PHOTO: Shutterstock.

The experiment presented here was designed to test the ability of experts to distinguish between various styles of dry Chenin blanc wines and describe them in two different sensory exercises, a free and a directed sorting.

South African dry Chenin blanc wines are made in a variety of styles, proof to the versatility of this cultivar. Even though generally accepted as Fresh and Fruity (FF), Rich and Ripe Unwooded (RRuW), and Rich and Ripe Wooded (RRW), the styles are not clearly defined (e g by a list of descriptors or by specific winemaking procedures). It is believed that the South African wine experts have an understanding of the styles’ characteristics and should be able to differentiate between them.

The experiment was designed to test the ability of experts to distinguish between various styles and describe them in two different sensory exercises:

  • Sort and characterise a number of Chenin blanc wines according to own criteria (undirected/free sorting), leading to the free description of the wines presented.
  • Sort and characterise the same wines into pre-defined groups according to style name (directed sorting). This part of the experiment allowed us to test preconceived ideas linked to Chenin blanc style attributes, even in the absence of official pre-defined style descriptors.

In total, 16 commercial Chenin blanc wines (including two blind duplicates) were chosen according to our previous experience to represent the styles. The wines were tasted by 15 panellists, experienced with Chenin blanc and familiar with general and specific attributes.


Did the experts describe the same wines the same way both times?

To answer this question, we compared the results for grouping and descriptors. From the statistical analysis, in both exercises the wines were sorted into three groups. The blind repeats grouped well in both cases, indicating that the panel was reproducible in their evaluation. The same wines were placed in the three groups, with one exception. One of the wines moved from the RRW group (free sorting) to RRuW group (directed sorting). This can be due to the wine itself, for example to the level of wood contact that the specific wine had, making it difficult to clearly place it in one of the groups and, therefore, to the number of times the sample was placed into both groups by the tasters. The descriptors used for the three groups were similar in the two sorting exercises. The Correspondence Analysis factor map showed that FF wines were generally described as ‘grapefruit’, ‘pineapple’, ‘guava’, ‘passionfruit’ and ‘green apple’, RRuW as ‘honey’ and ‘marmalade’, and RRW as ‘toasted’, ‘caramel’, ‘vanilla’ and ‘oaky/wooded’.

Delving a bit more into the details for the frequency of attributes, there was a higher overall frequency of citation for the directed sorting. Panellists may have felt the need to use more descriptors when limited to only three groups. On the other hand, there were more different descriptors for the undirected/free sorting.

The differences in the descriptors for the groups were also minimal between the two sorting exercises. It seems that panellists treated this wine sample set similarly in the two cases. FF group had very similar attributes, with ‘grapefruit’, ‘pineapple’ and ‘guava’, featuring in both sorting task experiments. We observed that ‘RRuW’ was cited for the wines in the FF category, meaning that wines from the FF category were also placed into the RRuW group, which shows the difficulty panellists have sorting between these two styles.

RRuW attributes were more different between the sorting exercises. While in both sorting experiments, ‘honey’, ‘guava’ and ‘pineapple’ featured often, in the undirected sorting the wines were perceived more as ‘grapefruit’, but in the directed sorting, the panellists were more likely to use ‘honey’ and ‘yellow apple’. In the directed sorting, the use of the descriptor ‘ripe/ripe fruit’ increased. Again, there was a lot of overlap in the two unwooded styles, as ‘FF’ was frequently cited for these wines.

RRW attributes are very similar in the two sorting exercises, sharing the descriptors of ‘oaky/wooded’, ‘vanilla’, ‘honey’ and ‘toasted’. Panellists did seem more likely to use ‘ripe fruit’ in the directed sorting, maybe due to the wording of the styles.

These effects (i e number of descriptors, similarity in description, and the use of specific attributes for certain groups) could be due to a top-down effect, logical reasoning and preconceived ideas playing a big role, panellists expecting certain categories and subconsciously using words associated with that.

In conclusion, this experiment showed again that even with a sample set chosen to span the sensory space and fit the three wine styles, there is still difficulty for some samples to be clearly attributed. These finding are supported by chemistry for this data set and by previous research. The wines were grouped similarly and were described similarly in both exercises. The number of descriptors was higher for undirected sorting, while frequency of citation was higher in the directed sorting. In other words, the tasters used more different attributes to describe the wines in the first case, while the wines were described by more of the same attributes in the second. We noticed though that some descriptors like ‘ripe’/’ripe fruit’ were associated with the RRuW once this style was indicated as a group type in the directed sorting. Combined with the fact that most of the panel grouped the wines into three groups even in the free sorting, this raises the question: Is there the possibility of product (over)-familiarity for our tasters?



The results indicated that, in the absence of restrictions linked to grouping according to style, the general cultivar characteristics predominated in the descriptions, while during the directed sorting exercise, some attributes considered as style-specific, were used by the assessors. The judges could, generally, indicate which of the wines belonged to the wooded group/RRW style.


FIGURE 1. Group description in the free (top) and directed (bottom) sorting exercises. Left: FF; middle: RRuW; right: RRW. Size of words scaled to the most frequently cited attribute in the respective group.



Buica, A., Brand, J. & Wilson, C., 2017. Evaluating South African Chenin blanc wine styles using an LC-MS screening method. Studia Universitatis 1 – 10.

Bester, I., 2011. Classifying South African Chenin blanc wine styles. Institute for Wine Biotechnology, Department of Viticulture and Oenology. MSc Thesis, Stellenbosch University.


– For more information, contact Astrid Buica at or Jeanne Brand at


You may like to read these:

Go Back