch 1994), some researchers have used more spontaneous methods whereby panelists, selected on the basis of appropriate expertise (Ballester et al. 2005), use their own vocabulary (Perrin et al. 2007).
This work aims to characterize wines from the four regions where Pinot noir is locally important: Central Otago, Marlborough, Martinborough, and Waipara. These regions span several degrees of latitude (40 to 45°S) and among them contain many different soils and distinct mesoclimates, factors known to be important for wine sensory attributes (Bohmrich 1996, Jones 2006, van Leeuwen et al. 2004, Vaudour 2002). In common with other recent studies of New Zealand Sauvignon blanc (Parr et al. 2007, 2010), evaluations were carried out by wine professionals who had not been specifically trained for this study. Overall, the results of this study benefit the New Zealand wine industry by providing information to promote regional products and for consumer education. The use of wine professionals as tasters is also of interest as they are influential within the industry and provide information to consumers and other wine industry personnel.
Materials and Methods
Wines.
Wines from each of the main Pinot noir producing regions (Central Otago, Marlborough, Martinborough, and Waipara) and two vintages (2007 and 2008) were selected (32 wines in total). These wines were considered to be representative of each region based on the professional opinion of winemakers from the specified regions and the region of origin classification on the label. All grapes used in the production of each wine were from the specified region. Most wines (30) were sealed with screwcaps and two wines sealed with corks. All wines were stored at room temperature (20°C).
Panelists.
The tasting panel was composed of experienced wine professionals (11 male, 10 female) from the Marlborough wine industry, ranging in age from 28 to 53 with an average age of 35. Of the tasters, 14 were winemakers, two were viticulturist, and two were engaged in wine and grape research. Panelists were not trained for this study. Wine-tasting experience ranged from 6 to 36 years with an average of 17 years. All tasters considered themselves experienced wine tasters, six considered themselves expert tasters, and five had experience as wine judges. Fourteen of the tasters had participated in a wine sensory panel prior to this experiment and seven of the tasters had prior sensory training, although only one person had training specifically for Pinot noir.
Preliminary tasting.
A subset of six panelists participated in a preliminary tasting session to choose sensory attributes to be included in the main experiment. Panelists listed wine aroma and palate descriptors for each of the wines. Subsequently, the main differences among wines were discussed. A final list was compiled by the experimenters from these attribute lists and previous sensory work on Pinot noir (Aubry et al. 1999, Guinard and Cliff 1987). Twenty-five attributes describing aroma (15), in-mouth flavor (4), and mouthfeel (6) were selected.
Experimental design.
The tastings took place during two separate periods. The 2007 wines were tasted in May/June 2009 and the 2008 wines were tasted in May/June 2010. The timings of the sensory evaluations were chosen to ensure that all wines were the same age when evaluated and all tastings were implemented during the same 5-week period in each year. All the 2007 wines were assessed in triplicate by all of the 21 panelists in three replicated tasting sessions using a randomized complete block design. Availability of panelists was restricted in the following year and so the 2008 wines were assessed using a balanced incomplete block (BIB) design (Ball 1997, Masuoka et al. 1995, O’Mahony and Goldstein 1986). Each panelist assessed 10 of the wines on each occasion and the design used ensured that all wines were assessed an equal number of times across the three replicated tasting sessions. Other than this difference in design between the two years, there were no other differences in the methodology. Wines were presented in a different random order at each tasting session to minimize any possible order effects and all attributes were assessed on each occasion.
Sensory evaluation.
Wine samples (50 mL) were poured into black, ISO standard tasting glasses (International Organization for Standardization 1977). Each glass was coded with a random three-digit number and covered with a watch glass. A number of bottles from the same wine batch were used during the sensory evaluation. No single bottle was used for more than two days. Opened bottles were sparged with argon at the end of the first day, resealed, and stored at 20°C. After the second day, opened bottles were discarded.
Panelists evaluated wines in a wine-tasting room illuminated with a mix of artificial and natural light and where the temperature was controlled to a constant 20°C. Tastings took place over five weeks. At each evaluation a single panelist occupied the tasting room. On average, each tasting was approximately one hour, although panelists were allowed as much time as was required to complete the task. The intensity of each attribute was marked on a 100 mm horizontal visual analogue scale (VAS) on a paper questionnaire. Each attribute had a minimum and maximum word anchor. The word anchors for all aroma and in-mouth flavor attributes were “none” and “extreme.” The word anchors for the mouthfeel attributes are listed in Table 1. Panelists first evaluated the aroma attributes followed by in-mouth flavor and mouthfeel attributes; they were instructed not to taste the wine until after evaluating the aroma attributes. Panelists expectorated and rinsed with water between samples.
Data analysis.
Intensity ratings were quantified using a number between zero and 100 that corresponded to the distance from the left end of the VAS to the scribed mark. Restricted maximum likelihood (REML) estimation was used to determine the variance of the different random effects for each attribute as a measurement of consistency within and between the panelists and to calculate the (mixed-model predicted) means of the attributes. Unlike analysis of variance (ANOVA), REML does not require balanced data and allows for separate specification of random (“nuisance”) and fixed (treatment) effects. Here the random effects were panelist (P), replicated tasting session (T), and wine within region and vintage (W). Fixed effects were region and vintage. A post-hoc test, whereby a factor representing the previously evaluated wine was incorporated in the REML model, was used to determine any carryover effects between samples. No effect of previous wine was detected (p > 0.05).
Canonical variate analysis (CVA) was carried out on all data with region as the classification variable. All data were standardized and CVA was carried out using the individual scores from each panelist/tasting session combination. In contrast to principal component analysis, which describes similarities between individual wines, the purpose of CVA is to produce a dimensional representation that highlights as accurately as possible the differences that exist between the subsets of data (Darlington et al. 1973, Heymann and Noble 1989), in this case regions. Results from CVA consisted of linear combinations of the attributes that accounted for the greatest proportion of the variance between the regions. The regions were positioned on the CVA using the centroids for the wines. Attributes that had the greatest contribution to each linear combination were used to explain the main differences between the regional wines. Between-region distances were calculated from the average positions of each region on the axes defined by the canonical variates. All statistical analyses were carried out using GenStat 12.2 (VSN International, Hemel Hampstead, UK).
Results
Panelist performance.
REML estimation of each sensory evaluation experiment (for 2007 and 2008 wines) showed that the effect of panelist was significant (p < 0.05) for all attributes: that is, averaged over wines and tasting sessions, attribute scores differed among panelists. In contrast, tasting session was not significant for most attributes: that is, averaged over wines and panelists, scores for each tasting session were not different. The wine effect was also significant (p < 0.05), displaying panelist ability to discriminate mong the wines. Evaluation of panelist performance, using panelist, tasting session, and wine interactions for each sensory attribute, can be found in Table 2.
For the 2007 wines, panelists had difficulty evaluating both chocolate and strawberry descriptors. Significant interactions were found for chocolate for P × T, P × W, and P × T × W and all interactions were significant for strawberries; thus, these two attributes were removed from the CVA. A significant P × T interaction was found for jam, oak, smoky, and dark fruit. None of the mouthfeel attributes had any significant interactions. For 2008 wines, no significant T × W or P × T × W interactions were found. There were significant P × T interactions for jam, oak, raspberries, fruit density/concentration, and graininess and significant P × W interactions for savory and graininess.
Regional separation.
The attributes chocolate and strawberries were removed from CVA due to the significant interactions found in REML analysis. Clear regional separation was found using CVA (Figure 1). All three variates were statistically significant at p < 0.05 and accounted for 44%, 37%, and 19% of the total variance, respectively. Correlation coefficients of attribute vectors with the canonical variates are given in Table 3. The barnyard and herbal aroma attributes and the balance mouthfeel attribute were most strongly correlated with CV1; the smoky aroma attribute and the graininess and oak tannins mouthfeel attributes were most strongly correlated with CV2; and oak, smoky, and violet aroma attributes, spice in-mouth flavor attribute, and balance, body, and graininess mouthfeel attributes were most strongly correlated with CV3.
The loadings of the sensory terms (Figure 1) imply that Marlborough Pinot noir was differentiated by greater raspberry and red cherry aromas, red fruit in-mouth flavor, and longer finish length and more harmonious balance. Martinborough Pinot noir was differentiated by greater black cherry, oak, and spice aromas and oak tannin mouthfeel. Waipara Pinot noir was differentiated by greater barnyard, herbal, and violet aromas and in-mouth fruit density/concentration. Central Otago Pinot noir was differentiated by fuller body.
The between-region distances indicated how similar or dissimilar each region was from the other regions. Central Otago and Martinborough were perceived as the most similar and Marlborough and Martinborough were perceived as the most dissimilar (Table 4). There was a vintage effect (data not shown) between the two years. However, the CVA results show the same regional separation for each year. For example, Marlborough wines had more intense red cherry, raspberry, and red fruit aromas and flavors when compared to the other regional wines for both 2007 and 2008 vintages.
Discussion
Panelist performance.
The REML analysis of the sensory evaluation experiments indicated that panelists were generally in agreement for each wine across each tasting session. In summary, 6 of the 25 attributes resulted in significant P × T interactions in 2007 and 5 out of 25 attributes in 2008, and only 4 out of 50 possible P × W interactions were significant. The cause of these significant interactions was unclear. A postanalysis investigation showed no significant carryover effect between samples. In addition, the error rate (p = 0.05) was set for each attribute and, consequently, a certain number of significant effects might be expected as the number of attributes investigated increases. Overall, it seems reasonable to consider these few significant effects as relatively inconsequential.
In reporting the results from consensus training methods, other authors have excluded panelists or attributes considered inconsistent on the basis of significant P × W interactions (Cliff and Dever 1996, Guinard and Cliff 1987). In a broader discussion of the limitations of intensity scoring when describing the aroma characteristics of complex mixtures (such as wine), Campo et al. (2010) considered that alternative approaches are required. They suggest that similarity-based approaches are better suited to deal with complex perceptions, although they provide less detail than conventional descriptive analysis. In the current study, no aroma and flavor standards were used in the sensory evaluation experiments; rather, assessments relied on skills associated with wine expertise, namely the ability to constantly recognize and evaluate specific notes in a wine (Parr 2007). However, there was a possibility that the meaning ascribed to particular terms might vary among panelists or between the groups of tasters (Gawel 1997, Stone et al. 1974).
Overall, similar attributes seemed to be grouped in terms of their correlations with the canonical variates. Red fruit aromas and in-mouth flavor (raspberry, red cherry, and red fruit) were all positively correlated with CV1; dark fruit and savory aroma and in-mouth flavors (black cherry, dark fruit, oak, savory, smoky, spice) were negatively correlated with CV2. This lends support to the work of Urdapilleta et al. (2011), who suggested that hierarchal organization of sensory properties may facilitate cognitive-based evaluation by reducing load while allowing for specific distinguishing information. In this study, although significant interactions have been noted, only those relating to the aroma attributes chocolate and strawberries resulted in a significant three-way P × T × W interaction and only in the first year involving the 2007 wines.
There appeared to be a high degree of agreement among these untrained wine professionals in assessing regional Pinot noir wines from New Zealand. Additionally, similar descriptors have been observed at informal wine tastings and by winemakers, although prior to this study no scientific evaluation of commercial wine had been undertaken. Marlborough wines have previously been described as red fruit, Martinborough wine as smoky, dark berry fruit, with savory notes, and Waipara wines as herbal (Cooper 2008).
Regional separation.
Canonical variate analysis of the data from the untrained wine professionals resulted in clear separation among regions for this set of 32 New Zealand Pinot noir wines (Figure 1). In other studies of Pinot noir, a total of 28 wines from two vintages were selected to represent the Carneros, Sonoma, and Napa regions of California (Guinard and Cliff 1987), and a total of 14 wines from two vintages and 10 wineries were selected to represent the production of British Columbia (Cliff and Dever 1996). In this evaluation, although all three of the canonical variates were statistically significant, regional separation could most clearly be seen in the two dimensions defined by CV1 and CV2. Within this space, three directions defined the differences between the regions. Thus, Marlborough wines were characterized with greater intensity of red cherry and raspberry aromas, greater red fruit in-mouth flavor, and greater balance and finish length mouthfeel; Martinborough wines were characterized with greater intensity of black cherry, oak, and spice aromas and greater oak tannins mouthfeel; Waipara wines were characterized by greater intensity of barnyard, herbal, and violet aromas and greater fruit density/concentration in-mouth flavor; and Central Otago was intermediate among the other three regions.
In this study panelists were not trained with specific aroma or flavor standards, but rather their descriptions relied on terms related to knowledge of shared prototypes (Urdapilleta et al. 2011). Despite this approach, and with the exception of chocolate and strawberry aromas, the characteristics associated with Pinot noir wines from each region appear to be reasonably well defined. However, these descriptions relate to relative differences rather than their presence in some wines and absence in others. In a similar study, Carneros wines were characterized by intense fresh berry, berry jam, cherry, and spicy aromas and differed from those of Napa and Sonoma (Guinard and Cliff 1987). Another study identified four major styles for British Columbia: complex with high fruit and berry flavors; intensely red, oaky, astringent, and spicy; herbaceous; and lacking fruit/berry character, astringency, and body (Cliff and Dever 1996).
The emphasis in both of the above studies was to identify regional or winery differences, although some factors that could influence specific sensory attributes were noted (fruit maturity and exposure, fermentation temperature, malolactic fermentation, use of new barrels) (Cliff and Dever 1996, Guinard and Cliff 1987). Similarly, in the current study, there were likely to be a range of factors that might influence wine style but that might not be considered “regional” in its strictest sense. Reynolds et al. (1996) demonstrated that the fruit environment plays a large role in determining important aromas and flavors in Pinot noir wines. Sauvageot and Vivier (1997) found that malolactic fermentation could somewhat alter the aromatic profile of Chardonnay and Pinot noir wines. Specific data were not collected on viticulture and winemaking techniques in the current study, but informal discussion with the wineries involved suggested that important factors (training system, acid adjustment, malolactic fermentation, proportion, and time in new oak barrels) were common. Anecdotally, “nonregional” factors that may be important are vine age and the relative proportions of available clones, but no published data for these parameters are available. Crop yields and harvest dates vary across the regions, but these should be considered as regional influences as they ultimately represent the influence of regional climate on management.
Conclusion
These results suggested that a panel consisting of wine professionals from New Zealand could evaluate a number of sensory attributes of New Zealand Pinot noir wines with a high degree of agreement. In two vintages, 19 of the 25 attributes had no significant interactions. All but two of the interactions were expected according to panelist performance found in other sensory work. Therefore, these results indicated that experienced but untrained wine tasters can be used to characterize stylistic differences, although any study that does not use consensus training should be careful to investigate those sensory terms with statistically significant interactions.
Pinot noir wines from the four main producing regions of New Zealand were differentiated using CVA, showing that the relative intensity of specific sensory attributes were characteristic of specific wine regions. The important differentiating attributes concur with those that wine professionals have used to describe Pinot noir wine from these regions in the nonscientific literature. In selecting wines for this study, the only consideration was that the grapes be sourced strictly from the specified region and that the wine was produced as a regional product. Thus, despite probable variations in grapegrowing and differences in production within a region, it appeared wines were broadly similar within a region and distinct from those of other regions. This information is extremely important for those attributes promoting “place of origin” uniqueness, as the information can be used for consumer education, marketing, and tourism.
Acknowledgments
Acknowledgments: This work was financially supported by the New Zealand Foundation for Research, Science, and Technology, Pernod Ricard (NZ) Ltd., and Lincoln University.
The authors thank Roger Harker (Plant and Food Research, NZ) for useful comments on the draft manuscript.
- Received November 2012.
- Revision received March 2013.
- Accepted May 2013.
- Published online December 1969
- ©2013 by the American Society for Enology and Viticulture