Food Quality and Preference 30 (2013) 169–179

Contents lists available at SciVerse ScienceDirect

Food Quality and Preference

journal homepage: www.elsevier.com/locate/foodqual

Quantitative sensory description using the Optimized Descriptive

Profile: Comparison with conventional and alternative methods for
evaluation of chocolate
Rita de Cássia dos Santos Navarro da Silva a,⇑, Valéria Paula Rodrigues Minim a,
João de Deus Souza Carneiro b, Moysés Nascimento c, Suzana Maria Della Lucia d, Luís Antônio Minim a
a
Departamento de Tecnologia de Alimentos, Universidade Federal de Viçosa (UFV), Zip Code: 36570-000 Viçosa, Minas Gerais, Brazil
b
Departamento de Ciência dos Alimentos, Universidade Federal de Lavras (UFLA), Zip Code: 37200-000 Lavras, Minas Gerais, Brazil
c
Departamento de Estatística, Universidade Federal de Viçosa (UFV), Zip Code: 36570-000 Viçosa, Minas Gerais, Brazil
d
Departamento de Engenharia de Alimentos, Universidade Federal do Espírito Santo (UFES), Zip Code: 29500-000 Alegre, Espírito Santo, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: Conventional descriptive methods require extensive training time of the sensory team to ensure that the
Received 24 January 2013 vocabulary and rating scales are used consistently. Seeking to reduce the sensory analysis time, the Rank-
Received in revised form 17 May 2013 ing Descriptive Analysis (RDA) was developed in the last decade which enabled a satisfactory reduction in
Accepted 18 May 2013
time, but allowed only for a qualitative assessment of sensory attributes. Accordingly, the Optimized
Available online 29 May 2013
Descriptive Profile (ODP) was proposed which also sought to reduce the time needed for descriptive anal-
ysis and at the same time estimate the magnitude of sensory difference between the products. This tech-
Keywords:
nique was applied in characterization of creamy cheeses and showed similar results to the conventional
Ranking descriptive analysis
Conventional profile
method. In an attempt to perform a more detailed study of the ODP, the authors saw the need to apply
Time reduction this sensory characterization method to products with smaller magnitudes of difference in sensory attri-
Discrimination of products butes. Thus, the objective of the present study was to use the ODP to obtain the sensory profile of choc-
olate, as well as compare the sensory description acquired by this method to the profile obtained by the
conventional (CP) and alternative methodologies (RDA). For this purpose three teams of judges partici-
pated in the sensory evaluation of chocolates, where each team utilized an assessment protocol according
to the three different methodologies. It was found that the three descriptive techniques presented similar
sensory profiles for the chocolates. The ODP and CP showed greater discrimination in relation to the RDA
for the four chocolate formulations. The form in which the unstructured scale was used by the judges pre-
sented different behaviors in the two quantitative techniques, where in the ODP the scores were more
homogeneous in comparison to the PC. Thus, sensory characterization by means of the ODP was similar
to the alternative and conventional methods, but with higher discrimination power for the formulations.
The ODP permitted a 50% reduction in sensory time compared to the CP without loss of information and
with high discrimination power of chocolates and quantitative description of the sensory attributes.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction & Sieffermann, 2002; Delarue & Sieffermann, 2004; Labbe, Ritz, &
Hugi, 2004; Richter, Almeida, Prudencio, & Benassi, 2010; Rodri-
Conventional descriptive methods present long execution times gue, Guillet, Fortin, & Martin, 2000; Silva et al., 2012). According
of the sensory technique, since they require extensive training of to Delarue and Sieffermann (2004), the main limitation of the con-
judges to ensure that the vocabulary and evaluation scales are used ventional descriptive analysis is the time consumed and often this
consistently, and that the judges present consensus, ability to dis- leads to non-execution of this valuable quality tool. This fact
criminate samples and repeatability of results. The training phase underscores the need for a rapid assessment in the practical con-
is extensive and the time demand is considered a limiting factor text of industry.
for its practical application in industry (Cartier et al., 2006; Dairou Several researchers have been working to develop alternative
methodologies to conventional descriptive methods, which require
less time for acquisition of the sensory profile, so as to permit its
⇑ Corresponding author. Tel.: +55 31 38993810.
use in the food industry. Found in recent publications are method-
E-mail addresses: [email protected] (R.C.S.N. Silva), [email protected] (V.P.R. Minim),
[email protected]fla.br (J.D.S. Carneiro), [email protected] (M. Nascimento), ologies that utilize the ranking procedure to evaluate products in
[email protected] (S.M. Della Lucia), [email protected] (L.A. Minim). relation to descriptive attributes, seeking to facilitate the evalua-

0950-3293/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.foodqual.2013.05.011
170 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179

tion procedure performed by the judges. In these methods it is nec- in the food industry. Found in recent literature are methodologies
essary only to compare the intensity of the products instead of that utilize the ranking procedure to evaluate products with re-
‘‘calibration’’ of the sensory memory. One methodology that uses spect to descriptive attributes, aiming to facilitate the evaluation
this classification/ordering process has been proposed recently, procedure performed by the judges. In these methods it is only
the ranking descriptive analysis (Richter et al., 2010; Rodrigue necessary to compare the intensity of the products instead of ‘‘cal-
et al., 2000). This method showed satisfactory reduction in test ibration’’ of sensory memory.
time for sensory description, but there are limitations in its appli- The ranking descriptive analysis is one methodology that uses
cability due to qualitative evaluation of the samples. the classification/ranking process, which has been proposed re-
The Optimized Descriptive Profile (ODP), proposed by Silva et al. cently in response to the demand for faster descriptive methods
(2012), seeks to meet the demand for fast descriptive methods, and (Richter et al., 2010). The first use of the ranking technique as a
at the same time provide quantitative information regarding the descriptive test was performed by Rodrigue et al. (2000) in the sen-
sensory attributes present in foods. The ODP focuses on quantita- sory characterization of sweet corn. Richter et al. (2010) standard-
tive characterization of the sensory attributes, permitting the anal- ized the methodology and termed the technique as the ‘‘ranking
yses of stability, quality control, optimization of formulations and descriptive analysis’’ (RDA). In this methodology it was recom-
correlations between sensory and instrumental measurements mended to use qualitative references to avoid interpretation errors
are performed. of the attributes. In this study, the authors compared the sensory
Thus, in order to focus on this interesting alternative method of profile obtained by ranking of the samples using an untrained team
sensory description, the authors saw a need to apply the ODP tech- (RDA) and according to the classical methodologies quantitative
nique to another food matrix, this time in a food with a lower mag- descriptive analysis and free profile. Sensory characterization of
nitude of difference in the intensity of the sensory characteristics. the products by means of the three methodologies was similar.
In this sense, the contribution of the present study was to apply the This alternative methodology meets the need for faster methods
ODP for obtaining the sensory profile of chocolates which present for the context of industry, but there are limitations to the applica-
small magnitudes of difference in sensory characteristics, and to bility of this method due to the qualitative evaluation of the sam-
compare the sensory description of this methodology with the con- ples. By means of the techniques presented, it is not possible to
ventional profile (CP) and the alternative method ranking descrip- identify the magnitude of the difference between the samples,
tive analysis (RDA). and therefore these methods are not recommended for stability
analyzes, quality control and optimization of formulations.
2. Literature review: presentation of the descriptive
methodologies
2.3. Optimized Descriptive Profile
Traditional descriptive sensory tests involve the detection and
qualitative and quantitative description of the sensory attributes The Optimized Descriptive Profile (ODP) proposed by Silva et al.
of foods, by evaluation of a team of trained judges (Meilgaard, Ci- (2012) seeks to meet the demand for fast and descriptive methods,
ville, & Carr, 2006). Specific methods reflect different sensory ap- while providing quantitative information on the sensory attributes
proaches (Lawless & Heymann, 1998), however the generic in food. In this method it is recommended to use the unstructured
descriptive analysis, which can combine different approaches of 9 cm scale anchored at the extremes by the terms ‘‘weak’’ and
all these methods, is often used during practical application in or- ‘‘strong’’ which are represented by the reference materials.
der to meet the specific objectives of the study (Murray, Delahunty, In this method, the judges evaluation of the products utilizing
& Baxter, 2001). the attribute-by-attribute protocol. In this type of evaluation,
judges receive all samples at once and are instructed to compare
2.1. Conventional profile
the samples with respect to a single attribute, and indicate the
intensity of the samples on an unstructured scale anchored at
The generic descriptive analysis adapts and/or combines differ-
the extremes by reference materials (weak and strong). In this
ent approaches of descriptive methods to meet the specific objec-
evaluation protocol re-tasting of the products and re-evaluation
tives of the research. In this method, all stages of the sensory test,
of the grades can be performed until concluding that the scores
from recruitment of volunteers until the data analysis must be con-
represent the correct ‘‘spacing’’ between intensities. It allows
ducted according to the recommendations of the descriptive liter-
judges with a low degree of ‘‘calibration’’ to consistently assess
ature methods. Generic methodologies were used by Dairou and
product intensity (Ishii, Stampanoni & O’Mahony, 2008). When
Sieffermann (2002), Delarue and Sieffermann (2004), Blancher
evaluating the products, reference materials of the sensory attri-
et al. (2007), Sinesio, Peparaio, Moneta, and Comendador (2010)
butes are presented together with the samples so as to permit
and Silva et al. (2012) and defined as the ‘‘conventional profile’’.
the judges to compare the samples with the references, facilitating
According to Murray et al. (2001), the descriptive sensory anal-
the allocation of the attribute intensity on the unstructured scale.
ysis of foods presents itself as the most comprehensive, flexible
In development of the methodology, this technique was applied
and useful among the sensory methods, providing detailed infor-
to the sensory description of five creamy requeijão cheese formula-
mation on all the attributes that make up a food. During this mil-
tions. It was found that the ODP allowed for acquisition of a sen-
lennium, it is expected that the descriptive analysis be increasingly
used and involve an even wider range of applications than has
been performed previously. Therefore, continued investment in Table 1
the development of descriptive techniques that challenge tradi- Type and quantity of chocolate utilized in processing of the formulations.
tional ideas is vital to ensure that the optimum potential is ac-
Formulations* Type and quantity (g) of chocolate
quired with this method in the future.
Milk Semisweet Bittersweet
2.2. Ranking descriptive analysis F1 9.6 2.4 –
F2 9.6 – 2.4
Several researchers have sought to develop alternative method- F3 – 12.0 –
F4 6.0 – 6.0
ologies to conventional descriptive methods which require less
time for obtaining the sensory profile, so as to allow for their use *
Each unit contains 12 g of chocolate.
 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179 171

sory profile very similar to the evaluation performed by the trained uating the products. Thus, a team of fifteen judges evaluated the
team and also enables a reduction of 54% in the sensory testing chocolates using the RDA protocol, another fifteen judges accord-
time without loss of information. The ODP is focused on quantita- ing to the ODP and ten judges using the CP.
tive characterization of the sensory attributes, enabling segmenta- In the three techniques the judges were recruited using struc-
tion of the consumers according to their preferences, quality tured questionnaires. After recruitment these judges were pre-se-
control, optimization of formulations and the correlation between lected based on discriminatory capacity of two chocolates (A and
sensory and instrumental measurements. B). For this purpose a series of four triangular tests were performed
and the judge that correctly responded to at least 75% of the
3. Material and methods assessments was selected. The chocolates used in processing of
the formulations were: (A) 90% milk chocolate and 10% semisweet
3.1. Samples chocolate; (B) 50% milk chocolate and 10% semisweet chocolate.
After selection of the participants based on their discriminatory
Chocolate formulations were prepared with three different power, the attributes which describe the test chocolates were eval-
chocolate types of the same brand (GarotoÒ). In preparation of uated by means of the previous test technique together with an
the formulations different mixtures of milk, semisweet and bitter- open discussion session, as recommended by Damásio and Costell
sweet chocolate were utilized. The quantities of each type of choc- (1991). The previous list was used from the study of Melo, Bolini
olate used in processing are described in Table 1. Formulations and Efraim (2009) to support open discussion. The attributes
were defined based on preliminary triangular tests so that they which described the chocolates were: brown color, cocoa mass ar-
present sensory difference (p < 0.001). oma, cocoa mass flavor, sweetness, residual bitterness, firmness,
In processing of the formulations, the chocolates were melted at spreadability, adhesiveness.
50 °C, mixed in the desired proportions, tempered (cooled to 25 °C
and then heated to 32 °C), shaped in polypropylene forms, cooled 3.3.1. Ranking descriptive analysis
to a temperature of 25 °C and packaged in laminated paper. The The fifteen preselected judges for the RDA team participated in
units used in the sensorial analysis of the formulations measured a familiarization session for the descriptive terms and their respec-
29 mm in diameter by 20 mm in height. tive reference materials. In this session the objective was to stan-
dardize the evaluation form and clearly state which sensory
stimulus referred to each term. In the RDA, the reference materials
3.2. Instrumental analysis
are qualitative, i.e., express the sensory intensity extremes of each
stimulus. After this stage of familiarization with the sensory stim-
3.2.1. Color analysis
uli, evaluations of the test chocolates were initiated. In each eval-
The color coordinates were determined using a Color Quest XE
uation session the judge received four chocolates (F1–F4) coded
colorimeter, which obtains measurements in a reading area of
with three random digits, which were presented in a randomized
8 cm. The equipment specifications were: natural daylight illumi-
and balanced mode. The evaluation form contained all the sensory
nant (CIE D65) and standard observer at 10° (CIE 1964). The read-
attributes and categorical scale (ranking) associated with each
ing system used was the CIE Lab, represented by the following
descriptive term. The judge was asked to rank the chocolates in
components: L (luminosity – black (0) to white (100)), a ( green
relation to the intensity of sensory stimuli listed on the evaluation
( 60), +red (+60)) and b ( b blue ( 60), +b yellow (+60)). All read-
form. Three repetitions were performed per judge, totaling three
ings were performed in triplicate.
sessions of product evaluation.
In analysis of the data, the grades attributed to each chocolate
3.2.2. Instrumental firmness analysis were summed in relation to each sensory attribute, and the results
The instrumental texture firmness property of the chocolate were analyzed using the Friedman test (Newell & MacFarlane,
formulations was evaluated by means of mechanical tests con- 1987).
ducted using a universal mechanical testing machine (Instron –
Series 3367, United States, 2005). Analyses were performed at 3.3.2. Optimized Descriptive Profile
25 ± 1 °C, representing the storage temperature of the chocolates. The fifteen pre-selected judges for the ODP team participated in
A 5 mm diameter probe was moved perpendicularly onto the choc- a familiarization session with descriptive terms and their reference
olate which measured 29 mm in diameter and 20 mm in height. materials. The objective of this session was to standardize the eval-
The test conditions were: load cell of 1 kN, compression distance uation form and clearly state which sensory stimulus referred to
of 20% the product height, velocity: 1.0 mm/s, and contact time each term, as well as anchor the ends of the unstructured scale
of 2 s. The force exerted on the product was automatically regis- (‘‘weak’’ and ‘‘strong’’). The judges were requested to read the def-
tered and the property of texture firmness (N) was calculated auto- inition of sensory stimulus and taste the references. In the ODP, the
matically by the software Blue Hill 2.0 (Instron, United States, reference materials are both qualitative and quantitative, therefore
2005) from the curves of force (N)  time (s) generated during they identify/define the sensory attribute and anchor the ends of
the test. Three measurements were obtained for each formulation. the rating scale. After this step, evaluations of the test chocolates
were initiated.
3.3. Sensory analysis Evaluation of the chocolates in the ODP was performed using
the attribute-by-attribute protocol so that only one attribute was
Sensory characterization of the chocolates was performed using evaluated per session. In each evaluation session the judge re-
three descriptive methodologies: ranking descriptive analysis ceived four chocolates coded with three random digits, which were
(RDA), Optimized Descriptive Profile (ODP) and conventional pro- presented in a randomized and balanced mode, and was asked to
file (CP). First, fifty judges were recruited, of which forty were evaluate the products in relation to a determined sensory stimulus.
pre-selected based on the discriminatory power of the products. The reference materials (‘‘weak’’ and ‘‘strong’’) from the sensory
All participants evaluated the descriptive terms to be evaluated stimulus evaluated were also presented to the judge together with
in the chocolates and defined the reference materials (‘‘weak’’ the samples. This permitted that the judge compares the samples
and ‘‘strong’’) for each sensory stimulus. After this procedure, the with each other and with the references before allocating the
judges were divided randomly into three different teams for eval- intensity of sensory attribute on the ranking scale.
172 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179

The evaluation form was organized by attributes and contained Table 2

a 9 cm unstructured scale associated with each formulation. The Means and standard deviations of the instrumental measures of color and texture of
the chocolates.
judge was asked to compare the samples with each other and with
the reference materials, and then allocate the stimulus intensity on Samples Color components Firmness (N)
the intensity scale. Evaluations were carried out according to a Bal- L a b
anced Block Design (BBD). Three repetitions were performed per F1 34.11 ± 0.08 9.00 ± 0.05 9.49 ± 0.13 70.4413 ± 3.7836
judge. Therefore, a total of 24 sessions were needed to evaluate F2 31.97 ± 0.28 7.89 ± 0.14 7.65 ± 0.23 79.2873 ± 7.1391
all attributes. F3 29.03 ± 0.20 5.92 ± 0.12 4.59 ± 0.09 130.4880 ± 8.7877
The results were analyzed by ANOVA with two sources of vari- F4 29.61 ± 0.47 6.38 ± 0.19 5.36 ± 0.30 82.1739 ± 2.8438

ation (formulation and judge) and interaction between the factors L, luminosity color component; a, green–red color component; b, blue–yellow color
formulation  judge. In the case of significant interaction, the F- component.
test for the formulation was calculated by using the Mean Square
of interaction (MSinteraction) as the denominator, as recommended
by Stone and Sidel (2004). The ANOVA was followed by Duncan’s of product groups (samples) considering the principal components
multiple range test (a = 0.10). of greatest explanation of the PCA, as used by Rodrigue et al.
(2000), Dairou and Sieffermann (2002), Delarue and Sieffermann
3.3.3. Conventional profile (2004), Cartier et al. (2006) and Richter et al. (2010).
Ten judges participated in evaluation of the chocolates by
means of the CP. Thus, after definition of the sensory attributes
3.4.2. Discrimination power of the treatments
and references, training was initiated. In this stage the reference
The criteria used to compare the discrimination power of the
materials were presented which anchored the extremes of the
methodologies was the significance level (p-value) of the variation
intensity scale and it was asked that the judges read the definition
source of interest (formulation), which is obtained in the ANOVA
of the sensory attribute and taste the references, trying to memo-
(probability of Fformulation). The power analysis of the F-test was
rize them. Training was conducted until the team no longer pre-
also performed to complement the evaluation of the discrimina-
sented difficulties in assessing the chocolates. A total of twenty-
tion power of the treatments for each descriptive methodology.
eight training sessions were needed for this purpose.
Power of the test was calculated by the probability function [1 –
To assess whether the judges were sufficiently trained, a test
P(type II error)] and represents the probability of correctly reject-
was performed to verify the individual performance of the volun-
ing the null hypothesis when it is false. In the present study, the
teers in relation to the discrimination power of the products and
power of the F test was calculated by fixing the significance level
repeatability of evaluations. For this purpose the judges were
(type I error) at 0.10. The formula for calculating the power of
served two chocolates (A and B) coded with three random digits
the statistical test is described in O’Brien and Lohr (1984). The
in individual booths. The judges were asked to evaluate the choc-
analysis described above was performed for the ODP and CP meth-
olates using the evaluation form of the CP. This form was organized
odologies, which quantitatively evaluate the sensory attributes and
per sample and contained the 9 cm unstructured scale associated
provide continuous data.
with each descriptive term. The performance test was conducted
Additionally, discrimination of the chocolate formulations was
using four repetitions. In analyzing the individual results of the
verified in each methodology reported by means of comparison be-
judges, the discrimination power of the chocolate formulations
tween the evaluation averages. For this purpose Duncan’s statisti-
and repeatability of the results was considered. Therefore, an AN-
cal test (a = 0.10) in the ODP and CP was used, and the Friedman
OVA was performed with two sources of variation (formulation
test (a = 0.10) was used to compare the sums of scores in the RDA.
and repetition) for each sensory attribute assessed by each judge.
Judges were selected that presented maximum probability of 50%
for discrimination of formulations (p.Fformulation < 0.50) and mini- 3.4.3. Study of the effect of interaction and use of the unstructured
mum probability of 50% for repeatability of results (p.Frepeti- scale in the Optimized Descriptive Profile and conventional profile
tion > 0.50). Based on these criteria, all judges presented good It was sought to compare the form of utilization of the 9 cm
performance and participated in the final evaluation of the unstructured scale by the trained (CP) and semi-trained teams
chocolates. (ODP). First, a study was performed on the existence of interaction
In the final evaluation of the chocolates, the formulations were between the judges and formulations by ANOVA. To analyze better
evaluated by the judges according to a Balanced Block Design. The the judge  formulation interaction a graphical evaluation was
chocolates coded with three random digits were served to the used in order to show the evaluations of each judge for each for-
judges in individual booths in a monadic and random mode. Eval- mulation in relation to the individual sensory attributes.
uation of the sensory characteristics was performed using an eval- A frequency analysis of using the scores for the two methodol-
uation form containing the 9 cm unstructured scale associated ogies (ODP and CP) was also performed. For evaluation of the fre-
with each descriptive term. quency distributions, histograms and the chi-squared test were
The results were analyzed by ANOVA with two sources of vari- used for comparing if the methodologies differ in the way in which
ation (formulation and judge) and the interaction between formu- the scale was used.
lation and judge. The ANOVA test was followed by Duncan’s
multiple range test (a = 0.10).
3.4.4. Correlation between sensory and instrumental measurements
3.4. Comparison of the descriptive methods Among the main applications of the descriptive analysis is the
ability to correlate sensory data with instrumental measurements,
3.4.1. Comparison between the graphical representations of the as well as the possibility to establish predictive models of sensory
descriptive data characteristics by means of instrumental measurements. Correla-
The proximity between the spatial configurations of the tion between the instrumental measurements of color and texture
descriptive data generated by the three methods (RDA, ODP and and the sensory measurements was determined using the Pearson
CP) was assessed by the correlations of the attributes (vectors) correlation coefficient, in ODP and CP, and Spearman correlation
with the principal components and the similarity in the formation coefficient in RDA method.
 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179 173

Fig. 1. Graphical representation of the samples and sensory attributes for data of the RDA (a), ODP (b) and CP (c).

3.5. Statistical analyses of the F-test was obtained using the JMP software of the institution
SAS. The Generalized Procrustes Analysis was performed using the
Analyses of variance, mean tests and principal component anal- function FactorMineR (Husson, Lê and Mazet, 2006) of the free R
yses were performed using the package SAS/STAT of the Statistical software (R Development Core Team, 2005) version 2.15.1;
Analysis System software (Institute Inc., North Carolina, USA, 1999) and the Sigma-Plot software was used for construction of the
version 9.1, licensed to the Federal University of Viçosa. The power graphs.
174 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179

Table 3
Estimates of the variances between treatments (MSformulation), variance due to interaction (MSinteraction) and residual variance (MSresidual).

Sensory atributtes ODP CP

MSformulation MSinteration MSresidual MSformulation MSinteration MSresidual
Brown color 442.8710 2.3007 0.8562 291.8376 5.0206 0.9504
Cocoa mass aroma 492.6452 2.4916 0.8325 294.6517 5.0444 1.7703
Cocoa mass flavor 506.1976 2.1089 0.8629 322.6944 4.7306 1.2263
Sweetness 445.1107 3.3547 1.4967 329.3612 5.0352 1.4859
Bitterness 407.6830 2.9891 1.2817 289.7917 6.6579 1.4762
Hardness 464.9520 2.5810 1.1984 296.9336 5.1160 1.6893
Spreadability 397.6071 2.3049 1.5318 265.5986 6.9692 1.7092
Adhesivity 397.3252 2.3164 2.0520 277.2911 5.7267 2.1003

4. Results
Table 5
Mean scores of the sensory attributes of chocolate in the Conventional Profile, tested
4.1. Instrumental characterization of color and texture of chocolates by means of Duncan’s multiple range test (a = 0.10).

Sensory atributtes Formulations

Table 2 presents the means and their respective standard devi-
F1 F2 F3 F4
ations for the instrumental measures of color and texture of the
chocolate formulations. Brown color 1.7c 4.1b 8.0a 7.9a
Cocoa mass aroma 0.9c 3.0b 7.2a 7.3a
The chocolate formulations presented coloration with tendency
Cocoa mass flavor 1.1c 3.1b 7.7a 7.5a
to red and yellow, due to the positive signal of components a Sweetness 8.1a 5.8b 1.5c 1.5c
(green–red) and b (blue–yellow). Formulation F1 showed greater Bitterness 0.6d 2.2c 6.2b 7.1a
intensity of the luminosity component and therefore presents a Hardness 1.6d 3.4c 8.6a 6.7b
lighter color. The inverse was found in formulation F3. Regarding Spreadability 7.8a 6.1b 1.5d 2.6c
Adhesivity 7.8a 5.9b 1.2d 2.5c
the components a (green–red) and b (blue–yellow), formulation
F1 showed greater intensity of the color components and chocolate Means followed by the same letter in the row do not differ at 10% probability.
F3 presented the lowest intensity. Regarding the property of
instrumental texture firmness of the chocolates, it was found that
formulation F3 showed greater firmness and formulation F1 ap- Table 6
Sums of rankings for sensory attributes of chocolate obtained by Ranking Descriptive
peared less firm. The four chocolate formulations presented little Analysis, tested by the Friedman test (a = 0.10).
variation in intensity of the color components and instrumental
firmness. Sensory atributtes Formulations
F1 F2 F3 F4
4.2. Comparison of descriptive methodologies: ranking descriptive Brown color 15b 30b 54a 51a
analysis Optimized Descriptive Profile and conventional profile Cocoa mass aroma 16c 29b 52a 53a
Cocoa mass flavor 17b 28b 53a 52a
Sweetness 59a 46a 20b 25b
The sensory profile of the chocolates was developed by means
Bitterness 17b 28b 54a 51a
of three descriptive techniques (RDA, ODP and CP). Hardness 17c 29b,c 59a 45a,b
Spreadability 56a 46a 19b 29b
4.2.1. Graphical representation of the descriptive data: principal Adhesivity 55a 46a,b 20c 29b,c
component analysis Minor sums of orders represent smaller sensory attribute intensities. Sums of
In the PCA, it was found that the first principal component ex- orders followed by the same letter in the row do not differ at 10% probability.
plained around 99% of total variance of the data for the three meth-
odologies, and thus only the first dimension was considered in the
graphical representation for interpretation purpose. This type of and F4. All attributes presented correlation with only the first prin-
analysis resulted in a bi-dimensional representation for easy visu- cipal component (p < 0.10).
alization of the sensory map (Fig. 1). In spatial separation of the
chocolate formulations, two groups were observed: one consisting 4.2.2. Discrimination power of the chocolate formulations in the three
of the F1 and F2 formulations and another formed by chocolates F3 descriptive methods
For the ODP and the CP there was a significant difference be-
Table 4 tween the formulations according to the F-test. For all sensory
Mean scores of the sensory attributes of chocolate in the Optimized Descriptive attributes in the ODP and the CP, the estimate of the F-test pre-
Profile, tested by means of Duncan’s multiple range test (a = 0.10). sented a probability less than 0.0001. For the type II error was
Sensory atributtes Formulations equal to 0.0001 for all sensory attributes in both descriptive meth-
odologies; and thus a power of 0.9999 was verified for the F-test.
F1 F2 F3 F4
Therefore, the methodologies ODP and CP showed no difference
Brown color 1.4c 3.0b 7.4a 7.3a
in relation to the power of the F-test.
Cocoa mass aroma 1.2d 2.2c 7.4a 7.1b
Cocoa mass flavor 1.0d 1.9c 7.2a 6.8b
The Table 3 presented the estimates of variances between treat-
Sweetness 7.8a 6.9b 2.0c 2.3c ments, variance due to interaction and residual variance for the da-
Bitterness 0.8c 1.5b 6.2a 6.2a tas of ODP and PC. In the ODP, there was greater variance among
Hardness 1.4d 2.5c 8.0a 6.2b the chocolate formulations and less variance due to interaction
Spreadability 7.5a 6.8b 1.7d 2.8c
when compared to the CP. The estimate of residual variance due
Adhesivity 7.5a 6.8b 1.7d 2.7c
to the random error was similar in both methods. Higher values
Means followed by the same letter in the row do not differ at 10% probability. of the Mean Square for formulations in the ODP indicate greater
 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179 175

Fig. 2. Utilization of the sensory scores in quantification of the chocolate sensory characteristics, where each judge is represented by a line on the graph. (a) Optimized
Descriptive Profile, (b) conventional profile.

discrimination between formulations in this technical sense when there was also the presence of non-discriminated means, i.e., lo-
compared to the conventional method. cated in the ‘‘confounding region’’. This region is characterized by
Because there was a significant difference among the formula- means which present two different letters, belonging to two
tions by the F-test, it was proceeded to apply the Duncan’s multiple groups simultaneously. Only the attribute cocoa mass aroma pre-
range test. The mean scores obtained in the sensory characteriza- sented discrimination of the treatments in three groups without
tion of chocolates by means of the ODP and CP are listed in Tables ‘‘confusion’’.
4 and 5, respectively.
In the ODP, the treatments were separated into four distinct 4.2.3. Study of the interaction and use of the unstructured scale in the
groups for five of the eight sensory attributes. For the CP, a similar Optimized Descriptive Profile and conventional profile
behavior to the ODP was verified for discrimination of the choco- From the F-test it was verified that interaction between the for-
lates. There was the formation of four distinct groups for half of mulations and the judges was significant (p < 0.10) for all attri-
the attributes. Increased discrimination of the chocolates was ver- butes in the ODP (except for adhesiveness) and CP. The existence
ified in the ODP for the attributes of aroma and cocoa mass flavor. of interaction indicates that at least one judge evaluated the for-
In contrast, greater discrimination was observed in the CP for the mulations differently from the team (Stone & Sidel, 2004).
sensory stimulus bitterness. The discrimination power of the two Interaction is an occurrence both common and difficult to con-
methods was similar. trol in sensory analysis, and its detection reflects the need for
On the other hand, in the RDA the discrimination power was graphical analysis of the interaction, in order to verify that this
evaluated by Friedman’s test when comparing the sums the scores occurrence is due to the use of different parts of the scale or if there
for each sensory attribute, which are described in Table 6. For five is a ‘‘severe’’ interaction effect. A ‘‘severe’’ interaction effect is
attributes, the four chocolate formulations were divided into two denominated when there is inversion in the perception of the
groups. For the three attributes, there was a greater discrimination intensity of sensory stimuli (Stone & Sidel, 2004). To verify if the
of chocolates, forming three distinct groups. Although there was formulation  judge interaction was ‘‘severe’’ a graphical evalua-
greater discrimination of the formulations for these attributes, tion was utilized (Fig. 2), considering the ODP (a) and PC (b).
176 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179

Fig. 3. Frequency distribution for sensory scores assigned to the descriptive attributes of chocolate. (a) Optimized Descriptive Profile, (b) conventional profile.

Table 7
Correlation between the sensory scores (RDA, ODP and CP) and the instrumental measures (color and texture) of the chocolates.

Instrumental properties RDA ODP CP

rs p-Value r p-Value r p-Value
Sensory brown color
L 0.95170* <0.0001 0.97276* <0.0001 0.98519* <0.0001
a 0.97333* <0.0001 0.98191* <0.0001 0.98257* <0.0001
b 0.97333* <0.0001 0.97893* <0.0001 0.98127* <0.0001
Sensory hardness
Instrumental firmness 0.95003* <0.0001 0.83099* 0.0008 0.82457* 0.0010
Sensory spreadability
Instrumental firmness 0.95003* <0.0001 0.79231* 0.0021 0.75800* 0.0043
Sensory adhesivity
Instrumental firmness 0.95003* <0.0001 0.77542* 0.0030 0.75427* 0.0046

L, luminosity color component; a, green–red color component; b, blue–yellow color component; r, Pearson’s correlation coefficient; rs, Spearman’s correlation coefficient; p-
value, significance level.
*
Significant at the 10% probability.

Both Fig. 2a (ODP) and b (PC) exhibit positive and negative Fig. 3 contains the histograms for frequency analysis of the use
slopes from one sample to another, so severe interactions are pres- of scores (ranges) in relation to each sensory attribute and for each
ent in both methods. An other hand, in the ODP was verified the descriptive methodology (ODP and CP). To verify if there were dif-
interactions are less pronounced, e.g. tighter clusters of lines. ferences in utilization of the parts of the scale by the teams (ODP
 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179 177

and CP), the chi-squared test was used to compare frequency dis- asymmetrical, not permitting the quantitative evaluation of sen-
tributions. In the chi-square test a significant difference was veri- sory stimuli (Richter et al., 2010; Rodrigue et al., 2000; Silva
fied between the distributions (p < 0.10) for all sensory et al., 2012).
attributes. Therefore, both the trained (CP) and semi-trained teams Regarding the significant effect (p < 0.10) of interaction, it was
(ODP) presented differentiated utilization of the unstructured scale found that this effect less pronounced in the ODP. The interac-
portions. tion effect has been observed in several studies using teams with
a high degree of training, including those of Cardello and Faria
(1998), 2000, Monrozier and Danzart (2001), Rocha, Minim, Della
4.2.4. Correlation of sensory data and instrumental measures
Lucia, Minim, and Coimbra (2003), Richter et al. (2010) and Silva
Correlation analysis was performed between the sensory color
et al. (2012). Moreover, the evaluation protocol of the ODP can
attribute ‘‘brown color’’ and instrumental color components Lab;
be presented as an alternative to reduce this effect, due to expo-
and for texture the sensory attributes ‘‘firmness’’, ‘‘spreadability’’
sure to all formulations in a single evaluation session, which can
and ‘‘adhesiveness’’ were correlated with instrumental firmness.
minimize the ‘‘inversion’’ effect in the perception of sensory
Table 7 shows correlation coefficients and the significance level
stimuli interactions. It is also highlighted that variation due to
associated with each coefficient for each attribute and descriptive
interaction (MSinteraction) was lower in the ODP when compared
methodology.
with the CP.
All sensory attributes showed significant correlation (p < 0.01)
A similar result was encountered by Silva et al. (2012) when
with instrumental measurements of color and texture of the choc-
studying the interaction between judges, who also confirmed the
olates in the three sensory techniques. For the measure of texture,
absence of interaction. In this study, four creamy requeijão
sensory data from the RDA showed higher correlation than the
cheese formulations were characterized utilizing the ODP and
ODP and CP methods. On the other hand, when comparing the cor-
CP techniques. The authors found that when the judges evalu-
relations between sensory and instrumental measures of texture in
ated the products using the ODP protocol there was no interac-
the ODP and CP, it was verified that in the ODP the correlations
tion between the judges and formulations (p > 0.10). On the
coefficient were greater and significant at a lower probability level.
other hand, when the requeijão cheeses were evaluated by CP
The alternative sensory description methods, RDA and ODP, per-
there was a significant interaction (p < 0.10). As in the present
mitted correlation with the instrumental measurements of color
study, where the products presented lower magnitude of in sen-
and texture of the chocolates, similar or higher than the conven-
sory characteristics, a significant interaction (p < 0.10) was veri-
tional method (CP).
fied in both the ODP and CP, but in the ODP the interaction
was of lesser magnitude (lower MSinteraction) when compared to
5. Discussion the PC.
Use of the differentiated intensity scale was also observed
In the PCA, the three sensory description techniques (RDA, ODP when comparing the two teams (ODP and CP). In the ODP, the
and CP) presented similar descriptive maps. Formulations with the full range of the unstructured scale was used, where it was ob-
highest concentrations of milk chocolate in the mixture (F1 and F2) served that all scores were used to express the intensity of the
showed higher intensity of the attributes sweetness, spreadability sensory attributes. In the CP use of the lower and upper extrem-
and adhesiveness, coherent with the ‘‘strong’’ reference material ities of the scale were used with greater frequency. This behavior
for these sensory stimuli, which was represented by milk choco- is probably due to the evaluation protocol of the treatments in
late. Moreover, these chocolates (F1 and F2) showed low intensity the two methodologies. In the ODP the judges received all formu-
of the attributes brown color, aroma and flavor of the cocoa mass, lations at the same time, which allows for comparison among the
firmness and bitterness, anchored by milk chocolate at the ‘‘weak’’ samples before marking the intensity scale. In addition, the refer-
extreme intensity. Sensory characterization of the chocolate for- ence materials that anchor the extremes of the scale are also
mulations obtained by means of the descriptive techniques RDA, served together with the formulations. For the CP, the evaluation
ODP and CP showed consistency with the reference materials. is performed for each formulation individually and the references
The three descriptive techniques presented similarity in the spatial of the intensity extremes are not consulted during the final eval-
configurations of the chocolate formulations in relation to the prin- uation of the formulations. Absence of the reference materials
cipal components. during the evaluation of the products may have led the judges
Cartier et al. (2006) found concordance when comparing the to evaluate the formulations with intensities closer to the ex-
trained and untrained teams, using the sorting procedure classifi- tremes. However it is important to consider that this behavior
cation methodology. Sinesio et al. (2010) also encountered a sim- did not affect characterization of the chocolates and acquisition
ilar profile when comparing the spatial configuration of creamy of the sensory profile in the CP.
soups obtained by two teams, where one team consisted of Alternative methodologies provided sensory profiles similar
trained judges and the other team with untrained, using the clas- and equivalent to the conventional method and presented reduc-
sification methodology. Richter et al. (2010) verified similarities in tions in the number of sessions required of 80% for the RDA and
spatial configurations between the sensory maps obtained in the 38% for the ODP. The time required in the product evaluation
evaluation of chocolate puddings by the RDA and QDA (quantita- sessions of the ODP is less than the time consumed in evaluation
tive descriptive analysis) techniques when using the GPA. Silva of the formulations in the CP protocol (sample-by-sample), since
et al. (2012) found two similar spatial configurations in the char- only one attribute is evaluated per session of the ODP. Thus, in
acterization of creamy requeijão cheese when comparing the ODP the ODP 38% fewer sessions were needed, but the time reduction
and CP methods. for the sensory test was even greater. Thus, it has been esti-
In the discrimination of samples analysis, the ODP and the CP mated that sensory characterization of the chocolates by the
were similar. The discriminations of samples in the ODP and CP PDO showed a 50% reduction in the time required in the CP. Ta-
were greater than RDA. In the RDA, where five of the eight attri- ble 8 provides a comparison between the sensory techniques in
butes were discriminated in only two groups and ‘‘confounding re- relation to the presence/absence of steps, number of sessions re-
gions’’ were detected. In the RDA, it is not possible to affirm if these quired, the number of judges and type of data obtained in each
groups present large or small differences in sensory characteristics. sensory methodology used to study the sensory profile of
This limitation is due to the evaluation scale, which is discrete and chocolates.
178 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179

Table 8
Comparison of the evaluation criteria of each of the described methods and characterization of the sample groups formed on the sensory profile of the chocolates.

Criteria Descriptive sensory methods

RDA ODP CP
Steps (number of sessions and protocols)
Recruitment 1 Session 1 Session 1 Session
Pre-selection 4 Sessions 4 Sessions 4 Sessions
Assessment of the 1 Session 1 Session 1 Session
descriptive terms
Presentation of the 1 Session 1 Session –
reference materials
Training – – 28 Sessions
Evaluation of judge – – 4 Sessions (p.FPRODUCT < 0.50 and
performance p.FREPETITION > 0.05)
Evaluation of the test 3 Sessions – one for each repetition (all 24 Sessions – one for each attribute and 12 Sessions – one for each product and
products samples only once) repetition (attribute-by-attribute repetition (sample-by-sample protocol)
protocol)
Estimated time required in 20 min per session 5 min per session 15 min per session
the evaluation sessions
Number of sessions 10 31 50
Number of judges 15 16 10
Groups formed in the Discrimination of the formulations in 2 or 3 Discrimination of the formulations in 3 or Discrimination of the formulations in 3
multiple comparisons test groups and detection of the ‘‘confounding’’ 4 groups or 4 groups
region
Percentage of explanation of 99.9% 99.9% 99.9%
the two principal
components
Number of groups identified 2 2 2
(clusters) (group1: F1 and F2; group 2: F3 and F4) (group 1: F1 and F2; group 2: F3 and F4) (group 1: F1 and F2; group 2: F3 and F4)
Characterization and composition of the groups
Group 1 Sweetness, adhesiveness and spreadability Sweetness, adhesiveness and Sweetness, adhesiveness and
spreadability spreadability
Group 2 Brown color, cocoa mass aroma and flavor, Brown color, cocoa mass aroma and Brown color, cocoa mass aroma and
firmness and bitterness flavor, firmness and bitterness flavor, firmness and bitterness

6. Conclusions References

The ranking descriptive analysis (RDA) and Optimized Descrip- Blancher, G., Chollet, S., Kesteloot, R., Nguyen Hoang, D., Cuvelier, G., & Sieffermann,
J.-M. (2007). French and Vietnamese: How do they describe texture
tive Profile (ODP), alternative methodologies, presented sensory characteristics of the same food? A case study with jellies. Food Quality and
profiles for the chocolates both similar and equivalent to the con- Preference, 18, 560–575.
ventional profile (CP), permitting that the sensory testing time be Cardello, H., & Faria, J. B. (1998). Análise descritiva quantitativa da aguardente de
cana durante o envelhecimento em tonel de carvalho (Quercus alba L.). Ciência
reduced. The ranking descriptive analysis showed greater reduc- Tecnologia de Alimentos, 18(2), 169–175.
tion in time, but only provided a comparative evaluation of formu- Cardello, H., & Faria, J. B. (2000). Sensory profile and physicochemical characteristics
lations, not permitting for estimating the magnitude of sensory of commercial Brazilian sugar cane spirits, both aged and non-aged. Brazilian
Journal of Food Technology, 3, 31–40.
difference between the chocolates. In contrast, the Optimized Cartier, R., Ritz, A., Lecomte, A., Poblete, F., Krystlik, J., Belin, E., et al. (2006). Sorting
Descriptive Profile generated a 38% reduction in the number of ses- procedure as an alternative to quantitative descriptive analysis to obtain a
sions (corresponding to a 50% reduction in time for conducting the product sensory map. Food Quality and Preference, 17, 562–571.
Dairou, V., & Sieffermann, J. M. (2002). A comparison of 14 jams characterized by
sensory test) compared to the Conventional Profile and provided
conventional profile and a quick original method, the flash profile. Journal of
quantitative evaluations of the sensory attributes, permitting that Food Science, 67, 826–834.
the magnitude of difference between the products be estimated. Damásio, M. H., & Costell, E. (1991). Análisis sensorial descriptivo: generación de
descriptores y selección de catadores. Revista Agroquimica de Technologia de
The ODP showed discrimination of the four chocolate formula-
Alimentos, 31(2), 165–178.
tions greater than the RDA and similar to CP. In ODP the entire Delarue, J., & Sieffermann, J. M. (2004). Sensory mapping using flash profile.
amplitude of the unstructured scale was used. This behavior was Comparison with a conventional descriptive method for the evaluation of the
not observed in the CP, which presented an elevated frequency flavour of fruit dairy products. Food Quality and Preference, 15, 383–392.
Husson, F., Lê, S., & Mazet, J. (2006). FactoMineR: Factor analysis and data mining
for utilization of the extremes. with R. R package version 1.14. <http://factominer. free.fr>.
The OPD is presented as an interesting alternative to the CP, Ishii, R., Stampanoni, C., & O’Mahony, M. (2008). A comparison of serial monadic
since it allowed for obtaining the sensory profile of the chocolates and attribute-by-attribute descriptive analysis protocols for trained judges.
Food Quality and Preference, 19(3), 277–285.
while requiring less time, and without loss of information. This Labbe, D., Ritz, A., & Hugi, A. (2004). Training is a critical step to obtain reliable
method showed to be superior to the RDA for discrimination of product profiles in a real food industry context. Food Quality and Preference, 15,
products and was also an alternative to decrease the interaction 341–348.
Lawless, H. T., & Heymann, H. (1998). Sensory evaluation of food: Principles and
between judges and formulations. Therefore, this method is pre- practices. New York: Chapman and Hall.
sented as a promising tool for sensory description of foods, both Meilgaard, M. C., Civille, G. V., & Carr, B. T. (2006). Sensory evaluation techniques (4th
in the context of research and in industry. ed.). Boca Raton: CRC Press.
Melo, L. L. M. M., Bolini, H. M. A., & Efraim, P. (2009). Sensory profile, acceptability,
and their relationship for diabetic/reduced calorie chocolates. Food Quality and
Preference, 20(2), 138–143.
Acknowledgements Monrozier, R., & Danzart, M. (2001). A quality measurement for sensory profile
analysis: The contribution of extended cross-validation and resampling
techniques. Food Quality and Preference, 12, 393–406.
The authors would like to acknowledge the CNPq and Fapemig Murray, J. M., Delahunty, C. M., & Baxter, I. A. (2001). Descriptive sensory analysis:
for their financial support. Past, present and future. Food Research International, 34, 461–471.
 R.C.S.N. Silva et al. / Food Quality and Preference 30 (2013) 169–179 179

Newell, G. J., & MacFarlane, J. D. (1987). Expanded tables for multiple comparison Rodrigue, N., Guillet, M., Fortin, J., & Martin, J. F. (2000). Comparing information
procedures in the analysis of ranked data. Journal of Food Science, 52, 1721–1725. obtained from ranking and descriptive tests of four sweet corn products. Food
O’Brien, R. & Lohr, V. (1984). Power analysis for linear models: The time has come. Quality and Preference, 11(1), 47–54.
In Proceedings of the ninth annual SAS User’s Group international conference SAS. (1999). Statistical analysis system. Institute Inc., North Carolina, USA, versão
(pp. 840–846). 9.1, licenciado para a Universidade Federal de Viçosa.
R Development Core Team (2005). R: A language and environment for statistical Silva, R. C. S. N., Minim, V. P. R., Simiqueli, A. A., Moraes, L. E. S., Gomide, A. I., &
computing. 3-900051-07-0. Vienna, Austria: R Foundation for Statistical Minim, L. A. (2012). Optimized Descriptive Profile: A rapid methodology for
Computing. <http://www.R-project.org>. sensory description. Food Quality and Preference, 24(1), 190–200.
Richter, V. B., Almeida, T. C. A., Prudencio, S. H., & Benassi, M. T. (2010). Proposing a Sinesio, F., Peparaio, M., Moneta, E., & Comendador, F. J. (2010). Perceptive maps of
ranking descriptive sensory method. Food Quality and Preference, 21, 611–620. dishes varying in glutamate content with professional and naive subjects. Food
Rocha, F. L., Minim, V. P. R., Della Lucia, F. D., Minim, L. A., & Coimbra, J. S. R. (2003). Quality and Preference, 21, 1034–1041.
Avaliação da influência dos milhos QPM nas características sensoriais de bolo. Stone, H., & Sidel, J. L. (2004). Sensory evaluation practices (3rd ed.). New York:
Ciência e Tecnologia de Alimentos, 23(2), 129–134. Academic Press.