Abstract
Sixteen California Cabernet Sauvignon wines were chosen to represent a spectrum of vegetal aroma characteristics. Sensory analysis by descriptive analysis profiling and experts’ groupings of the wines by similarity of aroma were completed for the wines. All wines were found to be significantly different from one another for all aroma terms rated by the descriptive panel, including the vegetal terms eucalyptus, bell pepper, olive, and cooked vegetable. The expert winemakers/enologists sorted the same wines into groups by three different similarity criteria: first, by similarity according to their own individual criteria; second, by intensity of vegetal aroma; and third, by intensity of sulfur defects. Comparison of the results from descriptive analysis and the expert panel sorting data revealed similarities in the classification of vegetal aromas for these wines regardless of the criteria used for sorting. Methoxypyrazines, typically associated with vegetal aromas, were also measured for these wines and were not correlated with any of the descriptive terms, suggesting that other classes of aroma compounds contribute to the vegetal aromas in wines. The results indicate a more complex interrelationship between chemical composition and sensory perception of vegetal aromas in Cabernet Sauvignon wines than had previously been hypothesized.
- Cabernet Sauvignon
- vegetal aroma
- bell pepper aroma
- descriptive analysis
- similarity analysis
- methoxypyrazines
At low levels, vegetative aromas such as bell pepper or asparagus contribute to the distinctive varietal aromas of Cabernet Sauvignon, Merlot, and Sauvignon blanc wines. However, at high levels, these vegetal notes may be considered undesirable or suggest possible defects. The bell pepper aroma in Sauvignon blanc (Allen et al. 1991) and Cabernet Sauvignon wines (Noble et al. 1995, Chapman et al. 2004) has been correlated with the concentration of 3-isobutyl-2-methoxypyrazine (IBMP). However, the general term vegetal can also be applied to a wide range of aroma notes. In addition, many sulfur-containing compounds elicit related aromas such as asparagus, cooked corn, cassis, boxwood, and rubber (Goniak and Noble 1987, Darriet et al. 1993, Tominaga et al. 1996, Butzke 1997, Bouchilloux et al. 1998, Swiegers et al. 2005). Another class of aroma compounds, the norisoprenoids, can also contribute vegetal-related aromas to wines, including the aroma of green, cut grass associated with a norisoprenoid-related compound, (E)-1-(2,3,6-trimethylphenyl) buta-1,3-diene (TPB), that has been recently reported in wines (Janusz et al. 2003, Cox et al. 2005).
Several studies have been conducted on the viticultural practices linked to increased IBMP in grapes and wines (Bayonove et al. 1975, Augustyn et al. 1982, Heymann and Noble 1987, Allen et al. 1989, 1991, Arnold and Bledsoe 1990, Morrison and Noble 1990, Noble et al. 1995, de Boubée 2000). In addition, a few studies have focused on effects of viticultural practices on terpene and norisoprenoid concentrations (Marais et al. 1992a, 1992b, Razungles et al. 1998, Marais et al. 1999, Hunter et al. 2004, Gerdes et al. 2002, Lee et al. 2007). Interestingly, factors such as increased light exposure (Morrison and Noble 1990, Noble et al. 1995) and temperature (de Boubée et al. 2000) that have been shown to decrease IBMP concentrations and bell pepper aromas, may actually increase norisoprenoid concentrations (Marais et al. 1992a, 1992b, Gerdes et al. 2002, Lee et al. 2007), although effects on sensory properties in the latter studies are unknown.
The focus of this research was to define the meaning of vegetal aroma in quantifiable sensory terms and to identify the criteria that winemaker “experts” use for identifying Cabernet Sauvignon wines as vegetal. Descriptive analysis was used to provide an objective description of specific sensory attributes for the wines. Expert wine-makers/enologists participated in sorting trials to define the criteria used to group and classify vegetal wines. The descriptive analysis and sorting data were then related to the concentrations of two pyrazines, IBMP and 3-isopropyl-2-methoxypyrazine (IPMP), to better understand the chemical compounds that may be associated with these complex vegetative aromas.
Materials and Methods
Wines.
Sixteen California Cabernet Sauvignon wines with a range of vegetal characters were identified (Table 1⇓). The wines were selected to represent this range of vegetal characters from a larger set of 45 commercial and University of California, Davis (UCD)-produced wines by a group of 27 UCD faculty and students in informal bench-top evaluations and sorting trials using similarity of aroma as the sorting criteria. All commercial wines were donated by the manufacturers, and standard wine-making procedures for the UCD wines have been described previously (Turbow 2003).
Sensory analysis.
Descriptive analysis.
Fifteen volunteer judges evaluated 16 wines over the course of six weeks during summer 2002. Judges were initially trained over a period of eight group meetings, the purpose of which was to develop a list of aroma descriptors that would describe and differentiate the 16 wines. Two different wines were introduced at each meeting, so panelists were familiar with all wines before beginning the formal evaluation. During the first session, panelists were asked to describe the orthonasal aroma of the wines presented to them, without consulting reference aroma standards. This allowed for the compilation of an unlimited list of descriptors. Panelists then were asked to evaluate a range of aroma standards that are typical of Cabernet wines and vegetal aromas and that have been used previously in the UCD laboratory; the panelists determined how those standards should be modified, what standards could be dropped, and what standards should be added. An effort was made to introduce a high number of vegetal standards, but many were ultimately determined to be too specific to describe this range of wines. As the training sessions continued, the list of descriptors was gradually refined to include only the terms that reappeared in several of the wines. Following these sessions, the panelists had a final vote to determine the definitive list of descriptors (Table 2⇓). Finally, the panelists conducted two training sessions in sensory booths to familiarize themselves with the computerized data acquisition process.
For the formal evaluation sessions, standard references (Table 2⇑), made fresh daily, were evaluated at the beginning of each session before panelists rated the specified wines. Individual aroma attributes were rated (orthonasally only) on a 9-point scale anchored at the ends with the terms low and highest and in the middle with moderate. All wines were rated for a given aroma attribute before the panelists could continue on to the next attribute. To avoid fatigue, panelists were asked to sniff water between analyses of individual attributes. All sessions were conducted under red light and at 20°C.
Wines (20 mL) were presented in coded black, tulip-shaped glasses, with four wines presented per session in an order of presentation determined using a randomized complete block design; three replications (12 total sessions) were done by each panelist for each wine. FIZZ software (Biosystemes, Couternon, France) was used to collect data for all descriptive analysis sessions.
Expert sorting trials.
Sixteen volunteer panelists, comprised of winemakers, assistant winemakers, and enologists from California, conducted sorting trials based on aroma only. Panelists evaluated 15 of the 16 Cabernet Sauvignon wines previously selected for descriptive analysis. Wine p was not included because of limited supply. Panelists were presented all 15 wines in black, tulip-shaped glasses and in random order and were asked to initially sort the wines into two or more groups based on similarity of aroma (each group had to contain two or more samples). Panelists sorted into groups according to their own individual/unspecified criteria and were asked to note the descriptors they used to distinguish each group.
Following the initial sort, panelists moved to a different station where half of the panelists sorted into three groups: none, some, or high vegetal character. The other half sorted into none, low, or high categories based on intensity of sulfur aromas. The panelists then rotated stations and completed the other sort. At each station, the same wines were presented in a different random order with different codes in black wineglasses. Expert judges were not told that the wines were the same in all sorting exercises.
Pyrazine analysis.
Instrumentation.
A Hewlett-Packard (HP) 5890 Series II GC interfaced to a quadrupole HP 5973 mass selective (MS) detector (Agilent Technologies, Little Falls, DE) was used for the analysis; the GC-MS was equipped with a HP 7673B auto injector/sample tray (Agilent Technologies) and a DBL TPRD splitless liner. A DB-5 MS capillary column, 30 m x 0.25 mm i.d. x 0.25-μm film thickness (J&W Scientific, Folsom, CA) was used for the separation. The carrier gas was helium at 1 mL/min at 25°C.
The injection port temperature and the MS transfer line temperature were set at 200°C and 230°C, respectively. The initial oven temperature was set at 50°C for 5 min and then ramped to 100°C at a rate of 2.5°C/min and held for 0.1 min. It was subsequently ramped to 250°C at a rate of 35°C/min and was held at this final temperature for 3 min. Total run time was 32.4 min.
Sampling and analysis.
All wine samples were extracted and analyzed in duplicate. The two pyrazines were extracted by passing 25 mL of wine sample through an activated C-18 SPE cartridge (Diagnostix, Mississauga, ON, Canada) by gravity f low. The cartridge was then dried under vacuum and subsequently the pyrazines were eluted from the column by gravity with 0.5 mL dichloromethane (Caledon, Georgetown, ON, Canada). Following extraction, 1 μL of extract was injected into the GC. The presence of the individual pyrazines was determined using selective ion monitoring using a target ion of m/z 152 amu and qualifying ions of m/z 137 amu and 124 amu for IPMP and a target ion of m/z 124 amu and qualifying ion of m/z 151 amu for IBMP. Typical retention times for IPMP and IBMP were at 12.30 min and 16.62 min, respectively. The absolute concentration of both pyrazines was determined by comparison to authentic standard concentrations of the pyrazines (Sigma-Aldrich, Oakville, ON, Canada) spiked in ethyl acetate (Caledon).
The limit of detection (LOD) and limit of quantitation (LOQ) were 2 ng/L and 5 ng/L respectively for both pyrazines. The method has very good linearity up to 200 ng/L, and imprecision was measured as the relative standard deviation (RSD%) for multiple analyses at 10 ng/L and was <8.0%. The recovery for IPMP and IBMP at 5, 10, and 15 ng/L was between 80 and 92% and between 76 and 87%, respectively.
Data analysis.
A mixed model ANOVA was performed on descriptive analysis (DA) data, treating judges as a random effect, using SAS version 8.2 (SAS Institute, Cary, NC). The general linear model procedure (proc GLM) was then used to determine the significance of aroma attributes across all wines. Sorting data were analyzed according to the frequency with which each wine was grouped with other individual wines, using SAS proc multidimensional scaling (MDS) (SAS Institute). Two dimensions were sufficient to explain data variance.
Principal component analysis (PCA) was performed using SAS proc factor (SAS Institute) on the mean intensity scores for all DA attributes that differed significantly across the wines, using the covariance matrix and no rotation. Generalized Procrustes analysis (GPA) was performed on the combined sensory data (wine p was not included in this analysis) and on the DA and pyrazine data. All GPA was performed using SensTools for Windows, version 3.0 (OP&P, Utrecht, The Netherlands).
Partial least squares (PLS) regression was performed on the pyrazine data and mean rating scores for descriptive terms using Unscrambler (CAMO Software, Oslo, Norway). The X variables included the mean concentration of each pyrazine and were the indicator variables. The Y variables consisted of the descriptive panel aroma attributes. PLS 1 and 2 were performed to evaluate the predictive capability of the model for individual sensory attributes as well as for the overall sensory data. All variables for pyrazines were normalized while the sensory terms were not adjusted. Leverage correction was used. Linear regressions between individual sensory attributes and pyrazine concentrations were done in Excel (Microsoft, Redmond, WA).
Results
Sensory analysis.
Descriptive analysis.
All sensory attributes varied significantly across the 16 wines ( p < 0.05; Table 3⇓). Principal component analysis was used to illustrate the interrelationships among the wines and descriptive terms, and, together, the first two dimensions accounted for 80.7% of the variance in these wines (Figure 1⇓). Principal component 1 (PC 1; the x-axis) was clearly divided into vegetal versus nonvegetal/fruity descriptors, while principal component 2 (PC 2; the y-axis) was divided into wines high in eucalyptus versus those low in eucalyptus. The bell pepper attribute also contributed significantly to PC2. Several of the wines were located close to the center of the graph; the wines in the center of the graph may be considered more balanced in the intensities of the various aroma attributes. Vegetal wines n, o, and p were negatively loaded on PC 1 (located on the left half of the x-axis), while fruity wines e, f, k, and L were positively loaded (located on the right half of the x-axis). Wine p was highest in cooked vegetable aroma (mean intensity score = 4.9), whereas wine n was highest in eucalyptus aroma (mean intensity score = 3.78) and second highest in bell pepper aroma (mean intensity score = 3.08). Wine k was highest in cherry aroma (mean intensity score = 5.07) and second highest in dry fruit aroma (mean intensity score = 4.12).
Although cocoa, vanilla, and dry fruit all appeared closely related in the first dimension, consideration of the third dimension of this PCA (PC 3, which accounts for an additional 8.75% of the variance) reveals that these terms can be distinguished (data not shown). In addition, PC 3 distinguishes between the bell pepper and pepper attributes, with samples a and I characterized as more intense in pepper character than the other wines.
The vegetal versus nonvegetal division in these wines did not correlate to any particular growing region or county. There were also no clear relationships between vintage year and oak regimes on vegetal perception (data not shown).
Expert sorting panel.
An expert panel was used to group wines by similarity of aromas. Initially, each panelist was asked to independently group the aroma of the wines according to his/her own perception of similarity and to note the criteria used for each grouping. The expert panelist criteria included the following descriptors: bell pepper, green aromas, band-aid/veggie, pyrazines/smoky, green bean/cooked spinach, dill, stemmy, herbal/sweaty/reduced, veg, weedy, carrot/pumpkin pie, old cabbage/broccoli/cooked veg, asparagus (fresh versus canned), pickled veg/canned, herbaceous/tea/earthy, mushroom/barnyard, geranium/floral/weedy, pyrazine, and olive. Fourteen out of 16 of the panelists used at least one vegetal term as a sorting criterion.
Multidimensional scaling (MDS) was used to provide a preliminary analysis of the sorting data (Figure 2⇓). Although the descriptors used for the independent/unspecified sorting encompassed a wider range of aromas than those selected for the DA, many of the groupings between the two methods were similar. For example, wines I, n, and o, characterized by DA as high in vegetal characters such as bell pepper and olive, were also grouped together by the sorting procedure. Similarly, wines, f, k, L, and m, characterized with fruity descriptors by DA, were grouped together in two related groupings by the sort procedures (Figure 2⇓). In addition, very similar clusters of wines with vegetal and fruity characters were obtained when the criteria for sorting was vegetal aroma or sulfur defects (data not shown). Therefore, regardless of sorting criteria, similar groupings were obtained.
Generalized Procrustes analysis (GPA) was used to examine more fully the sorting data and to relate these data to the DA profile (Figure 3⇓). The wines were grouped in similar patterns regardless of sensory technique. The first dimension of the vegetal sort (VEG dim 1) was most similar to the first two dimensions of the DA mean scores (PCA dim 1 and PCA dim 2), which is not surprising, as both exercises were directed by vegetal criteria. The sulfur sort (SUL dim 1) and the general/unspecified sort (Gen dim 1) were nearly identical in the first dimension, which suggests one of two things. Sulfur aromas may have been the primary factor distinguishing these wines, or the sulfur sort may have been an unclear criterion, and thus panelists may have resorted to using the same criteria for both the independent and sulfur sorting exercises. Some distinctions among the sorting methods and the DA are apparent in the second and third dimensions of the GPA plot, indicating some differences in the sensory procedures used to characterize some wines (data not shown). Differences in fruity aroma perception were determined in the DA, but no sorting trials using fruity perception as a sorting criterion were performed. Therefore, it is possible that fruity perceptions led to some of the differences observed in the analyses by the various sensory methods; however, further studies are needed to fully characterize these effects.
Pyrazine analysis.
Two pyrazines were identified and quantified (Table 3⇑). IBMP is characterized by a bell pepper aroma, has an aroma threshold of 2 ng/L in water, and ranged in concentration from ~2 to 24 ng/L in these wines. IPMP, which has a potato, earthy aroma (Seifert et al. 1970), has an aroma threshold of 1 ng/L in water, and ranged in concentration from ~2 to 16 ng/L in these wines. Based on thresholds and measured concentrations, both of these compounds could have contributed to the sensory properties of these wines.
Relationship of chemical profiles to sensory attributes.
Partial least squares regression, GPA, and linear regression were used to relate the vegetal aroma perception (determined by DA) to the pyrazine concentrations; however, no significant relationships were obtained for any of the sensory attributes measured in the wines (Figure 4⇓; data shown for linear regression analysis between pyrazines and bell pepper intensity only). For example, wine b had the highest concentration of IBMP (24 ng/L); however, the vegetal attributes assessed by the descriptive panel were only of average intensity relative to the other wines (Table 3⇑).
Discussion
Definition of vegetal aromas in quantifiable sensory terms.
Trained sensory panelists using defined reference standards were able to distinguish qualitative differences in vegetal related aromas. Most notably panelists distinguished between a more “fresh” vegetal aroma characterized by bell pepper and a “cooked” vegetal aroma defined here by a mixture of canned vegetables which included potatoes, corn, green beans, and asparagus. The olive attribute, defined in this study by a combination of green and black olives, was located between the fresh and cooked vegetal aromas in the PCA (Figure 1⇑). Other attributes, such as eucalyptus and black pepper, also appeared to be related to the bell pepper aroma, but were clearly discernable by the panelists.
The wines used in this study were selected to provide as broad of a range of vegetal characteristics as possible. However, the vegetal attributes used here may not cover the entire spectrum of specific vegetal characteristics that may be present in other wines. Therefore, caution should be exercised in extending these results to a broader set of wines. However, some generalization may be made. For example, while qualitative differences in vegetal aromas can be defined and quantified, it also appears that the major factor distinguishing these wines is the contrast between the combined vegetal characteristics and the non-vegetal/fruity characteristics. Interestingly, this contrast appears to be remarkably consistent among Cabernet Sauvignon wines across years and different sensory panels. For example, DA of Cabernets from a range of different viticulture regions showed a largely similar PCA profile with vanilla, berry, and fruit attributes grouped together and opposing the bell pepper, vegetal, green bean, and soy attributes (Heymann and Noble 1987). Similar results have been observed (Chapman et al. 2004). This contrast between vegetal and fruity aromas seems apparent even in wines of different varieties such as Zinfandel and Pinot noir (Noble and Shannon 1987, Guinard and Cliff 1987), indicating, at least in those wines where vegetal characters are present, that vegetal aromas may be a dominant factor influencing sensory perceptions.
Winemaker “experts” criteria.
By comparing sorting data obtained from winemakers with the results of the trained DA panel, we observed that the primary criteria for classifying the wines were similar for both the expert winemaker panelists and the trained DA panel regardless of the sensory techniques used. These results also indicated that distinct but related aromas may be considered vegetal when characterizing wines. Therefore, use of the general vegetal term is complex, consisting of many related aromas, and precise terminology for vegetal-related characteristics is necessary when communicating about wine sensory properties. It is interesting that the expert panelists did not use any fruit-related criteria for the sorting trials, although they were used by the DA panel to differentiate among the wines. The reasons for this discrepancy are unclear.
These results also highlight some of the difficulties in characterizing perceptual differences in complex odor mixtures. For example, Laing (1991) has demonstrated that it becomes increasingly difficult to identify individual odor qualities as the number of chemicals in the mixture increases above three. When put into the context of the results from our studies, it appears that panelists may have readily distinguished two distinct and perceptually different attributes such as vegetal and fruity (or sulfur and fruity attributes), while the perceptually more similar vegetal and sulfur attributes were grouped together. However, we did not have the expert panelists sort the wines based on similarities of fruity attributes, and it would be interesting to determine if similar groupings would have been obtained had this been a sort criterion. In addition, it would be interesting to determine whether the expert panelists could distinguish between vegetal and sulfur attributes if asked to sort the wines between these two categories.
A comparison of results from DA and sorting experiments also leads to interesting considerations regarding the application and interpretation of these two sensory methodologies. In our study, similar results were obtained with DA and sorting experiments, which seems to be consistent with experiments that compared intensity ratings and similarity ratings for citrus and woody odors in complex mixtures (Lawless 1999). Lawless raised several possibilities that may explain the similarity in results from the two different types of sensory tests: (1) that intensity judgments (as we used for DA analysis) and similarity judgments (as were used for the expert sort analyses) are closely linked, (2) that subjects may substitute one type of judgment for the other, (3) that other decision processes may underlie both types of ratings, and (4) that subjects may ignore the instructions and respond as expected to task demands. Clearly, more work is needed to evaluate how the choice of sensory methodologies may influence the results that are obtained when measuring complex odor mixtures.
Chemical composition influencing vegetal perception.
Differences in vegetal perception among the wines used in this study may be due to several factors, including (1) higher concentrations of chemicals that contribute directly to vegetal aromas, (2) the absence of compounds that contribute fruity aromas, or (3) a masking of vegetal aromas by fruity aromas and vice versa when compounds associated with fruity aromas are also present. A recent study indicated that fruity aromas may significantly decrease perception of bell pepper aromas in wines, even when the concentration of the bell pepper aroma compounds does not change (Hein 2005). In the present study, no compounds associated with fruity aromas were analyzed, so we cannot evaluate their contributions to wine flavor in these wines.
Here we focused only on one category of aroma compounds, the pyrazines (IBMP and IPMP) that have been most closely linked to vegetal aromas in Cabernet. Many other compounds have vegetal aroma characteristics, including sulfur-containing volatiles and norisoprenoids (Goniak and Noble 1987, Darriet et al. 1993, Tominaga et al. 1996, Butzke 1997, Bouchilloux et al. 1998, Janusz et al. 2003, Swiegers et al. 2005, Cox et al. 2005); however, these compounds have not been as widely studied as pyrazines in Cabernet Sauvignon. In the present study, concentrations of pyrazines alone were not correlated with any of the sensory attributes, indicating that other volatiles also affect sensory perception of vegetal characters. Sensory attributes identified during the initial screening and during the DA, as well as results of the sorting trials (e.g., the similarity in sorting when sulfur was used as a sorting criteria compared to vegetal criteria), indicated the possibility that sulfur-containing compounds may play a role in the vegetal aromas of these wines. Recent studies have indicated that dimethyl sulfide, which by itself has a cooked corn/vegetal aroma, when present in mixtures can enhance fruity aromas (Segurel et al. 2004, Escudero et al. 2007); however, the overall impact is dependent upon the dimethyl sulfide concentration and the other aroma compounds that are also present. Clearly, further analysis of a range of volatile sulfur compounds is necessary to fully understand the effects of these compounds on vegetal aroma perception in Cabernet Sauvignon wines.
Our study indicates that analysis of pyrazine concentrations alone may give an incomplete picture of factors that can impact vegetal aromas. This is particularly important to consider, given that many aroma compounds are present in grapes and produced during winemaking. Therefore, even when viticultural practices are optimized to result in low IBMP levels, vegetal wines may still result.
Conclusions
While the findings of this study cannot be extended to all Cabernet Sauvignon wines, they do serve to underline the complexity of vegetal aromas in wine. In general, specific qualitative differences in vegetal aromas (e.g., bell pepper versus cooked vegetal) can be defined and used by trained panelists to distinguish among a set of Cabernet Sauvignon wines. However, the contrast between generic vegetal aromas and nonvegetal/fruity aromas appears to be a main criterion for characterizing Cabernet Sauvignon wines.
Results from DA panels and sorting trials gave similar results, independent of the sorting criteria for grouping the wines. While winemaking experts used a wide range of descriptive terms to classify vegetal wines, the general groupings into vegetal or nonvegetal categories were similar. These results indicate that many related aromas may be considered vegetal by winemakers. Whether results from these sensory trials can be extended to perception of vegetal attributes by consumers is not known.
Although the vegetal aroma in Cabernet Sauvignon is commonly associated with IBMP, this study reveals that other chemicals are likely influencing vegetal aroma as well; therefore, a more detailed study of a range of volatile compounds in Cabernet Sauvignon is needed. In addition, while much progress has been made in our understanding of the impact of viticultural and enological parameters on vegetal aromas in wines, many of these studies have not focused on the complex interactions among many different chemical compounds. Our results can now be used to design additional experiments that can take these interactions into account and improve our overall understanding of the factors that impact the chemical composition and sensory perception of vegetal characters in wines.
Footnotes
Acknowledgments: This project was partially funded by the American Vineyard Foundation and the California Competitive Grant Program for Research in Viticulture and Enology.
The authors thank the collaborating winemakers and wineries who donated wines, Kevin Byrne of E&J Gallo Winery for technical assistance, John Thorngate of Constellation Wines for helpful discussions, and the sensory judges who participated in the DA and sorting trials.
- Received July 2007.
- Revision received February 2008.
- Copyright © 2008 by the American Society for Enology and Viticulture