le même processus pour tous - Université de Bourgogne

s.chollet@isa-lille.fr
32
What are consumers able to perform
in beer tasting?
Chollet Sylvie 1 , Lelièvre-Desmas Maud 12 , Abdi Hervé 3 , & Valentin Dominique 2
1 Institut Supérieur d’Agriculture, Université Catholique de Lille, France
2 UMR CSG 5170 CNRS, Inra, Université de Bourgogne, Dijon, France
3 The University of Texas, Dallas, United States
Perception
Aroma detection task
Method Results
Limits ascendant method
10
9
8
1
7
(Isoamyl acetate)
8
2
4 Determination of
detection threshold
6
5
4
3
2
1
p < .09
16
0
Consumers Trained
Consumers have a higher threshold than trained assessors (15 h. of
training), but the difference is not significant.
1
Discrimination task
Method Results
Same/different task
= ? ≠
6
Sorting task
Method Results
1 1
5
Method
Memory
Recognition task
Results
Learning step Recognition step
8 beers
16 beers
Targets
9
3
2
7
1 2
8
Fig. 5: Four dimensional compromise maps for 13 trained (E) and 3 groups of 18 untrained
assessors (A, B, C) for the set 3.
Distractors
4
1 0
Have you ever tasted
these beers ?
Targets
RV coefficients
1
0,9
0,8
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
Accuracy (A')
mg/L
Set 1 Set 2 Set 3
1
0,8
0,6
0,4
0,2
Consumers Trained
0
Familiar Unfamiliar
Fig 7. Mean A’ as a function of expertise level for
learned and new beers. A’ is an index recognition
accuracy varying from 0 to 1; 1 represents a perfect
recognition and 0.5 an answer at the chance level.
Trained assessors have 36h. of training.
For unfamiliar beers, consumers have the same performance as trained
assessors after 36h of training.
RV coefficients
*
1
0,9
0,8
0,7
0,6
0,5
0,4
0,3
0,2
0,1
*
Fig. 3: Similarity between consumer and
trained assessors’ sorts for 3 different
sets of beers (RV coefficients between
the MDS configurations). The stars
indicate a p-value < .05.
Fig 1. Group detection threshold of
consumers and trained assessors (with
15 h. of training).
Consumers Trained
* * *
* * * * *
* *
* *
0
Set 1 Rep 1-2 Set 3 Rep 1-2 Set 3 Rep 1-3 Set 3 Rep 1-4 Set 3 Rep 2-3 Set 3 Rep 2-4 Set 3 Rep 3-4
Fig. 6: Similarity between different repetition sorts for consumers and
trained assessors for 2 sets of beers (RV coefficients between the MDS
configurations). The stars indicate a p-value < .05.
Consumer and trained assessors’ MDS maps are similar, except for set 1.
Consumers are less consensual than trained assessors.
The 3 different groups of 18 consumers behave similarly and their results
are similar to those of trained assessors.
Globally, consumers and trained assessors are repeatable.
%
%
100
90
80
70
60
50
40
30
20
10
0
100
90
80
70
60
50
40
30
20
10
0
Inter-assessors' variability
Familiar beers
Consumers Trained
Unfamiliar beers
Consumers Trained
Consumers have the same discrimination performance than trained
assessors with a short training (15 h. of training) or with unfamiliar beers.
1 2 3 4 5 6 7 8 9 10 11 12
%
100
90
80
70
60
50
40
30
20
10
0
Familiar beers
Consumers Trained
Fig 2. Percentages of correct responses
for consumers and trained assessors after
15h of training (on the left) and 72 h. of
training (on the right).
20
18
16
14
12
10
8
6
4
2
0
Consumers Trained
Fig. 4: Inter-assessors variability for
consumers and trained assessors.
Verbalization
Aroma identification task
Method
Results
Identification of aromas in beers
?
?
% of correct identification & rejection
50
45
40
35
30
25
20
15
10
5
0
Consumers Trained
Correct identifications Correct rejections
Consumers make fewer correct identifications than trained assessors,
whereas they make more correct rejections.
Consumers are as confident on their answers as trained assessors.
Banana !
Oh yes ! This one
smells banana!
Communication task
Method Results
Banana ? OK !
I have got it !
It’s this one !
Consumers have fewer correct matches
than trained assessors.
Consumers use a proportion of specific
terms as large as trained assessors, but
they also use an important proportion of
global and intensity terms.
For consumers, taste as well as hedonic
terms are efficient, whereas for trained
assessors, global terms are inefficient.
Method
Step 1: Description
Odeur: Très forte de fleu r
Lilas
Plutôt dé sagréable
Goût : Goût trè s lilas
Confère un goût amer
Plutôt dé sagréable
Très pe rsi stant en bouche
Step 2: Matching
Odeur: Agrume
Odeur (de citr on) ou plu tôt Citro nnelle
Très forte, Fleur d’oranger
Goût : Goût agrume
R appelle la fleur d’oranger
Fait plutôt a rtificiel
Goût as sez persi stant
Spoken matching task
?
Odeur: Très forte de fleu r
Lilas
Plutôt dé sagréable
Goût : Goût trè s lilas
Confère un goût amer
Plutôt dé sagréable
Très pe rsi stant en bouche
100
90
80
70
60
50
40
30
20
10
0
% correct matching
Familiar commercial beers
**
*
Fig 8. Percentages of correct responses (correct
identifications and correct rejections) in aroma
identification task for consumers and trained assessors
(32h. of training).
Correct matches
Matching task
Results
Exp/Exp Nov/Nov Exp/Nov Nov/Exp
Conclusion
12
10
8
6
4
2
Consumers Trained
0
11 H 44 H 61 H
Fig 10. Trained assessors’ and consumers’ performance for the
communication task. The green line represents the average trained
assessors’ number of correct matches in each session. The blue
line represents the average consumers’ number of correct matches
computed on the three sessions.
Familiar supplemented beers
Concerning perceptual performance, consumers discriminated between
beers as well as moderately trained assessors or highly trained assessors
when beers are not learned by trained assessors.
Moreover, consumers and trained assessors categorized beers similarly in
sorting tasks.
Concerning memory performance, consumers are as good as highly
trained assessors when beers are not learned by trained assessors.
Concerning verbal performance, although a large part of the terms used
is common between consumers and trained assessors' beer descriptions,
the communicative value of the vocabulary used by consumers is lower
than the one of trained assessors.
100
90
80
70
60
50
40
30
20
10
0
% correct matching
*** **
Exp/Exp Nov/Nov Exp/Nov Nov/Exp
Efficiency of used terms
Unfamiliar supplemented beers
► Consumers are able to perform perceptual tasks as well as moderately
trained assessors. Nevertheless, when language is needed, consumers
show lower performance than trained assessors.
100
90
80
70
60
50
40
30
20
10
0
% correct matching
***
Exp/Exp Nov/Nov Exp/Nov Nov/Exp
Fig 12. Percentages of correct matches for commercial, learned supplemented, and new supplemented beers. The horizontal axis
represents the four matching conditions: Exp/Exp (experts matching experts’ descriptions), Nov/Nov (novices matching novices’
descriptions), Exp/Nov (experts matching novices’ descriptions), Nov/Exp (novices matching experts’ descriptions).
Trained = consumers = chance level with consumers’ description.
Trained > consumers > chance level with trained assessors’ description.
Trained assessors’ descriptions could be understood by consumers.
% of efficacity
% of citation
90
80
70
60
50
40
30
20
10
0
50
45
40
35
30
25
20
15
10
5
0
2,8
2,7
2,6
2,5
2,4
2,3
2,2
2,1
Score
Nature of used terms
Consumers Trained
Taste Specific Global Somesthesic After-taste Intensity Hedonic
Consumers Trained
Consumers Trained
Confidence
Fig 9. Degree of confidence in aroma
identification task for consumers and trained
assessors (32h. of training).
Taste Specific Global Somesthesic After-taste Intensity Hedonic
Fig 11. Nature and efficiency of terms used by consumers
and trained assessors (average on the 3 sessions).
**
***