Wine quality is difficult to define, but can be evaluated based on the sensory characteristics of the product using the sensory skills of experts, which include, but are not limited to, vision (clarity and color), gustation (taste), and olfaction (aroma). Wine aroma may be the most important factor in assigning quality, as faults, like microbial contamination and lack of typicality of style, are often detected by olfaction [1
]. Frequently used indicators of the potency of aroma compounds associated with these faults include odor detection threshold (ODT) and odor activity value (OAV) measurements. Sensory thresholds are ill-defined in theory [2
], but are based on the lowest concentration of a compound at which individuals can perceive a difference in sensory quality, relative to a control.
The ODT, which may be mg/L (ppm), µg/L (ppb), or ng/L (ppt), is numerically equal to the minimum concentration at which 50% of the judges succeed in differentiating a sample from a control [3
]. The triangle test has been widely used for determining the detection threshold of volatile compounds important to wine aroma, including: Diacetyl [4
]; oak lactone [5
]; 3-isopropyl-2-methoxypyrazine [6
], and rotundone [7
]. The International Standards Organization (ISO) 13301 standard (encompassed in the American Society for Testing and Materials (ASTM) Standard Practice E679-04 [8
]) has been used to establish detection thresholds for 2,4,6-trichloroanisole (TCA) [9
] and ethyl hexanoate in solution with tannin and mucin [10
]. The OAV is defined as the concentration of a compound present in a matrix divided by the ODT for that compound in that specific matrix. Generally, the larger the OAV, the more likely that compound would contribute to the overall odor of a complex odor mixture [10
]. In theory, ODTs and OAVs may be determined this way for compounds of interest [10
], but, in practice, a large number of issues influence the perception of these measures. Matrix effects, for example, have been shown to influence ODT perception in wine [11
], and would therefore influence OAVs. The presence of other compounds in solution can also change the perception of target aromas, not just above the ODT (supra-threshold), but also below it (i.e., at sub- or infra-threshold). For example, the impact of infra- and supra-threshold concentrations of ethylphenols on wine was investigated [14
], and it was found that both sub- and supraliminal concentrations of off-flavors not only changed the “hedonic valence” (positive or negative character) of the perception, but had a masking effect on fruity notes.
The composition of the panel used will also influence the success of sensory threshold studies. A panelist’s ability to detect an odor will vary depending on their level of fatigue, health status, and matrix effects in the study medium [8
]. The reliability of the results can be increased by enlarging the panel and by replicating tests, but this will not negate matrix effects in another medium. Due to the unique nature of the product, sensory studies on aroma in wine are very specific to particular cultivars and styles, and frequently suffer from a lack of repeatability. For example, if research on the ODT of 4-ethylcatechol is reviewed, Larcher et al. [15
] determined a detection threshold of between 100 and 400 µg/L in white and red wine, while another study determined a considerably higher odor threshold of 774 µg/L [16
]. The Australian Wine Research Institute noted that the odor threshold of this compound increased to 1131 µg/L in “green” red wine, and 1528 µg/L in oaky red wine [16
]. It would therefore seem obvious that ODTs are confirmed in the matrix in which a study is to be carried out, but evidence for confirmation is lacking in the literature.
In this study, five odor-active compounds were selected that had been associated with specific taint issues in wine aroma, namely guaiacol, o
-cresol, 4-ethylphenol (4EP), 3-isobutyl-2-methoxypyrazine (IBMP), and 2,4,6-trichloroanisole (TCA). Aroma descriptors for these compounds cover a continuum from “burnt, smoky” and “medicinal” to the “moldy, dusty” flavors associated with cork taint (TCA), and “herbaceous, green” aromas associated with IBMP. All of these attributes have been associated with South African red wines [13
]. In order to progress with work on interactions between these compounds in a partially de-aromatized red wine, it was necessary to confirm odor thresholds that had been previously determined (Table 1
), and test sensitivity of potential panelists.
Guaiacol exhibited the lowest odor detection threshold of all the phenols tested in a red wine matrix by Boidron et al. [26
]. It forms part of the “woody” family of descriptors [22
], and is extracted from pyrolyzed (toasted) oak during wood maturation of wine. The detection threshold of 23 µg/L for guaiacol confirmed by Parker et al. [21
] was chosen as it was determined in a relatively recent formal sensory study at the Australian Wine Research Institute in an Australian “base red wine”.
It is very difficult to find any sensory information in the primary literature about the cresols. There has been some work done in whisky, but the only current wine-related references concern its appearance in wine after smoke-events in the vineyard [27
]. Boidron et al. [26
] described o
-cresol as “tar” (bitumen)-smelling, and gave an ODT of 0.8 mg/L. As the cited study was investigating wooded wines, and o
-cresol is a known pyrolysis product of oak wood [28
], this relatively high ODT seems reasonable. Parker et al. [22
] reported a threshold, through a formal sensory process with 22 assessors, of 62 µg/L (standard error = 0.8) for o
-cresol in “base red” wine. Due to the rubbery, tarry, and phenolic character of the compound, it was included in our study, and Parker’s [22
] threshold used, as it is the only recent threshold available in the literature carried out in an unwooded base wine.
Associated with “brettiness” and “medicinal, phenolic” smells, as well as those of “leather/horse” and “bacon/meatiness”, 4EP has complex effects in wine. The detection threshold of 605 µg/L in red wine determined by Chatonnet et al. [23
] is commonly cited. According to Escudero et al. [29
], 4EP falls in the same semantic category as woody odorants, and in wooded wines, the wood character may mask the aroma character of the compound, making it more difficult to detect as a specific attribute (leather/horse/medicinal). This is in agreement with the findings of Curtin et al. [16
], and seems to explain the high threshold found by Boidron et al. [26
]. For the purposes of this study, Boidron’s published detection threshold was used, as it has been cited by a number of other researchers, and is based on recognized sensory methodology.
IBMP is a potent aroma-active compound often found at higher, above-thresholds concentrations that are detrimental to red wine quality [24
] due to its “green/herbaceous/bell pepper” characteristic. Information on the determination of odor thresholds for pyrazines in the literature is limited. IBMP, specifically, is a very strong-smelling compound in its pure form, and its odor is pervasive and difficult to disperse. It is also known to cause skin, eye, and respiratory irritation [31
]. Shibamoto [32
] determined odor thresholds of 46 pyrazines in water, but did not test IBMP. French studies showed that IBMP was the main contributor to vegetal aroma in red Bordeaux and Loire wines from different vintages and cultivars [24
], and the ODT was determined by comparing IBMP concentrations with the intensity of the green bell pepper character. Through this, the threshold value, which seems to be a recognition threshold, and not a perception threshold, was estimated to be 15 ng/L. This does not seem to have been confirmed by any further formal sensory studies.
TCA is a compound associated strongly with cork taint in wines and is described as having a “moldy” and “damp cardboard” odor. Tempere et al. [33
] noted that amongst wine defects, 2,4,6-trichloroanisole has a specific impact on wine perception. In addition to giving to the wine an unpleasant odor, it has a strong masking effect on fruity notes. It is difficult to find information on formal sensory detection threshold determinations of TCA, but in a study on the effects of TCA on consumers’ wine perception [25
], a level of 0.13 ng/L TCA was considered ”low” and 5 ng/L was considered “high” based on concentrations frequently found in contaminated red wines. TCA detection threshold depends on the wine organoleptic characteristics, and the person perceiving it, as wine professionals had very varying abilities towards TCA detection in wine. These authors tested detection of TCA with a panel selected based on their sensitivity to this compound, in a range of red and white wines, some of which were wooded. They found TCA was detected, in red wines, at 10–15 ng/L. Prescott et al. [34
] established “consumer rejection thresholds” of TCA to be 3.1 ng/L and 3.7 ng/L in Chardonnay, depending on the panel. Although Mazzoleni et al. [9
] found much higher levels for Italian red wines, and Cravero et al. [35
] gave an identification threshold of 7 ng/L (in Barolo), the wine in this study was unwooded and partially de-aromatized, and therefore the detection threshold of 4 ng/L, as indicated by Prescott, was chosen as suitable for the matrix.
The formal procedures for ODT determination are outlined in standard methods, specifically ASTM E679 (Determination of Odor and Taste Thresholds By Ascending Forced-Choice (AFC) Concentration Series Method of Limits) [8
]. Standard practice for defining and calculating individual and group sensory thresholds are outlined in ASTM E1432-04 [3
]. Normally, individual threshold calculations would require 20 to 40 AFC presentations per panelist, with data sets taken at five or more concentration scale steps (typically six or seven), with pretesting to ensure individual thresholds fall within the testing range. Obviously, this is laborious, and as the ASTM standard itself states, the costs and availability of panelists places serious limitations on what can be covered by experimenters who, typically, are limited to panels of five to 15 individuals. Even with a limited panel, a simple threshold determination will involve 100–600 presentations, which would not be achievable for most researchers. This study looked at a simple, pragmatic approach to ODT sensitivity testing for the purposes of selecting panelists for a study.
From Table 1
and the discussion above, it is clear that ODTs found in water are not comparable to those determined in ethanol solution, and thresholds in white wine are different to those found in red wine. Differences between ODTs for various compounds within the categories of red wines (different cultivars [36
], wooded/unwooded [29
], for example) have also been shown to exist. Even on the rare occasions when thresholds are properly determined in wine according to the official E-679 method, they are affected by the specific wine matrix and style [11
], and therefore do not necessarily have relevance in other matrices. It is also evident from the literature that different people have varying responses to compounds based on their culture, experience, and age, and that training may make a difference to a person’s ability to perceive an aroma [39
For these reasons, the ODTs of compounds of interest to our work were investigated in a de-aromatized red wine matrix, which we planned to use in interaction studies. The sensitivity of potential panelist to compounds at ODT level was tested using a simplified version (three sets of samples per compound at the ODT) of the ASTM E-679-04 method. This study should thus provide an interesting perspective to researchers when they consider confirmation of ODTs in other matrices, as this is a pragmatic approach given the fairly complex and time-consuming standard practice. This research also set out to establish if panelists had particular sensitivities to compounds belonging to diverse aroma families, and whether previous training in sensory evaluation of wine influenced their ability to distinguish threshold differences between samples and controls.
To explore a pragmatic approach to judge sensitivity for five taint compounds (guaiacol, o-cresol, 4EP, TCA, and IBMP) in red wine, three triangle tests (two controls, one sample spiked at the selected ODT for each compound) were carried out for 34 judges. Spiked samples were correctly differentiated from controls in the following order: o-cresol (62% correct), 4EP (59% correct), guaiacol (56% correct), TCA (44% correct), and IBMP (41% correct). What could be deduced from an analysis of correct identification of spiked compounds was that correlations in perception between pairs of compounds was very weak. If a particular compound was detected by the judges, it did not have a bearing on another group, but judges that had higher scores generally were more likely to detect all compounds across both families. As IBMP and TCA were the last two compounds to be assessed of the five, and also the two compounds that showed the least correct responses or differences between experienced and inexperienced judges, it is probable that the assessors experienced sensory fatigue. This is a limitation of this strategy, and if the study were to be repeated, it would be advisable for each compound to be tested on a different day at different levels, or presented in a random order to the judges. Ideally, a full ODT study should be carried out to determine ODTs for a specific matrix, but the simplified method will be a frame of reference for further work. In this study, the responses of the judges confirmed that the ODTs for guaiacol, and 4-ethylphenol in de-aromatized Shiraz wine, but an additional, more comprehensive sensory study to establish whether the levels could be reduced would be useful to determine the actual thresholds of detection in this matrix.
The concentration of o-cresol could certainly be reduced in future work in this matrix, as almost two thirds of the panelists were able to detect spiked samples, and concentrations of IBMP and TCA could be increased, as only 44% and 41% of judges could distinguish these compounds spiked at accepted ODT levels in this matrix. This may also have been due to sensory fatigue, and is a limitation of testing this number of compounds on one day.
One interesting aspect that emerged from this work was the ability of judges to detect compounds more easily if they had some experience of winetasting, showing that training can possibly sensitize people to attributes that they might not have noticed previously. The confirmation of ODTs in a partially de-aromatized red wine matrix has opened the way for further research and the possibility of training industry members to sensitize them to these off-flavors. The unique and complex interactions between language, training, and olfaction should definitely be assessed in future studies. This study highlights how important it is to understand the detection thresholds of compounds in different products, and has relevance for many foods and beverages where sensory thresholds from the literature may be taken for granted and not tested in the study matrix.