Está en la página 1de 15

FREE CHOICE PROFILE

SENSORY ANALYSIS
UNIVERSIDAD SIMN BOLVAR MXICO.
FREE CHOICE PROFILE .

Mastachi P. Laura S.
Mendieta B. Dylan X.

Ingeniera en Alimentos.

6to Semestre.

Maestro: I.A. Ivn Mendez G.


FREE CHOICE PROFILE
2

Contents.

Introduction.

Free Choice Prole sensory testing .

Characteristics evaluated.

Product tested.

Screening and selection of judges.

Training of judges.

Scales/Scorecard and replications.

Facilities and testing conditions.

Statistical Analysis.

Procrustes analysis.

Description.

Theory of Procrustes analysis.


1. An easily geometrical look.
2. Transformations.
3. The Level-Effect : Translation.
4. The interpretation-effect: Rotation/ Reection.
5.The range-effect: Isotropic Scaling.

Applications of Procrustes analysis to sensory data.

Conclusions.

References.

SENSORY ANALYSIS
FREE CHOICE PROFILE
3
Introduction:

Descriptive analysis, also referred to as quantitive sensory, consensus,or xed choice proling, is
the method by which information about products has been obtained by sensory scientists for many
years.These tests uses the most sophisticated tools of the sensory scientist and involve the
detection, discrimination ,and description of both the qualitative and quantitative sensory
components of a consumer product by trained panel of judges.
The qualitative aspects of a product include all aroma, appearance, avor ,texture, after taste and
sound properties of a product, which distinguish it from others. Sensory judges then quantify these
product aspects in order to facilitate description of the perceived product attributes.
Free-choice profile (FCP), developed in the 1980s, is a sensory analysis method that can be
carried out by untrained panels.
The participants need only to be able to use a scale and be consumers of the product under
evaluation. The data are analysed by sophisticated statistical methodologies like Generalized
Procrustean Analysis (GPA) . To facilitate a wider use of the free-choice profiling procedure,
different authors have advocated simpler methods based on principal components analysis (PCA)
of merged data sets.
The most important characteristic of the proposed method is that quality responsible managers
could use this methodology without any scale evaluation. Only the free terms generated by the
assessors are necessary to apply the script, thus avoiding the error associated with scale
utilization by inexpert assessors.
About the implication for descriptive tests we found:
There is no consensual language to describe the sensory sensations elicited by a food and the
challenge of descriptive tests is to cope with this lack of consensus.
Flavor Profile & Profile Attribute Analysis
Quantitative Descriptive Analysis (QDA)
Texture profile and Sensory spectrum
Free-Choice Profile.

SENSORY ANALYSIS
Table 1. Comparison about Descriptive Analysis.
FREE CHOICE PROFILE
4

Free Choice Prole sensory testing .



Free-Choice proling, is a relative new technique used to obtain information about the product.
Free Choice proling differs from conventional descriptive proling in that product are evaluate by
members of a panel who describe perceived qualities of that product using their own individual list
of terms, rather a common scorecard .
Terms are not shared or used collectively by panel members and may be mutually exclusive. The
numbers of terms used by panelist may also vary, based on panelist experience and familiarity with
the product.Terms must be dened and understood by their originator to ensure consistency in
their use. Free Choice Proling is similar to traditional descriptive to traditional descriptive methods
in that panelist must be able to detect differences between similar products, verbally describe
specic attributes of products, and quantify those attributes.
This test was able to use free word association successfully for market research purposes,
however, was developed by William and coworkers (1981) and applied by William and Langron
(1984) for the evaluation of commercial port wines.
This technique has been further described by Arnold and William (1986).

Free Choice proling has a number of procedures in common with descriptive analysis and,
therefore, can be considered another type of descriptive analysis .
Panelist are recruited, selected ,and trained in scale usage, impartiality on judgments, and
consistent term usage so that reproducible results may be obtained. However, the degree of
training is less, because difculties associated with achieving agreement among panelist do not to
have to be confronted.
Since Free-Choice Proling is fairly new, it has had only limited use and therefore procedural
guidelines have not been stablished or standardized. Standardization of guidelines comes only
from extensive testing of established methods obtained though sound research and,
nally ,acceptance by potencial users as an adequate technique for obtaining product information.

Characteristics evaluated.

Similar to other descriptive analysis methods, FCP can be used to describe a product in terms of
one or more characteristics. This method has been used to describe products in terms of a single
characteristic, in the evaluation of the aroma of coffees, or in the evaluation of texture of cheeses,
similar to texture or avor proling. Free Choice Proling can be used to describe a product in
terms of a number of characteristics, such as appearance, avor, aroma , texture, or any
combination of these, similar to the Spectrum or Quantitative Descriptive Analysis methods.
For example, several investigator have used free-choice proling for the evaluation of appearance
in addition to aroma, avor, or texture.
The characteristic being judged may be restricted by the sensory analyst, but the number of terms
used by each panelist is limited only by their perceptual an descriptive skills.

Product tested.

Although some traditional descriptive methods may be restricted to the evaluation of specic
characteristic of a product, texture an avor proling, none restricted to a specic class of products,
neither is free-choice proling, this one studies have dealt with a variety of products FCP has been
used most extensively in the evaluation of beverages, such as wines, coffees, whiskies, and meat
products for example pork roasts. In addition to beverages and meats, others products evaluated
includes chocolates, cheeses, and yogurts.

SENSORY ANALYSIS
FREE CHOICE PROFILE
5
Screening and selection of judges.

Traditionally, judges for descriptive analysis are selected on a number of criteria depending on the
descriptive method to be used. The criteria could include ability to discriminate between products ,
communicate perceived perceptions, product usage ,task comprehension ,availability, interest,
health, attitude, level of condence , etc.
Based selection of Free Choice Prole judges on availability and willingness to participate, as well
as ability to discriminate and verbally communicate perceptions.
It also used criteria above, such as availability and health, and others to screen and select free
choice proling judges. Individual were a given a short questionnaire concerning likes ,dislikes and
the extent of their texture vocabulary, panelist were selected based on their responses to questions
pertaining to food, interested, availability and health.
These subjects were then asked to complete a series of acuity tests including odor matching,
basic taste identity, texture ranking, and series of triangle tests. Judges were further screened
based on their performance. In other studies ,however, a screening process was not used and
selection was based on past experience of the judges.
Studies with free-choice proling have routinely used from 8 to 20 panelist in evaluation products.
These numbers are comparable to conventional descriptive analysis techniques. For example Guy
in 1989, on the other hand, recruited 100 subjects to study the usefulness of free-choice proling
by consumers. It is possible to develop panels of a larger size than used in other descriptive
methods because free-choice proling does not requiere extensive training .
However it has not been established that larger number are necessary or desirable because of the
difculty in interpreting the data.

Training of judges.

The training of judges is the major factor to distinguishes descriptive analysis from the others
sensory testings methods and is common to all conventional descriptives techniques. A training
period of about 6 months is not uncommon for most descriptive methods, and the time is usually
product dependent.
A avor prole panel, for example, may require approximately 60 hours of training and 100 hours
of practice per judge. About 4 months is required for training a Spectrum judgment . It is
suggested for training a Quantitative Descriptive Analysis panel that six to eight one-hour
sessions are needed for the development of terms or attributes , as well as additional sessions to
orient judges to new products.
For most descriptive analysis methods, the training procedures are well standardized. Standards
or references are routinely used to acquaint judge with the product attributes and intensities and
the ranges that might be encountered during testing.
In comparison, FCP studies vary considerably in training protocols. In some studies, judges
received no training, while other investigators used extensive training and held numerous practice
sessions. Reference samples were sometimes presented to train or familiarize free-choice prole
panelist with the products or ingredients.
Even investigators who frequently use free-choice proling do not fallow xed training procedures.
In the evaluation for example of wines by expert and nonexpert wine tasters, judges were
introduced to the concept of free-choice proling and scaling with no mention of any training
(Williams and Langron, 1984).
Judges were asked to identify different attributes of wine singularly or in combination with other
attributes and assign intensity values to them.
Unlabeled samples were introduced, and judges were asked to assign their own descriptive terms.
Terms used by each individual were dened and categorized into appearance , aroma, and avor
groups.
A list of terms used by three experts is given.

SENSORY ANALYSIS
FREE CHOICE PROFILE
6

Descriptive Terms used by three inexperienced wine tasters.


These two groups used different terms to describe the same wines, based on their experience.
Clearly, more terms were developed by the more experienced taster. Similarly, judges had no
Expert Appeareance Aroma Flavor
Taster 1 Depth
Fresh
Brightness

Cleanness
Freshness
Fruitiness
Richness
Smoothness
Concentration
Cleanness
Fresh
Body
Grip
Round
Aftertaste
Taster 2 Tawny
Ruby
Purple
Cloudy

Clean
Fruity
Green
Body
Firmness
Coarse
Tannin
Hard
Crisp
Sour
Taster 3 Red
Blue
Brown
Intensity
Fruit
Esters
Wood
Oloroso
Spirit
Burnt
Tannin
Acid
Sweetness
Chocolate
Body
Green
Table 2. Source: Williams and
Langron, 1984
Nonexpert Appeareance Aroma Flavor
Taster 1 Clarity
Intensity
Redness
Yellowness

Fruitiness
Woodiness
Alcoholic Strength
Overall intensity
Cleanness
Fresh
Body
Grip
Round
Aftertaste
Taster 2 Intensity
Red
Brown

Burnt Astringent
Acidy
Taster 3 Color
Red
Brown
Smell Taste
Acidity
Tannin
Softness
Table 3.Source: Williams and
Langron, 1984
SENSORY ANALYSIS
FREE CHOICE PROFILE
7
formal training in the evaluation of aromas of six varieties of coffee or in appearance, avor, and
texture of ve samples of milk chocolate.
In the majority of free-choice proling studies, the judge develop terms individuality with not
suggestion of terms being added or eliminated by an outside mediator, however, added two terms,
smoothness and maturity, to all consumer free-choice proling ballots for whiskies.
This was done to assess whether consumers could evaluate important attributes even if they did
not generate the description themselves. What consumers characterized as maturity of whiskies
the trained panel characterized as smooth, sweet, vainilla, and malty, in addition, the consumers
perception of smooth, mellow, and sweet in the whiskies were characterized as estery and woody
by the trained descriptive panel.
This emphasizes the necessity for panelist to use their own terms, even if they are not as specic
as the investigator would like.

Scales/Scorecard and replications.

Scales and scorecards used in descriptive analysis are usually very specic and are dependent
upon the method used. For example ,Quantitative Descriptive Analysis uses a 6-inche
unstructured scale, while Spectrum scales are similar, being 15 cm in length. A 15-cm
unstructured line scale is adopted in many descriptive techniques. Final scorecards are derived by
consensus whit guidance of a group leader.
In Free Choice Prole, no standardized scaling method or scorecard has emerged. The number of
terms on an individual scorecards is idiosyncratic and reects the subjects ability to describe the
subjects ability to describe the product attributes.
Category and unstructured lines scales (6.5-15 cm long) are commonly used, the diversity of
scales and scorecards in free-choice proling is evident.

SENSORY ANALYSIS
Fig.1 Example Free Choice Prole Questionnaire.
FREE CHOICE PROFILE
8
Three or four replications by each judge for each treatment is normally recommended for any
sensory testing to ensure reliable results but the actual number applied is the choice of the
sensory analysts.
Complex products whit minor differences may require more explications. Similarly ,the number of
samples presented at one time is dependent on the products characteristics and the judgment of
the analyst.
The free-choice proling reect the same diversity in numbers of samples and replications seen in
other testing methods. Since free-choice judges are no extensively trained, the number of samples
presented at one time usually limited to two o three. Samples are often presented monadically or in
pairs.Generally the number of replications is three or four.

Facilities and testing conditions.

It is accepted in descriptive analysis to present samples to panelist using standard procedures


under conditions in which lighting, temperature, humidity, odors, and sounds can be controlled to
minimize distractions and potencial biases. Similar practices have been implemented in many of
the free-choice proling studies. Testing of samples by panelist was done under controlled
conditions utilizing individual sensory booths to ensure independent judgments. However, in the
evaluation of whiskies by consumers trained in the free-choice prole technique, samples were
evaluated at their homes and ballots returned by mail.

Statistical Analysis.

The procedures used in the aforementioned studies vary considerably, illustrating the lack of
uniformity in the FCP at the past time. Unlike more conventional descriptive analysis techniques
the data collected cannot be analyzed and compared using univariate or traditional multivariate
analysis because of the lack of uniformity in scorecards among judges. Thus, the information must
be examined differently.
Analysis of free-choice proling data differs from associated with descriptive analysis in that it is
nicely interpreted using Procrustes analysis. Individual judges responses to a set of samples can
be geometrically arranged into a single conguration where scores for each sample can be
dened as a point in space, with each attribute dening a different dimension for each attribute in
space.
Congurations of individual judges responses to each sample can then be matched and
compared using Procrustes analysis.

Procrustes analysis.

The advantage and disadvantages of using FCP or more conventional descriptive analysis
techniques have been suggested above. Since it is inception, Free Choice Proling has been
thought to have some advantages over conventional descriptive proling. FCP is less time
consuming, since judges require less training.
Judges may feel more at ease with their judgments because is not necessary for them to agree
upon a common list of descriptors. In addition , the frustration associated with trying to force
agreement among judges is thought to be alleviated.
Similar to descriptive analysis , free-choice proling is not example from dealing with potentially
large variations that can be associated with any method that uses individuals as analytical tools.
However, by using Procrustes analysis to interpret Free Choice proling data, many of these
judge-to-judge variations can be reconciled. As with any statistical tool , these are certain
limitations and disadvantages to Procrustes analysis.

SENSORY ANALYSIS
FREE CHOICE PROFILE
9
Description.

In conventional descriptive analysis as well as free-choice proling, judges evaluates samples by
assigning to them intensity values for various attributes. The values can be placed in a matrix in
which samples and attributes are represented by rows and columns, respectively.
Each element in the matrix is the intensity value perceived for that sample attribute. A single matrix
is single matrix is created judge as seen in next data matrix.
Matrix Attribute 1 Attribute 2 Attribute 3
Sample 1 Intensity values Intensity values Intensity values
Sample 2 Intensity values Intensity values Intensity values
Sample 3 Intensity values Intensity values Intensity values
Sample 4 Intensity values Intensity values Intensity values
Table 4. Data matrix. Matrix conguration of simulated data for judge 1. Rows represent samples and columns represent attributes.
The number (Intensity value) within each column and row represent the intensity values assigned by a single judge.
SENSORY ANALYSIS
Attributes
Samples Hardness Cohesiviness Springiness
A 5.9 7.1 6.9
B 4.2 3.1 7.0
C 6.0 7.0 2.9
D 3.9 2.9 3.1
Table 5. Matrix example. Simulated data set for judge 1 in the evaluation of whole wheat
bread formulations.
Fig.2 Original conguration. The original conguration of simulated data set
presented in three-dimensional space.
Shapes represented breads (A,B,C,D) for judge 1. Axes represent attributes.
FREE CHOICE PROFILE
10

The matrices for several panelist can be represented geometrically, with each sample dening a
point in space, and each attribute corresponding to different dimension. A judges responses to a
set of samples then denes a conguration of points in that space.
Procrustes analysis is a multivariate statistical tool in which congurations of individual judges can
be matched and compared in this way Procrustes analysis uses geometrical transformations,
whereby the congurations are adjusted to match each other as closely as possible.
Transformations allow for reexpresion of the congurations. The closeness of the matrices to one
another can be characterized by summing of squares distances between the points (matrices) and
common centroid (target of the consensus) matrix. This sum of squares is referred to as the
Procrustes statistic.
The data from conventional proling experiments can be seen as 3-mode data structure built from
N products, M attributes and K assessors.

The (N x M x K) data block consist of K layers, each with the (N x M) data matrix of one assessor.
Other slices of this block may be analyzed but Procrustes analysis focusses on the agreement of
the K matrices from the individual assessors.
In Free Choice Prole the assessor are free to come up with their own attributes, which they use
for judging the products. So between the assessors there is no agreement about attributes. As a
result it is impossible to average the individual data, because it makes no sense to combine
different attributes. The data from Choice proling experiments must be analyzed by individual
difference methods, or rather K-sets methods of which generalized Procrustean analysis is one.
Unlike conventional proling data, FCP data cannot be rearranged in some kind of 3-mode data
structure.
Because each assessor k=1,., K may have a different number of attributes (Mk), furthermore
the j th attributes of the assessors are not the same the next gure shows the structure of FCP
data set.
SENSORY ANALYSIS
Fig. 3. 3-Mode data structure representing conventional proling data: N
products are judged by K judges using M attributes.
FREE CHOICE PROFILE
11

Theory of Procrustes analysis.

In this section Procrustes analysis is introduced in geometrical way in order to describe easily the
analysis. A ve steps are provide for this statistical analysis in which explained the relationship with
FCP.

1. An easily geometrical look.

Initially, each assessors data matrix Xk, consist of N rows with scores on Mk attributes. This data
matrix contains elements Xijk.
Where:

i = Is the index over the N products


j= 1,,Mk The number of attributes of the k th assessor
k= 1,K The number of assessor

The scores in assessor`s datamatrix describes N objects using M attributes. Geometrically the N
points can be seen as to lie in an M-dimensional space for example with M= 2 attributes we can
draw a plane with the N points in it, but in general M will be (much) larger.
Next plot shows an example about conguration of N points from the data of an assessor judging
on only 2 attributes. ( We are using only 2 attributes because the dimensional space in this
projection give to us an imaginable lower dimensional space, often two dimensions, so it can be
plotted on paper ).

SENSORY ANALYSIS
Fig. 4. Data structure representing free choice prole data. N products are
judged by K judges using Mk attributes.
FREE CHOICE PROFILE
12

We have M-dimensional congurations of N points for all K assessors. Suppose that we deal with
assessors, to keep this example. We can draw two different conguration of N=4 points (look at
Fig.5). The objective of Procrustes analysis has been mentioned in this document like try ti get the
same objects as close to each other as is possible by shifting entire congurations ,rotating them
and reecting them if necessary.
The important underlying assumption is that distances between the N objects for one assessor
may not changed during these transformations. When the congurations are also allowed to
stretch or shrink the relative distances between the objects reect the relations between the
objects. Objects close together are similar, objects far apart are different . The reason to keep the
distances invariant is that in the process of matching, the relations between the N objects of one
assessor should not change. Similar objects must remain similar, different objects must remain
different.

2. Transformations.

The transformation mentioned above, shifting, rotating, reecting and stretching or shrinking, that
make up a Procrustes analysis, turn out to correct for a number of assessor effects.

3. The Level-Effect : Translation.

The so-called level-effect manifest itself by the different average scoring position on a line scale of
different assessors. One assessor may give all N products scores that lie between, say, 5 and 25
and another assessor may use scores from 60 to 100 (assuming a 1 to 100 line-scale score).
These two extreme assessors could very well perceive the objects identically, and would perhaps
agree with one another completely, had not they possessed such different scaling behavior. This
level-effect can easily by corrected for by expressing the scores as deviations from the average
score of an assessor on attribute . Geometrically this result in translating the entire conguration of
SENSORY ANALYSIS
Fig. 5. Two congurations with points representing scores on four products from
assessor 1 (A1, B1, C1, D1) and from assessor 2 ( A2, B2,C2,D2) with their
centroids M1 and M2
FREE CHOICE PROFILE
13
an assessor such that the centre of the N objects-points coincides with the origin of the space look
at gure 6.
The centres M1 and M2 in gure 5 re shifted onto each other and this point is labeled C in gure 6.
Mathematically this translation operation is known as column-centring, in Analysis of Variance
terms the assessor main-effect is removed.

4. The interpretation-effect: Rotation/ Reection.

The transformations which allows for the fact that the attributes do not have to be the same (the
interpretation-effect) for all assessors are rotation and reection.The entire conguration of an
assessor can be rotated to bring the N objects-points in agreement with the N points of the other
congurations. If necessary the conguration can be reected in a particular dimension too.
As can be seen from gure 6, the object-points are not very close yet, the lines between the pairs
of points (A1,A2), (B1,B2) etc. Indicate the distance that is to be minimized. Mathematically and
rotation and (reection) are represented in rotation matrix Hk for the conguration of assessor k.

SENSORY ANALYSIS
Fig. 7. Congurations after centring and rotation. This
gure shows the two example congurations after
rotation. Note that the N points actually are closer
(A1to A2, B1 to B2, C1 to C2, D1 to D2) than in gure
6.
Fig. 6. Centred congurations of two assessors.
FREE CHOICE PROFILE
14

5.The range-effect: Isotropic Scaling.

Another individual scaling effect. This is shown by the different ranges of scoring that the
assessors use. One assessor may give scores ranging between 10 and 95 another assessor uses
from 60 to 80. This difference in range is another unwanted effected caused by individual
differences in scoring behavior.
The underlying perception is believed no to depend on these differences in scale range, so the
effect is controlled for. The correction is used is called isotropic scaling, which means that a
conguration is shrunk or stretched in its entirety alike in all directions of space.
A different scaling range shows as a different extensiveness of the congurations. Figure 7
showed the two example congurations after centring and after rotation. It can be seen that the
second conguration is contained within the rst. The second assessor must have used a smaller
range of the line-scale.
The thick lines can now be shortened by stretching the inner conguration a little bit. The result of
this operation is shown in the next gure.

Applications of Procrustes analysis to sensory data.

Procrustes analysis , can be an effective way of obtaining information about samples, their
attributes, and judges assessing them. Although Procrustes analysis has most often been used for
Free-Choice Proling data, it as other applications as well. In general this method is used for
multiple reasons like:

1) To assess judges performance.


2) To describes samples differences.
3) To relate instrumental data with sensory.
4) To compare methodologies.
5) For acceptance and preference testing.

SENSORY ANALYSIS
Fig. 8. The two example congurations after centring and isotropic scaling.
FREE CHOICE PROFILE
15

Conclusions.

By using free-choice proling and analyzing the resulting data using Procrustes analysis, sample
information, as well as judge information can be provided.
Judges responses can then be evaluated and outliers identied. One note of caution in using free-
choice proling and applying Procrustes analysis to data is that a false sense of security may be
obtained from its use.
The analyst must still examine the data judge whether it is meaningful within the context of the
test. Sensory professionals must not be misled by the power of this statistical tool. Training is
indeed very important in getting accurate and reliable information about a product.
Free Choice Prole judges still must be trained in the use of the scales, and to be consistent.
Procrustes analysis can be a valuable tool to sensory scientist who understands not only the way
in which it can be applied to sensory data, but also its limitations.

References.

Arnold, G.M. and Williams, A.A 1986. The use of generalized Procrustes techniques in sensory
analysis. In Statistical Procedures in Food Research ( J. R. Piggott, Ed.). Elservier Applied
Science, London.

Howard Moskowitz, Ph.D.Applied Sensory Analysis of Foods Valhalla, New York.

David Kilcast 2010. Sensory analysis for food and beverages quality control .A practical guide.,
CRC Press Boca Raton Boston New York Washington, DC.

Helbert Stone and Joel L. Sidel. Sensory Evaluation Practices, University of California, Tragon
Corporation Redwood City, California.

Multivariate Analysis of Data in sensory Science Ed. By T. Noes and E. Risvik 1996 Elsevier
Science B.V. All rights reserved.

Food Quality and Preference 1989 1 (2) 69-73 (c) Longman Croup UK Ltd 1989
0950-3293/89/012040691Received 5 August 1988 Accepted 30 December 1988.

IFAPA, Centro de Palma del R ! o (A


!
rea de Tecnologa Postcosecha e Industrias
Agroalimentarias), Avda Rodrguez de la Fuente s/n, 14700 Palma del R ! o (Crdoba),Gower, J.
C. and Dijksterhuis, G. B. (2004) Procrustes Analysis. Oxford: Oxford University Press.

Gower, J. C., Dijksterhuis, G. B., Wieringa, J. and van Perlo, F. (2007) An optimal scaling version
of generalised Procrustes analysis. To be published.

SENSORY ANALYSIS

También podría gustarte