The TDS-bands plot : a new graphical tool for Temporal

The TDS-band plot : a new graphical tool for
Temporal Dominance of Sensations data
C. Monterymard, M. Visalli, P. Schlich*
INRA, Centre des Sciences du Goût et de l’Alimentation, Dijon, France
* [email protected]
INTRODUCTION & OBJECTIVES
In Temporal Dominance of Sensations (TDS ; Pineau et al., 2009), panelists are asked to describe the sequence of sensory dominances perceived during the time spent for testing
a food or a beverage. Practically, they put the product into their mouth and successively click on the attributes they perceive as dominant. Computer records durations and
intensities (optional) for each attribute elicitation. An attribute is considered dominant until another attribute is selected.
TDS data for a given product are usually represented as a set of curves which plots the dominance rate of each attribute against time. The dominance rate of an attribute at a
given time gives the proportion of panelists who found that attribute dominant over the others at that time.
Visual inspection of TDS curves is quite often focusing on the attribute which dominance rate peaks at the highest level time by time. However, when several attributes have
almost the same level of dominance rate over a period of time, this visual inspection becomes cumbersome.
The aim of this work is to propose the TDS-band plot as a mean for solving this problem.
The TDS-band plot of a product gives at each time the attribute having the highest dominance rate within an interval of time centered on this time (moving average process).
MATERIALS & METHODS
RESULTS
0.25
SYR-3%
BITTER
PUNGENT
RED FRUITS
SPICES
SWEET
WOODY
According to TDS curves, “Syrah (-3%) is first dominated by acid, then
bitter and red fruits, next bitter and astringency then red fruits
sensations.”
0.20
However, between 5 and 10 seconds, confusion is possible, because
of curves smoothing and superimposition. The dominant attribute
could be either acid, or bitter or woody. A decision must be taken.
ACIDASTRINGENT
0.15
significance_level
HEATSWEET
WOODY
BLACKCURRANT
SPICES
PUNGENT
0.10
Wines were provided and dealcoholized (reverse osmosis)
by UE Pech Rouge (INRA, Narbonne, France) :
ACID
ASTRINGENT
BITTER
BLACKCURRANT
HEAT
RED FRUITS
dominance_rate
• The TDS data come from the study of the impact of partial
dealcoholization on the sensory perception of red wines (Meillon et al.,
2009).
• 16 trained panelists TDS-profiled 4 wines in triplicate (in 3 sessions)
using 10 attributes of mouth and flavor sensations.
0.30
TDS CURVES
DATA USED AS EXAMPLE
chance_level
Thanks to TDS bands, a choice is made. The same sequence is found
with more precisions :
“Syrah (-3%) is first dominated by acid, then woody and bitter. Then
acid, bitter and red fruits are dominant. Final is heat, with short
peaks of red fruits and blackcurrant.”
0.05
Standard Syrah
13.5% (SYR_TEM)
-3%
0.00
- 1.5 %
TDS_CURVES - product : SYR_3
Syrah -3%
10.5%
(SYR_3)
Syrah -1.5%
12%
(SYR_1_5)
0
20
40
60
TDS BANDS
80
100
120
time
TDS-band plot simplifies TDS curves interpretation by deciding
on the best sequence of attribute dominances and by adding
the intensity information (if available).
+3%
Realcoholized Syrah
13.5%
(SYR_REC)
TDS_CURVES - product : SYR_TEM
0.20
0.25
80
100
…
0
4
6
6
3
4
3
1
…
0
3.3 5.3
5
5
3.3 3.7 …
4.3 3.3 2.7
2
…
(4+6+6)/3=5.3
SYR_3
SYR_3
20
40
60
80
100
time
TDS_CURVES - product : SYR_3
TDS_CURVES - product : SYR_REC
0.30
time
120
SYR_REC
SYR_REC
0.25
SYR_REC
discretization = 1.5
WS = 4.5
ASTRINGENT
BITTER
0.20
WOODY
HEAT
SPICES
RED FRUITS
ACID
HEATSWEET
WOODY
BLACKCURRANT
SPICES
PUNGENT
chance_level
0.15
significance_level
dominance_rate
ACIDASTRINGENT
0.15
dominance_rate
…
14
12
10
3
SYR_1_5
0
BITTER
0.20
2.3 4.7 5.3
8
6
4
2
0
chance_level
0.10
dominance_rate
120
0.10
5
SWEET
SPICES
PUNGENT
BLACKCURRANT
0
significance_level
20
40
60
80
Bandes-Temps
DTS
de chaqueTime
produit et(s)
intensités
100
120
temps
SWEET
chance_level
SYR_TEM
0.05
1
SYR_TEM
SYR_TEM
0.00
4
SYR_TEM
PUNGENT
BLACKCURRANT
WOODY
ASTRINGENT
HEAT
0.05
60
SYR_3
0.10
4
significance_level
0.00
40
0.05
7
0.15
0.20
0.15
0.10
0.00
20
0.00
5
BITTER ACID
2
0
…
14
12
10
8
6
4
2
0
BITTER ACID
chance_level
RED FRUITS
0
ACID
SYR_1_5
0.25
Mean number of elicitations of
attributes (panel)
moving average
process
WS
SPICES
0
For a given product, at each time, the dominant attribute is the one
with the maximum number of elicitations evaluated by the moving
average process.
Discretization (s)
significance_level
0.05
dominance_rate
Parameters defining the moving average process :
− Discretization converts continuous time measures in time intervals.
Default interval is the first decile of the distribution of all durations of
dominance.
− Window Size (WS) corresponds to the median of the distribution of
all durations of dominance.
Bandes-Temps DTS
de chaque produit et intensités
SYR_1_5
BITTER
RED FRUITS
PUNGENT
SWEET
RED FRUITS
ASTRINGENT
BLACKCURRANT
WOODY
ACID
TDS BANDS
TDS_CURVES - product : SYR_1_5
SYR_TEM
HEATBITTER
CONSTRUCTION OF BANDS
Number of elicitations
of attributes (panel)
TDS CURVES
0
20
40
60
80
100
120
0
20
40
60
80
100
120
SYR_1_5
SYR_1_5
time
In the above sample, BITTER is dominant at time 2, 4 and 10 seconds.
The resulting sequence of dominances is represented as a time-band
of rectangles whose lengths are proportional to durations of
dominance. Colors denote the attributes. Rectangles borders are plain
if dominance rate is greater than significance level* (panelist
accordance), dotted otherwise.
* Significance level expresses the smallest value of the proportion being significantly (5 %)
higher than the chance level (1/(number of attributes)) in a binomial proportion test .
time
Panel leader can change the parameters of the moving
average process in order to get a smoother sequence
with fewer changes. Consequently, the sequence shows
the main dominances only.
SYR_3
SYR_3
SYR_REC
SYR_REC
The bands of every product from a given study are
drawn on a single graph, allowing direct visual
comparison among products.
discretization = 3
WS = 9
0
20
40
60
80
100
120
Time
(s)
temps
CONCLUSION
The TDS-band plot is a useful simplification of TDS curves which helps interpretation of data.
On a single graph, TDS-band plot provides for each product : its sequence of dominant sensations, together with an assessment of significance, at panel level, of each sensation and
the evolution of their mean intensities.
The moving average process replaces TDS-curve smoothing and provides different levels of interpretation accuracy thanks to the discretization and window size parameters.
References :
Pineau, N., Schlich, P., Cordelle, S., Mathonnière, C., Issanchou, S., Imbert, A., Rogeaux, M., Etiévant, P., Köster, E. (2009). Temporal Dominance of Sensations : construction of the TDS curves and comparison with time-intensity. Food Quality and Preference, 20, 450-455
Meillon, S., Urbano, C., Schlich, P. (2009) Contribution of the Temporal Dominance of Sensations (TDS) method to the sensory description of subtile differences in partially dealcoholized red wines. Food Quality and Preference, 20, 490-499
2nd Conference of the Society of Sensory Professionals, 27-29 October 2010, Napa, CA, USA