http://www.foodproductdesign.com/archive/1992/0592CS.html
An Inventory of the Toolbox
May 1992 -- Cover Story
By: Scott Hegenbart
Editor*
*April 1991-July 1996
Being a fan of Star Trek: The Next Generation, I
often dream about what it would be like to be a food
product designer in the 24th century. It would be
pretty simple, I imagine, thanks to having a trusty
food replicator at my disposal.
"Computer, I'd like to create a new product."
"State product description," the dutiful computer
would respond.
After I'd entered the idea for the product, the
replicator would just whip out a perfect prototype,
right? Not quite...
"Insufficient data. Please state product
parameters in recognized units."
Egad, I guess even 24th century technology has
its limitations.
In our own century, formulating products to meet
specific consumer-based requirements for taste and
texture is a common challenge. Obtaining objective
data through sensory analysis is one way to keep the
project on track. In addition to aiding in new product
development, sensory analysis also can help with quality
control, process changes and in determining shelf-life.
Unfortunately, misconceptions and a lack of understanding
often leave sensory analysis underutilized.
"Often, there's no appreciation for how we try to
make sensory evaluation a scientific test," says
Kathleen Rutledge, sensory analyst for Provesta Corp.,
Bartlesville, OK. "All of the controls and consistencies
are a mystery unless seen and appreciated."
To maximize its capabilities as a development
tool, food product designers must know more about what
sensory analysis is and what it can do.
Definition difficulty Some of the misunderstanding regarding sensory
analysis is based on the fact that it means different
things to different people.
"It's really hard to get a handle on an exact
definition because there is such a fine line between
sensory analysis and market research," explains
Jeffrey Kroll, Peryam & Kroll Research Corp., Chicago.
"Some say the difference starts when you ask for an
opinion."
Because sensory testing data is useful in such a
broad range of applications, it tends to be a
chameleon that changes its purpose depending on the
objective. In one instance, sensory data might provide
the ability to communicate information for a
particular product in a way that is understood by the
business managers. On the other hand, a major use for
sensory testing is in quality control.
Whatever the application, the common denominator
in sensory analysis itself is that it is the measurement,
analysis and interpretation of a food's characteristics as
perceived by the human senses. The detector, in this case,
is not a mechanical instrument, but a sensory panel.
To make the results of sensory testing most
useful, they have to be consistent. To achieve
consistency, the sensory panel must be carefully
selected or the information may be skewed. For
example, if you were to select a panel from among your
fellow food scientists, they'll be much more critical
than the typical consumer. Even non-food scientists at
a company can throw a panel off because they have a
specific interest in the company's product line.
For this reason, sensory analysts believe better
results are obtained from panelists selected from
outside the company. At the very least, in-house
panelists should come from a department that is far
removed from the R&D function Admittedly, this is
more effective at larger companies.
As far as the degree of training a panel should
receive, many possibilities exist. Trained panels tend
to be more consistent -- a critical element in
descriptive tests where a unified language and
reproducible profile results are essential. Because
training is costlier, some analysts choose to use a
general panel of only partially trained people for
most other tests.
Sometimes, completely untrained panels are best
because not everyone responds well to the training. If
you only focus on people that do respond well, you
risk unbalancing the results in preference or
acceptance testing.
Whatever the selection and training process, the
goal is to be able to enforce controls for
consistency. This should be true no matter what test
methodology is used.
Methods to the madness With a panel in place, the "sensory instrument"
has its detector. Next, test procedures are required
to make use of the information provided by the
detector. Sensory testing methods fall into three
general categories: discriminative, descriptive and
affective. Discriminative and descriptive analyses
provide analytical results while affective tests seek
out the acceptance of the sample to the panel.
-- Discriminative tests require panelists to find
specific differences in samples. This sort of test can
be quick and the results valuable. According to Ellen
Daw, sensory analyst for J.M. Smucker Co., Orrville,
OH, discriminative tests are perfect for screening
samples to eliminate variations that aren't
significantly different. The same methods also are
useful for making sure products remain consistent when
scaling up.
Among the most commonly used difference tests is
the paired-comparison. Here the panelists evaluate two
coded samples. In a simple difference, the panelists
must choose whether or not the samples are different.
A directional difference requires panelists to select
the sample that is stronger in a given characteristic.
Other tests may be used to determine the
significance of a difference. One is the duo-trio
where the panelist identifies one sample out of two
that matches a standard. The triangle test does not
identify a standard and the panelist much choose the
different sample from among three coded selections.
Statistics may be applied to the results of either of
these tests in order to determine the significance of
any difference should one be detected. Analysts often
prefer the triangle test because the odds of a
panelist choosing the odd sample by chance are only
3-to-1 versus2- to-1 for the duo-trio.
At times, it's necessary to compare more than one
variation. Ranking is one test that can determine
differences between several samples at the same time.
All samples, including a control, if any, are ranked by the
panelist by the intensity of a given characteristic. The
analysis averages the total scores for each sample and
determines the differences statistically.
Another multiple variation test is the scalar
difference. As with ranking, panelists are given
several samples. This time, the control is identified
and the panel rates each unknown on a scale that can
range from "no different from control" to "somewhat
different" to "extremely different." Statistics again
can determine the significance of any differences
found.
One common application for difference testing is
the determination of threshold levels for flavors. For
this, a series of paired comparison, duo-trio or
triangle tests will be administered to the panel. The
threshold value usually is determined by a
pre-determined percent majority of correct responses
by the panel.
-- Descriptive testing looks beyond simple differences
and seeks to identify and quantify specific
characteristics. These characteristics may include
various flavor notes, product texture, physical
sensations ("feeling factors") such as heat or cold.
This form of test even more closely emulates the idea
of a human chromatograph and requires the panelists to
be trained. Many methods of descriptive analysis have
been created over the years, but most are variants of
profile analysis and quantitative descriptive
analysis.
Quantitative descriptive analysis (or QDA) testing
produces a graphic representation of a products
sensory attributes in the form of the familiar QDA
"spider" chart. A panel of at least six to 10 members
rates different product attributes and gives each a
rating for relative intensity. The mean intensity
score for each attribute is used to assemble the
graph. Product attributes are scored by the order in
which they are first perceived. This order is
maintained in the chart reading clockwise from the
top.
Flavor profile analysis (FPA) differs both in how
intensities are assigned and the subsequent graphical
presentation of the data. Panelists first work individually
to characterize different product attributes and assign
intensity ratings on a continuous scale. A leader then
conducts an open discussion and the group merges their
individual results into the final profile.
As with QDA, product attributes are evaluated by
the order of perception in flavor profile analysis.
When graphed, flavor profile results are a series of
bars with the height (or amplitude) indicating
intensity. The order of perception is read from left
to right.
In addition to special training, descriptive
panels also create the terminology for describing
characteristics. Both the training and the terminology
are geared toward a specific type of product. Although
statistical treatment of the data is not required in either
form of descriptive testing, it may be done to determine
significant differences in certain attributes between
different samples.
-- Affective testing is used to determine preference
and acceptance of samples -- usually in terms of
qualitative factors such as like-dislike. The tests
themselves may involve a simple choice, a ranking of
samples by intensity of liking or a continuous hedonic
rating. While the data from these tests is geared to
show preference, a product's degree of acceptance
usually can be inferred from the results.
By modifying the objective of a paired comparison,
panelists can perform a simple, rapid acceptance test.
In paired preference, the panelist again is given two
coded samples, but is asked which one they prefer,
rather than if there is a difference.
Another useful and rapid application of this is to
combine the paired comparison/preference tests. A
panel is given two samples. After determining if there
is a difference they then are asked to select their
preference if they perceive a difference.
To obtain comparative results for several samples,
the sensory analyst may recommend a ranking preference
test. As in difference testing, the panelist receives
a series of samples. Rather than ranking them by the
intensity of a specific characteristic, though, the
panelist ranks the samples based on their degree of
liking. Although useful, the data from a ranked
preference test must be used cautiously because the
results only compare liking with respect to the other
samples presented.
To get a better idea of the overall degree of
liking, analysts use rating tests. These can be
designed not only to determine just how much a
panelist likes a sample, but what attributes about
that sample contribute to that liking.
Perhaps the best known of these tests is the
hedonic rating scale and its many variations. Most
common is the nine-point scale where the panelist
rates specific attributes on a scale from "dislike
extremely" to "like extremely." In addition to getting
a feel for the liking of specific attributes, simple
preference and acceptance for a sample may be
statistically inferred from the overall scores.
Hedonic tests are very flexible and can vary in
design to determine preference for many different
attributes. The tests may involve a single sample
(monadic testing) or several. For the latter,
different formulations of a new product may be
compared for liking, or a sample product may be tested
against either an in-house standard or a competitive
target.
The human factor is a limitation in hedonic scale
testing that must be considered in the test design.
Keep in mind that only so many samples may be tested
in a given period and that panelists may not be able
to focus on specific attributes as an expert would.
Questions of correlation
Many in the industry believe that rating tests
dance on an invisible fence between sensory analysis
and consumer testing. The side of the fence on which
the test falls depends on the questions that are
asked. For example, asking a panelist whether they
like a certain product attribute is considered sensory
testing, but asking if that attribute would affect the
frequency with which that same panelist would buy a
product is a consumer test.
"This is a gray area," says Beverly Kroll, Peryam
& Kroll. "The sensory discipline is intimately
involved in many different types of testing. I feel it
is more 'product testing using sensory skills.' "
In the real world of product design, the answers
to both types of questions carry tremendous weight. It
would make sense from both a time and cost standpoint
to combine the two. After all, if you have a panelist
in the booth to determine preference, why not ask to
what degree that preference extends to buying choices.
A few independent research firms are taking such
an approach to product testing using sensory skills. A
client can, for example, test a sample for acceptance
and overall liking while obtaining valuable data on
potential purchase frequency. By using the controls
and smaller panels of the sensory discipline, usable
results are available much more quickly and for less
cost than many full-scale consumer tests. With
thoughtful design to the experiment, this sort of
testing can be a valuable developmental tool.
Correlating sensory evaluation with consumer tests
not only is a time saver, but may very well have a
promising future. Possibly by the 24th century, food
scientists will be able to obtain a full-scale product
taste analysis by computer. But will the computer be
able to say whether the product will sell? I don't
think so.
The Evolution of Sensory Analysis
As a science, sensory evaluation is comparatively
young. This is one contributor to why lower credence
often is given to sensory data. On the positive side,
is that one can easily study the history of sensory
analysis because, unlike most other sciences, many of
the pioneers still are alive to document and recall
the evolutionary process. One of these is David
Peryam, senior psychologist at Peryam & Kroll Research
Corp.
According to Peryam, sensory analysis actually was
used in a more informal way as early as the 1920s. The
1930s even provide evidence of a movement toward more
formal studies. Most of this work, however, used
trained experts. "A company expert or expert panel was
widely been used by the dairy industry, for example,"
says Peryam. "This person was responsible for assuring
the good quality of every batch."
With the continued expansion of product lines and
product categories, it eventually became impossible
for an expert to remain an expert in everything.
About this time, tobacco companies were using
blind taste tests. While this was done primarily for advertising
purposes, it showed that people could use the sense of
taste to make identifications. Studies conducted at Cornell
further demonstrated people's ability to evaluate different
tastes through the evaluation of oxidized flavors in milk.
Studies such as this soon provided the first inklings of modern
sensory science.
The non-expert factor also was explored by a group
at the University of California, Davis. Here, wine taste-testing
was common, but only by enologists (wine making experts.)
U.C.-Davis researchers became interested in a non-trained
opinion. The Department of Agriculture even was experimenting
with paired comparison taste tests, although not extensively
until after World War II.
The earliest practice that resembles the
scientific methods we now use started in the liquor
industry at the James Seagram company in 1938. In the
liquor industry, consistency is paramount (unlike
wines which vary by year of vintage.) Maintaining this
consistency year after year can be challenging since
batches of spirits take years to age. To assure consistency,
the aged results were blended and checked through
paired-comparison tests.
This work eventually lead to the creation of the
duo-trio, and triangle difference test methods. Many
of the other basic tests used today actually go back
about a century and have their basis in psychology.
"Getting into psychological feelings of pleasure
provides the capacity to evaluate hedonics," adds
Peryam.
Around 1943, the army quartermaster got involved
in sensory because even the most nutritious diet is
useless if it goes uneaten. To maintain the fighting
health of the troops, the quartermaster sought to
assure acceptance of the army diet to the soldiers.
"To a soldier, the two most important things are
mail and food," says Peryam.
Here, the modern hedonic scale was developed to
determine preferences among soldiers. Preference and
acceptability pre-qualifying samples were eventually
required to be sent to the quartermaster for acceptability
testing prior to awarding of any contract for supplies.
The government further became involved in sensory
testing through the work that resumed around 1950 at
the regional USDA laboratories.
The 1960s and the 1970s brought greater realization
of the growing importance of the consumer. This
prompted companies to rely less on expert panels and
to focus on the opinions of the people who actually
would use the product. Sensory testing finally came
into its own. The major change in sensory analysis in
the last 20 years has been increased awareness and
greater credence assigned to sensory data.
As far as the future, Peryam is concerned about
sensory analysis becoming too popular.
"Many people are doing good work," says Peryam.
"There's a general sharpening of the methods and in
the interpretation of the results. Many, though, are
just going through the motions."
Still, Peryam admits that colleges and
universities are becoming more aware. Consequently,
more courses in sensory analysis are being offered
making students more aware. Some, such as U.C.-Davis,
even offer degrees in sensory analysis.
"The more people become trained, the better they
will be. These people will elevate the status of both
the procedures and the results," says Peryam. " People
will take sensory analysis even more seriously."
© 1992 by Weeks Publishing Company
3400 Dundee Rd. Suite #100
Northbrook, IL 60062
Phone: 847-559-0385
Fax: 847-559-0389
E-mail: info@foodproductdesign.com
Website: www.foodproductdesign.com
Recent Comments