3.1. Objectives

An expert panel was used to ensure no wine used in the experiments contained faults or obvious quality issues and for the panel to undertake a sorting trial to examine whether the experts could perceive any similarities between the mainstream and emerging varietal wines and provide an initial sensory profile of the wines.

3.2. Participants

An expert panel composed of eight individuals (three females, five males) was assembled. All panellists had 5+ years wine industry employment and experience with the varietals under study and were either winemakers, oenology and sensory academics, or wine professionals (Parr et al., 2002).

Previous studies have utilised between 9 and 15 experts to conduct sorting tasks to obtain wine sensory maps (Bécue-Bertaut and Lê, 2011; Cartier et al., 2006) and indicate that the more expertise plus type of product set can reduce the number of evaluations required to ensure the stability of the sensory space (Blancher et al., 2012).

3.3. Procedure

Experts evaluated the wines at individual benches in an open-plan focus group room. The 38 wines (30 ml each) were presented to the experts in a randomised order (Torgerson and Torgerson, 2003) in clear International Standards Organisation (ISO) approved 215 ml wine glasses coded with 4-digit codes and covered with a petri dish. Experts were asked to sort the wines, based on wine aroma and flavour similarity, into as many groups as they deemed appropriate, disregarding colour. The use of black glasses to disregard colour was not possible due to the number of available glasses and transport issues with black glasses to Australia. Experts allocated wines into their groups noted on large pieces of white paper, which were inputted manually into a computer per expert for future analysis. The experts next completed a variant of the RATA methodology (group-RATA) to obtain the sensory aroma and flavour profiles of their groups. The experts selected from a supplied list only those attributes that applied to each group and rated the intensity on a 5-point scale ranging from “Low” to “High”. Sensory attributes were chosen from the red wine lexicon used in other studies (Danner et al., 2018; Ferrero-del-Teso et al., 2020; Kontkanen et al., 2005) with additional attributes that have been shown to be prevalent in these specific emerging varieties (Bonello et al., 2018; Cravero et al., 2012; Falqué et al., 2004; Mezei et al., 2021). Sensory attributes consisted of 29 for aroma and flavour, three for taste and eight for mouthfeel (Supplementary data, Table S3). Descriptors were presented to experts as combined flavours and aromas due to time constraints, and the panel had previously been trained on mouthfeel attributes.

This sorting task assisted in identifying three faulty wines that were subsequently removed from the remainder of the study.

4.1. Objectives

Trained RATA panellists were required to generate detailed wine sensory profiles, along with a preliminary hedonic liking “score” for the individual wines, to both examine similarity amongst wines and facilitate the selection of a subset of wines for the consumer trial.

4.2. Participants

A RATA panel (n = 36) was assembled consisting of wine sensory panellists formally trained in wine sensory evaluation, including aroma, taste, flavour and mouthfeel and who had sat on numerous wine RATA panels. RATA participants included wine science staff members and higher degree students recruited from the University campus wine science department.

4.3. Procedure

The RATA trials took place over three weeks and consisted of three sessions of approximately one hour. Participants were required to attend each session and evaluate the same wines used in the expert trial but with the three faulty wines removed: 35 wines in total, with 12, 12 and 11 wines in sessions one, two and three, respectively. Sessions took place in the sensory laboratory in individual booths under fluorescent lighting at a constant temperature (21 °C).

With the exception of the 15 panellists in session one, panellists were presented with each wine monadically. All tasters had an enforced 1-minute break between samples and a 10-minute break between wines six and seven. Samples (30 ml) were presented randomly in clear, covered ISO-approved wine glasses labelled with 4-digit codes.

For each wine, participants were first asked to rate their liking on a 9-point hedonic scale (ranging from “dislike extremely” to “like extremely”). Next, panellists were instructed to rate the intensity of only the sensory attributes that they perceived in the wine, using a 7-point rating scale (ranging from “extremely low” to “extremely high”). The flavour and aroma descriptors used were the same as the experts.

Data was collected through Red Jade software (2016, Redwood City, USA)

5.1. Objectives

The consumer trial aimed to assess the participants’ preference towards wines made from emerging varietals and whether they deemed the emerging varietal wines similar sensorially to the corresponding major varietal wines.

5.2. Wines

From the trained RATA panel data, two wines were selected from each emerging varietal bracket and one wine from each corresponding major varietal. Each two emerging varietal wines were chosen based on the favourable preliminary hedonic responses and flavour profiles obtained through the trained RATA trial and possessed similarities to the benchmarked major varietals chosen yet were different to each other (Supplementary data, Table S1).

5.3. Participants

The consumer panel consisted of 116 participants recruited via email from the University of Adelaide’s wine consumer database and through social media. To participate, consumers were required to have consumed red wine in the last month, be over the age of 18, and not have any formal wine qualifications or wine industry experience.

5.4. Procedure

These trials took place over three weeks, taking approximately one hour to complete. Participants were seated in individual sensory booths and undertook an online questionnaire consisting of demographic questions and completed the Fine Wine Instrument (FWI) (Johnson and Bastian, 2015), which is a psychographic segmentation approach used to group wine consumers into clusters based on wine knowledge, engagement with wine and fine wine behaviour. The FWI has uncovered consumer clusters with different wine-associated behaviours, knowledge and wine taste preferences and emotions (Danner et al., 2020). Consumers then had to indicate their liking of the wines and their perception of the similarity between emerging varietal wines and their corresponding major varietals.

Wines were presented two at a time, one emerging variety and its purported corresponding major varietal (Montepulciano with Shiraz, Nero d’Avola with Grenache, Touriga Nacional with Cabernet-Sauvignon). Consumers were asked to rate their liking of each on a 9-point hedonic scale and were then asked to rate how similar the two wines were on a 100-mm line scale with 7-word anchors from No Difference (10 %) to Moderate Difference (50 %) to Extremely Large Difference (90 %). This was repeated six times, one for each emerging varietal. Wines were assigned random 4-digit codes, and the order of wines presented was randomised both between and within each emerging and major varietal pairing. A break of one minute was enforced between each pair of wines.

Data was collected through Red Jade software (2016, Redwood City, USA)

2.1. Preliminary hedonic results

Prior to RATA sensory profiling, the 36 participants were asked to rate their hedonic responses to the wines. One-way ANOVA (Supplementary data, Table S4) demonstrated that all wines, except for MON2 (mean liking score 4.89), scored greater than 5 on a 9-point scale, indicating participants moderately liked all wines. TOUR8 (mean liking score 6.47) from the Riverland was the most liked wine and was liked significantly more than 17 other wines.

2.2. Trained RATA panel sensory results

To examine the sensory attribute similarities and differences between individual wines from a trained RATA panel perspective, ANOVA was conducted on the RATA profiling intensity ratings, and a PCA was performed using the statistically significant attributes that differentiated the 35 wines (DF = 68; F = 8.522; Pr > F < 0.0001). Of the 67 RATA attributes used, 46 significantly differentiated the wines at a 95 % confidence interval (P < 0.05). The biplot of the PCA with the preliminary hedonic response overlaid as supplementary data is presented in Figure 3.

In the first and second-dimension biplot (Figure 3a), PC1 contributed to 31.62 % of the variance in the data. Wines were distinguished as showing aromas and flavours of red fruits, floral, blue flowers and confectionery on the left-hand side, with wines on the right-hand side showcasing aromas and flavours of leather, savoury, earthy, meaty/salami, pepper and tobacco, flavours of dried fruits and greater mouthfeel alcohol/heat and longer aftertaste.

PC2 contributed 12.37 % to the variance of the data. Wines in the upper half possessed aromas of vegetal, herbaceous, green, mint, dried herbs and eucalypt, and vegetal, minty, cooked vegetables and green flavours. Wines in the bottom half had greater jammy, sweet oak and chocolate aromas, flavours of caramel/butterscotch and sweet oak, and were perceived as tasting sweet.

PC3 contributes 11.80 % to the variation of the wines in the sensory space. Wines present in the upper half were more predominant in dark fruit, minty, eucalypt and dark fruit flavours and aromas, with a positive correlation towards blue flower/violet aromas. Wines were also perceived as having greater astringency and body than most other wines. Wines in the lower half contained aromas and flavours of caramel/butterscotch, earthy, meaty/salami attributes and sweet taste.

Figure 3a indicated primarily Grenache (GREN1, GREN2) and Nero d’Avola (NERO1, NERO3, NERO4, NERO5, NERO6, NERO8 and NERO9) showing greater red fruit, floral and confectionary attributes, similar to what was described by the expert panel (Figure 2). The upper right quadrant of Figure 3a housed primarily green, dried herbs, cooked vegetables and herbaceous aroma and flavour attributes associated with the three Cabernet-Sauvignon wines.

The Touriga Nacional wines were relatively evenly dispersed across the sensory space occurring in each quadrant, thus reflecting that numerous wine styles made from this variety were apparent in this sample set. However, TOUR3 and TOUR7 saw the strongest positive correlation towards green and herbaceous characteristics compared with other Touriga Nacional samples in the first and second dimension biplot (Figure 3a) yet were not as strong as Cabernet-Sauvignon was towards these attributes. The first and third-dimension biplot (Figure 3b) indicated a greater positive correlation towards green and herbaceous characteristics for more Touriga Nacional replicates (TOUR1, TOUR4 and TOUR6), indicating that there may indeed be some similarities between these varieties.

The three Shiraz wines (SHZ1, SHZ2 and SHZ3) plus MON8, MON4 and MON6, TOUR1, TOUR4 and TOUR6 appear in the bottom right-hand quadrant in Figure 3a, associated with increased body and longer aftertaste, dark fruit aromas and flavours plus oak, sweet oak and chocolate oak aromas and flavours, plus pepper flavour and more bitter and sweet taste. Along PC3, SHZ2 and 3, plus MON6, 5, 4, 8, 1, 2 and 9 are more associated with dark fruits and eucalyptus, minty, blue flowers and more astringent styles of wine versus SHZ2, MON1, 2 and 9, which were more savoury, dried fruit, earthy, meaty/salami and caramel butterscotch styles.

Figure 3. Principal component analysis of the significantly different (α ≤ 0.05) aroma and flavour attributes used to profile the 35 wines by RATA panellists using preliminary liking as supplementary data with A) displaying PC1 and PC2 and B) displaying PC1 and PC3. Asterisks indicate mainstream wines.

Save View full size Expand inline Collapse inline

Figure 3a in the first and second dimensions highlighted that preliminary liking was primarily driven by confectionary and blue flower flavours and aromas, with floral, red fruit, caramel/butterscotch, sweet oak and sweet taste also driving liking. Figure 3b in the first and third dimensions similarly showed that liking was driven by floral, blue flowers and confectionary aromas and flavours. The most liked wine in the tasting, TOUR8, was shown to be strongly correlated with these flavour attributes. Conversely, wines that were least liked (MON2, TOUR3, TOUR7) were associated with savoury, meaty/salami and cooked vegetable attributes.

Wines analysed by the trained RATA panel showed similar flavour and sensory profiles to those described by the expert panel. The MFA analysis indicated that there was a significant (P < 0.001) relationship between the RATA data generated by the two panels, returning a RV coefficient value of 0.448. This suggests that the expert and trained RATA panels, despite undertaking different tasks (sorting followed by group-RATA in an open plan sensory lab and monadic wine sample RATA in individual booths), were describing the wines in a relatively similar way and provided confidence when summarising the generic sensory profiles of the wines made from the distinct varietals.

Wines were colour-coded into groups based on results obtained from the expert AHC (Figure 1). Figure 3a closely resembled the group separation denoted in Figure 2.

The basic chemistry measurements (Supplementary data, Table S2) were overlayed on the sensory results in Supplementary Figure S6. Basic chemistry attributes of alcohol (% ABV) and residual sugar (RS) were in line with sensory attributes. Mouthfeel body was strongly correlated with % ABV, whilst sweeter attributes of taste and caramel correlated with RS.

2.3. Consumer liking scores and similarity responses

The consumer panel was a convenience sample of South Australian red wine drinkers. Females comprised 55 % of the consumer panel, with 44 % male and 1 % non-identified, 50.8 % of whom were over the age of 55, with the remainder below. 79.3 % of the consumers held tertiary education qualifications, and 52 % had an annual household income of over AUD$100,000. Most individuals consumed red wine frequently, with 81 % of consumers drinking wine at least once a week and 70 % purchasing their wine for $25 or less per bottle (Supplementary data, Table S5). Each of the two emerging varietals chosen were based on the favourable hedonic responses and flavour profiles obtained through the trained RATA trial (hedonic scores of 5.5–6.47), and possessed a broad range of flavour profiles (including wines higher in leather, tobacco, oak and alcohol; wines with more intense red fruit, confection, blue flower, minty, herbaceous and vegetal attributes with low to moderate astringency; and wines with more dark fruit, jammy, pepper, chocolate and sweet oak characters, with more prominent astringency and fuller-bodied), as well as possessing similarities to the benchmarked major varietal chosen, yet were different to each other (Supplementary Table S1) from multiple (5) regions across South Australia.

Considering these factors, the following wines were chosen for the consumer trial: GREN1, CABS3, SHZ3, MON1, MON8, NERO5, NERO8, TOUR6 and TOUR8 (Supplementary data, Table S1).

Table 1 presents the consumer liking data and overall similarity scores for the nine wines in the trial (Hedonic one-way ANOVA; DF = 11; F = 1.69; Pr > F < 0.07). Liking was assessed using a 9-point liking scale, with all wines scoring above 5, indicating consumers liked all wines equally (Table 1).

Consumers were also asked to indicate their perceived similarity (Similarity one-way ANOVA; DF = 5; F = 12.575; Pr > F < 0.0001) of wine style between the emerging variety and its corresponding major varietal. There was no information as to what the consumers were tasting. Consumer results indicated that one pairing, NERO8 and GREN1, was perceived as significantly more dissimilar than all other pairings, including NERO5 and GREN1. The remaining pairings were considered as having moderate to slight differences, indicating perceived similarity. MON8 and, to a slightly lesser extent, MON1 showed the greatest amount of similarity with their proposed counterpart wine, SHZ3.

2.4. Psychographic influences on consumers’ liking

Following the protocol outlined in Johnson and Bastian (2015), the FWI was used to segment the consumer panel into three segments: Wine Enthusiasts (WE), Aspirants (ASP) and No Frills (NF) based on their wine involvement, knowledge and wine connoisseur behaviour. Demographic information on the segments is displayed in Supplementary data, Figure S5.

Table 2 indicates the overall FWI segment's hedonic response towards the wines, highlighting no statistical differences in wine liking between the WEs and the ASPs. NF consumers liked MON8 and TOUR6 significantly less than the other segments. Table 2 also demonstrates that all segments reported NERO8 and GREN1 to be the most dissimilar pairing. Significant differences were only observed for the perceived similarity of the NERO5 and GREN1 pairing, with the WEs finding this pairing significantly more dissimilar than the other segments.

Correlation analysis of the FWI consumer segment hedonic scores and trained taster RATA sensory attributes are displayed in a correlation coefficient graph (Figure 4). Absolute values of > 0.4 were used to determine the importance of attributes driving segment wine liking or disliking (Mezei et al., 2021). Thirty attributes were significant in driving wine liking or disliking for consumer segments. Dark fruit, dried fruit, chocolate, sweet oak and oak aromas, dark fruit, dried fruit, jammy, blue flower and sweet oak flavours, bitter taste, fuller body, longer aftertaste and more intense alcohol mouthfeel characteristics contributed to WEs liking of wine; whilst aromas of red fruit and earthy drove their wine disliking (Figure 4). For the ASPs, dark fruit, dried fruit, meaty/salami, tobacco, leather, chocolate, sweet oak and oak aroma; dark fruit, dried fruit, jammy, savoury, tobacco, leather, sweet oak and oak flavours, bitter taste, fuller body and higher alcohol and longer aftertaste mouthfeel were the attributes that drove their liking of the wines. Aromas of red fruit and floral, and flavours of red fruits, confectionary, floral and barnyard/horsey drove their dislike of the wines (Figure 4). Finally, the NF consumers saw aromas of savoury, tobacco, leather, chocolate and oak, flavours of tobacco, leather and oak and more body, higher astringency and alcohol, plus longer aftertaste as drivers of their dislike towards the wines (Figure 4).

Figure 4. Sensory Attributes with an absolute correlation coefficient that indicates sensory drivers of consumer liking for the FWI Segments.

Save View full size Expand inline Collapse inline

Blue bars represent Wine Enthusiasts, red Aspirants, and green No Frills. A = Aroma, F = Flavour, M = Mouthfeel, T = Taste.

3.1. Wine expert point of view towards emerging and main-stream wine varieties similarity

A panel of wine experts performed a sensory sorting trial to form groups of similar wines to reveal if any emerging varietal wines were similar to wines made from mainstream varieties. The panel was not informed about the varieties and was told they were Australian dry red table wines. The expert dendrogram (Figure 1, blue cluster) showed that all three Grenache wines were grouped together with seven out of the ten Nero d'Avola wines. Thus, these wines were interpreted to be perceived as similar, as both the expert and trained panel (Figures 2a,b and 4a,b) consistently noted the same flavour and mouthfeel attributes, with both varieties producing wines with floral, red fruit and spice attributes, with lower body and astringency. These findings are consistent with the results of (Mezei et al., 2021).

Extensive overlap of the sensory profiles of Shiraz and Montepulciano wines was also found. Previous literature has highlighted similar flavour profiles of Shiraz (Bastian et al., 2010; Herderich et al., 2012; Robinson et al., 2013) and Montepulciano (Kalleske, 2016; Suzzi et al., 2012). Expert data obtained in this study alludes to an observable similarity between these varieties, indicated by the group clustering of these varietals in the AHC dendrogram (Figure 1, orange cluster). Additionally, the expert AHC dendrogram showcased no clusters containing Cabernet-Sauvignon and Touriga Nacional together. Dendrogram (Figure 1, green cluster) also separated Touriga Nacional from Cabernet-Sauvignon (with the exception of TOUR1 in the orange cluster), contradicting the indication of similarity existing between these varietals. Compared to the literature, this appears representative of wines from this variety in Australia. The difference may have been because the Mezei et al. (2021) study used both European (Old World) varieties and fewer examples from each variety compared to the current study.

3.2. Wine consumer similarity response to emerging and main-stream wine variety

The reluctance observed in consumer purchase of unfamiliar wine indicates the hedonistic, non-utilitarian nature of this beverage (Lacey et al., 2009). In other words, consumers are less inclined to take a perceived risk (Johnson and Bruwer, 2004). This orthodoxy may be countered by providing clear, unconfrontational product information to reassure consumers (Manske and Cordua, 2005). This notion is supported by studies determining that passing quality information to consumers through trusted agents, such as retailers and waiting staff, can lessen purchasing uncertainty (Hilger et al., 2011). Since a strategy to grow the emerging varietal wine market may be by educating consumers as to what familiar wines the emerging varietal wines are similar to, one of the purposes of this study was to understand whether consumers perceived similarity between emerging and major varietals. Consumer understanding will enable the promotion of these varietals in the future and reduce fears and anxiety around the consumption of wines made from unfamiliar varieties (Mueller et al., 2009). This could be achieved by communicating to the consumer the similarities in sensory characteristics of emerging wine varieties to the wines they usually consume, increasing their product choice as well.

The study by Mezei et al. (2021) concluded that certain emerging varieties could act as replacement varieties in the future. Montepulciano and Shiraz were shown to be similar in their flavour profiles in the Mezei et al. (2021) report; AHC (Figure 1) data in the present study supports this. However, the AHC data (Figure 1) for Montepulciano illustrates its potential as a highly versatile grape variety with an ability to make varying wine styles due to its presence across all three clusters, as suggested previously (Dry et al., 2017). Consumer opinions towards the similarity of Montepulciano and Shiraz yielded the strongest perception of similarity in the study. Notably, there was no statistically significant difference in similarity reported between the different FWS, confirming that despite differences in understanding and knowledge of wine, all consumers generally agreed that the Montepulciano sample set appeared to be similar to Shiraz. Both Montepulciano and Shiraz wines were derived from the Barossa Valley. They were described as fuller-bodied and heavier in style, with an average alcohol reading of 14.5 % v/v across the three wines (Supplementary Table 2). Montepulciano grown in other regions will likely show similarity to Shiraz from the same region, with future research necessary to further support the findings. Montepulciano, made as a full-bodied red wine reminiscent of Shiraz, is liked by consumers in the present study and perceived as being similar and, therefore, may have the potential to act as a substitute varietal.

Based on the data obtained from the expert and trained tasters, Nero d’Avola and Grenache showed the greatest potential for the similarity between varietals. Consumer evaluation, however, varied. Consumers as a cohort found one of the Nero d’Avola wines to be significantly different to the Grenache. However, as individual segments, ASPs and NFs found NERO5 to be significantly more similar to GREN1, whilst WEs did not. WEs’ negative response to similarity for NERO5 could potentially be due to their greater skills with wine knowledge and involvement, indicating they may be able to pick up subtle differences indicating dissimilarity; however, this needs to be tested further with more examples.

The observed dissimilarity was likely related to the presence of more blue flower characters in the NERO8 example or the colour of the Nero d'Avola compared with the Grenache. Colour intensity data indicated that the Grenache used was lighter than the Nero d'Avola (Supplementary Table 1). Literature reports that colour influences an individual's perception of wine flavour and aroma (Parr et al., 2003). Whilst consumers were indicated to disregard colour when analysing the similarity between wines, this may have played a role in their decision-making. Black glasses may have negated this issue with consumers; however, consumers normally use clear glasses when drinking wine.

Based on the results from the RATA panel, NERO8 showed a difference in perceivable flavours compared to NERO5 and GREN1. Additionally, NERO5 was derived from the McLaren Vale region, the same region as GREN1, while NERO8 was derived from the Barossa Valley. It appears Nero d’Avola’s similarity to Grenache is like that between Montepulciano and Shiraz, where wines derived from the same region may invoke a greater similarity potential and could act as a replacement varietal.

Considering attribute differences between NERO5 and NERO8, it is unsurprising that a negative response to similarity, valued above 50 (Table 2), was associated between NERO8 and GREN1 for the consumer trial. However, it reduces the validity of a consistent, observable similarity between Nero d'Avola and Grenache from a consumer perspective. Future research may wish to examine Nero d'Avola and Grenache further, as the expert data showed a clear similarity in flavour profiles. ASPs and NFs found NERO5 and GREN1 to be significantly more similar than the WEs. NERO5 was noted as displaying increased red fruit attributes, which WEs and ASPs found to negatively affect their liking of the wine, whilst NFs did not find red fruit a driver of dislike towards the wine, but has been shown as a positive driver in previous research (Mezei et al., 2021). Whether consumer hedonic responses influenced perceived similarity needs further investigation.

Confirmation that Nero d’Avola and Grenache display similarity from a consumer’s perspective cannot, therefore, be confirmed conclusively from this research and would require a more detailed study. Wines more closely correlated in flavour to NERO5 and GREN1 (such as NERO1, NERO6 and NERO9) may have elicited a different result.

A similarity between wines made from Cabernet-Sauvignon and Touriga Nacional was shown by Mezei et al. (2021). Dissimilarity was seen between TOUR6 and CABS3 for both ASPs and NFs, yet the statistical difference between these segments and the WEs, who indicated similarity between this pairing, did not occur. When observing the overall similarity score between TOUR6 and CABS3, the similarity was denoted, indicating that consumer data found the two Touriga Nacional wines somewhat like Cabernet-Sauvignon. Despite this observation, expert AHC (Figure 1) illustrated the observed separation between Cabernet-Sauvignon and Touriga Nacional between groups, except for TOUR1 and CABS1 and CABS3 in group 2 (orange), highlighting divergence in flavour and aroma similarities.

Therefore, we conclude that Touriga Nacional cannot be identified as a clear substitute varietal for Cabernet-Sauvignon. Results indicated that consumers can clearly perceive the similarity between Touriga Nacional and Cabernet-Sauvignon. However, a lack of confidence from wine experts suggests that further research is warranted before accepting similarity.