Discrepancy Between the Ability of Wine Experts and Consumers in Identifying Grape Smoke Exposure in Different Wine Matrices in California

ABSTRACT

Volatile phenols (VPs) are present in fresh smoke from wildfires due to thermal degradation of wood lignin and can absorb into grape berry skins where they are rapidly glycosylated. VPs and their glycosides can impact the sensory profile of wines that are characterized by smoky, medicinal aromas and a retro‐nasal ashy “aftertaste.” Anecdotal evidence indicates 15–20% of consumers cannot distinguish the retro‐nasal ashy characteristic of smoke taint. It is difficult to mimic smoke‐tainted wine through spiking with related smoke marker compounds due to their complex synergistic interactions with the wine matrix. This study compared smoke‐impacted and non‐smoke‐impacted wines of Cabernet Sauvignon, Malbec, and Syrah varieties made from the same location (Napa and Sonoma). Smoke‐impacted grapes were intentionally smoked on drying tables or naturally exposed to smoke during wildfires. The wines were blended through serial dilutions, achieving different levels of smoke impact. The free, total, and bound VPs of all wines were determined by GC‐MS and LC‐QTOF‐MS, respectively. Wines were sensorially evaluated by 62 wine experts and 58 regular wine consumers separately; using a series of hedonic, Just‐About‐Right, and Check‐All‐That‐Apply questions. Correspondence analysis determined that different levels of smoke impact were detected differently by wine experts and consumers. Experts could distinguish smoke impact at lower levels, as determined by VP analysis, and compared to consumers who were generally only able to distinguish smoke‐impacted wines at higher levels. Further difference testing with an additional group of wine consumers, saw an increase in the sensitivity towards smoke when consumers were trained to identify/recognize the ashy attribute.

Practical Applications

This research provides winemakers with valuable insights into how wildfire smoke impacts wine flavor and the levels at which both expert and non‐expert consumers can detect smoke taint. By understanding these sensory thresholds, wineries can make informed decisions on whether and how to treat smoke‐affected wines—particularly through blending strategies. This enables them to manage quality in smoke‐exposed vintages and transform wines that might otherwise be unsellable into acceptable, marketable products.

1. Introduction

When a fire event happens, volatile phenols (VPs) are released in fresh smoke from the thermal degradation of lignin in wood (Szeto et al. 2020). VPs are absorbed into grape berries where they are rapidly glycosylated as part of the plant's defense mechanism (Hayasaka et al. 2010; Ristic et al. 2016). Wines made from these smoke‐affected grapes could have smoky orthonasal and retronasal aromas which are described as smoky, burnt, medicinal aromas and a retro‐nasal ashy “aftertaste” (Kennison et al. 2007; Krstic et al. 2015; Ristic et al. 2016). Wines of such nature are referred to as “smoke tainted” and are undesirable to winemakers who may choose not to make the wines from grapes affected by smoke for economic or branding reasons. The economic losses to the California wine industry were estimated to be around $3.7 billion for the fires that occurred in 2020 (Andrew Adams 2021; Downey Brand LLP 2021). This impact is not limited to California, but rather affects the global wine industry through North America, South America, Australia, and Europe (Mirabelli‐Montan et al. 2021).

The smoke‐related descriptors mentioned above are not limited to describing smoke‐impacted wines. Some smoky flavors could also be present in non‐impacted wines due to grape characteristics where varietals such as Syrah can produce a smoky impression (Krstic et al. 2015)or processing conditions such as barrel aging in heavy toast oak barrels which were fire toasted (Koussissi et al. 2009; Prida and Chatonnet 2010). VPs such as guaiacol, syringol, and 4‐methylguaiacol which are produced during the toasting process can impart smoky characters to the wines (Pollnitz et al. 2004; Prida and Chatonnet 2010). The chemical compounds that are responsible for the smoky characteristics in smoke‐impacted wines are complex. Currently, VPs and phenolic glycosides are the marker compounds used to determine the smoke impact of a wine (Parker et al. 2012; Hayasaka et al. 2013; Jiang et al. 2021). VPs such as guaiacol, 4‐ethylguaiacol, 4‐methylguaiacol, o‐cresol, m‐cresol, p‐cresol, syringol, and 4‐methylsyringol impart smoky aromas and flavors, while phenolic glycosides such as methylsyringol gentiobioside, phenol rutinoside, guaiacol rutinoside, and syringol gentiobioside are precursor markers of smoke impact (Wilkinson et al. 2011). These glycosides can breakdown during fermentation, barrel aging, and bottle aging to release VPs, as well as in the mouth by the enzymes present in the saliva which are thought to contribute to the lingering retronasal ashy aftertaste (Hayasaka et al. 2010; Parker et al. 2012; Mayr et al. 2014). Ideally, for a study of smoke impact, it would be best to spike base wine with the relevant VPs and phenolic glycosides. However, due to the synergistic effects of the wine matrix, it is difficult to mimic the effects of smoke impact through spiking of the wines (Crews et al. 2022; Coulter et al. 2022). There is also a lack of robust chemical models that relate the chemical composition of the wines to the intensity of smoky attributes. Parker et al. (2023) showed some relationship between smoke flavor intensity and VPs and phenolic glycosides. Tan et al. (2024) developed predictive models to determine overall smoke taint using VPs. Previous work also showed that highly smoke‐impacted grapes result in wines with strong smoky aromas with high levels of VPs and phenolic glycosides. Mid to low smoke‐impacted grapes may not always be smoky, rather there could be a quality impact such as the wine being perceived as thin and watery (Mayr et al. 2014; Bilogrevic et al. 2023). As such, sensory analysis is still the determining tool for winemakers or wine companies to decide if the wine meets expectations for sale to the market or if the degree of smoke taint is beyond that which experts or consumers will find acceptable.

To determine the extent of smoke impact on the wines, sensory analysis is needed. Normally, a descriptive analysis would be used such that the intensities of each attribute can be quantified (Lawless and Heymann 2010). However, descriptive analysis takes significant amounts of time, resources, and expertise of sensory scientists which may not be available to winemakers. Rapid sensory assessments such as modified descriptive analysis, rate‐all‐that‐apply, flash profiling, and pivot profile are methods that are preferred in the industry today as these methods are more efficient, easier to run, and present similar data reliability to descriptive analysis (Muñoz et al. 2017; Danner et al. 2018; Liu et al. 2018; Lestringant et al. 2019; Pearson et al. 2020; Nishida et al. 2021). Even with such rapid methods, a trained panel is needed to assess the products which may not be practical for winemakers. The trained panel can assess the intensity of attributes but does not assess the suitability of the attributes for the product in terms of its overall liking or adequacy of the attributes. When considering overall differences between products, difference testing (i.e. triangle test, duo‐trio or 2‐ or 3‐alternative forced choice) can be used to determine how much of a difference there is between two products. Degree of difference may also be used in quality control to determine if there are differences between a test and control product (Pecore et al. 2006). In this study, we used difference testing to find out where consumers were able to find differences between the control wine and the different treatments through serial dilutions to identify their threshold levels of perception (Lawless and Heymann 2010). Naïve consumers are not normally as sensitive to changes in a product as an expert is, as they lack the technical expertise to assess small differences. Experts have a greater understanding of technical issues in a wine and are able to better explain these small differences and changes (De Mets et al. 2017; Basalekou et al. 2023). A directional difference test such as the 2‐alternative forced choice (2‐AFC) or 3‐alternative forced choice (3‐AFC) test can be used to assess if consumers can pick up which product is higher or lower for a specific attribute. This is often used in threshold testing for a single compound. The upside of using a difference test is that it considers the product tested in its entirety. However, it does not consider the overall balance and suitability of attributes which are key drivers in assessing a quality‐driven product like wine. Other methods in the realm of consumer testing can be used to get data for overall liking and adequacy of attributes of a product through hedonic liking, preference mapping, and check‐all‐that‐apply (CATA) (Delgado and Guinard 2011; Francis and Williamson 2015). These consumer‐driven methods consider the preferences of the consumers and give information on descriptive attributes. Using hedonic liking measurements, we can infer from the overall liking scores of the product, its overall sensory quality, that is high hedonic ratings would indicate a highly acceptable sensory quality. For consumers, their assessment of the quality of a product is dependent on both how much they like it and how well it meets their expectations (Everitt 2009). Yet, among both consumers and experts, there is anecdotal evidence of the consistency of ratings in each group based on their overall liking (Moskowitz 1996; Ruiz‐Capillas and Herrero 2021). Experts usually are good at rating one set of products, either red wine or white wine, and there was a consistency improvement in the scoring when scores of a small team of tasters were combined (Gawel and Godden 2008). It is important to understand the differences between how experts and consumers rate products.

In the recent study by Bilogrevic et al. (2023), there was alignment between consumers and a trained smoke panel in the perception of smoke‐like characteristics. Despite this, there is still little known about how the perception of different smoke aromas and flavors, if they were not classified as a taint, would affect the consumer acceptance of these wines. Conversely, in studies of different faults/ off‐flavors in wine such as “Brett” and “oxidation,” low levels had no effect on the acceptability of the wines to consumers (Francis and Williamson 2015). On the other hand, some faults, such as cork taint at perceptible levels, impacted wine acceptability by consumers (Prescott et al. 2005). These studies show that there is no yardstick to determine the acceptability of faults in wine by consumers. In smoke taint assessment, studies where smoke‐related compounds were spiked into base wine, there was a lower level of acceptance of the product when substantial amounts of the related compound were added (Bilogrevic et al. 2023). There is some evidence that 15–20% of the overall population may be insensitive to smoke in wine, specifically (Mayr et al. 2014; Bilogrevic et al. 2023), but there is no evidence to date of how the winemaking population may be anosmic towards smoke. Wine experts can pick up defects in a wine from its organoleptic cues due to the matrix, where a decrease in wine quality is an indication of smoke taint to them (Parker et al. 2012). This ability is quintessentially important as it would relate to further decisions by the winemaker if a wine could be salvaged through treatment, especially if it was more of a quality impact than a smoke impact. It is also unknown whether consumers can be trained to pick up a taint in a wine matrix instead of classifying that taint as complexity in a wine.

This research tested the hypothesis that experts and (non‐expert) consumers would differ in their ability to differentiate smoke‐impacted (naturally or intentionally) and non‐impacted wines across different red wine varietal matrices—Cabernet Sauvignon, Syrah, and Malbec using serial dilutions between the smoke‐impacted and non‐impacted wines. A secondary hypothesis we tested was that the ability of (non‐expert) consumers, to perceive smoke taint in a single wine matrix could be increased upon familiarization with the sensory attributes of smoke taint.

2. Materials and Methods

2.1. Wines

Wines used for the study were pairs of non‐smoke impacted (NST) and smoke impacted (ST) wines (Table 1). We tested a subset of the wines used in the descriptive analysis by Lim et al. (Lim et al. 2024). CS_1_ST, MA_ST, and SY_ST were wines made from grapes that were naturally exposed to the wildfires in California in 2020. In 2021, a single lot of Cabernet Sauvignon grapes was split into two lots, CS_2_NST and CS_2_ST. The CS_2_ST grapes, approximately 100 kg, were intentionally smoked on drying tables in a purpose‐built tent (3 m × 1.5 m × 1.5 m) using plastic poles and six‐millimeter polyethylene sheeting (Frost King & Thermwell Products Co., Inc., Mahwah, NJ, USA). These grapes were smoked for 1 h using a Z Grills pellet smoker (Z Grills Inc., Ontario, CA, USA) with 100 grams of hickory wood pellets from Traeger (Traeger, Salt Lake City, UT, USA). Standard winemaking practices were used, as described in Lim et al. (2024).

TABLE 1.

Summary table of base wines used for blends extracted from Lim et al. 2024. Smoke exposure was determined through a post‐smoke event air quality index (AQI) analysis from PurpleAir network.

Wine Identifier	Year	Varietal	AVA	County	Taint	Smoke exposure period (AQI)
CS_1_NST	2021	CS	Dry Creek Valley	Sonoma	NST	—
CS_1_ST	2020	CS	Dry Creek Valley	Sonoma	ST	8/19/20–9/21/20 (5 days >150)
CS_2_NST	2021	CS	Oakville	Napa	NST	—
CS_2_ST	2021	CS	Oakville	Napa	ST	Intentional Smoking
MA_NST	2021	MA	Dry Creek Valley	Sonoma	NST	—
MA_ST	2020	MA	Dry Creek Valley	Sonoma	ST	8/19/20–9/21/20 (5 days >150)
SY_NST	2021	SY	Dry Creek Valley	Sonoma	NST	—
SY_ST	2020	SY	Dry Creek Valley	Sonoma	ST	8/19/20–9/21/20 (5 days >150)

NST‐ indicates non‐smoke impacted wine, ST‐ indicates 100% smoke impacted wine.

AQI that was above 150 is in the unhealthy range and was used as a determination of high smoke‐exposure in the area.

The concentration of the serial dilutions of the wines was decided by a bench tasting (n = 3). The ST wine was sequentially blended with the NST wine in a series of 50% dilutions, starting from 100% ST. At each step, the proportion of ST was halved (i.e. 100%, 50%, 25% etc.) by volume until the ashy character was no longer perceptible in the blend. One further dilution was then carried out. The serial dilutions are summarized in Table 2. The choice of this set of dilutions was to have a wine between the non‐impacted wine and the lowest non‐ashy wine that was tasted in the experimental design. The blended wines were bottled in antique green bottles under screw cap (Saranex liner. Amcor, Zurich, Switzerland) and stored at 14°C until sensory analysis.

TABLE 2.

Serial dilutions of smoke‐impacted wine.

CS_1 & CS_2		MA & SY
Identifier	% ST wine	Identifier	% ST wine
A	0.00	A	0.00
B	1.56	B	3.12
C	3.12	C	6.25
D	6.25	D	12.50
E	12.50	E	25.00
F	25.00	F	50.00
ST	100.00	ST	100.00

Dilution levels were determined by a bench tasting, diluting the smoke impacted wine by a factor of 2 until no smoke characteristics were observed. One further dilution was made for the complete set of dilutions.

2.2. Chemical Analysis

2.2.1. Chemicals

The solvents (HPLC grade) and chemicals (analytical grade) were purchased from Sigma‐Aldrich and Merck (Darmstadt, Germany). Deuterium‐labeled standards were purchased from C/D/N Isotopes Inc. (Quebec, Canada), Toronto Research Chemicals (Toronto, Canada), and EPTES (Vevey, Switzerland).

2.2.2. Free VPs, Acid‐labile (Total) VPs and Phenolic Glycoside Analysis

The concentrations of VP glycosides were determined using SPE isolation(Caffrey et al. 2019) followed by liquid chromatography tandem mass spectrometry (LC‐MS/MS) using an Agilent 1290 Infinity UHPLC (Agilent Technologies, Santa Clara, CA) equipped with a binary pump, temperature‐controlled autosampler, and a thermostated column compartment. The procedure used was as described in Lim et al. (2024).

For the free VPs, liquid‐liquid extraction with pentane‐ethyl acetate (1:1) was used to quantify the guaiacol, creosol (4‐methylguaiacol), o‐cresol, phenol, 4‐ethylguaiacol, p‐cresol, m‐cresol, 4‐ethylphenol, syringol, and 4‐methylsyringol in samples. For total VPs (VP’s measured after hydrolysis), harsh acid hydrolysis followed by liquid‐liquid extraction with pentane‐ethyl acetate (1:1) as mentioned above in the free VPs was used to quantify the VPs (Oberholster et al. 2022). Calibration curves were constructed for all VPs (free and total). Deuterated VPs were used as internal standards.

For CS_2, MA and SY, an Agilent 7890B gas chromatograph equipped with a 5977B high‐efficiency source mass spectrometer (Agilent Technologies, Santa Clara, CA, USA) quantified the compounds above following the procedure described in Oberholster et al. (Oberholster et al. 2022).

For CS_1, an Agilent 7890A gas chromatograph coupled to an Agilent 7000B triple quadrupole mass spectrometer with an MPS 2 autosampler (Gerstel, Inc., Linthicum, MD) was used with the following conditions. A DB‐WAXetr fused silica capillary column with the dimensions 30 m length × 0.32 mm i.d. × 1.0 µm film thickness (Agilent Technologies, Santa Clara, CA, USA) was fitted onto the gas chromatogram. The inlet temperature was held constant at 220°C. Oven program was held at 75°C for 1 min initially. It was increased to 180°C at a rate of 15°C/min. It then increased to 230°C at a rate of 10°C/min and held for another 1 min. Finally, temperature was increased to 250°C at a rate of 50°C/min and held for 3 min. Total run time was 17.4 min. The GC and MS interface was held at 220°C. Pulsed splitless mode was used. The split vent was opened at 1 min with a flow of 50 mL/min. Helium carrier gas was used in constant flow mode at 2.0 mL/min. An electron ionization source, 70 eV, was fitted onto the triple quadrupole mass spectrometer.

The source was held at 230°C and reagent gas, helium was introduced to the source at 1 mL/min. The solvent delay was 7.5 min. Multiple reaction monitoring (MRM) quantitative and qualitative transitions and collision energies were chosen for each compound based on signal‐to‐noise ratios. The dwell times were set with 15 scans over each peak to ensure quantitative peak integration. The helium quench gas and nitrogen collision gas were fixed at 2.25 mL/min and 1.5 mL/min, respectively.

The limit of detection (LOD) and limit of quantification (LOQ) values are reported in Table T1 for volatile phenols and Table T2 for individual bound glycosides.

2.2.3. Basic Wine Analysis

The wines were analyzed at the UC Davis Teaching and Research Winery Laboratory. The titratable acidity (TA) in tartaric acid equivalents was measured using a Mettler‐Toledo DL50 titrator (Mettler‐ Toledo Inc., Columbus, OH, USA); pH was measured using an Orion 5‐star pH meter (Thermo Fisher Scientific, Waltham, MA, USA); ethanol content % (v/v) was measured with an alcohol analyzer (Anton Parr, Ashland, VA, USA); acetic acid, malic acid, and residual sugar (RS) were determined by enzymatic analysis using the Gallery automated analyzer (Thermo Fisher Scientific, Waltham, MA, USA).

2.3. Sensory Analysis

2.3.1. Expert and Consumer Recruitment

A total of 62 wine experts (29 females, 33 males, 21–65 years old) and 58 regular wine consumers (non‐experts, 32 females, 24 males, 2 non‐binary, 21–65 years old) participated in the sensory test in separate sessions. Wine experts took part in the sensory test as part of a research symposium. These experts were primarily winemakers, enologists, wine academics, and wine professionals, all of whom had experience in professional winetasting. Recruitment of non‐experts for the consumer test was carried out with Qualtrics survey software (Qualtrics International Inc., Provo, UT, USA) from the general population in and in proximity of Davis, California. Screening criteria were to be 21 years and older, to consume red wine at least once a week, to have no allergies, and to be available for the test time slots. Separately, fifty‐five non‐expert consumers (32 females, 23 males, 21–65 years old) who did not participate in the consumer test above were then recruited with the same criteria as above, for the difference test. Informed consent was sought from each participant before the commencement of the study.

The study and recruitment were conducted in accordance with the guidelines of the Declaration of Helsinki, with approval from the Institutional Review Board (IRB) of the University of California, UC Davis IRB Protocol 1288072‐1.

2.3.2. Expert and Consumer Test Administration and Questionnaire

Wine experts took part in the consumer test as part of a research symposium at the UC Davis Conference Center. Experts were seated spaced out about 1.5m from each other. Consumers evaluated the wines in the Silverado Vineyards Sensory Theater at the Robert Mondavi Institute for Wine and Food Science at UC Davis. The sensory theater is a pressure‐positive room and dividers were placed between consumers to create isolated booths and maintain the independence of evaluations. Both participant groups were served six labeled glasses of wine at the same time, arranged from left (lowest dilution‐identifier A) to right (highest dilution‐ identifier F) in the fixed order of tasting for all participants. The glasses were labeled A to F, denoting the order of tasting. This was to reduce the carry‐over effects of the “ashy” aftertaste of the wine. Approximately 30 mL of each wine was poured in a black Riedel wine glass (Riedel Crystal of America, Edison, NJ, USA). Participants were required to expectorate each sample after tasting. They were provided with unsalted saltine crackers (Nabisco, East Hanover, NJ, USA), plain water, and a 4 g/L dextrose solution. To minimize the carry‐over effects, each participant was instructed first to use the dextrose solution to rinse their mouths and then to wait 2 min between samples (Fryer et al. 2021; Oberholster et al. 2022). Plain water was used as needed by the participants.

The questionnaire was administered using Redjade (Redjade, Redwood City, CA, USA). Participants used their personal electronic device such as smartphones, laptops, or tablets to fill out the survey. Tablets were available to participants should they have required one. For each wine, participants first rated overall opinion, and liking of flavor (taste and aroma), mouthfeel, and alcohol level using the 9‐point hedonic scale (Jones et al. 1955). Next, they used a 5‐point Just‐About‐Right (JAR) scale to evaluate the adequacy of attributes such as flavor intensity (too weak—too strong), acidity (too low—too high), sweetness (too low—too high), bitterness (too low—too high), body/viscosity (too little—too much), and astringency (too low—too high) (Lawless and Heymann 2010). Next a Check‐All‐That‐Apply (CATA) list of 18 attributes (red fruit, dark fruit, jam fruit, dried fruit, bell pepper, cooked vegetable, baking spice, black pepper, cigarette smoke, musty, medicinal/Brett, liquid smoke, sweet BBQ, menthol, solvent, tar, ashy aftertaste, alcohol hotness) was presented to participants and they were asked to check the attributes that described the wine. This list of 18 attributes was selected from a previous descriptive analysis of the non‐impacted and 100% smoke‐impacted wines (Lim et al. 2024). Finally, participants were asked two open‐ended questions on what specifically they liked or disliked about the wine.

After evaluating the six samples, participants completed an exit survey in which they answered questions about their demographics. Wine experts were also asked about their years of wine‐tasting experience. Consumers were asked questions regarding wine consumption habits, and where they regularly bought wine.

2.3.3. Difference Test Administration and Questionnaire

Difference testing was conducted with (non‐expert) consumers on the Malbec wines in individual red‐light booths in a pressure‐positive sensory lab. Malbec was chosen as the variety as it was the only wine that had sufficient volume available for the test. Ultimately, we sought to use this as a proof of concept. Consumers were recruited following the same criteria as stated above for the consumer test. Participants only knew they were doing a wine consumer test and could not be participants in the previous test. When participants arrived, they were first briefed on the overall flow of the session. First, a series of four triangle tests was performed, followed by a short break, and then a series of four 3‐Alternative Forced Choice (3‐AFC) tests. In both the triangle test and 3‐AFC test, a warm‐up set was included. During the short break, participants were trained using the “ashy” standard prepared by using burnt leeks and water to get an “ashy” solution, as stated in Lim et al. (Lim et al. 2024). This was to familiarize and expose them to the “ashy” attribute which was needed to help them identify smoke‐impacted wines. The “ashy” attribute was described as the retronasal ashtray character after tasting or consuming a smoke‐tainted wine (Fryer and Tomasino 2022). It was chosen as it appears to be the single unique characteristic of smoke‐tainted wines (Parker et al. 2020). All the wines were served in black Riedel wine glasses of 30 mL per glass with 3‐digit blinding code. Each participant had saltines, plain water, and the dextrose rinse (4 g/L) provided. There was a 2‐min enforced break between sets of wine before participants were given the next set of samples to prevent carryover effects (Fryer et al. 2021).

For the triangle test, a balanced triangle test design was used where all six combinations were presented at random. For each combination shown, there was always the non‐impacted wine; either two non‐impacted and one test (impacted) wine or one non‐impacted and two test (impacted) wines (MacRae 1995). The order of presentation of the wines in each triangle test was randomized across participants. Participants were served the test triangles in an increasing concentration of the smoke‐impacted wines for a total of four sets of wines (Rogers 2017; Marques et al. 2022). A warm‐up exercise was conducted first using two different commercially available boxed red wines.

In the 3‐AFC test, participants were asked which wine was the “most smoky/most ashy.” Participants always had two non‐impacted wines among the three wines served. The serving order of each set was the same for each participant. Test samples were similarly presented in an increasing order of smoke impact over four sets of wines. Similarly, a warmup exercise was done using regular mineral water and a 4 g/L dextrose solution, asking which was the “most sweet.”

Finally, an exit survey was filled out which asked the same questions as in the consumer test described above.

2.4. Statistical Analysis

All statistical analyses were conducted using R, version 4.2.1 “Funny‐Looking Kid” (R Core Team 2022). Type I error α = 0.05 was used throughout.

2.4.1. Analysis of Variance and Post‐Hoc Tests

First a multivariate analysis of variance (MANOVA) was performed across overall liking, and liking of flavor (taste & smell), mouthfeel, and alcohol level to determine if there were differences in liking across the wines and among consumers using the “Wilks Lamda” test (Todorov and Filzmoser 2010). Next univariate ANOVAs was performed to determine differences in liking of the wines for overall liking, and liking of flavor (taste & smell), mouthfeel and alcohol level, respectively. Using the “agricolae” package, Fisher's least significant difference (LSD) was used to assess individual differences in liking among the wines, when there was a significant difference in the univariate ANOVA.

Univariate ANOVA and Fisher's LSD were used to determine differences in the chemical differences of the wines for basic chemical composition, free VPs, total VPs, and VP glycosides.

2.4.2. Analysis of JAR Data

Penalty analysis was used to relate overall liking to the JAR attributes of flavor intensity, acidity, sweetness, bitterness, body/viscosity, and astringency. First, the average overall liking score of the JAR and the non‐JAR levels on the 9‐point hedonic scale were computed. The “penalty” was computed as the difference between average scores of the JAR and non‐JAR levels. The percentage of consumers selecting each non‐JAR category was calculated next. And then, the “penalty” was plotted against the percentages of the non‐JAR attributes (Rothman and Parker 2009). Generally, any attribute with a penalty of 1 point or more, judged as non‐JAR by more than 20% of the consumer population is considered as an important driver of liking.

2.4.3. Analysis of CATA Data

A sensory map of the wines was constructed using correspondence analysis (CA) of the CATA attributes selected by the experts and non‐experts. To determine the associations between the wines and the CATA descriptors, chi‐squared distances were calculated. Cochran's Q‐test was also used to determine which CATA attributes were significant among the wines tested for each test (Halim et al. 2020). The CATA attributes were also linked to the overall liking using a penalty/lift analysis (Meyners 2016).

2.4.4. Preference Mapping

An internal preference map was constructed from the matrix of the experts or non‐experts overall liking scores across the wines evaluated. This was a principal component analysis (PCA) based on the covariance matrix, where the experts/non‐experts were the variables and the wines were the objects (Guinard 2002; Lawless and Heymann 2010).

2.4.5. Difference Test Data

The total number of correct responses was computed for each sample in both the triangle test and 3‐AFC test. The p‐value was calculated for a one‐tailed test (Lawless and Heymann 2010).

3. Results

3.1. Chemical Analysis

The different dilution wines—A to F, as shown in Table 2, had small differences in their wine chemical makeup across all sets (Table 3) which were within levels that would not have a significant sensory impact (King and Heymann 2014). The wines were all dry (residual sugar <2.0 g/L), alcohol content differences were within 0.5% v/v of each other, and titratable acidity and pH were similar to each other. The wines evaluated were not different chemically overall from a basic wine chemistry standpoint.

TABLE 3.

Basic chemical composition of the wines.

Wine	Taint %	Alcohol (v/v%)	pH	Titratable acidity, g/L	Residual sugar, g/L	Malic acid, mg/L	Acetic acid, g/L
CS_1
CS_1_A	0.00	14.68 ± 0.02 a	3.79 ± 0.01 b	5.98 ± 0.01 a	0.09 ± 0.07 b	49.50 ± 70.00 a	0.30 ± 0.00 b
CS_1_B	1.56	14.70 ± 0.03 a	3.80 ± 0.00 b	5.90 ± 0.00 b	0.15 ± 0.07 ab	0.00 ± 0.00 a	0.30 ± 0.01 b
CS_1_C	3.12	14.66 ± 0.01 ab	3.8 ± 0.01 b	5.90 ± 0.00 b	0.20 ± 0.00 a	0.00 ± 0.00 a	0.29 ± 0.01 b
CS_1_D	6.25	14.61 ± 0.03 b	3.8 ± 0.01 b	5.80 ± 0.00 c	0.20 ± 0.00 a	0.00 ± 0.00 a	0.30 ± 0.01 b
CS_1_E	12.50	14.49 ± 0.01 c	3.80 ± 0.00 b	5.80 ± 0.00 c	0.20 ± 0.00 a	0.00 ± 0.00 a	0.31 ± 0.01 ab
CS_1_F	25.00	14.24 ± 0.01 d	3.82 ± 0.00 b	5.70 ± 0.00 d	0.20 ± 0.00 a	0.00 ± 0.00 a	0.32 ± 0.00 a
CS_1_ST	100.00	14.15 ± 0.01 *	3.94 ± 0.01 *	5.18 ± 0.07 *	0.25 ± 0.01 *	7.00 ± 9.90 *	0.45 ± 0.04 *
CS_2
CS_2_A	0.00	15.3 ± 0.04 ab	3.49 ± 0.01 a	6.60 ± 0.00 a	0.20 ± 0.00 a	25.00 ± 35.36 a	0.34 ± 0.01 a
CS_2_B	1.56	15.27 ± 0.01 b	3.49 ± 0.00 a	6.60 ± 0.00 a	0.20 ± 0.00 a	0.00 ± 0.00 a	0.35 ± 0.00 a
CS_2_C	3.12	15.28 ± 0.01 b	3.49 ± 0.00 a	6.65 ± 0.07 a	0.20 ± 0.00 a	25 ± 35.36 a	0.34 ± 0.00 a
CS_2_D	6.25	15.27 ± 0.01 b	3.49 ± 0.00 a	6.60 ± 0.00 a	0.20 ± 0.00 a	25 ± 35.36 a	0.35 ± 0.00 a
CS_2_E	12.50	15.31 ± 0.01 ab	3.50 ± 0.01 a	6.60 ± 0.00 a	0.20 ± 0.00 a	50.00 ± 0.00 a	0.35 ± 0.01 a
CS_2_F	25.00	15.33 ± 0.01 a	3.49 ± 0.01 a	6.60 ± 0.00 a	0.20 ± 0.00 a	6.60 ± 0.00 a	0.35 ± 0.01 a
CS_2_ST	100.00	15.44 ± 0.02 *	3.46 ± 0.00 *	6.82 ± 0.11 *	0.18 ± 0.01 *	101.50 ± 3.54 *	0.34 ± 0.04 *
MA
MA_A	0.00	15.12 ± 0.01 d	3.31 ± 0.01 de	7.40 ± 0.00 a	0.40 ± 0.00 a	0.00 ± 0.00 a	0.25 ± 0.01 c
MA_B	3.12	15.17 ± 0.02 bc	3.34 ± 0.01 c	7.40 ± 0.00 a	0.40 ± 0.00 a	0.00 ± 0.00 a	0.27 ± 0.01 bc
MA_C	6.25	15.13 ± 0.00 cd	3.33 ± 0.01 cd	7.40 ± 0.00 a	0.40 ± 0.00 a	0.00 ± 0.00 a	0.26 ± 0.00 bc
MA_D	12.50	15.10 ± 0.01 d	3.32 ± 0.01 e	7.40 ± 0.00 a	0.40 ± 0.00 a	0.00 ± 0.00 a	0.25 ± 0.00 c
MA_E	25.00	15.18 ± 0.02 b	3.38 ± 0.00 b	7.40 ± 0.00 a	0.40 ± 0.00 a	0.00 ± 0.00 a	0.27 ± 0.00 b
MA_F	50.00	15.25 ± 0.01 a	3.46 ± 0.01 a	6.90 ± 0.00 a	0.50 ± 0.00 a	0.00 ± 0.00 a	0.29 ± 0.00 a
MA_ST	100.00	15.53 ± 0.03 *	3.64 ± 0.01 *	6.62 ± 0.05 *	0.62 ± 0.00 *	0.00 ± 0.00 *	0.35 ± 0.00 *
SY
SY_A	0.00	13.83 ± 0.01 e	3.64 ± 0.01 a	6.40 ± 0.00 e	0.00 ± 0.00 a	0.00 ± 0.00 a	0.26 ± 0.00 e
SY_B	3.12	13.86 ± 0 d	3.63 ± 0.00 a	6.50 ± 0.00 d	0.00 ± 0.00 a	0.00 ± 0.00 a	0.27 ± 0.00 d
SY_C	6.25	13.86 ± 0.01 d	3.62 ± 0.00 b	6.50 ± 0.00 d	0.10 ± 0.00 a	0.00 ± 0.00 a	0.26 ± 0.00 e
SY_D	12.50	13.91 ± 0.01 c	3.61 ± 0.01 c	6.65 ± 0.07 c	0.10 ± 0.00 a	0.00 ± 0.00 a	0.28 ± 0.00 c
SY_E	25.00	14.00 ± 0.00 b	3.58 ± 0.00 d	6.80 ± 0.00 b	0.20 ± 0.00 a	0.00 ± 0.00 a	0.29 ± 0.00 b
SY_F	50.00	14.16 ± 0.01 a	3.51 ± 0.00 e	7.05 ± 0.07 a	0.40 ± 0.00 a	0.00 ± 0.00 a	0.34 ± 0.01 a
SY_ST	100.00	14.76 ± 0.03 *	3.42 ± 0.01 *	7.56 ± 0.06 *	0.99 ± 0.03 *	219 ± 90.51 *	0.44 ± 0.01 *

Wines analysis were done in triplicate reps (n = 3) with standard deviations reported, Fisher's LSD for means comparison of all samples across each chemical parameter. Letters that are different indicate a significant difference, p ≤ 0.05 in each set of wines.

^*Was not included in the L.S.D test of significant difference.

3.2. Free and Total Volatile Phenols, and Individual Bound Glycosides

Across all the free VPs (Table 4), total VPs (Table 5), and individual bound glycosides data (Table 6), the wines tested, A‐F were significantly different among their own set. There was an increasing concentration of free VPs, total VPs, and individual‐bound glycosides across all wines with increasing percentages of ST wine included. In the individual bound glycosides, CS_2 was observed to have low concentrations of the 8 tested glycosides as compared to the other 3 wines, CS_1, MA, and SY. Yet, the concentration of total VPs and free VPs in CS_2 were about the same or higher than CS_1, MA and SY.

TABLE 4.

Free volatile phenol mean concentrations (µg/L) of the wines.

Wine	Taint %	Guaiacol	4‐methylguaiacol	o‐cresol	Phenol	4‐ethylguaiacol	p‐cresol	m‐cresol	4‐ethylphenol	Syringol	4‐methylsyringol
CS_1
CS_1_A	0.00	0.784 ± 0.029 f	0.111 ± 0.004 f	0.555 ± 0.022 f	2.925 ± 0.287 e	0.015 ± 0.003 e	0.792 ± 0.016 f	0.427 ± 0.022 f	0.163 ± 0.011 d	11.349 ± 0.487 e	<LOD
CS_1_B	1.56	1.124 ± 0.013 e	0.203 ± 0.002 e	0.972 ± 0.011 e	3.475 ± 0.133 e	0.043 ± 0.003 d	1.206 ± 0.034 e	0.925 ± 0.022 e	<LOD	14.116 ± 0.158 d	<LOD
CS_1_C	3.12	1.646 ± 0.003 d	0.325 ± 0.004 d	1.347 ± 0.002 d	5.131 ± 0.051 d	0.046 ± 0.001 d	1.547 ± 0.006 d	1.421 ± 0.037 d	<LOD	14.513 ± 0.136 d	<LOD
CS_1_D	6.25	2.610 ± 0.016 c	0.538 ± 0.006 c	1.984 ± 0.019 c	7.935 ± 0.233 c	0.060 ± 0.002 c	2.136 ± 0.016 c	2.323 ± 0.021 c	0.329 ± 0.003 c	15.168 ± 0.051 c	<LOD
CS_1_E	12.50	4.643 ± 0.034 b	0.963 ± 0.013 b	3.298 ± 0.039 b	14.365 ± 1.010 b	0.087 ± 0.003 b	3.278 ± 0.084 b	4.086 ± 0.076 b	0.527 ± 0.006 b	16.03 ± 0.127 b	<LOD
CS_1_F	25.00	8.261 ± 0.067 a	1.753 ± 0.008 a	5.757 ± 0.007 a	24.713 ± 0.123 a	0.141 ± 0.001 a	5.429 ± 0.024 a	7.479 ± 0.027 a	0.874 ± 0.009 a	17.454 ± 0.099 a	<LOD
CS_1_ST	100.00	33.203 ± 0.122*	5.592 ± 0.116*	20.925 ± 0.228*	72.418 ± 0.232*	0.437 ± 0.005*	16.704 ± 0.129*	22.169 ± 0.203*	2.924 ± 0.032*	18.851 ± 0.066*	1.803 ± 0.019*
CS_2
CS_2_A	0.00	0.736 ± 0.013 f	<LOD	0.522 ± 0.003 f	2.209 ± 0.031 e	<LOD	0.219 ± 0.011 d	0.257 ± 0.002 f	<LOD	9.875 ± 0.141 f	<LOD
CS_2_B	1.56	1.375 ± 0.007 e	0.373 ± 0.026 e	0.801 ± 0.013 e	2.222 ± 0.014 e	0.042 ± 0.001 e	0.239 ± 0.005 c	0.746 ± 0.007 e	<LOD	12.208 ± 0.068 e	0.400 ± 0.024 e
CS_2_C	3.12	2.291 ± 0.013 d	0.786 ± 0.010 d	1.141 ± 0.006 d	2.804 ± 0.018 d	0.121 ± 0.002 d	0.274 ± 0.007 b	1.216 ± 0.004 d	<LOD	13.985 ± 0.054 d	0.959 ± 0.039 d
CS_2_D	6.25	3.864 ± 0.013 c	1.640 ± 0.031 c	1.787 ± 0.014 c	3.136 ± 0.012 c	0.226 ± 0.004 c	0.286 ± 0.003 b	2.107 ± 0.031 c	0.109 ± 0.005 c	17.826 ± 0.291 c	1.982 ± 0.014 c
CS_2_E	12.50	7.391 ± 0.008 b	3.407 ± 0.016 b	3.159 ± 0.003 b	4.466 ± 0.035 b	0.477 ± 0.004 b	0.309 ± 0.004 a	4.089 ± 0.02 b	0.388 ± 0.017 b	25.781 ± 0.202 b	4.211 ± 0.030 b
CS_2_F	25.00	14.132 ± 0.061 a	6.537 ± 0.027 a	5.656 ± 0.037 a	7.382 ± 0.026 a	0.933 ± 0.002 a	0.249 ± 0.009 c	7.559 ± 0.028 a	0.907 ± 0.022 a	40.790 ± 0.804 a	8.171 ± 0.046 a
CS_2_ST	100.00	57.529 ± 0.017*	28.099 ± 0.02*	21.555 ± 0.300*	23.753 ± 0.064*	3.912 ± 0.034*	0.571 ± 0.015*	29.743 ± 0.006*	4.583 ± 0.035*	135.206 ± 0.235*	34.462 ± 0.032*
MA
MA_A	0.00	0.920 ± 0.010 f	0.058 ± 0.001 f	0.455 ± 0.022 f	1.300 ± 0.039 f	<LOD	0.352 ± 0.011 f	0.286 ± 0.008 f	< LOD	11.654 ± 0.333 f	<LOD
MA_B	3.12	1.41 ± 0.005 e	0.238 ± 0.006 e	0.688 ± 0.005 e	2.340 ± 0.023 e	<LOD	0.620 ± 0.018 e	0.637 ± 0.018 e	<LOD	13.535 ± 0.167 e	<LOD
MA_C	6.25	1.934 ± 0.041 d	0.429 ± 0.014 d	0.969 ± 0.023 d	3.710 ± 0.059 d	<LOD	0.851 ± 0.006 d	0.956 ± 0.038 d	<LOD	14.190 ± 0.133 d	<LOD
MA_D	12.50	2.787 ± 0.047 c	0.782 ± 0.021 c	1.377 ± 0.008 c	5.791 ± 0.029 c	0.010 ± 0.002 c	1.225 ± 0.01 c	1.509 ± 0.010 c	0.109 ± 0.005 c	15.238 ± 0.156 c	0.021 ± 0.001 c
MA_E	25.00	4.836 ± 0.014 b	1.595 ± 0.016 b	2.391 ± 0.001 b	10.986 ± 0.037 b	0.053 ± 0.002 b	2.148 ± 0.005 b	2.842 ± 0.017 b	0.349 ± 0.017 b	17.851 ± 0.035 b	0.103 ± 0.007 b
MA_F	50.00	7.227 ± 0.015 a	2.670 ± 0.025 a	3.602 ± 0.016 a	17.042 ± 0.1 a	0.111 ± 0.002 a	3.255 ± 0.004 a	4.544 ± 0.037 a	0.678 ± 0.025 a	19.581 ± 0.091 a	0.219 ± 0.005 a
MA_ST	100.00	16.245 ± 0.103*	6.475 ± 0.039*	8.326 ± 0.011*	38.567 ± 0.403*	0.302 ± 0.002*	7.212 ± 0.091*	9.962 ± 0.136*	2.008 ± 0.009*	31.712 ± 0.532*	0.617 ± 0.003*
SY
SY_A	0.00	8.020 ± 0.018 f	0.012 ± 0.001 f	0.931 ± 0.041 f	1.381 ± 0.011 f	<LOD	0.338 ± 0.006 f	0.220 ± 0.004 f	<LOD	9.866 ± 0.083 f	<LOD
SY_B	3.12	8.820 ± 0.020 e	0.177 ± 0.004 e	1.159 ± 0.009 e	3.047 ± 0.026 e	<LOD	0.541 ± 0.008 e	0.565 ± 0.011 e	<LOD	10.730 ± 0.035 e	<LOD
SY_C	6.25	9.298 ± 0.030 d	0.326 ± 0.003 d	1.366 ± 0.022 d	4.551 ± 0.023 d	<LOD	0.688 ± 0.02 d	0.844 ± 0.017 d	0.030 ± 0.003 d	11.282 ± 0.207 d	<LOD
SY_D	12.50	10.498 ± 0.146 c	0.648 ± 0.012 c	1.843 ± 0.017 c	7.616 ± 0.014 c	<LOD	1.103 ± 0.008 c	1.514 ± 0.027 c	0.134 ± 0.004 c	12.908 ± 0.101 c	0.05 ± 0.007 c
SY_E	25.00	12.755 ± 0.176 b	1.262 ± 0.014 b	2.736 ± 0.038 b	13.674 ± 0.084 b	0.008 ± 0.001 b	1.880 ± 0.017 b	2.670 ± 0.014 b	0.320 ± 0.016 b	16.058 ± 0.007 b	0.112 ± 0.003 b
SY_F	50.00	17.232 ± 0.083 a	2.624 ± 0.021 a	4.463 ± 0.028 a	24.995 ± 0.051 a	0.045 ± 0.003 a	3.473 ± 0.019 a	5.311 ± 0.032 a	0.738 ± 0.038 a	22.565 ± 0.231 a	0.282 ± 0.004 a
SY_ST	100.00	25.273 ± 0.315*	4.777 ± 0.068*	7.188 ± 0.048*	46.601 ± 0.472*	0.112 ± 0.002*	5.948 ± 0.110*	9.169 ± 0.006*	1.546 ± 0.039*	30.432 ± 0.609*	0.544 ± 0.008*

Wines analysis were done in triplicate reps (n = 3) with standard deviations reported, Fisher's LSD for means comparison of all samples across each chemical parameter. Letters that are different indicate a significant difference, p ≤ 0.05 in each set of wines.

<LOD indicates compound is below the detection limit of the instrument. ^*was not included in the L.S.D test of significant difference.

TABLE 5.

Acid‐labile (total) volatile phenol mean concentrations (µg/L) of wines as determined by acid hydrolysis.

Wine	Taint %	Guaiacol	4‐methylguaiacol	o‐cresol	Phenol	4‐ethylguaiacol	p‐cresol	m‐cresol	4‐ethylphenol	Syringol	4‐methylsyringol
CS_1
CS_1_A	0.00	6.184 ± 0.046 d	0.938 ± 0.020 b	2.389 ± 0.002 e	25.453 ± 0.168 e	0.585 ± 0.013 d	4.166 ± 0.072 e	3.340 ± 0.009 f	17.537 ± 0.243 b	22.760 ± 1.023 f	0.32 ± 0.001 e
CS_1_B	1.56	6.916 ± 0.239 d	0.871 ± 0.050 b	4.003 ± 0.075 d	37.314 ± 2.574 d	0.663 ± 0.021 c	6.991 ± 0.448 d	5.878 ± 0.018 e	17.815 ± 0.414 b	33.854 ± 0.145 e	11.211 ± 0.419 d
CS_1_C	3.12	7.799 ± 0.227 d	0.813 ± 0.036 b	4.680 ± 0.067 d	34.601 ± 1.165 d	0.603 ± 0.010 cd	7.484 ± 0.280 d	7.317 ± 0.135 d	15.162 ± 0.807 c	42.073 ± 0.735 d	15.222 ± 1.347 d
CS_1_D	6.25	10.394 ± 0.643 c	0.823 ± 0.023 b	6.632 ± 0.282 c	44.597 ± 5.648 c	0.626 ± 0.011 cd	10.283 ± 1.161 c	10.877 ± 0.831 c	15.518 ± 1.206 c	56.236 ± 0.208 c	26.443 ± 5.007 c
CS_1_E	12.50	15.617 ± 0.317 b	0.841 ± 0.018 b	10.840 ± 0.230 b	61.854 ± 2.555 b	0.842 ± 0.062 b	15.339 ± 0.692 b	18.613 ± 0.509 b	18.539 ± 1.356 b	85.209 ± 0.651 b	44.515 ± 2.818 b
CS_1_F	25.00	25.934 ± 2.142 a	12.154 ± 0.417 a	19.337 ± 1.702 a	92.294 ± 3.989 a	1.035 ± 0.057 a	24.016 ± 0.939 a	31.893 ± 1.177 a	24.050 ± 0.162 a	141.071 ± 1.280 a	85.939 ± 0.693 a
CS_1_ST	100.00	84.382 ± 1.042 *	19.660 ± 1.009 *	33.417 ± 1.502 *	112.163 ± 4.062 *	1.298 ± 0.043 *	33.455 ± 1.752 *	47.758 ± 2.541 *	14.135 ± 1.703 *	415.379 ± 3.397 *	211.291 ± 0.557 *
CS_2
CS_2_A	0.00	2.14 ± 0.013 f	0.02 ± 0.001 f	0.978 ± 0.025 f	5.128 ± 0.022 f	0.053 ± 0.002 f	5.419 ± 0.324 a	1.115 ± 0.024 f	1.082 ± 0.016 f	16.999 ± 0.112 f	0.638 ± 0.015 f
CS_2_B	1.56	3.045 ± 0.019 e	0.403 ± 0.003 e	1.392 ± 0.023 e	5.814 ± 0.029 e	0.126 ± 0.005 e	6.326 ± 0.381 a	1.823 ± 0.003 e	1.674 ± 0.031 e	20.487 ± 0.364 e	1.776 ± 0.030 e
CS_2_C	3.12	4.173 ± 0.018 d	0.912 ± 0.010 d	1.812 ± 0.02 d	6.735 ± 0.018 d	0.193 ± 0.003 d	6.037 ± 0.684 a	2.372 ± 0.058 d	1.845 ± 0.042 d	22.397 ± 0.349 d	3.220 ± 0.043 d
CS_2_D	6.25	5.878 ± 0.022 c	1.768 ± 0.028 c	2.495 ± 0.024 c	6.940 ± 0.053 c	0.295 ± 0.027 c	6.063 ± 0.526 a	3.343 ± 0.036 c	2.068 ± 0.051 c	26.976 ± 0.231 c	5.576 ± 0.043 c
CS_2_E	12.50	9.851 ± 0.027 b	3.587 ± 0.048 b	3.975 ± 0.028 b	8.261 ± 0.024 b	0.556 ± 0.004 b	6.132 ± 0.617 a	5.446 ± 0.044 b	2.388 ± 0.028 b	37.694 ± 0.287 b	10.549 ± 0.054 b
CS_2_F	25.00	16.785 ± 0.055 a	7.142 ± 0.011 a	6.598 ± 0.008 a	11.105 ± 0.019 a	1.074 ± 0.032 a	5.954 ± 0.435 a	8.940 ± 0.033 a	2.988 ± 0.030 a	55.017 ± 0.154 a	19.434 ± 0.073 a
CS_2_ST	100.00	65.113 ± 0.137 *	28.289 ± 0.043 *	22.891 ± 0.276 *	28.557 ± 0.306 *	4.544 ± 0.028 *	4.926 ± 0.081 *	30.052 ± 0.133 *	6.257 ± 0.073 *	169.813 ± 1.001 *	78.203 ± 0.242 *
MA
MA_A	0.00	3.530 ± 0.044 f	0.829 ± 0.016 f	1.143 ± 0.022 f	6.074 ± 0.075 e	0.071 ± 0.00 f	5.109 ± 0.285 e	1.170 ± 0.042 f	2.261 ± 0.056 f	19.999 ± 0.151 f	0.946 ± 0.023 f
MA_B	3.12	5.917 ± 0.519 e	1.889 ± 0.261 e	1.804 ± 0.180 e	11.556 ± 2.389 d	0.157 ± 0.004 e	5.229 ± 0.760 e	2.246 ± 0.049 e	3.072 ± 0.274 e	24.981 ± 0.411 e	1.930 ± 0.078 e
MA_C	6.25	7.941 ± 0.027 d	2.943 ± 0.047 d	2.435 ± 0.020 d	13.083 ± 0.026 d	0.267 ± 0.002 d	6.998 ± 0.779 d	3.019 ± 0.008 d	4.027 ± 0.057 d	30.594 ± 0.373 d	3.176 ± 0.076 d
MA_D	12.50	11.286 ± 0.060 c	5.072 ± 0.022 c	3.321 ± 0.017 c	17.118 ± 0.119 c	0.306 ± 0.003 c	8.650 ± 0.468 c	4.369 ± 0.010 c	4.424 ± 0.117 c	36.868 ± 0.183 c	4.817 ± 0.036 c
MA_E	25.00	18.761 ± 0.044 b	9.029 ± 0.154 b	5.242 ± 0.025 b	25.497 ± 0.072 b	0.490 ± 0.002 b	12.014 ± 0.299 b	7.119 ± 0.100 b	5.265 ± 0.110 b	54.859 ± 0.264 b	8.823 ± 0.053 b
MA_F	50.00	30.479 ± 0.210 a	15.117 ± 0.005 a	8.219 ± 0.012 a	39.053 ± 0.058 a	0.710 ± 0.001 a	15.697 ± 0.419 a	11.317 ± 0.218 a	6.771 ± 0.010 a	77.995 ± 0.769 a	14.328 ± 0.288 a
MA_ST	100.00	61.368 ± 0.072 *	30.832 ± 0.217 *	15.789 ± 0.084 *	73.290 ± 0.341 *	1.448 ± 0.026 *	23.327 ± 0.308 *	20.817 ± 0.135 *	10.459 ± 0.066 *	153.718 ± 0.406 *	31.328 ± 0.069 *
SY
SY_A	0.00	44.11 ± 0.209 f	0.676 ± 0.015 f	2.418 ± 0.008 f	6.556 ± 0.017 f	0.024 ± 0.002 f	5.835 ± 0.105 f	0.920 ± 0.011 f	2.374 ± 0.051 f	17.730 ± 0.147 f	0.964 ± 0.010 f
SY_B	3.12	49.129 ± 0.277 e	1.805 ± 0.013 e	3.373 ± 0.019 e	9.354 ± 0.009 e	0.132 ± 0.002 e	7.384 ± 0.143 e	2.002 ± 0.027 e	3.435 ± 0.060 e	21.944 ± 0.154 e	1.985 ± 0.127 e
SY_C	6.25	51.325 ± 0.421 d	2.645 ± 0.048 d	3.832 ± 0.032 d	11.555 ± 0.054 d	0.176 ± 0.004 d	8.141 ± 0.142 d	2.658 ± 0.034 d	3.785 ± 0.034 d	25.349 ± 0.342 d	2.797 ± 0.068 d
SY_D	12.50	55.641 ± 0.352 c	4.498 ± 0.041 c	5.008 ± 0.024 c	16.867 ± 0.044 c	0.209 ± 0.001 c	9.725 ± 0.129 c	4.137 ± 0.033 c	4.327 ± 0.081 c	33.026 ± 0.148 c	4.716 ± 0.018 c
SY_E	25.00	65.298 ± 0.194 b	8.700 ± 0.113 b	7.349 ± 0.020 b	26.671 ± 0.184 b	0.301 ± 0.007 b	12.948 ± 0.144 b	7.403 ± 0.165 b	6.411 ± 0.034 b	49.179 ± 0.450 b	8.186 ± 0.145 b
SY_F	50.00	81.183 ± 0.538 a	16.079 ± 0.165 a	10.951 ± 0.205 a	43.481 ± 0.437 a	0.449 ± 0.005 a	18.353 ± 0.431 a	13.052 ± 0.414 a	8.508 ± 0.034 a	77.722 ± 0.285 a	14.960 ± 0.117 a
SY_ST	100.00	111.218 ± 3.657 *	26.282 ± 0.529 *	16.300 ± 0.231 *	77.367 ± 3.005 *	0.612 ± 0.010 *	21.51 ± 0.471 *	19.372 ± 0.102 *	9.639 ± 0.531 *	126.483 ± 0.348 *	25.791 ± 0.333 *

Wines analysis were done in triplicate reps (n = 3) with standard deviations reported, Fisher's LSD for means comparison of all samples across each chemical parameter. Letters that are different indicate a significant difference, p ≤ 0.05 in each set of wines.

<LOD indicates compound is below the detection limit of the instrument.

^* Was not included in the L.S.D test of significant difference.

TABLE 6.

Individual bound glycoside mean concentrations (µg/L) of wines.

Wine	Taint %	Guaiacol gentiobioside	Syringol gentiobioside	Guaiacol glucoside	Phenol rutinoside	Guaiacol rutinoside	4‐methylsyringol gentiobioside	Cresol rutinoside	4‐methylguaiacol rutinoside
CS_1
CS_1_A	0.00	0.019 ± 0.002 f	1.335 ± 0.087 f	0.549 ± 0.005 f	1.278 ± 0.002 f	0.319 ± 0.017 f	0.107 ± 0.006 f	1.245 ± 0.084 f	0.496 ± 0.032 f
CS_1_B	1.56	0.064 ± 0.002 e	6.582 ± 0.048 e	1.091 ± 0.029 e	2.041 ± 0.040 e	0.916 ± 0.053 e	0.866 ± 0.044 e	1.884 ± 0.077 e	0.924 ± 0.053 e
CS_1_C	3.12	0.193 ± 0.012 d	13.136 ± 0.189 d	1.843 ± 0.043 d	3.402 ± 0.060 d	1.406 ± 0.066 d	1.784 ± 0.085 d	2.921 ± 0.039 d	1.384 ± 0.058 d
CS_1_D	6.25	0.241 ± 0.013 c	24.133 ± 0.451 c	2.894 ± 0.046 c	5.235 ± 0.160 c	2.632 ± 0.037 c	3.327 ± 0.105 c	4.028 ± 0.053 c	2.073 ± 0.055 c
CS_1_E	12.50	0.342 ± 0.018 b	48.174 ± 1.033 b	5.180 ± 0.154 b	8.744 ± 0.175 b	4.696 ± 0.057 b	6.601 ± 0.044 b	6.426 ± 0.231 b	3.352 ± 0.091 b
CS_1_F	25.00	0.740 ± 0.036 a	97.713 ± 1.332 a	9.576 ± 0.123 a	15.835 ± 0.468 a	8.656 ± 0.183 a	13.028 ± 0.294 a	11.554 ± 0.223 a	5.910 ± 0.103 a
CS_1_ST	100.00	3.034 ± 0.085 *	448.29 ± 3.398 *	33.678 ± 0.199 *	53.508 ± 0.453 *	32.841 ± 0.162 *	48.610 ± 0.563 *	42.495 ± 0.482 *	20.88 ± 0.293 *
CS_2
CS_2_A	0.00	0.059 ± 0.007 d	2.208 ± 0.043 c	0.405 ± 0.02 f	2.252 ± 0.075 bc	0.529 ± 0.037 ab	0.160 ± 0.007 e	1.121 ± 0.042 c	0.313 ± 0.008 e
CS_2_B	1.56	0.010 ± 0.001 f	1.887 ± 0.033 d	0.67 ± 0.011 e	2.036 ± 0.036 cd	0.505 ± 0.018 b	0.164 ± 0.008 e	1.092 ± 0.020 c	0.377 ± 0.002 d
CS_2_C	3.12	0.018 ± 0.001 e	2.283 ± 0.038 c	1.043 ± 0.034 d	2.408 ± 0.176 ab	0.538 ± 0.011 ab	0.250 ± 0.009 d	1.108 ± 0.040 c	0.402 ± 0.025 d
CS_2_D	6.25	0.067 ± 0.003 c	2.498 ± 0.092 b	1.673 ± 0.067 c	2.513 ± 0.18 ab	0.509 ± 0.007 ab	0.313 ± 0.023 c	1.214 ± 0.035 ab	0.447 ± 0.017 c
CS_2_E	12.50	0.085 ± 0.004 b	2.318 ± 0.102 c	2.413 ± 0.041 b	1.864 ± 0.112 d	0.501 ± 0.049 b	0.388 ± 0.009 b	1.173 ± 0.007 b	0.484 ± 0.031 b
CS_2_F	25.00	0.129 ± 0.001 a	3.331 ± 0.078 a	5.504 ± 0.080 a	2.287 ± 0.191 ab	0.569 ± 0.046 ab	0.704 ± 0.005 a	1.23 ± 0.007 ab	0.574 ± 0.024 a
CS_2_ST	100.00	0.593 ± 0.0180 *	7.713 ± 0.097 *	21.897 ± 0.369 *	2.375 ± 0.076 *	0.711 ± 0.013 *	2.672 ± 0.045 *	1.476 ± 0.026 *	1.299 ± 0.023 *
MA
MA_A	0.00	0.186 ± 0.012 e	2.957 ± 0.047 f	1.073 ± 0.037 f	1.854 ± 0.120 f	1.280 ± 0.023 f	0.189 ± 0.015 f	2.301 ± 0.028 f	2.064 ± 0.072 f
MA_B	3.12	0.196 ± 0.013 e	5.258 ± 0.019 e	2.362 ± 0.029 e	2.627 ± 0.042 e	1.673 ± 0.013 e	0.472 ± 0.008 e	2.903 ± 0.029 e	2.225 ± 0.044 e
MA_C	6.25	0.327 ± 0.010 d	8.956 ± 0.020 d	4.350 ± 0.022 d	4.434 ± 0.006 d	2.508 ± 0.028 d	0.978 ± 0.012 d	3.736 ± 0.042 d	2.716 ± 0.008 d
MA_D	12.50	0.632 ± 0.004 c	14.062 ± 0.077 c	7.171 ± 0.015 c	6.864 ± 0.055 c	3.661 ± 0.039 c	1.696 ± 0.070 c	5.121 ± 0.007 c	3.369 ± 0.025 c
MA_E	25.00	1.111 ± 0.021 b	27.125 ± 0.185 b	13.577 ± 0.044 b	12.624 ± 0.293 b	6.240 ± 0.009 b	3.368 ± 0.074 b	9.047 ± 0.224 b	5.669 ± 0.038 b
MA_F	50.00	2.198 ± 0.018 a	50.230 ± 0.196 a	25.041 ± 0.089 a	24.108 ± 0.122 a	10.822 ± 0.073 a	6.773 ± 0.131 a	16.494 ± 0.166 a	9.779 ± 0.032 a
MA_ST	100.00	3.898 ± 0.033 *	110.034 ± 1.82 *	48.343 ± 0.165 *	41.247 ± 0.701 *	20.834 ± 0.346 *	17.743 ± 0.203 *	29.965 ± 0.572 *	18.61 ± 0.272 *
SY
SY_A	0.00	4.938 ± 0.055 f	2.691 ± 0.022 f	31.479 ± 0.194 f	1.183 ± 0.060 f	7.345 ± 0.182 f	0.283 ± 0.003 f	2.344 ± 0.113 f	1.889 ± 0.045 f
SY_B	3.12	5.132 ± 0.022 e	4.979 ± 0.058 e	34.544 ± 0.248 e	3.655 ± 0.048 e	8.411 ± 0.052 e	0.825 ± 0.009 e	3.105 ± 0.048 e	2.077 ± 0.011 e
SY_C	6.25	5.797 ± 0.034 d	7.350 ± 0.053 d	36.318 ± 0.119 d	5.623 ± 0.077 d	9.162 ± 0.094 d	1.293 ± 0.015 d	4.579 ± 0.054 d	3.080 ± 0.068 d
SY_D	12.50	6.805 ± 0.064 c	12.52 ± 0.059 c	41.537 ± 0.101 c	10.278 ± 0.134 c	11.232 ± 0.008 c	2.420 ± 0.017 c	8.050 ± 0.071 c	4.903 ± 0.044 c
SY_E	25.00	9.838 ± 0.186 b	23.301 ± 0.188 b	52.973 ± 0.068 b	19.490 ± 0.097 b	15.812 ± 0.225 b	4.529 ± 0.029 b	15.458 ± 0.031 b	9.394 ± 0.037 b
SY_F	50.00	14.798 ± 0.074 a	46.283 ± 0.197 a	75.544 ± 0.466 a	37.699 ± 0.176 a	23.593 ± 0.083 a	8.824 ± 0.053 a	31.916 ± 0.095 a	16.775 ± 0.076 a
SY_ST	100.00	27.207 ± 0.415 *	104.578 ± 0.948 *	121.403 ± 0.721 *	65.235 ± 1.355 *	40.267 ± 0.075 *	22.984 ± 0.249 *	59.671 ± 1.086 *	34.895 ± 0.298 *

Wines analysis were done in triplicate reps (n = 3) with standard deviations reported, Fisher's LSD for means comparison of all samples across each chemical parameter. Letters that are different indicate a significant difference, p ≤ 0.05 in each set of wines.

^*Was not included in the L.S.D test of significant difference.

3.3. Expert and Consumer Test

Across Figure 1, the wines in each of the respective correspondence analyses (CA) of the CATA selections are mainly described by variations along the first dimension accounting for 78% (Figure 1A), 83.2% (Figure 1B), 61.6% (Figure 1C), 43.1% (Figure 1D), and 54.4% (Figure 1E) of the variance, respectively. Across both dimensions, 88.7% (Figure 1A), 90.9% (Figure 1B), 80.0% (Figure 1C), 68.3% (Figure 1D), and 71.7% (Figure 1E) of the variance was explained, respectively. Generally, the fruit and non‐smoke‐related attributes, “red fruit,” “dark fruit,” “dried fruit,” “black pepper,” and “baking spices” weighed in one direction of the CAs, while smoke‐related attributes, “cigarette smoke,” “liquid smoke,” “sweet BBQ,” “ashy aftertaste,” and “tar” weighed in the other. In Figure 1A,B for the experts, the wines are distributed from left to right on the correspondence analysis across the first dimension which is reflective of the amount of smoke impact on the wines (Table 2). The point where change is observed appears to be around wines CS_1_D (Figure 1A) or CS_2_C (Figure 1B). This is where the wines are described as “solvent” and “medicinal,” moving away from the fruit attributes and towards the smoke related attributes. This is further confirmed by the overall liking scores received by the wines (Figure 2) where were 2 LSDs away from the NST. The expert ratings for CS_1 (Figure 2A) showed the change in liking at CS_1_D based on Fisher's LSD. For CS_2, the overall liking score was observed to be different at CS_2_C which confirms the change in the wine position described by CATA attributes in the CA.

FIGURE 1.

Correspondence analysis biplot derived from CATA selections by experts and consumers. There was a rotation on the x‐axis for (A) expert correspondence analysis‐CS_1 and (B) expert correspondence analysis‐CS_2 to improve readability across figures. Non‐expert correspondence analysis shown for (C) consumer correspondence analysis‐CS_2, (D) non‐expert correspondence analysis‐SY and (E) consumer correspondence analysis‐MA.

FIGURE 2.

Mean hedonic ratings for all wines rated by experts and consumers, respectively, with Fisher's LSD at significance level, α = 0.05, denoted by the letters above each column. Wines were compared in their respective sets. Expert overall liking shown for (A) CS_1, and (B) CS_2. Consumer overall liking shown for (C) CS_2, (D) SY, and (E) MA.

If one examines at which smoke taint dilution a given wine transitioned in the correspondence analysis biplots of CATA ratings from the smoke attributes side of the plots to the fruity attributes side, this expectedly happened at a higher smoke taint level for the consumers, that is with wine CS2_D (Figure 2C) and a lower smoke taint level for the experts, that is with wine CS2_C (Figure 2B).

There were differences in the preference maps of the expert and consumer groups. In the expert preference map (Figure 3A,B), the majority of experts preferred wines A, B, C, and D. Most consumers main direction of liking was along the right side of dim 1 of the preference map. Comparing this to the consumer preference maps (Figure 3C–E), there was no real preference for the products based on the maps shown. This is typical in a preference map of consumers of the products where their overall liking for the products are normally distributed across all wines. For the wine CS_2 specifically, the effect can be seen between experts (Figure 3B) and consumers (Figure 3C) where experts had one‐sided preferences and the consumers’ liking were scattered.

FIGURE 3.

Internal preference map based on overall liking of wines among experts and consumers. The expert preference map is shown in (A) CS_1 and (B) CS_2, where liking is relatively skewed in one direction across the x‐axis. The consumer preference map is shown in (C) CS_2, (D) SY, and (E) MA, where the liking of the wines were distributed evenly over all wines.

In the penalty/lift analysis (Figure 4), it is observed that fruit attributes (“red fruit,” “jam fruit,” “dark fruit,” and “dried fruit”), and spice attributes (“black pepper” and “baking spice”) contributed positively towards the overall liking scores across both groups, experts, and consumers. Conversely, smoke‐related attributes (“musty,” “ashy aftertaste,” “tar,” “medicinal /Brett,” “liquid smoke,” “sweet BBQ,” “cigarette smoke”) were all negative contributors towards the overall liking scores of the wines. One exception is the “sweet BBQ” and “liquid smoke” attributes for the MA (Figure 4E) where it contributed positively to the overall liking instead for the consumers. Experts and consumers were generally aligned for the positive and negative contributors of the attributes that impact the overall liking score of the wines tasted.

FIGURE 4.

Penalty lift analysis relating CATA terms to the overall liking score for all wines rated by experts and consumers respectively. Terms that contributed positively towards the overall score is pointed towards the right and terms that contributed negatively to the overall score is pointed towards the left. The expert penalty lift analyses were shown in (A) CS_1 and (B) CS‐2. The consumer penalty lift analyses were shown in (C) CS_2, (D) SY, and (E) MA.

In the penalty analysis (Figure 5), experts (Figure 5A,B) and consumers (Figure 5C–E) both penalized flavor being too weak and too strong, and bitterness and astringency being too high. In addition, consumers (Figure 5C–E) also penalized sweetness being too low and acidity being too high. Thus, perceived adequacy of sweetness and acidity marked the difference between expert and consumer ratings.

FIGURE 5.

Penalty analysis relating JAR ratings to the overall score for each wine rated by experts and consumers. The top right quadrant is the main penalty where more than 20% of the responses and drop in hedonic score is more than one point. The expert penalty analysis are shown in (A) CS_1 and (B) CS‐2. The consumer penalty analysis are shown in (C) CS_2, (D) SY, and (E) MA.

3.4. Difference Test With Consumers

A summary of the difference test results is shown in Table 7. In the triangle test, with no knowledge of the nature of the difference between the wines (i.e. in an overall difference test), and prior to being exposed to an “ashy” standard, consumers were able to discriminate between the non‐smoke impacted wine MA_A and the smoke‐impacted wine, MA_E (P = 0.011). In the 3‐AFC test, following training with the “ashy” standard, consumers discriminated between non‐smoke impacted wine MA_A and smoke‐impacted wine MA_C (P = 0.022). This shows that the consumers could distinguish between the non‐smoke‐impacted and smoke‐impacted wine at a lower percentage of smoke‐taint following training with the ashy standard. However, it is noted that this could also have been due to the sensitivity of the methods used.

TABLE 7.

Difference test results. The reference wine used was MA_A, with the number of participants n = 55.

Wine	MA_B	MA_C	MA_D	MA_E
Triangle test, number of correct responses	18	16	21	27
Triangle test, P‐value	0.588	0.790	0.265	0.011
3‐AFC, number of correct responses	19	26	32	41
3‐AFC, P‐value	0.475	0.022	0.000	0.000

4. Discussion

There was a difference between the ratings of experts and consumers in distinguishing smoke‐impacted and non‐smoke‐impacted wines as would be expected from previous research. Sensory differences were primarily driven by varietal characteristics, with these distinctions becoming more pronounced under conditions of smoke exposure. The smoke impact was also quantified using VP and individual bound glycoside analysis (Simoneit 2002; Parker et al. 2023).

The individual‐bound phenolic glycosides tested were only a list of 8, that is the current standard for smoke analysis in grapes (Favell et al. 2022). These are the only available compounds that can be identified and quantified with high confidence as authentic standards. In general, wines with high free and total VP's also contained high levels of the individual bound glycosides. The exception was the smoke‐impacted Cabernet Sauvignon, CS_2 wines. The low concentrations of the CS_2 individual bound glycosides and the high concentrations of total VPs and free VPs could be attributed to the initial glycosylation of VPs to other intermediaries or glycosides present (Szeto et al. 2020), which were not among the eight tested glycosides. This could be because unlike all the other smoke‐impacted wines Cabernet Sauvignon (CS_1), Malbec (MA), and Syrah (SY) that were exposed to natural smoke during separate events in August 2020 and September 2020, the CS_2 wines were made from grapes that were intentionally smoked post‐harvest using hickory pellets. The post‐smoking was proven to be an effective approach to mimic naturally smoke‐impacted wines by Culbert et al. where similar levels of VPs and individual glycosides were found in both natural smoke events and post‐harvest smoking (Culbert et al. 2021). Hickory pellets were chosen as the VPs released during burning thereof were similar to those released from ash collected during the 2020 wildfires (data not shown) in the surrounding areas. After intentional smoking, the grapes were stored overnight at room temperature to allow for glycosylation to take place. However, the time period for VP evolution in the berry was much shorter, which could have resulted in a different glycosylation profile. The fuel source was also different. Total VPs are determined by the acid hydrolysis of individual bound precursors and will include the VPs released from a wide range of precursors, extending beyond the eight individual bound glycosides measured. In all the chemical results shown in Tables 4, 5, 6, there is agreement that there is an increase in the individual bound glycosides, free VPs, and total VPS as the percentage of the smoke‐impacted wines increased. There was a linear relationship between the free VP, total VP, and individual‐bound glycosides which increased with the amount of smoke‐impacted wine in the blend. This linear relationship would be a useful tool in smoke‐impact prediction.

Experts and (non‐expert) consumers differed in their judgment of the wines mainly in their overall liking scores as seen in Figure 2. Experts rated the wines lower on the 9‐point hedonic scale than consumers, in particular the CS_2 wines which were evaluated by both experts and consumers (Figure 2B,C). Overall liking scores from experts and consumers were analyzed using a two‐way ANOVA (Type [expert vs. consumer] × Wine), which revealed significant differences between the groups (p < 0.01). Per Fisher's LSD, only wine CS2_A did not differ significantly between experts and consumers, whereas all other wines showed significant differences (Figure S2). The expert scores were below five points on the 9‐point hedonic scale for all wines except CS2_A, showing a dislike for the wines, while the consumer scores were above five points except for CS2_F, showing like for the wines. Overall, experts “disliked” the wines more than the consumers. Hedonic ratings by experts could be used to infer the overall quality of a wine (Everitt 2009). We attribute these low ratings from the experts to the smoke taint, of course, but also to the fact that the wines for this research were produced in stainless steel and did not see any oak aging, something that the experts were not made aware of. They were told to rate this as a finished wine which could account for the lower ratings the wines received. Experts are generally more sensitive to faults and overall balance of a wine. They would heavily penalize the quality of the wine based on these small imperfections in the wine which would then be reflected in the overall liking score of the wines. Experts are also more consistent in their ratings of the wines among each other. This is observed in the preference maps in Figure 3 where experts were consistent in their liking of the wines. The consumers liked the wines more equally. This similar trend has been seen in other studies comparing expert and consumer preferences of a single product (Delgado and Guinard 2011). Interestingly, both experts and consumers were equally proficient in their abilities to describe the wines using the CATA terms. As seen in Figure 1, the CA maps were similar in the distribution of the terms between both groups. Dimension 1 of the correspondence analysis for the experts accounted for 83.2% of the variance in the data versus 61.6% for the consumers, which indicates that differences were mainly on the first dimension. Wines CS2_E and CS2_F were generally to the right of the figures, associated with the smoke character, while wines CS2_A and CS2_B were on the fruity side. The experts described CS2_C and CS2_D as “menthol” and “musty,” whereas consumers saw CS2_C was closer to the fruity characteristics of CS2_A and CS2_B. Consumers only saw CS2_D as “musty” and “menthol.” This is further observed in the CA maps (Figure S1) where only significant attributes were projected, experts saw the fruit terms to be significant in describing low‐impact wines A and B from their respective sets. Consumers, on the other hand, (Figure S1) were less able to discriminate among the wines and could only distinguish these wine based on smokiness. This could have been due to the non‐discriminatory descriptors picked for the wine which were not significant in the Cochran's test. Yet, both groups were similar in how the CATA terms impacted their overall scoring as seen in the penalty‐lift analyses (Figure 4), and how JAR attributes impacted the overall liking score in the penalty analysis (Figure 5). In Figure 4, except for the Malbec (MA) wine for consumers, the selection of CATA fruit attributes contributed positively, and smoke attributes contributed negatively towards the overall liking of the wines. This indicates that both experts and consumers generally were in agreement over terms that described the wines positively and negatively. However, the penalty observed here for the experts was close to two points while the penalty for the consumers was close to one point; the lift for both groups was close to being the same at one point. In Figure 4A,B, the experts tended to give larger penalties towards attributes (“cigarette smoke,” and “ashy aftertaste”) that impact the wine quality, and consumers a smaller penalty which is indicative of the experts’ hedonic ratings being reflective of their concept of quality. In the penalty analysis (Figure 5), there was agreement regarding the attributes that penalized the wines. However, consumers also penalized wines for being not sweet enough or too acidic which would be reflective of consumer preferences (Sena‐Esteves et al. 2018). Generally, sugar reduces acidity perception (Noordeloos and Nagel 1972), and consumers prefer slightly “sweet” wines which would naturally be more palatable. These wines not significantly different in residual sugar but had small differences in acidity (Table 1). The slight increase in acidity, combined with low sugar levels could have caused these wines to not be sweet enough to the consumers, resulting in a penalty for sweetness being too low. Wine experts were able to use more objective approaches on the suitability of attributes to describe the wine where the concern for them would be the balance of flavors, acid, and sweetness. As such, the penalization of the wine not being sweet enough highlights the differences in the perception between experts and consumers.

The consumer hedonic ratings of the Malbec (MA) wines did not differ significantly (Figure 2E) and the CA map (Figure 1E) also saw little discrimination among the wines based on the selection of sensory attributes in the CATA list. From the penalty‐lift analysis (Figure 4E), it was also observed that “sweet BBQ” and “liquid smoke” contributed positively towards the overall liking score. This observation was unexpected as in a previous DA conducted on the wines by Lim et al. for the non‐impacted Malbec wine (MA_A) and 100% smoke‐impacted wine (MA_ST) (Lim et al. 2024), the two wines were vastly different. MA_A was described as ‘red fruit and “cooked fruit” with an ashy score of 10.8, whereas MA_ST was described as “ashy,” “tar,” “cigarette smoke,” and “liquid smoke” with an ashy score of 45.5. This highlights the very different level of smoke impact between these two wines. We speculate that during the serial dilutions, there was a larger than expected matrix effect which did not show the smoke impact in MA as obviously compared to the other wines dilutions of Cabernet Sauvignon, CS_1 and CS_2 respectively. This was further confirmed by Cochran's Q test which showed no differences across the MA wines for the CATA attributes selections. In the Syrah (SY) wine, there were no differences between the overall liking of the wines (Figure 2D) but the CA map (Figure 1D) saw a clear separation of wines SY_E and SY_F from the rest of the wines. However, in the correspondence analysis for only the significant attributes determined by Cochran's Q test (Figure S1D), SY_F was grouped with SY_C and described as “liquid smoke.” SY_A, SY_B and SY_E were described as “ashy aftertaste” and “hotness alcohol.” There was little differentiation among wines, which largely overlapped with each other.

The consumer test, with non‐expert (consumer) wine drinkers, showed small differences which were non‐significant statistically, mainly in the overall liking scores among the wine dilutions of SY and MA respectively, and wines were not distinguished from each other based on the percentage of ST wine included in the blend. The wines did show subtle differences in the CA in Figure 1. When a higher p‐value was used in the Cochran's Q test (data not shown), there were more differences observed for the SY and MA wines. In the next step, discrimination/difference testing was used to determine whether there was a perceptible difference between the products, and if so, what the nature of that difference was. However, due to wine limitations, only Malbec (MA) was used in this part of the study. In the overall difference (triangle) test, participants were only asked to identify which wine was different in the set. From the results above, it was only at MA_E that there was a detectable difference to consumers. The calculated threshold, Tables 8 and 9, using the geometric mean of the concentrations of smoke‐impacted wine was 17.68% of smoke‐impacted wine where consumers could detect a difference at p < 0.05. In the 3‐AFC test a significant difference was observed between MA_B and MA_C, after training with the “ashy standard”. The calculated geometric mean of 4.42% of smoke‐impacted wine, in Tables 8 and 9, can be used as the recognition threshold for smoke taint for the consumer population we tested, upon familiarization with the sensory concept of smoke taint. This amount of smoke‐impacted wine in the blend is closer to that of CS_1_D (6.25%) and CS_2_C (3.125%) which experts rated in their sensory test. This suggests consumers can be trained to identify smoke impact in wine and be equally discriminating as experts (Oppermann et al. 2017). Upon training, their sensitivity was closer to that of an expert group of tasters, although a proportion of specific anosmia to the smoke aromas could be expected in any randomly selected panel (Mainland et al. 2013; Bilogrevic et al. 2023).

TABLE 8.

Calculated geometric means of MA wines free and total VPs to determine threshold concentrations (µg/L).

Type	Product	Taint %	Guaiacol	4‐methylguaiacol	o‐cresol	Phenol	4‐ethylguaiacol	p‐cresol	m‐cresol	4‐ethylphenol	Syringol	4‐methylsyringol
Total	MA_Triangle	17.68	14.551	6.767	4.172	20.891	0.387	10.194	5.577	4.827	44.972	6.519
Free	MA_Triangle	17.68	3.671	1.116	1.814	7.976	0.023	1.622	2.071	0.195	16.493	0.047
Total	MA_AFC	4.42	6.681	2.262	2.035	11.538	0.206	6.298	2.587	3.605	27.776	2.501
Free	MA_AFC	4.42	1.651	0.319	0.817	2.946	0.000	0.726	0.781	0.000	13.859	0.000

TABLE 9.

Calculated geometric means of MA wines individual bound glycosides to determine threshold concentrations (µg/L).

Type	Product	Taint %	Guaiacol gentiobioside	Syringol gentiobioside	Guaiacol glucoside	Phenol rutinoside	Guaiacol rutinoside	4‐methylsyringol gentiobioside	Cresol rutinoside	4‐methylguaiacol rutinoside
Bound	MA_Triangle	17.68	0.838	19.530	9.867	9.309	4.780	2.390	6.807	4.370
Bound	MA_AFC	4.42	0.253	6.862	3.205	3.413	2.048	0.679	3.293	2.458

The determination of smoke impact is a combination of both chemical analysis and sensory evaluation. The chemical analysis would offer data on free VP, total VP, and individual bound glycoside levels in a wine which well relates to the sensory characteristics present in a wine. In the sensory evaluation conducted, comprising of hedonic, JAR, and CATA tasks, experts were able to distinguish the wines solely based on their liking scores for the CS_1 and CS_2 wines. We relate overall hedonic liking to the acceptability of the wines. VPs in Cabernet Sauvignon wines CS_1 and CS_2 which saw a drop in liking (Figure 2) were higher than in the non‐impacted wine (Tables 4 and 5). Comparing consumers and experts, the consumers were able to tell the difference at a level (CS2_D, 6.25% ST wine) just one above the experts (CS2_C, 3.12% ST wine). Evidently, in the eyes of the expert, this small change in the level of VPs affected the wine detrimentally, which was observed for the wine based on their liking score. The linear relationship of the amount of VPs (Tables 4 and 5) present with the perceived impact though the liking score by experts is symbiotic; where the increase in VPs is accompanied by a decrease in the liking score, which confirms previous studies (Bilogrevic et al. 2023). A similar relationship was also observed in the consumers for CS_2, although the impact was less evident with the change in their liking happening at a higher level of smoke taint. The liking score may not be the most ideal metric for smoke taint assessment. As trained professionals, the wine experts were better able to express their concerns about the quality (complex, free of faults, well balanced) of the wines through their hedonic ratings (Hopfer and Heymann 2014). This was seen in penalty lift analysis in Figure 4B,C where the level of penalty associated with “medicinal/Brett,” “ashy aftertaste,” “liquid smoke” for experts were closer to negative two while that for consumers was around negative one. In the CS_1 and CS_2 wines, the differences too between the non‐impacted and where the change was detected was generally for the total VPs around 2–4 µg/L guaiacol, 0.5 µg/L 4‐methyl guaiacol, 1–4 µg/L o‐cresol, 1.5‐20 µg/L phenol, 0.15 µg/L 4‐ethyl guaiacol, 1.5‐6 µg/L p‐cresol, 1.10‐7 µg/L m‐cresol, 0.8‐2 µg/L 4‐ethyl phenol, 3.5‐6 µg/L syringol and 5–25 µg/L 4‐methyl syringol, summarized in Table 10. The large range in the differences showed that even in a single wine matrix, the amount of VP present affecting the liking of the wine varies greatly. In the MA and SY wines, the non‐experts were not able to pick out this difference clearly using their overall liking score. This could have been due to potential reasons such as a more complex matrix effect in the wine, where the smoke‐impact was masked by other factors in a wine matrix, that is stronger fruit aroma, other aromas, residual sugar and acidity. In the MA Wines, between the non‐smoke‐impacted wine and MA_C wines, the amount of total VPs differences lies around 4 µg/L guaiacol, 2 µg/L 4‐methyl guaiacol, 1 µg/L o‐cresol, 7 µg/L phenol, 2 µg/L 4‐ethyl guaiacol, 2 µg/L p‐cresol, 2 µg/L m‐cresol, 1.8 µg/L 4‐ethyl phenol, 10 µg/L syringol, and 2.2 µg/L 4‐methyl syringol (Table 5) for the sensory difference seen in the MA wines in the difference test (Tables 8 and 10). Evidently, here the numbers above the baseline for MA differ from the CS wines which would show that there is an even larger matrix effect that primarily drove the threshold determination of smoke impact in wine (Coulter et al. 2022).

TABLE 10.

Calculated difference of total VPs concentrations (µg/L) between non‐impacted wine (NST) and the threshold detection level to determine the estimated threshold concentrations levels above baseline.

Product	Guaiacol	4‐methylguaiacol	o‐cresol	Phenol	4‐ethylguaiacol	p‐cresol	m‐cresol	4‐ethylphenol	Syringol	4‐methylsyringol
CS1	4.210	NA a	4.243	19.144	0.041	6.117	7.537	NA a	33.476	26.123
CS2	2.033	0.892	0.834	1.607	0.140	0.618	1.257	0.763	5.398	2.582
SY	11.531	3.822	2.590	10.311	0.185	3.890	3.217	1.953	15.296	3.752
MA	3.151	1.432	0.895	5.468	0.136	1.188	1.417	1.345	7.776	1.551

^aIndicates no best estimate as the threshold value was lower than the base value. Leading to a negative sum.

As expected, preference mapping revealed some differences in liking of both the consumers and the experts, but we felt that the limited number of subjects in both populations did not warrant a full preference segmentation analysis. That should be a focus of future studies. Indeed, it is likely that some level of smoke taint may add to the complexity of some wines and be liked by some consumers.

This study explored the sensitivities and preferences of both wine experts and non‐expert consumers. Future work will include creating smoke‐impacted wine from other varieties to better understand the specific matrix effects of the varietals on their smoke taint threshold levels and VPs marker compound composition. On the side of the subject experiencing the wine, whether expert or non‐expert, it is likely that specific anosmia to guaiacol and other smoke‐related compounds could be playing a role, as well as training and preference (Mainland et al. 2013; Flaherty and Lim 2016; Parker et al. 2020).

5. Conclusion

This study showed that there are differences between the abilities of experts and non‐expert consumers in identifying smoke impact. Experts are more sensitive in picking up faults as they are generally more sensitive towards these faults and are more adept towards changes in a wine matrix, especially those that affect it negatively. These experts also often have prior knowledge of a wine which may cause experts to unnecessarily fault a wine as smoke impacted when there may be no smoke impact. The hedonic aspect of the sensory experience might be more important to consumers than it is for expert professionals who may be more accustomed to focusing on the analytical aspect of the sensory experience (“how does the wine taste?” rather than “how do I like it?”). But this research confirms that a bit of training or simple familiarization with the sensory trait of interest, that is smoke taint in this instance, can go a long way to give sufficient ‘expertise’ to consumers. The results from this study could be used by the wine industry to make informed winemaking decisions in the context of the increasing incidence of wildfires potentially affecting vineyards worldwide.

Author Contributions

Lik Xian Lim: conceptualization, methodology, formal analysis, writing – original draft, writing – review and editing. Cristina Medina‐plaza: conceptualization, methodology, formal analysis, writing – review and editing. Ignacio Arias‐perez: methodology, formal analysis, writing – review and editing. Yan Wen: formal analysis. Bishnu Neupane: formal analysis. Larry Lerno: methodology. Jean‐Xavier Guinard: writing – review and editing, supervision, methodology. Anita Oberholster: conceptualization, methodology, writing – review and editing, supervision, project administration, funding acquisition.

Disclosure

A poster of the results was presented at the 15th Pangborn Sensory Symposium 2023‐ Meeting new challenges in a changing world (PSSS 2023):S01. Available at SSRN: https://ssrn.com/abstract = 4550233.

Ethics Statement

The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board (or Ethics Committee) of the University of California, Davis (protocol code 1288072‐1, June 26, 2018).

Consent

Informed consent was obtained from all subjects involved in the study.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Materials: jfds70621‐sup‐0001‐SuppMat.docx

Lim, L. X. , Medina‐Plaza C., Arias‐Perez I., et al. 2025. “Discrepancy Between the Ability of Wine Experts and Consumers in Identifying Grape Smoke Exposure in Different Wine Matrices in California.” Journal of Food Science 90, no. 10: e70621. 10.1111/1750-3841.70621