Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: The objective of this work was to propose an alternative use to ranking method, as descriptive test, here
Received 3 September 2009 named Ranking Descriptive Analysis (RDA). RDA was compared with Free-Choice Profiling (FCP) and
Received in revised form 18 March 2010 Quantitative Descriptive Analysis (QDA). Four chocolate puddings were used as samples. A group of
Accepted 18 March 2010
assessors performed FCP. Another group of selected assessors developed a list of attributes and their cor-
Available online 27 March 2010
responding definitions. These assessors were divided into two groups: a panel was trained to perform the
QDA and other panel was familiarised with the RDA procedures. Sample discrimination was similar using
Keywords:
the three techniques. The RDA panel showed better consensus than the other two ones. The QDA showed
Ranking
Descriptive analysis
the best correlation with the instrumental analysis of color and texture. Despite the larger number of
Free-Choice Profiling assessors, RDA has the advantage of minor costs associated with the requirement of fewer sessions
Quantitative Descriptive Analysis and a smaller amount of product than required by other techniques.
Generalized Procrustes Analysis Ó 2010 Elsevier Ltd. All rights reserved.
Pudding
0950-3293/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodqual.2010.03.011
612 V.B. Richter et al. / Food Quality and Preference 21 (2010) 611–620
Ranking seems to be a facilitating procedure. There are situations (50 g) (Table 1). The formulations, based on Iop, Beleia, and Silva
in which the interest is not in the value on the scale, but whether (1999), were developed to present differences in the characteristics
there are significant differences among products and if so, in which of color, texture, aroma, and flavor. In preliminary tests was also
attributes and in which ‘‘ranking order”. considered that the formulations should reflect commercial prod-
Dairou and Sieffermann (2002) proposed a variant of FCP, ucts that were studied in previous researches (Oliveira & Benassi,
named Flash Profile (FP), and compared its efficiency with a con- 2010; Oliveira, Frasson, Almeida, & Benassi, 2004). The puddings
ventional profile method. FP was a combination of free-choice were prepared with water (1 L) and cooked under constant agita-
terms selection with a ranking method based on simultaneous pre- tion until boiling (12 min). Then, they were removed from heat,
sentation of the products set. Delarue and Sieffermann (2004) com- agitated for 2 min, and conditioned in proper containers according
pared FP and a conventional descriptive method in the evaluation to the analysis. The samples were covered, to prevent drying, and
of flavor of two kinds of fruit dairy products: yogurts and chesses. stored under refrigeration (7 ± 2 °C) for approximately 20 h before
Similar profiles were obtained by the two methods and the main analysis.
benefits of FP were its rapidity and easy of use. However, the
authors pointed that the semantic consensus obtained in the con-
2.2. Instrumental evaluation of color and texture
ventional profile allowed a more accurate description of the
products.
For texture evaluation, the puddings were conditioned in
Rodrigue et al. (2000) also suggested that a ranking test could
smooth plastic cups of 50 mL, 4.8 cm in diameter and 4 cm in
be used in the description of a product. After four training sessions,
height. The samples were analysed (15 repetitions) using a textu-
a trained panel of eight assessors evaluated 10 attributes of sweet
rometer TAX-T2 (Stable Micro Systems) (Surrey, England), using
corn using the conventional method. An untrained panel of 20
cylindrical stainless probe P35, distance of penetration of 3 mm
assessors performed the same task on the same samples using a
and force of 0.05 N. Hardness (N), cohesiveness (dimensionless),
ranking procedure, but only one session was conducted to familiar-
springiness (dimensionless), adhesiveness (Ns), and gumminess
ise assessors with samples and ranking procedures. The results
(N) were evaluated.
from both methods were similar in terms of overall product dis-
For color analysis, samples were conditioned in Petri plates of
crimination, but slight discrepancies were found in the discrimi-
9 cm in diameter (three plates for sample). The readings were
nating attributes between panels. The findings suggest that when
made (three repetitions for plate) using colorimeter Minolta
time is insufficient to train a panel, the use of an untrained panel
CR10 (Tokyo, Japan), with a reading area of 8 mm, lighting CIE
and ranking test should be considered.
D65, and observer standard CIE 10° angle. The colorimeter directly
Although Rodrigue et al. (2000) and Delarue and Sieffermann
supplied the values of L* (lightness), a* (red–green component),
(2004) have described a good performance by a ranking panel, they
and b* (yellow–blue component) and the hue (H* = arc tang
also considered that other studies would be necessary to verify
(b*/a*)) was calculated.
whether ranking tests could be used fruitfully or adapted to a
A randomised design was applied and the results were submit-
wider range of products and a greater number of attributes. An-
ted to the analysis of variance (ANOVA) and Tukey test (p < 0.05)
other consideration is that in the study described by Rodrigue
(STATSOFT STATISTICA for Windows – Program manual, 1995).
et al. (2000), the attributes were pre-defined and not developed
by the panels, although they were validated by the panel. For un-
trained and completely naive assessors, grasping the relation be- 2.3. Sensory analyses
tween attribute and perception in one session could be a difficult
task. A standardisation of samples’ characteristics, assessors’ sensi- 2.3.1. Testing procedure and glossary development
bility, and the development of attributes by the panels would allow Samples (40 mL) were served cold (taken immediately from the
a better comparison of the efficiency of the methods. refrigerator) in covered 100 mL transparent plastic cups, labelled
Chocolate pudding was chosen as sample in the current work with random-three digit codes. White light was used and the pre-
because it is easy to prepare, it allows definition and control of dif- sentation order was randomised by session.
ferences in appearance, aroma, flavor, and texture attributes and it Attribute terms for the evaluation of samples in the three
has been frequently used in descriptive methods studies in the lit- methods were developed by the panel using the methodology de-
erature (Kilcast & Clegg, 2002; Lethuaut et al., 2005; Weenen, Jell- scribed in Moskowitz (1983). Assessors were requested to record
ema, & Wijk, 2005; Wijk, Gemert, Terpstra, & Wilkinson, 2003; the similarities and differences between each sample. Two ses-
Wijk, Prinz, & Janssen, 2006). sions were conducted. In each session, a pair of samples was pre-
The objective of this study was to propose a descriptive sensory sented by session, in order to obtain the highest possible number
method of simple and fast application for situations in which the of attributes with regard to appearance, flavor, texture, and aro-
magnitude of attributes or the distance between categories is less ma. One pair was composed of samples B and C and the other
relevant. The method, here called Ranking Descriptive Analysis
(RDA), was compared to two traditional descriptive methods,
Table 1
Quantitative Descriptive Analysis and the Free-Choice Profiling, Other ingredients of pudding formulations (in gram).
using chocolate puddings with sugar and different sweeteners as
Ingredients Samples
the sample.
A B C D
(g) (g) (g) (g)
pair, samples A and D. A detailed description of the glossary ardised with skimmed milk and brown food colourants. After the
development for each method is presented on the next sections tests, the group was composed of 33 selected assessors.
(2.3.2 and 2.3.3).
In the three studied methods, during evaluation sessions, asses- 2.3.3.2. Definition of the descriptive terminology. After attributes
sors received the developed glossary with the correspondent score development, assessors selected the most frequent terms to com-
sheet. For each panel, a protocol with instructions was offered pose the score sheet. In a second session, assessors defined attri-
reminding how to observe appearance, odour, flavor, texture, and butes and developed a consensus evaluation procedure using the
mouthfeel after swallowing. suggested samples and market products as qualitative references
For texture, it was emphasized that the assessor would need to (Table 2).
cut the sample with the spoon, to compress a portion of the sample A final session was conducted to verify the consensus on the list
between the tongue and the palate, and ‘‘to chew” the sample until of attributes, glossary, and protocol of instructions. After this
ready to swallow. The protocol was fixed in the booth during all of phase, the assessors were randomly divided into two panels, one
the evaluation sessions. For the FCP, the protocol was only used to panel with 21 people for the RDA and the other with 12 people
facilitate the evaluation. For QDA and RDA, the protocol was the for the QDA. Training and samples evaluation of each panel were
group consensus’ result. conducted separately.
Table 2
Attributes, definitions and references used by QDAa,b and RDAb panels to describe the sensory properties of chocolate puddings.
increasing intensity. The results were evaluated using Friedman 3.2. Sensory characterisation
test (Newell & MacFarlane, 1987) to evaluate sample differences
for each attribute. In order to compare the configuration obtained 3.2.1. Evaluation of assessor’s performance: comparison among panel’s
with those observed in the two other methods, data was also eval- behaviour
uated by GPA, as described for the FCP. Analysis of the general configuration and residual variance of
assessors of each panel demonstrated that there was consensus
3. Results in the three panels, without behaviour discrepancies among mem-
bers (Figs. 1 and 2).
3.1. Instrumental characterisation of texture and color The QDA panel (Figs. 1B and 2B), which received quantitative
training, and the FCP panel (Figs. 1A and 2A), which did not receive
Pudding samples showed different texture and color profiles training, showed a greater dispersion of the assessors in the gen-
(Table 3). Sample D was characterised as more firm and gummy eral configuration and in the residual variance (maximum of 1.2%
than puddings B and C and it was also identified as more clear and 1.5%, respectively) than the RDA panel; which showed less dis-
and yellowish than the others. Pudding C was differentiated as persion and a low residual variance (0.5%) (Figs. 1C and 2C).
the darkest sample, less hard and gummy, and more elastic than
samples A and B. Sample B was characterised as the least adhesive
3.2.2. Samples evaluation
and presented intermediate values of hardness, lightness, and hue.
Fig. 3A and B show configuration of samples consensus in the
Sample A was the most reddish sample.
FCP and the QDA; the products are represented by triangles that
indicate the repeatability (the higher the distance between the ver-
Table 3 tices, the less repeatability). Both panels presented good repeat-
Instrumental characterisation of textureA and color.B,C ability for all samples.
The attributes better correlated with the first two dimensions
Attributes Samples
for each assessor were represented for FCP (Table 4), QDA (Table
A B C D
5) and RDA (Table 6) panels. To facilitate the evaluation and com-
ab b c
Hardness (N) 1.7 ± 0.2 1.6 ± 0.2 1.4 ± 0.1 1.8 ± 0.1a parison between methods, there were considered the attributes
Adhesiveness (Ns) 0.4 ± 0.1a 0.2 ± 0.2b 0.4 ± 0.5a 0.4 ± 0.1a
that presented higher correlation (minimum |0.5|) in each dimen-
Springiness 0.86 ± 0.05b 0.86 ± 0.05b 0.89 ± 0.0a 0.90 ± 0.01a
Cohesiveness 0.57 ± 0.02a 0.57 ± 0.01a 0.57 ± 0.02a 0.57 ± 0.12a sion for each assessor.
Gummines (N) 0.99 ± 0.12ab 0.93 ± 0.09b 0.81 ± 0.07c 1.04 ± 0.06a Considering the consensus configuration, 38% of the observed
Lightness 36.5 ± 0.2b 36.5 ± 1.2b 35.2 ± 0.6c 38.5 ± 0.4a variability was explained for the FCP. Dimension 1, responsible
Hue 46.4 ± 0.4c 47.9 ± 0.4b 47.9 ± 0.6b 49.5 ± 2.1a for 25% of the variance, could be explained (in the negative direc-
A
Average of 15 analyses ± standard deviation. tion) for the attributes brown color, sweet taste, and chocolate fla-
B
Average of nine analyses ± standard deviation. vor. These terms had high correlation for most of the panel
C
Different letters in the same line indicates significant differences (p < 0.05). members (nine assessors). Pudding D, located at the right side,
V.B. Richter et al. / Food Quality and Preference 21 (2010) 611–620 615
Fig. 1. Graphic representation of the assessors dispersion along the two first
dimensions: general configuration of assessors in FCP (A), QDA (B) and RDA (C)
panels. Numbers (1–21) indicates assessors. Fig. 2. Residual variance (%) associate with each assessor in FCP (A), QDA (B) and
RDA (C) panels.
was considered the clearest sample, with less sweet taste and
chocolate flavor. The samples configured at the left side, B, A, and butes of brown color (12 assessors), chocolate flavor (10), choco-
C (in this sequence), were characterised as presenting more ac- late aroma (9), creaminess (8), and brightness (6), and, in the posi-
cented color and flavor of chocolate, and also more sweet taste. tive direction, by attribute consistency (5). Dimension 2 was
Dimension 2 (13% of the variance) separated samples primarily correlated negatively to sweet taste and positively to residual bit-
by bitter taste and residual bitter taste. Puddings C and D, located ter taste. Therefore, samples located at the upper side were charac-
in the superior part of the graph, were characterised by higher terised by a higher intensity of residual bitter taste and lower
intensity of bitter taste (Fig. 3A, Table 4). sweet taste. Samples configured at the left side were described as
In the QDA method, the first two dimensions were responsible more characteristic of chocolate (brown color, aroma and flavor
for 46% of the variance (32% for dimension 1 and 14% for dimension of chocolate) and those located towards the right side were charac-
2). Dimension 1 was explained, in the negative direction, by attri- terised as more consistent (Fig. 3B, Table 5).
616 V.B. Richter et al. / Food Quality and Preference 21 (2010) 611–620
4. Discussion
Table 4
Attributes better correlated (r) with the first two dimensions for each assessora on FCP.
Aa Dimension 1 Dimension 2
1 Brown color ( 0.78); sweet taste ( 0.95); chocolate flavor ( 0.78) and brightness ( 0.61) Bitter taste (0.72); crust thickness (0.50) and crust firmness (0.51)
2 Brown color ( 0.85); sweet taste ( 0.92); chocolate flavor ( 0.70) and brightness ( 0.69) Bitter taste (0.77) and chocolate aroma ( 0.54)
3 Brown color ( 0.65) and sweet taste ( 0.50) Bitter taste (0.52); residual bitter taste (0.50); chocolate aroma
(0.50) and syneresis (0.57)
4 Chocolate color ( 0.57); sweet taste ( 0.57); chocolate flavor ( 0.50); consistency Bitter taste (0.83); sour taste (0.86) and consistency ( 0.53)
(appearance) ( 0.56) and firmness (cut) (0.74)
6 Sweet taste ( 0.82); chocolate flavor ( 0.81) and bubbles ( 0.63) Bitter taste (0.69); residual bitter taste (0.71) and chocolate aroma
(0.51)
7 Brown color ( 0.76); sweet taste ( 0.50) and chocolate flavor ( 0.64) Bitter taste (0.59); chocolate aroma (0.52) and creamy texture
( 0.61)
8 Brown color ( 0.59) and sweet taste ( 0.88) Smooth surface (0.57); consistency (cut) (0.53) and
homogeneousness (0.53)
9 Sweet taste ( 0.74); chocolate flavor ( 0.72) and consistency (cut) (0.69) Surface crust (0.59)
10 Chocolate flavor ( 0.73); chocolate aroma (0.59) and homogeneous surface ( 0.69) Sweet taste ( 0.69)
11 Chocolate color ( 0.87); sweet taste ( 0.83); chocolate flavor ( 0.77); chocolate aroma Consistency (mouth) ( 0.50)
( 0.79) and homogeneousness ( 0.91)
13 Brown color (0.64); chocolate flavor ( 0.79); chocolate aroma ( 0.60) and consistency Sweet taste ( 0.65)
(cut) (0.76)
14 Brown color (0.60) and smooth surface (0.57) Bitter taste (0.69); sweet taste ( 0.71); chocolate flavor ( 0.58)
and viscosity (0.51)
a
Assessors were identified by numbers 1–14.
Table 5
Attributes better correlated (r) with the first two dimensions for each assessora on QDA.
A Dimension 1 Dimension 2
1 Brown color ( 0.88); chocolate aroma ( 0.78); chocolate flavor ( 0.67) and residual Creaminess ( 0.63); consistency (0.65) and sweet taste ( 0.55)
bitter taste (0.67)
2 Brown color ( 0.76); chocolate aroma ( 0.61) and chocolate flavor ( 0.71)) Homogeneousness (0.51) and sweet taste ( 0.73)
3 Brown color ( 0.73); chocolate aroma ( 0.75); creaminess ( 0.87); consistency (0.77) Chocolate flavor (0.73) and sweet taste ( 0.64)
and residual bitter taste (0.84)
4 Brown color ( 0.76); brightness ( 0.80); chocolate aroma ( 0.52); creaminess ( 0.64); Sweet taste ( 0.74) and residual bitter taste (0.87)
consistency (0.77) and chocolate flavor ( 0.50)
5 Brown color ( 0.87); brightness ( 0.51); chocolate aroma ( 0.69); creaminess ( 0.70); Sweet taste ( 0.51)
consistency (0.51); chocolate flavor ( 0.50) and residual bitter taste ( 0.68)
6 Brown color ( 0.85); brightness ( 0.82); homogeneousness ( 0.74); chocolate aroma Sweet taste ( 0.52); consistency ( 0.51) and residual bitter taste ( 0.50)
( 0.64)
and chocolate flavor ( 0.82)
7 Brown color ( 0.83); brightness ( 0.58); chocolate aroma ( 0.69); creaminess ( 0.85);
consistency (0.64); chocolate flavor ( 0.68); sweet taste ( 0.83) and residual bitter
taste (0.67)
8 Brown color ( 0.71); brightness ( 0.75); chocolate aroma ( 0.51); creaminess ( 0.60) Sweet taste ( 0.67) and residual bitter taste (0.51)
and chocolate flavor ( 0.71)
9 Brown color ( 0.72) and brightness ( 0.61) Residual bitter taste (0.53)
10 Brown color ( 0.78); creaminess ( 0.79); consistency (0.77) and chocolate flavor Sweet taste ( 0.65) and residual bitter taste (0.94)
( 0.72)
11 Brown color ( 0.81); creaminess ( 0.67); chocolate flavor ( 0.88) and sweet taste Residual bitter taste (0.64)
( 0.79)
12 Brown color ( 0.56); chocolate aroma ( 0.57) and creaminess ( 0.82) Sweet taste ( 0.81) and residual bitter taste (0.61)
a
Assessors were identified by numbers 1–12.
These results agreed with those described by Rodrigue et al. that considered D as presenting the least intensity and C as the
(2000) in a study with two panels: one trained for descriptive anal- most intensity of brown color (Fig. 3 and Tables 4–6). Samples A
ysis and another not trained for the ranking test. Results from the and B, which were configured close in sensory methods with re-
study showed similarities between the trained and untrained pan- gard to dimension 1 (high correlation with attribute brown color)
els. These authors suggest that when time is insufficient to train a (Fig. 3 and Tables 4–6), did not present a difference in L* value (Ta-
panel, the use of an untrained panel and a ranking test should be ble 3).
considered. However, we observed it was important to train a pa- In a general approach, similar profiles were observed in the
nel in order to obtain good descriptor conceptualisation and great- samples configuration by GPA for the three methods (Fig. 3, Tables
er panel consensus. 4–6). The three studied methods described similar intensity of
Consensus across the characterisations of the samples obtained sweet taste and chocolate flavor in samples A and B, the highest
the descriptive analysis methods (Fig. 3, Tables 4–6) and the intensity of these attributes in sample C, and the lowest in sample
instrumental evaluation of color and texture (Table 3) was D. Samples C and D were described as presenting a higher residual
observed. bitter taste than samples A and B.
With regard to appearance, instrumental measure of color de- The QDA and RDA panels presented similar results for aroma
scribed sample D as the clearest sample and C as the darkest sam- attribute, both characterised sample C with intense aroma of choc-
ple (Table 3). It agreed with the observed in all sensorial methods olate (Tables 7 and 8).
618 V.B. Richter et al. / Food Quality and Preference 21 (2010) 611–620
Table 6
Attributes better correlated (r) with the first two dimensions for each assessora on RDA.
A Dimension 1 Dimension 2
1 Brown color ( 0.99); homogeneousness ( 0.86) and creaminess ( 0.86) Brightness ( 0.84); consistency (0.92); chocolate flavor ( 0.82) and
residual bitter taste (0.82)
2 Brown color ( 0.75); brightness ( 0.99); creaminess (0.76); consistency ( 0.76) and Chocolate aroma ( 0.82); residual bitter taste (0.51); creaminess
chocolate flavor ( 0.99) ( 0.65) and consistency (0.65)
3 Brown color ( 0.99); brightness ( 0.99); homogeneousness ( 0.99); consistency Creaminess ( 0.84) and residual bitter taste (0.51)
(0.86) and chocolate flavor ( 0.52)
4 Homogeneousness ( 0.86); brightness ( 0.76); chocolate aroma ( 0.76); creaminess Brown color ( 0.84) and consistency (0.67)
( 0.86); sweet taste ( 0.99); chocolate flavor ( 0.86) and residual bitter taste ( 0.99)
5 Brown color ( 0.75); homogeneousness ( 0.86); creaminess (0.76) and consistency Chocolate aroma ( 0.92); sweet taste ( 0.92) and residual bitter taste (0.92)
( 0.76)
6 Brown color ( 0.75); brightness (0.99); creaminess ( 0.76); sweet taste ( 0.86) and Consistency (0.67)
residual bitter taste (0.86)
7 Brown color ( 0.99); brightness (0.99); chocolate flavor ( 0.76) and sweet taste Chocolate aroma ( 0.92); creaminess ( 0.57) and consistency (0.57)
( 0.86)
8 Brown color ( 0.86); homogeneousness (0.86); chocolate aroma ( 0.99); creaminess Brightness ( 0.92) and residual bitter taste (0.82)
(0.99); consistency ( 0.99); chocolate flavor ( 0.76) and sweet taste ( 0.99)
9 Brown color ( 0.99); chocolate aroma ( 0.99); creaminess (0.75); consistency Brightness ( 0.57)
( 0.75); chocolate flavor ( 0.99); sweet taste ( 0.99) and residual bitter taste (0.99)
10 Brown color ( 0.86); brightness (0.76); homogeneousness (0.86); chocolate aroma Residual bitter taste (0.92)
( 0.86); creaminess (0.75); consistency ( 0.76); chocolate flavor ( 0.99) and sweet
taste ( 0.86)
11 Brown color ( 0.99); chocolate aroma ( 0.76); creaminess (0.99); consistency Brightness ( 0.67) and residual bitter taste ( 0.57)
( 0.99); chocolate flavor ( 0.99) and sweet taste ( 0.52)
12 Brown color ( 0.86); brightness (0.86); creaminess (0.75); consistency ( 0.75) Homogeneousness ( 0.92); chocolate aroma (0.57); chocolate flavor
and sweet taste ( 0.86) (0.82) and residual bitter taste (0.51)
13 Chocolate flavor ( 0.99) and sweet taste ( 0.86) Brown color ( 0.84); brightness (0.84); chocolate aroma ( 0.84);
creaminess (0.51); consistency ( 0.51) and residual bitter taste (0.51)
14 Brown color ( 0.99); homogeneousness ( 0.75); chocolate aroma ( 0.99) and Brightness (0.92); sweet taste ( 0.92); residual bitter taste (0.82);
chocolate flavor ( 0.76) creaminess ( 0.57) and consistency (0.57)
15 Brown color ( 0.99); brightness ( 0.99); homogeneousness ( 0.86); chocolate aroma Creaminess (0.84) and consistency (0.84)
( 0.99); chocolate flavor ( 0.99); sweet taste ( 0.99) and residual bitter taste (0.99)
16 Brown color ( 0.99); brightness ( 0.86); homogeneousness ( 0.75); chocolate aroma Creaminess ( 0.67)
( 0.99); chocolate flavor ( 0.86) and sweet taste ( 0.86)
17 Brown color ( 0.99); chocolate flavor ( 0.86) and sweet taste ( 0.99) Homogeneousness ( 0.82); brightness (0.82); chocolate aroma (0.82);
creaminess ( 0.82); consistency (0.82) and residual bitter taste (0.92)
18 Brown color ( 0.99); chocolate aroma ( 0.86); consistency ( 0.99) and chocolate Brightness ( 0.82); creaminess ( 0.67); sweet taste ( 0.84) and
flavor ( 0.76) residual bitter taste (0.82)
19 Homogeneousness ( 0.99); chocolate aroma ( 0.86); creaminess (0.75); consistency Brown color ( 0.92); brightness ( 0.92); chocolate flavor (0.82) and
( 0.75) and sweet taste ( 0.86) residual bitter taste (0.82)
20 Brown color ( 0.99); brightness ( 0.99); homogeneousness (0.86); chocolate aroma Residual bitter taste (0.92)
( 0.99); creaminess (0.76); consistency ( 0.76); chocolate flavor ( 0.86) and sweet
taste ( 0.86)
21 Brown color ( 0.99); chocolate aroma ( 0.76); creaminess (0.99); consistency Homogeneousness ( 0.92) and residual bitter taste (0.92)
( 0.99); chocolate flavor ( 0.99) and sweet taste ( 0.75)
a
Assessors were identified by numbers 1–21.
Table 7 Table 8
Values of sensory attribute intensities analyzed by QDA.A,B Characterization of samples by RDA.A,B
The greatest differentiation between the methods’ perfor- hardness and gumminess less intense than sample D (Table 3).
mance was observed in texture evaluation. The QDA demon- These parameters cold be associated with the sensory attribute
strated better correlation with the instrumental results, of consistency (Table 2). In this way, it seems that the qualitative
considering sample B less consistent than D (Table 7). The RDA training (use of reference standards and glossary) was not suffi-
panel considered sample A more consistent than B and D (Table cient for the consensual use of texture attributes by QDA and
8). In the instrumental profile of texture, sample B presented RDA panels.
V.B. Richter et al. / Food Quality and Preference 21 (2010) 611–620 619
Table 9
Comparison of the requirements in the descriptive sensory methods studied.
Comparing instrumental data with sensory methods, texture (qualitative aspect) and final analysis using the ranking procedure
attributes analysed by QDA showed good agreement. The assessors for each attribute, allowed the discrimination of the samples with
of QDA and RDA were submitted to the same pre-selection tests and efficiency similar to that displayed by the descriptive methods of
they all participated in the attributes development, and they had a the QDA and FCP.
‘‘qualitative training” with the samples of reference. Therefore, all The facility of the procedure for ranking samples (ordinal scale),
of the assessors were considered to have the same perception for in comparison with the use of interval scales used in the traditional
each attribute. In this work, the better performance for texture descriptive techniques, showed that the RDA panel presented
attributes of QDA panel could be attributed to the training with greater consensus. However, for some attributes, the training of
scale endpoints. Although the three methods studied show efficient the QDA panel facilitated the agreement of the definitions. A more
samples discrimination and consensus among assessors, the QDA intense qualitative training for RDA would allow more consistent
training facilitated the understanding of the attributes definitions. results, primarily for complex attributes of texture.
A comparison of the applicability of the sensory methods (num- Despite the larger number of assessors required, RDA has the
ber of assessors, number of samples and time required) could be advantage of minor costs associated with the requirement of fewer
observed in Table 9. The data, however, do not consider the addi- sessions than QDA and a smaller amount of product than required
tional difficulty in the definition and preparation of the products by other sensory techniques.
used in the stages of pre-selection and standards for training
(QDA and RDA). The FCP was the fastest method (seven sessions), References
since some stages conducted in the QDA and the RDA were elimi-
nated from the FCP (pre-selection, consensus for attributes, train- Barylko-Pikielna, N., Matuszewska, I., Jeruszka, M., Kozlowska, K., Brzozowska, A., &
ing, and selection), implying an economy of samples in relation Roszkowski, W. (2004). Discriminability and appropriateness of category
scaling versus ranking methods to study sensory preferences in elderly. Food
to the QDA. Due to the procedure of evaluation (three repetitions), Quality and Preference, 15, 167–175.
even with fewer assessors, the FCP required a larger amount of Dairou, V., & Sieffermann, J.-M. (2002). A comparison of 14 jams characterized by
samples when compared with the RDA. conventional profile and a quick original method, the flash profile. Journal of
Food Science, 67, 826–834.
The initial procedures for the QDA and RDA were similar. In Delarue, J., & Sieffermann, J.-M. (2004). Sensory mapping using flash profile.
these two methods, the assessors underwent a pre-selection test Comparison with a conventional descriptive method for the evaluation of the
and qualitative training sessions to improve their perception and flavour of fruit dairy products. Food Quality and Preference, 15, 383–392.
Dijksterhuis, G. B., & Gower, J. C. (1991). The interpretation of generalized
to consensually define the attributes; this required the analyst to
procrustes analysis and allied methods. Food Quality and Preference, 3, 67–87.
have experience in conducting these initial procedures. In RDA, Gower, J. C. (1975). Generalized procrustes analysis. Psychometrika, 40, 33–51.
due to the facility of application technique in the final evaluation Hein, K. A., Jaeger, S. R., Carr, T. B., & Delahunty, C. M. (2008). Comparison of five
common acceptance and preference methods. Food Quality and Preference, 19,
of the samples, a reduction in the number of sessions and, conse-
651–661.
quently, the number of samples required compared with the Iop, S. C. F., Beleia, A. P., & Silva, R. S. F. (1999). Formulation and evaluation of dry
QDA, was observed. However, it should be considered that since dessert mix containing sweetener combinations using mixture response
evaluation using RDA method was made only once for each sam- methodology. Food Chemistry, 66, 167–171.
Kilcast, D., & Clegg, S. (2002). Sensory perception of creaminess and its relationship
ple, it did not allow to check repeatability of panel performance. with food structure. Food Quality and Preference, 13, 609–623.
All results were analysed using GPA; however, ANOVA can also Kim, K., & O’Mahony, M. A. (1998). New approach to category scales of intensity:
be used for the QDA and the Test of Friedman for the RDA. This rep- Traditional vs. rank-rating. Journal of Sensory Studies, 13, 241–249.
Lee, H. S., Hout, D., & O’Mahony, M. (2007). Sensory difference tests for margarine: A
resents an advantage in relation to the FCP in view of the small comparison of R-indices derived from ranking and A–Not A methods
availability GPA programmes. considering response bias and cognitive strategies. Food Quality and
The requirement, in quantitative terms, of assessors, sessions, Preference, 18, 675–680.
Lee, H. S., & O’Mahony, M. (2005). Sensory evaluation and marketing: Measurement
and samples will no doubt vary in relation to specific situations of a consumer concept. Food Quality and Preference, 16, 227–235.
(e.g., the requirement for assessor retraining or new assessor selec- Lethuaut, L., Brossard, C., Meynier, A., Rousseau, F., Llamas, G., Bousseau, B., et al.
tion in QDA). (2005). Sweetness and aroma perceptions in dairy desserts varying in sucrose
and aroma levels and in textural agent. International Dairy Journal, 15, 485–493.
Liem, D. G., Mars, M., & Graaf, C. (2004). Consistency of sensory testing with 4- and
5. Conclusions 5-year-old children. Food Quality and Preference, 15, 541–548.
McEwan, J. A. (1999). Comparison of sensory panels: A ring trial. Food Quality and
Preference, 42, 161–171.
The method of Ranking Descriptive Analysis (RDA), using the Meilgaard, M. C., Civille, G. V., & Carr, B. T. (1999). Sensory evaluation techniques.
pre-selection of assessors, attributes development and definition Boca Raton: CRC Press.
620 V.B. Richter et al. / Food Quality and Preference 21 (2010) 611–620
Moskowitz, H. R. (1983). Product testing and sensory evaluation of foods – Marketing STATSOFT STATISTICA for Windows – Program manual, 1995. Tulsa: Statsoft, Inc.
and R&D approaches. Westport: Food and Nutrition Press Inc. Stone, H., & Sidel, J. L. (1998). Quantitative descriptive analysis: Developments,
Murray, J. M., Delahunty, C. M., & Baxter, I. A. (2001). Descriptive sensory analysis: applications, and the future. Food Technology, 52, 48–52.
Past, present and future. Food Research International, 34, 461–471. Villanueva, N. D. M., Petenate, A. J., & Da Silva, Maria A. A. P. (2005). Performance of
Newell, G. J., & MacFarlane, J. D. (1987). Expanded tables for multiple comparison the hybrid hedonic scale as compared to the traditional hedonic, self-adjusting
procedures in the analysis of ranked data. Journal of Food Science, 52, 1721–1725. and ranking scales. Food Quality and Preference, 16, 691–703.
Oliveira, A. P. V., & Benassi, M. T. (2010). Avaliação sensorial de pudins de chocolate Weenen, H., Jellema, R. H., & Wijk, R. A. (2005). Sensory sub-attributes of creamy
com açúcar e dietéticos por Perfil Livre. Ciência e Agrotecnologia, 34, 146–154. mouthfeel in commercial mayonnaises, custard desserts and sauces. Food
Oliveira, A. P. V., Frasson, K., Almeida, T. C. A., & Benassi, M. T. (2004). Aceitação de Quality and Preference, 16, 163–170.
sobremesas lácteas dietéticas e formuladas com açúcar: teste afetivo e mapa de Wijk, R. A., Gemert, L. J., Terpstra, M. E. J., & Wilkinson, C. L. (2003). Texture of semi-
preferência interno. Ciência e Tecnologia de Alimentos, 24, 627–633. solids, sensory and instrumental measurements on vanilla custard desserts.
OP & P Product Research (1998). Senstools Versão 2.3. Utrecht: OP & P Product Food Quality and Preference, 14, 305–317.
Research. Wijk, R. A., Prinz, J. F., & Janssen, A. M. (2006). Explaining perceived oral texture of
Penna, E. W. (1980). Evaluacion sensorial: una metodologia actual para tecnología de starch-based custard desserts from standard and novel instrumental tests. Food
alimentos. Santiago: Universidade do Chile. Hydrocolloids, 20, 24–34.
Rodrigue, N., Guillet, M., Fortin, J., & Martin, J. F. (2000). Comparing information Williams, A. A., & Langron, S. P. (1984). The use of free-choice profiling for the
obtained from ranking and descriptive tests of four sweet corn products. Food evaluation of commercials ports. Journal of the Science of Food and Agriculture,
Quality and Preference, 11, 47–54. 35, 558–568.