Sei sulla pagina 1di 100

FOOD 3007 and FOOD 7012



Associate Professor Richard Mason

The University of Queensland
Stephen Nottingham
Sensory Evaluation


These notes form the basis of a practical workshop presented for personnel at Naresuan
University, Phitsanulok, Thailand in July, 2002. We would like to thank Michael O’Mahony
for his permission to include copies of the statistical tables from his book “Sensory
Evaluation of Food: Statistical Methods and Procedures” and material supplied by the Centre
for Food Technology, DPI, Brisbane.

COPYRIGHT 2 R L Mason and S M Nottingham

Sensory Evaluation


ACKNOWLEDGEMENTS ....................................................................................................2

PROGRAM ............................................................ ERROR! BOOKMARK NOT DEFINED.


THE HUMAN SENSES IN SENSORY EVALUATION .....................................................7

THE SENSES - AN INTRODUCTION .................................................................................7

SENSE OF SIGHT- .................................................................................................................9

THE SENSE OF SMELL......................................................................................................13

THE SENSE OF TASTE.......................................................................................................15

THE SENSE OF HEARING.................................................................................................19

THE SENSE OF TOUCH .....................................................................................................20

SENSORY INTERACTION .................................................................................................21

OPERATIONAL PRINCIPLES OF SENSORY TESTING .............................................23

DESIGN OF A SENSORY TESTING AREA.....................................................................31

STATISTICAL PRINCIPLES .............................................................................................34

SENSORY EVALUATION METHODS .............................................................................38

AFFECTIVE TESTS .............................................................................................................38

SPECIFIC TEST METHODS ..............................................................................................39

PAIRED PREFERENCE TEST.......................................................................................39
RANKING FOR PREFERENCE.....................................................................................41
RATING FOR PREFERENCE ........................................................................................44

SENSORY EVALUATION IN CONSUMER TESTING ..................................................46

ANALYTICAL SENSORY TESTS: ....................................................................................53

DIFFERENCE TESTING.....................................................................................................53

SIMPLE DIFFERENCE TEST ............................................................................................53

TRIANGLE TEST .............................................................................................................53
DUO-TRIO TEST..............................................................................................................55
TWO-OUT-OF-FIVE TEST.............................................................................................59
COPYRIGHT 3 R L Mason and S M Nottingham
Sensory Evaluation

“A” – “NOT A” TEST.......................................................................................................59

DIFFERENCE-FROM-CONTROL TEST (DFC) .........................................................59

DIRECTIONAL DIFFERENCE TESTS ............................................................................65

PAIRED COMPARISON TEST ......................................................................................65
RANKING TEST ...............................................................................................................67
RATING TEST ..................................................................................................................69

STATISTICS FOR SENSORY: DIFFERENCE TESTING .............................................73

DESCRIPTIVE TESTING ...................................................................................................76

STATISTICS FOR SENSORY: DESCRIPTIVE TESTING ............................................81

SELECTION, TRAINING AND MOTIVATION OF A PANEL .....................................86

REPORTING .........................................................................................................................91

SELECTED BIBLIOGRAPHY............................................................................................92

JOURNALS ............................................................................................................................94

STATISTICAL TABLES......................................................................................................95

COPYRIGHT 4 R L Mason and S M Nottingham

Sensory Evaluation


Sensory evaluation - A scientific discipline used to evoke, measure, analyse and interpret
reactions to those characteristics of foods and materials as they are perceived by the senses of
sight, smell, taste, touch and hearing.

Sensory evaluation was one of the earliest methods of quality control and it is still widely
used in industry. However, the level of application depends on the situation (e.g. beer and
wine tasting to operators sampling of products from production line).

Four variables affect sensory evaluation:

• The Food
• The People
• The Testing Environment
• Methods

Sensory evaluation terminology

• Sensory evaluation
• Sensory Analysis
• Organoleptic Analysis
• Taste Testing
• Psychophysics
• Subjective Evaluation


• Gives real answer regarding consumer quality

• Relatively cheap process (depending on how it is done)
• Rapid
• Many applications
• Objective methods are more reliable, accurate and reproducible. However, they must
be correlated to sensory evaluation to indicate a consumer response.


• Time consuming
• Expensive to run
• Method selection
• Analysis
• Interpretation

Industry applications of sensory evaluation

• Product development
• Product matching
• Product improvement

COPYRIGHT 5 R L Mason and S M Nottingham

Sensory Evaluation

• Process change
• Cost reduction
• New raw materials selection
• Quality control
• Storage stability
• Product grading / rating
• Consumer acceptance
• Consumer preference
• Panel selection / training
• Correlation subjective / objective

Sensory Standards

Aus Standard Year Title

AS 2542.0 1995 Sensory analysis of foods - Introduction and list of methods

Sensory analysis of foods - General guide to methodology -

AS 2542.1.1 1984
General requirements
Sensory analysis of foods - General guide to methodology -
AS 2542.1.2 1984
Types and choice of test
Sensory analysis of foods - General guide to methodology -
AS 2542.1.3 1995
Selection of assessors
Sensory analysis of foods - Specific methods - Paired
AS 2542.2.1 1982
comparison test

AS 2542.2.2 1983 Sensory analysis of foods - Specific methods - Triangle test

AS 2542.2.3 1988 Sensory analysis of foods - Specific methods - Rating

AS 2542.2.4 1988 Sensory analysis of foods - Specific methods - Duo-trio test

AS 2542.2.5 1991 Sensory analysis of foods - Specific methods - 'A not A' test

AS 2542.2.6 1995 Sensory analysis of foods - Specific methods - Ranking

AS 2542.3 1989 Sensory analysis of foods - Glossary of terms

COPYRIGHT 6 R L Mason and S M Nottingham

Sensory Evaluation



The sensory properties of foods are related to three major attributes:

• Appearance - colour, size, shape;

• Flavour - odour, taste; and
• Texture - mouth feel, viscosity and hearing.

These attributes are expressed as a continuum and not as finite properties. It is impossible to
rate each one individually unless special precautions are taken, e.g. blindfolds, nose clips,
coloured lights, purees.

COPYRIGHT 7 R L Mason and S M Nottingham

Sensory Evaluation

Humans possess about 30 different senses. However, the sensory properties of foods are
perceived through the senses of:

• Sight;
• Smell;
• Taste;
• Touch; and
• Hearing.


A stimulus is any chemical or physical activator that causes a response in a receptor, e.g. eye
is receptor for light, ear is receptor for sound.

An effective stimulus produces a sensation, the dimensions of which are:

• Intensity/strength;
• Extent/separation;
• Duration/retention; and
• Hedonics/like-dislike.


Receptors are the stimuli detecting cells of the sense organ, e.g. taste buds on tongue, light
receptors in retina of eye.


Perception is the psychological interpretation of sensations determined by comparison with

past experiences, e.g. the sour taste of lemons is the perception of the sensation received by
the receptors (taste buds) from a chemical stimulus (citric acid).

COPYRIGHT 8 R L Mason and S M Nottingham

Sensory Evaluation


The appearance of food

Stimuli = visible light

Receptor= retina of the eye
Perception=sight, vision, appearance

The appearance of foods is a major factor governing its acceptability and can be subdivided
into three main categories:
• Optical properties- colour, gloss and translucency
• Physical form-shape and size
• Mode of presentation-lighting packaging etc

Optical properties


Vision is a complex phenomena consisting of several basic components. A stimulus, light,

from an external source interacts with the object and is brought to focus on the retina of the
eye. The retina is the receptor of vision and contains two types of cells. The rods are
responsible for vision in dim light and the cones are responsible for colour vision. Light
incident on these cells causes a photochemical reaction that generates an electrical impulse
which is transmitted to the brain via the optic nerve. Colour blindness is caused by loss or
lack of colour receptor cells in the cones. Approximately 8% of the population have some
defect with relation to colour; mostly males.

COPYRIGHT 9 R L Mason and S M Nottingham

Sensory Evaluation


Visible light is that part of the electromagnetic spectrum which radiates between wavelengths
of 380 - 770 nm.

Different wavelengths produce different colours

380 - 450 nm =violet

450 - 475 nm =blue
500 - 575 nm =green
575 — 590 nm =yellow
590 — 770 nm = red

[NOTE: All electromagnetic radiations are physically the same. However, the optical system
of the eye is such that only the visible range of wavelengths is absorbed by the lens.]

Light sources

Incandescent lights consist of a tungsten filament which is heated in an inert gas. The higher
the temperature, the more light produced. Light from this source tends to be harsh and tends
to highlight the red end of the spectrum.

Fluorescent lights operate by electrical excitation of atoms that produces spectral lines at
specific wavelengths which then impinge onto fluorescent materials which convert the
incident light into light at a longer wavelength. Light produced is softer but can produce
colour distortion at particular wavelengths.

Natural light is too variable for use in evaluating appearance of foods.

Light - Object interactions

Light incident on an object may be:

• Absorbed;
• Reflected;
• Transmitted; and
• Refracted.

The relationship between and within each of these components is responsible for the colour
and gloss characteristics of the food. The main light/object interactions produced are:

Chroma/purity; and

Physical form

The second class of product appearance is physical form that can be subdivided into three
COPYRIGHT 10 R L Mason and S M Nottingham
Sensory Evaluation

• Shape;
• Surface texture; and
• Visual consistency.

Shape and size are important from a food technologist's point of view because these can be
altered during the manufacture of processed products. Some examples include:

• Sliced, diced, pieces whole

• Length of frozen French fries
• Cut of beans
• Extrusions

Surface texture can indicate product texture. Some examples include:

• Open dry structure of meat

• Wrinkling of peas
• Wilting of lettuce

Visual consistency can indicate product viscosity as in:

• Setting of a jelly
• Syrups of different concentrations
• Pastes and purees

Mode of presentation

This aspect should be considered from a marketing point of view and is important because it
influences sales. Mode of presentation is applicable on the supermarket shelf (at retail level)
and also in terms of presentation at the table (home and restaurant).

Factors to be considered are:

• Product description - name, price, ingredient, etc;

• Packaging - shape, design, colour;
• Contrast - phenomena of adjacent colours; and
• Illumination - affects apparent product colour.

COPYRIGHT 11 R L Mason and S M Nottingham

Sensory Evaluation


Appearance is an important aspect of food quality as it is the first subjective evaluation made
of food quality. The product has to pass the visual assessment before the consumer can or
will consider the other parameters such as taste and texture.

Factors that should be considered in evaluating product appearance include:

• use of standard conditions:

• light source (type, intensity, colour);
• background; and
• style of presentation (unless tested).
• selection of appearance attribute(s) for inclusion on scoresheet;
• using appearance to reduce tasting load;
• should be masked to eliminate unwanted interactions when assessing parameters
involving other senses; and
• colour charts/standards help rating.

COPYRIGHT 12 R L Mason and S M Nottingham

Sensory Evaluation



Stimuli = volatile chemicals

Receptors= olfactory cells in the nose
Perception=smell, odour, aroma, flavour

Smell is one of our most primate senses. Supposedly prehistoric people were more
influenced by smell than other senses.

The human nose is capable of detecting thousands of different odour substances. However,
our sensitivity is much less than other animals. (Animals use smell - food, mating, territory

Smell is detected both before and during eating. Smell is an important aspect of flavour.
There are 20x106 olfactory receptors, but only about 1000 taste receptors.

Odour description requires the development of an odour/flavour memory, e.g. fishy, flowery,
woody. This is the basis of flavour/odour memory development by wine judges and
milk/cheese graders. Individuals vary a great deal in their sensitivity to different

Anatomy of olfactory system

From the diagram it can be seen that most of air misses the olfactory area. Only 5-10% of
inspired air passes over olfactory receptors. However, this amount can be increased by
sniffing harder; obviously the more air which passes over the receptors the better the
COPYRIGHT 13 R L Mason and S M Nottingham
Sensory Evaluation


The large number of olfactory receptors (20x106) enable detection of :

• More odours than tastes;

• A greater variety of odours; and
• Odours at much lower concentration (10 molecules/mL).

In order for odour to register:

• Substance needs to be volatile enough to get into air in the sensory region.
• Substance needs to be partially soluble in mucus covering of receptors.
• Minimum number of odorous molecules need to be present.
• Need to be in contact with receptors for minimum time.

Olfactory intensity

Human nose is about 10-100 times more sensitive to odours than any physico-chemical
analysis (e.g. gas chromatography). It has been demonstrated that human nose is capable of
detecting ethyl mercaptan at a concentration of 0.01 mg/230m3 of air, which is equivalent to
about 8 molecules/receptor.

Olfactory threshold

Detection threshold is the concentration where smell is detected.

Recognition threshold is the concentration where the smell is recognised.

Olfactory interactions

Nature of the response may change with concentration (e.g. perfumes at low concentration
are pleasant but at strong concentration may be unpleasant).

Interaction of odours:
• Additive - increase intensity;
• Suppressive - decrease intensity; and
• Blending - when new odour unrelated to originals.

Olfactory adaptation

Initial sensation maybe strong - but weakens and makes identification difficult; this is due to
adaptation of olfactory receptors.

In testing we therefore need to allow for this by:

• Taking first impression of odour and/or
• Waiting between tests to allow receptors to recover.

COPYRIGHT 14 R L Mason and S M Nottingham

Sensory Evaluation


• Smell is a major component of food flavour.

• Human nose is much more sensitive than analytical instruments.

• Foods contain numerous compounds of varying volatility that can make analytical
interpretation difficult (e.g. strong peaks may produce weak odour whereas weak
peaks may produce a strong odour).

• Smell measures perception of a mixture; analytical testing does not.



Stimuli = soluble chemicals or chemicals which are solublised during chewing

Receptors= taste buds in mouth
Perception=taste, flavour

What is commonly referred to as taste/flavour is actually a combination of:

• Taste;
• Smell;
• Touch; and
• Temperature.

Strictly speaking taste involves only those sensations mediated by the Gustatory Nerve Fibres
and these sensations have five (5) basic qualities:

• Salt;
• Sweet;
• Sour; and
• Bitter.
• Umami

Taste stimuli

Taste response requires an aqueous solution of the substance (stimulus) to contact the taste
buds. Therefore, saliva secretions are important in terms of ensuring contact between the
product and the taste buds. Saliva production is generally stimulated by chewing, as well as
the appearance and odour of the food. The tongue is important as it brings the food into
contact with the taste buds and also provides a mixing action which enables an even
distribution of food about the taste buds as well as preventing the development of
concentration gradients.

COPYRIGHT 15 R L Mason and S M Nottingham

Sensory Evaluation

Taste receptors

The receptors for taste are the taste buds and these are mounted on papillae (folds in the skin
of the tongue). The area of greatest response is the top of the tongue. Other areas in the
mouth and throat where taste buds are situated include: palate, pharynx, larynx, tonsils,
epiglottis, lips, cheeks, underside of tongue and floor of mouth.

Taste buds are mainly located at the tip, sides and rear of tongue. There is very little response
in the centre of the tongue. Different areas of the tongue are most responsive to different

• Tip sweet
• Sides - salty
• Sides - sour
• Rear - bitter

Taste cells constantly degenerate and regenerate. Their life cycle is 10 days and they are
easily destroyed by heat.

The tongue itself is important as it brings the food into contact with the taste buds and also
provides a mixing action which enables an even distribution of food about the taste buds as
well as preventing the development of concentration gradients.

COPYRIGHT 16 R L Mason and S M Nottingham

Sensory Evaluation

The five basic tastes

A basic taste is one for which specific taste buds have been identified as being
physiologically responsible for the particular taste sensation.


This is the simplest taste as only acids (H+) produce sourness and as the (H+) increases the
sourness increases

However there are some anomalies to this:

• organic acids are more acidic than expected.

• sourness of aliphatic organic acids relates to chain length.
• some amino acids are sweet (aspartane)
• picric acid is bitter
• sugar may enhance/depress sourness
• sourness is also affected by pH and acid
• presence of buffers affects sourness


The common substances that produce the sweet taste are the sugars and other hydroxy
compounds such as alcohols and glycols. Other substances such as lead salts, amino acids,
proteins, non-nutritive sweeteners (cyclamates, saccharin and aspartame ) also taste sweet.


Many crystalline water-soluble salts yield a salty taste, but only sodium chloride gives a pure
salty taste. Other substances taste salty but also bitter, alkaline, sweet and salt in various


Many chemically different compounds have a bitter taste. However, bitterness is mainly
associated with alkaloids such as caffeine, quinine, strychnine and nicotine. Originally it was
thought that bitterness was an indication of danger (poison). However, many alkaloids are
used as drugs (e.g. codeine) and many other bitter substances are harmless (glycosides, esters
and aldehydes and tannins in wines and tea).

Bitterness is generally perceived at very low concentration and a relationship appears to exist
between sweet and bitter as many sweet substances produce a bitter aftertaste (saccharin).
Bitterness is the taste which most people have difficulty in detecting and response level varies
greatly from individual to individual.

COPYRIGHT 17 R L Mason and S M Nottingham

Sensory Evaluation


Umami is the taste that has been shown to be associated with substances that contain
glutamate. The most notable example is mono-sodium glutamate (MSG). MSG is well
known as a flavour enhancer and can cause adverse reactions in some sensitive individuals.
However, there are many other compounds which contain glutamate and which are capable of
producing the savoury, spicy, brothy taste associated with MSG. Many foods contain
naturally high levels of glutamate.

Taste interactions

Having described the 5 basic tastes it is obvious that foods are a very complex system which
contain many different taste compounds and therefore many different tastes. The fact that
there are only 5 basic tastes and yet we are able to detect hundreds of different taste
sensations is due to a series of complex taste interactions that can range from simple 2 way
interactions to complex 5 way interactions

Interactions between the 4 basic tastes were previously described simplistically by the taste

Adaptation and fatigue

During exposure to a stimulus, sensitivity decreases due to adaptation and fatigue. This loss
in sensitivity varies considerably with the taste (sweet, sour, salty or bitter) and also with the
compound. For example, tasting a series of acids causes the sensitivity to be reduced by the
preceding acids. However, recovery is usually rapid because most common organic acids are
very soluble.

Taste thresholds and sensitivity

There is great variability between individuals in their levels of sensitivity. Sensitivity is

affected by:
• Temperature;
• Sleep;
• Hunger;
• Age; and
• Sex.

Absolute/Detection threshold - Concentration of stimulus at which a subject can detect a

difference between two samples in a paired test.

Recognition threshold - Concentration at which the specific taste can be identified.

Recognition threshold is generally higher than detection threshold.

Both absolute threshold and recognition threshold will vary between individuals. Most
people can detect taste within 0.2 - 0.6 seconds and therefore if there is no response within
this time the level is sub-threshold. However, recognition times vary between the basic tastes

COPYRIGHT 18 R L Mason and S M Nottingham

Sensory Evaluation

• Salt = 0.3s
• Sweet = 0.4s
• Sour = 0.5s
• Bitter = 1.0s
• Vision = 0.02s
• Hearing = 0.01s
• Touch = 0.005s

Reaction times also relate to retention times for example; bitterness has the longest reaction
time (1.0s) and the sensation lingers considerably after tasting.


• ·Five types of taste receptors - salt, sweet, sour, bitter and umami.

• ·Different areas of the tongue respond to different sensations.

• ·Substances must be dissolved for taste buds to detect them.

• ·Flavour of the food is a complex interaction of different tastes and odours.

• ·Sensitivity to taste varies between individuals and is affected by their physiological




Stimuli = physical movement of sound waves in a medium (air)

Receptor= ear drum
Perception=sound, hearing


Sound is the perception by humans of vibrations in a physical medium (air). The sound of
food when it is being eaten is an important aspect in determining quality.

Positive aspects:

• Snap, crackle and pop;

• Fizz of champagne or beer;
• Crispiness of lettuce or celery; and
• Tapping a melon for quality.

COPYRIGHT 19 R L Mason and S M Nottingham

Sensory Evaluation

Negative aspects:

noisy environment may distract tasters or mask product sounds.


(Texture, Kinesthetics)

Stimuli = physical contact between the food and body tissue

Receptors= muscles and nerves in mouth and fingers
Perception=touch, feel, texture, viscosity

Texture usually relates to solid food while viscosity relates to homogeneous liquid foods and
consistency relates to non-homogeneous liquid foods.

Instrumental methods only measure one aspect of "texture" and again cannot relate the
complex interactions which produce the perception of food texture.

Finger feel

Firmness/Softness indicates the eating quality of some food products:

• Ripeness level of fruit such as avocado and mango;

• Crumb texture of bread;
• Firmness of cheese; and
• Spreadability of butter or spread.

Juiciness can be used as a subjective quality index (eg the “thumbnail” test for corn).

Mouth feel


• Viscosity - thin to viscous, e.g. milk, cream.

• Consistency - thin to thick, e.g. fruit yoghurts.


Classification of textural characteristics - assessed mainly by chewing.

Textural Terminology Mechanical Characteristics

Hardness Soft, firm, hard, e.g. fruit ripeness, cheese maturity.

Brittleness Crumbly, crunchy, brittle, e.g. muesli bars and biscuits

Chewiness Tender, chewy, tough, e.g. meat.

Grittiness Gritty, grainy, coarse, e.g. stone cells in fruit, "sand" in

COPYRIGHT 20 R L Mason and S M Nottingham
Sensory Evaluation

Fibrousness Fibrous, cellular, e.g. string/fibre in vegetables.

Moistness Dry, moist, wet, e.g. cracker biscuit, cheeses, water


Oiliness/Greasiness Oily, greasy, fatty, e.g. french fries, chips.


As has been indicated previously when eating or tasting food there is a continuous
relationship between the senses and unless steps are taken to separate the individual senses or
stimuli, interactions may occur. It is not known whether interactions occur at the receptor
site or the brain. However, the second option would appear to be more likely.

Interaction between senses

This is the ability of a response from one modality to influence or affect the response from
another. There are two aspects of this:

Positive - interactions giving clues to possible identity, e.g. pink milkshake being strawberry

Negative - If clues are not correct this may lead to confusion and a wrong judgement, e.g.
pink milkshake with pineapple flavour.

Types of sensory interactions

Taste - odour

Receptors for these two senses are very close so that interactions between these senses are
highly likely and these may be important in classifying a particular taste.

Taste - tactile

The taste threshold for sugar, salt, caffeine have all been shown to be lower in water than in
tomato sauce. This may be due to the fact that in more viscous solutions the chemicals do not
react with the receptors as easily as in pure solutions.

Taste - sight

This is a very important aspect because vision is the first sense affected and appearance of a
product will have a major influence on absolute quality. Bright colours indicate strong
flavours whereas dull colours indicate mild flavours.
Other interactions include:

• Odour - Sight
• Odour - Tactile
• Taste – Hearing
COPYRIGHT 21 R L Mason and S M Nottingham
Sensory Evaluation

• Odour - Hearing

Multiple interactions

Multiple interactions between more than two modalities are also possible.

Example: Tasting food pureed, blindfolded and with nose clips gives a different response
than when interactions are allowed.

Interactions between stimuli

These interactions are more difficult to define and measure but are just as important as
interactions between the senses. Some examples include:

• suppression of one flavour by another, e.g. sweetness is suppressed by acidity. This is

the basis of ensuring brix/acid ratio for fruit juices are constant;

• neutralisation of one flavour by another;

• blending to produce a totally different flavour, e.g. garlic flavoured cheeses;

• partial blending producing a new flavour and the original flavours;

• no effect; original flavours are distinct and separate, e.g. fruit in cheese;

• intensification resulting in enhancement of flavours, e.g. salt and MSG on food

improves the natural flavours.

Similar situations may exist for all other stimuli.


Interaction must be considered when designing sensory panels. If only one sense or stimulus
is to be evaluated then all others must be masked. However, if interactions are required then
ensure this can be achieved by means of sample preparation.

COPYRIGHT 22 R L Mason and S M Nottingham

Sensory Evaluation


When evaluating properties of foods using people as measuring instruments it is important to

control the methods and conditions of testing as rigidly as possibly. This helps to eliminate
the numerous errors or biases that can be caused by psychological and physiological factors.

The mental attitude and physical condition of a taster, and the atmosphere of the testing
environment all influence their judgements. There are therefore a number of basic rules
which should always be applied, as stringently as circumstances allow, when running taste
panels. These relate to:

• Selection of panellists;
• Preparing the testing environment;
• Designing the experiment;
• Preparing samples;
• Serving samples.

General principles that should always be followed are:

Never ask anyone to taste food they do not like;

Make sure that the "correct" panellists are selected (see section on panel selection and
training) and that they know in advance when they will be required.

Keep a strict control over all variables except those being tested (e.g. sample size and

Make sure the environment gives optimum opportunity for concentration. Tasting properly is
a difficult job. Train panellists to be silent while tasting. This prevents panellists from
influencing one another.

Make tasting interesting and desirable. Use rewards to motivate taters, vary these and choose
foods that contrast with those being tasted. Motivated tasters are more efficient. Give
feedback on results whenever possible.

Avoid giving any unnecessary information to panellists that may influence their scores.
Tasters usually find what they expect to find; e.g. in a storage test they expect to find samples

Plan your experiment in advance. Which will be the best test to use? Consider all aspects
including how you will get the information required from your results (statistics). Run
preliminary tests, i.e. practise and choose the best method for:

Sample size - adequate but not excessive;

Serving temperature - standardise for all samples. It must be maintainable, and be an
acceptable temperature for the food;
Serving vessels;
Eating utensils.

Sample preparation and serving

COPYRIGHT 23 R L Mason and S M Nottingham
Sensory Evaluation

Serve tasters promptly and make sure they have everything they need.

Run a taste panel as you would expect a good restaurant to be run, i.e. give courteous
friendly service, be efficient, and serve good food.

Keep accurate records of any cooking or preparation methods used. Record temperatures and
size of samples served and any special conditions (e.g. coloured lighting).

It is important that panellists do not see the samples being prepared as this may indicate
quality difference.

Sample preparation should be uniform:

• Temperature
• Cooking
• Thawing
• Size and shape (provided this is not a variable)

Sample should be randomly allocated to:

• Avoid bias
• Overcome any non—uniformity

Sample size should be adequate:

• 30g solids
• 30mL liquids

Samples should be served immediately after preparation to reduce:

• Flavour loss
• Discoloration
• Textural changes

Sufficient samples should be prepared to allow for seconds

COPYRIGHT 24 R L Mason and S M Nottingham

Sensory Evaluation

Containers for presentation

Containers for presentation and tasting should be:

• Clean
• Identical for all samples and sessions
• Disposable containers or re—usable
• Coloured to mask product appearance (if required)
• Relevant to product

Serving temperature

• Serve at room temperature where possible

• Preference tests use normal temperature
• Difference tests may alter temperature to accentuate flavours/odours
• Do not overheat:

ƒ too hot to taste

ƒ drying out
ƒ off flavours
ƒ browning

Dilutions and Carriers

Most foods should be served in the way they are normally eaten. However, some products
such as spices, chillies, alcohol, onions, etc. may require dilution before testing. If dilutions
are used they must be uniform in terms of diluent and concentration.

Carriers are substances that are added to assist tasting of certain products. Carriers are a
problem because they can be:

• Expensive
• Time consuming
• Variable quality
• Difficult to control product/carriers ratio uniformity.

For example: developing a cake icing individually may not allow for interaction with flavour
or it may be incompatible with the cake (affects texture or falls off).

Number of samples

Samples / Sessions

The number of samples presented at any testing session will depend on:

• Type of product - strong flavour —> less samples

• Type of test
• Rating scale may require fewer samples
• Test dictates sample number eg: triangle test = 3 samples
COPYRIGHT 25 R L Mason and S M Nottingham
Sensory Evaluation

• Type of panel — trained / experienced -> more

• Experimental design

As a general rule usually not more than 6 samples/sessions.

Sessions / Trials

Before starting your scheduled tasting sessions run two preliminary sessions. These will
familiarise your panel with the scoresheet, the products to be tested and the procedures you
wish them to follow. It also gives you practice at preparing and serving the quantity of
samples needed, and a last chance to iron out any unforeseen problems.

In calculating the number of sessions consider the following:

• Total number of samples for tasting

• Statistical design
• Taster fatigue
• Motivation
• Type of panel (trained/untrained)

Phsiological factors in taste testing

Time of Tests

• Monday and Friday are recognised as being bad days for tasting
• Normally taste 1 hour before meals and 1 - 2 hours after
• Sometimes this becomes difficult in practice due to:
• Unavailability of tasters
• Number of sessions

Smoking / Taste Affecting Substances

As indicated earlier, smoking affects sensitivity to flavours —therefore should either:

• Not use smokers

• Ensure they do not smoke for at least 1 - 2 hours before tasting
• Chewing gum, mints and spices etc may also influence taste


Sensitivity of people suffering from illness is reduced -particularly those with colds or flu
(physical and psychological)

Likes / Dislikes

In preference testing a series of treatments within a specific product type, it is legitimate to

eliminate people who dislike the product (or those who are not discriminatory).

Palate Clearing
COPYRIGHT 26 R L Mason and S M Nottingham
Sensory Evaluation

It is a good idea to get panellists to cleanse their palate:

• Before tasting to remove any lingering tastes

• Between samples to reduce adaptation of taste buds.
• Warm water, biscuits, bread, apples may be used as a palate cleaning agent.

Palate clearing can be optional but whatever is done must be constant.

The time between samples should also be kept constant if possible

Perfumes / Spices

Ask panellists to refrain from wearing strong perfumes or breathing spicy odours wherever

Psychological factors

Because sensory evaluation is a subjective system, it is necessary to allow for any

psychological factors that may influence results and possibly lead to errors.

Good results can only be obtained from a co-operative, responsive panel. Tasting becomes a
chore when there are large numbers of samples/sessions involved. Motivating panellists by
can reduce this problem by:

• Stressing importance of work

• Stimulating company expansion
• Greater profits
• More pay
• Ensuring panellists know what is involved with the trial ie: sessions, products, when
and where tasting will be conducted
• Having adequate facilities
• Using effective methods and designs
• Publicising results obtained from work
• Rewarding panellists

Sample Coding

Remove possible bias or influence from samples codes. Do not use.

• Single digit numbers

• Consecutive letters
• Same codes at consecutive sessions

Randomly or statistically generated three digit number codes are best.

Order of Presentation

Always use either a random order of presentation or a statistically balanced design to avoid:

COPYRIGHT 27 R L Mason and S M Nottingham

Sensory Evaluation

• Donkey vote (first is best; last is worst)

• Position bias - in triangle tests middle one is different
• Contrast effect — good after bad appears better, or bad after good appears worse.

Devise your own system for remembering orders, e.g. 3 digit numbers - put in sequence of
one of digits. Keep it a secret!

Always work systematically in coding, labelling, setting up, e.g. as in reading a page

(1) Left to Right

(2) Top to Bottom

This provides an automatic check if something goes wrong.

Balance presentation of samples whenever possible. This avoids contrast effect.

ie. 2 samples A, B. - Half panel taste A first, other half taste B first.
- Half panel receive A on the left, other half
receive B on the left.

3 samples 6 different orders in which they are tasted. Use

every order the same number of times. Number
of tasters is a multiple of six.
- Position of samples on plate must also be

4 samples 24 different orders: use them all if possible (see

table on next page).

4 samples Generate random order. Write out set of cards

and shuffle them.

When you cannot use balance to eliminate bias, use randomisation.

COPYRIGHT 28 R L Mason and S M Nottingham

Sensory Evaluation

Four sample balanced orders

1 A B C D
2 A B D C
3 A C B D
4 A C D B
5 A D B C
6 A D C B
7 B A C D
8 B A D C
9 B C A D
10 B C D A
11 B D A C
12 B D C A
13 C A B D
14 C A D B
15 C B A D
16 C B D A
17 C D A B
18 C D B A
19 D A B C
20 D A C B
21 D B A C
22 D B C A
23 D C A B
24 D C B A

Expectation Error

Any information a panellist receives before a test will influence the results. This is called
expectation error. To overcome this:

• Do not give detailed information about treatments

• Do not use people on panel who know what the treatments are
• Sample coding and design can prevent expectation error

Logical / Stimulus Error

Tasters look for clues to get the “right” answer eg: a difference in sweetness may be
associated with sample differences such as size, shape and colour. This error can be
overcome by ensuring sample preparation is uniform or use masking.

Halo Effect

When more than one factor in a sample is evaluated at one time the result obtained may be
different than if factors evaluated separately. This can be overcome by tasting each aspect
separately. However, this is time consuming and would only be done if extremely accurate
results were required. Testing one aspect at a time in preference does not simulate the “real
COPYRIGHT 29 R L Mason and S M Nottingham
Sensory Evaluation

situation” ie: consumers do not taste every aspect separately.


Influence of other panellist may bias or influence results. This can be prevented by:

• Using booths
• Not allowing talking in tasting area
• Reducing outside distractions

Questionnaire design

Questionnaire design should be simple and easy to follow in terms of design and language
and make sure tasters know how to use it. You may need to include some instructions on the
scoresheet itself, but it is usually better to give instructions verbally to your panel first. The
questionnaire should generally not be more than one page and include:

• Name
• Date
• Time
• Product
• Sample codes
• Instructions
• Comments section

COPYRIGHT 30 R L Mason and S M Nottingham

Sensory Evaluation


The main considerations to keep in mind when preparing an area for sensory testing concern
the requirements for an atmosphere conducive to concentration, where conditions can be
controlled. Sensory panellists need somewhere comfortable and free from distractions if they
are to be able to "tune in" to the sensations triggered by the stimuli in the food products they
are tasting. Product characteristics can be markedly affected by temperature and humidity,
and appearance is affected by lighting intensity.

The conditions should be controlled in order to :

• Reduce bias
• Improve accuracy
• Improve sensitivity
• (compare to the conditions used in an analytical laboratory)

International standard (ISO 8589-1988)

The standard looks at the design of the testing area for both new and existing buildings. It
also specifies which recommendations are considered essential and which are only desirable.
Important points summarised from the standard are listed below. If designing an area that is
to be dedicated solely to taste panel work, these should be seriously considered.

Total area should include:

• Testing area with individual booths and a group area;

• Preparation area/kitchen;
• Office;
• Cloakroom;
• Rest room; and
• Toilets.

General testing area

• Easily accessible but in quiet position.

• Location - close proximity to preparation area, but separate entrance, and with
complete "close-off" capability.

• Temperature and relative humidity - constant, controllable, and comfortable.

• Noise - keep to a minimum, soundproof area as much as possible.

• Odours - keep area free from odours (air conditioner with carbon filters, slight
positive pressure).

• Use odourless materials in construction and decoration.

• Use odourless cleaning agents.

COPYRIGHT 31 R L Mason and S M Nottingham

Sensory Evaluation

• Decoration - use neutral, light colours for walls and furniture (e.g. off-white, light

• Lighting - ambient lighting must be uniform, shadow-free and controllable. For

consumer testing - as close to home conditions as possible.


Number - minimum three, normally five to ten - six is a useful number since it fits in well
with balanced ordering of 3 samples.

Space - allow sufficient space for movement of tasters and for serving samples.

Set-up - permanent booths recommended. Temporary acceptable. If adjacent to preparation

area include openings in the wall to pass samples through. Size and style specified. Consider
space for samples, utensils, spittoons, rinsing agents, scoresheets and pens, computerised
equipment. Include comfortable seats.

Lighting - uniform, shadow-free, controllable, adequate intensity for assessing appearance.

Devices to mask appearance (e.g. dimmers, coloured lights or filters).

Group work area

General Necessary for discussion and training purposes. Include large table and
several chairs. "Lazy Susan" useful. Include board for discussion notes, etc.

Lighting As for general area, with coloured lighting options like booths.

Preparation area

General Located close to assessment areas but no access to tasters. Design for efficient
work-flow. Well ventilated. Flexible services (i.e. plumbing, gas, electricity).

Equipment Depending on testing required. Include working surfaces, sinks, cooking

equipment, refrigerator, freezer, dishwasher, etc. Storage space for crockery etc. Crockery,
glassware etc for serving samples.

Office area

General Separate but close to testing area, reasonable size, desk, filing cabinet,
computer, bookcase. Photocopying service needed.

Additional areas

Useful to include rest room, cloak room and toilets.

Practical alternatives

The requirements specified in the International Standard (ISO 8589) will obviously provide a
suitable area, but they are not always feasible, either from the point of view of financial
COPYRIGHT 32 R L Mason and S M Nottingham
Sensory Evaluation

resources or physical space available.

Very few industries are able to start from scratch, designing new premises solely dedicated to
sensory analysis work. I therefore would like to abbreviate the list proposed in the standard
to one which I consider includes the bare essentials.

Minimum of 2 areas:

Preparation area and office area. If possible position these at opposite ends of the room to
avoid messy paperwork!

Testing area with entrance separate from preparation area.

Preparation area requires

• Adequate storage for utensils and equipment;

• Adequate working surfaces to set out samples;
• Washing up facilities - minimum double sink with hot and cold running water;
• Refrigerator - minimum 2 door with separate freezer, preferably at least auto-defrost;
• Cooking equipment - depending on sample requirements;
• Rubbish bin - large with liner bags;
• Source of boiling water;
• Hand washing facilities.

Testing area requires

• Comfortable chairs for panellists;

• Minimum space - 4 panellists;
• Table which can be easily divided into booths if required;
• All equipment likely to be needed while a panellist is tasting, e.g. pencils, spittoons,
toothpicks, tissues/serviettes;
• Well placed, efficient lighting;
• Waiting are with noticeboard - for tasters to wait for booths to become free and to
collect rewards after tasting.

A system using collapsible booths can work quite well if it is not possible to keep an area
solely for sensory work. These may be made of painted wood, heavy duty cardboard, or
"corflute". They can be made specifically to fit any available benches or tables and folded
and stored when not in use.

The type of facility will depend on:

• Finance
• Available space
• Frequency of use
• Tests conducted

COPYRIGHT 33 R L Mason and S M Nottingham

Sensory Evaluation


This section looks at the role of statistics in sensory evaluation and introduces some terms
and concepts required to correctly apply statistical methods in evaluating sensory type data.

Why do we need statistics in sensory evaluation?

When we measure something (eg salt level in cheese) we find there is variation in what we
are measuring. This variation is called natural variation or experimental error and implies
that there is some true measurement but because of our limitations we cannot reproduce the
correct readings every time. This is a fact of life and we have limited control over this sort of

Because of this variation there is some risk in making decisions about changing formulations
or introducing new products onto the market. Using statistics we have rules to estimate and
minimise the risk and enable us to extrapolate our results from an experiment to a more
general situation.

What is an experiment?

It is any process that generates raw data.

There are many sources of error in sensory data. Some of these include.

• differences between people, (likes and dislikes)

• differences within a person from time to time, eg adaptation
• differences among samples,
• differences in interpretations of scales
• and many more.

How can we describe our data?

Lets say we have collected some data from an experiment and we have 20 scores of flavour
acceptability in a mango sample rated on a 9 point hedonic scale. If we plot a bar graph
(histogram) using the score along the horizontal axis and the count for a particular score on
the vertical axis then we have a frequency distribution. An example is shown below.

COPYRIGHT 34 R L Mason and S M Nottingham

Sensory Evaluation


3 4 5 6 7 8 9
Flavour Acceptability

Looking at the graph or distribution we ask what is the best single estimate of the panels
score and what is a good measure of their variability? The best or most likely single
estimates are called measures of central tendency. The three most commonly used are:

mean - or average (sum of all data values divided by number of observations)

median - 50th percentile or middle value
mode - most frequent value, good for categorical data

Measures of variability include the range, standard deviation and variance. The range is
simply the difference between the smallest and the largest. The standard deviation is
probably the most common and is calculated by using the formula below.

∑ ( X − M )2
( N − 1)

where M is the mean or average of X scores and N is the number of scores. This formula
calculates the deviation of each score from the mean and squares it to take into account
positive and negative values and the square root is then taken to bring it back to the original
units. The variance is simply the square of the standard deviation and is used in a number of
statistical formulas.

COPYRIGHT 35 R L Mason and S M Nottingham

Sensory Evaluation

The normal distribution

Many things we measure about a group of people will be normally distributed. This means
they will form a bell shaped curve described by an equation usually attributed to Gauss.

How does the standard deviation relate to the normal distribution?

Standard deviations describe discrete percentages of observations at certain degrees of

difference from the mean. So for a normal distribution about 66% of our data will be within
one standard deviation of the mean and about 95% will be within two standard deviations.
For our mango flavour data with a mean of 6.0 and standard deviation of 1.89 then 66% of
our data lie between 4.11 and 7.89. If the standard deviation had been 1.00 then 66% of the
values would be between 5.0 and 7.0, a smaller range indicating less variability.

In addition any score, X can be described in terms of a z-value, which describes how far the
score is from the mean in standard deviation units.

Z = X-µ/σ

Since z-scores are related to percentages under the normal curve they can predict how far a
score is from the mean and how likely or unlikely it is. So the z-score can be converted to a
probability value or p – value. This p - value is found from the area under the curve outside
the z score and is the chance with which we would see a score of that size or greater. Tables
are often used to convert z - scores to p – values.

An important concept

When we do an experiment we are using results from a sample taken from a larger
population of possible results. Since we cannot take all possible results from the population
we infer from our sample results what should happen in the rest of the population. By
making this generalisation we often express our results in terms of probability or p- values.
This is our safety margin or level of confidence about our result. It is often quoted like this -
the flavour score for naturally ripened mango was significantly higher (P<0.05) than that for
artificially ripened mango.

But what does this mean?

We are at least 95% certain that based on our experimental conditions the naturally ripened
mango will have more flavour than artificially ripened mangoes. This conclusion will be
wrong about five times out of 100. Sometimes a 1 % value or 0.01 is used for greater

How does all this help?

We need to identify some more concepts before we can be confident in using statistics.
Experiments need to be planned and carried out correctly before we can use statistics and two
important principles are replication and randomisation.

Replication is the assessment of each treatment more than once. A treatment can be the
addition of sweetener to a product or the storage temperature of a fruit. With replication we
COPYRIGHT 36 R L Mason and S M Nottingham
Sensory Evaluation

can assess the natural variability and separate this from our variability due to treatment
differences. This is like a signal to noise ratio. Is our signal greater than the background
noise (natural variation)?

Random allocation of treatments to samples or products ensures each sample has an equal
opportunity of receiving any treatment, and that this chance is unaffected by the treatments
assigned to other samples. For example if two products are tasted by 24 tasters and they all
taste product A first then this may well bias the results, as the first product tasted may tend to
be preferred regardless of which it is. Subjective allocation of treatments in a haphazard way
is not a satisfactory alternative to randomisation.

COPYRIGHT 37 R L Mason and S M Nottingham

Sensory Evaluation


There are two main types of sensory methods:

Affective :tests which involve consumer preference or acceptance

Analytical : tests which are involved with analyzing specific product attributes in terms of:

• discrimination/difference
• description


Preference infers a preference for one product over another; either overall or in relation to a
particular parameter.

Acceptance infers actual utilisation/purchase of the product.

Panel selection

Select panel on basis of end use:

• Age
• Race
• Religion
• Sex or
• Random selection for overall

Panel training

No need for training, in terms of technique or ability. However, panellists should be

instructed/briefed in terms of:

• Method
• Questionnaire
• Length of trial
• Number of samples

Panel size

1. 20 to 100 people
2. 20 = pilot consumer panel
3. 100 = consumer panel

COPYRIGHT 38 R L Mason and S M Nottingham

Sensory Evaluation



(Reference: AS 2542.2.1; 1982.)

Application: to establish whether there is a preference between two samples.

Principle: a pair of samples (one may be a control) is presented to each assessor. The
assessors are asked to choose the sample they prefer.

This test is a ‘forced choice’ ie: the assessors must select one sample as being more
preferable. Responses indicating no preference are not permitted.

Statistically based on null hypothesis that there is no preference between the samples.

ie:PA = PB = 50%= 0 5

Bilateral Test - no expectation of preferences

Specimen Answer form for bilateral paired preference test

Which sample do you prefer?
Please examine code 349 first.
Please tick the appropriate box.

Code 922 349

Place tick



• no preference
• A preferred to B
• B preferred to A

Question — which of the two samples do you prefer? Count the number of replies citing one
of the two samples the more frequently.
Conclude that this sample is significantly preferred to the other if the number obtained is
greater than or equal to that shown in Table 4.


Two drinks ‘A’ and ‘B’, are offered to a panel of 30 assessors. The two samples are
presented under random number eg: ‘789’ and ‘379’. The test supervisor accepts a 5% level
COPYRIGHT 39 R L Mason and S M Nottingham
Sensory Evaluation

of significance (ie: P < 0.05%). It is not known which of the two samples contains more

Question - Which sample do you prefer?

Replies: 22 prefer ‘A’
8 prefer ‘B’

From Table 4 it can be concluded that Drink ‘A’ is preferred to Drink ‘B’.

Unilateral Test - expect one sample to be preferred.

Specimen Answer form for unilateral paired preference test

Do you prefer sample 186 to sample 592?
Please examine code 592 first.
Please tick the appropriate box.




• no preference
• the declared sample is preferred

Question — Do you prefer sample ‘A’ to sample ‘B’? Conclude sample A is preferred if
number of positive replies is greater or equal to the number shown in Table 3.


Two drinks, ‘A’ and ‘B’, are offered to a panel of 30 assessors. The two samples are
presented under a random number eg: ‘789’ and ‘379’. The test supervisor accepts a 1%
level of significance (ie: P < 0.01%). It is known that drink ‘A’ contains more sugar than
drink ‘B’.

Question - Do you prefer sample ‘A’ to sample ‘B’?

Replies: 23 Yes and 7 No.

From Table 3 it can be concluded that there is preference for drink ‘A’ over drink ‘B’.

COPYRIGHT 40 R L Mason and S M Nottingham

Sensory Evaluation

N.B. If uncertain always use the bilateral test.


• Simple test to conduct

• Suitable for children and consumer panels
• Easy to analyse (for > 100 assessors use t test or CHI squared)


• Only suitable for 2 products (note – multiple Comparisons can be used but other
preferences tests are more commonly used. See ASTM manual on sensory testing
method, STP 434; 1968)

• No magnitude of preference is given ie they both may be disliked but one can still be


• Product Development
• Product Matching
• Process Change


(Australian Standard 2542.2.6)


Judges are asked to rank two or more samples in order or preference ie: most preferred
sample is ranked first.

Ranking is a forced choice procedure ie no ties are allowed.

Specimen Answer form for ranking for preference.

Please taste the samples in the order presented, moving from left to right and rank
them in order of preference. You may retaste the samples to check the ranking.
Give the sample that you most prefer the a rank of 1 and the sample you prefer
next a rank of 2 etc.

You must give each sample a different rank. Equal ranks are not allowed.


COPYRIGHT 41 R L Mason and S M Nottingham

Sensory Evaluation

Statistical analysis

Kramer’s tables, which have been used in the past to analyses differences between rank sums,
should not be used due to questions of accuracy and statistical validity.

When there is no expectation of a specific rank order being made (eg when ranking
preference of new product prototypes) the Friedman Test should be used (see statistical
method s section for details).


Twelve households were presented with four samples of meat seasoning to be used in
cooking. They were asked to use the samples as directed and to rank them in order of
preference. The results are shown below:

COPYRIGHT 42 R L Mason and S M Nottingham

Sensory Evaluation

Rankings for the preference of four meat seasonings

HOUSHOLD Seasoning
1 1 3 2 4
2 2 1 3 4
3 1 4 2 3
4 1 4 2 3
5 2 3 1 4
6 3 4 2 1
7 3 4 2 1
8 3 4 1 2
9 1 2 3 4
10 1 2 3 4
11 1 2 3 4
12 1 3 2 4
Rank sums 20 36 26 38

The F value is calculated as follows:

F= (20 2 + 36 2 + 26 2 + 38 2 ) − 3 × 12(4 + 1)
12 × 4(4 + 1)


the calculated value is compared to the critical f value in table 7 (7.81 for 3 df). since 10.8 is
greater than 7.81, the experimenter can conclude that there is a significant (p<0.05) difference
between the rank sums.

Two samples will be significantly different if the absolute value of the difference between the
rank sums is greater than or equal to the following critical value:

12 × 4(4 + 1)
1.960 = 12.396
Sample A B C D
Rank Sum 20a 36b 26ab 38b

Rank sums that do not have a common superscript are significantly different (P<0.05)

COPYRIGHT 43 R L Mason and S M Nottingham

Sensory Evaluation


(Australian Standard 2542.2.30


Assessors are asked to evaluate one or more samples and indicate the degree of liking for the
product or some characteristic of the product.

When performing preference testing it is important to include as many panellists as possible.

Personal preferences in foods are being measured which are purely subjective, so the variance
in the data is large. This makes it more difficult to obtain statistically significant results. The
larger the panel, the more chance there is of obtaining a significant result. Only untrained
panellists are used and should be selected at random or from a targeted group related to the

Pilot consumer panel = 20-25

Consumer panel = 100

Types of response scale

Category scale/structured scale

The response scale is divided into categories or boxes.

The response scale is usually divided into an arbitrary number of categories - usually between
7 and 13
Category scales must be bipolar.
Verbal descriptors or facial expressions may be used to identify the levels of acceptance.

Hedonic category rating


Like extremely
Like very much
Like moderately
Like slightly
Neither like nor dislike
Dislike slightly
Dislike moderately
Dislike very much
Dislike extremely

COPYRIGHT 44 R L Mason and S M Nottingham

Sensory Evaluation

Facial hedonic scale

7 point facial hedonic scale


Graphic rating scale

• The response is recorded by marking a position on a line

• Also called visual-analogue scale, line mark scale or unstructured scale
• -Physical lengths 100-150 mm.
• This scale may also use facial expressions for measurement.
• Numbers and/or descriptors are usually attached to a rating scale.

Recording and interpretation of results

Ratings must be converted to numerical scores for analysis and interpretation. For category
scales, successive integers are assigned to successive categories and these are used in
analysis, e.g. with a 9-point scale, the integers 1-9 would be used. For graphic scales, the
distance, e.g. in mm, between the response mark and one end of the scale serves as the
response score.

The arithmetic mean and standard deviation, when obtained for each sample, serve as
measures of central tendency and variability, respectively. For statistical analysis, the
analysis of variance technique is appropriate (or a t-test in the case of one or two samples).

Correlation or regression analysis may be used for subjective/objective correlations.


• Test is relatively simple and easily understood;

• Indicates the degree of preference;
• Can be used to infer acceptance;
• Suitable for different age groups; and
• Can measure > one parameter at a time.


• Statistical analysis is required; and

• Results may be biased by type of assessors used.

COPYRIGHT 45 R L Mason and S M Nottingham

Sensory Evaluation


• storage trials
• product development
• consumer testing
• quality control
• subjective/objective correlations
• research


Three samples of frozen chicken casserole were presented to a 24 member panel who
assessed the appearance, flavour, texture and general acceptability of the products using a 13

The following results were obtained:

Appearance 10.8 a 8.5 d 10.3 b
Flavour 9.9a 9.4 b 9.2 b
Texture 10.4 a 9.6 b 9.1 b
General Acceptability 10.3 a 9.2 b 9.4 b

Scores within each row that do not have a suffix in common are significantly different.



The personal response by current or potential customers of a product, a product concept, or

specific characteristics of a product is collectively grouped under what we call consumer
testing. However, it is important to define the terms acceptance and preference often
associated with consumer testing. Acceptance refers to the degree of liking or disliking for a
particular product or the ability of the product to meet expectations of consumers while
preference refers to a choice made by panellists among several products on the basis of
liking or disliking. Unfortunately ‘preference’ is widely used as a generic term to describe
both acceptance and preference judgements. It is important to note for example in paired
preference testing that although one product may be preferred to another, neither product may
be liked to any degree. The term ‘hedonic’ is an adjective associated with degrees of
pleasure or displeasure and is applied to both acceptance and preference testing.

COPYRIGHT 46 R L Mason and S M Nottingham

Sensory Evaluation

Applications of Consumer Testing

The reasons for conducting consumer tests usually fall into one of the following categories:

• Product maintenance
• Product improvement/optimization
• Development of new products
• Assessment of market potential

Product maintenance

Research and development projects may involve cost reduction, substitution of ingredients,
process and formulation changes and packaging modifications without affecting the product
characteristics and overall acceptance. Usually difference tests would be used to determine
whether a difference was perceived or not but it is necessary to take the product out to the
consumer to determine if the reformulated product will achieve at least parity with the current

Product maintenance is also a key issue with quality control/quality assurance and shelf-life/
storage projects. Feedback on consumer response gives important information on those
sensory characteristics that are most important to consumer choice and which should
therefore be rigorously controlled. A combination of in-house profile testing on the
magnitude and type of change over time, condition, production site, raw material sources etc
can be used in conjunction with consumer testing to determine how large a difference is
sufficient to change the acceptance rating.

Product improvement/optimisation

The intense competition among consumer products drives companies to constantly improve
and optimise products so that they can deliver what the consumer is really looking for and
therefore increase market share.

In product improvement, prototypes are made, tested by a trained panel to verify that the
desired attribute differences are perceptible, and then tested with consumers to determine the
degree of perceived product improvement and its effect on overall acceptance or preference

For product optimisation, ingredients or process variables are manipulated and a trained panel
identifies the key sensory attributes affected and consumer tests are conducted to determine if
consumers perceive the change in attributes and if such modifications improve the overall

Development of new products

During the new product development from concept to a range of trial samples to a modified
sample range and finally a choice to launch, consumer testing should be used throughout in
conjunction with trained panel assessment.

COPYRIGHT 47 R L Mason and S M Nottingham

Sensory Evaluation

Assessment of market potential

In addition to the use of sensory evaluation to gather information about key attributes of a
product, typical marketing questions such as intent to purchase, purchase price, current
purchase habits, consumer food habits, effects of packaging, advertising and convenience are
critical for the acceptance of branded products. It is often convenient for these marketing
type questions to be included in a questionnaire presented to consumers when assessing the
sensory characteristics of the product.

Conducting Consumer Tests

There are a number of factors to consider when conducting consumer tests and these are:

• Test design
• Test subjects
• Test location
• Test questionnaire

Test Design

There are two main types of design, one is qualitative measuring subjective responses while
the other is quantitative determining the responses of a large group to a set of questions
regarding preference, liking, sensory properties etc.

Qualitative Tests include focus groups, focus panels and one-on-one interviews. Each of
these has their use in a particular situation depending on what is required and how sensitive
the topic is. Essentially small groups are used to uncover as much specific information from
as many participants as possible. It is frequently recorded either by video and or audio and a
summary is made.

Quantitative Tests

Essentially all the good practice principles used in sensory evaluation as described in the
difference and descriptive testing should be followed here such as 3 digit random codes for
product and presentation in a balanced order. Some typical designs used include:

• Monadic test where only one product is assessed which makes it fast and the least
expensive but is relatively insensitive and requires large numbers of consumers (at
least 200).
• Sequential monadic where one product is assessed, removed and then replaced by a
second product in a balanced design giving it greater sensitivity.
• Paired preference testing where two products are assessed simultaneously and a
direct comparison is made making it quite sensitive.
• Acceptability testing. Usually the nine-point hedonic scale is used to determine
consumers liking of a product and if required the relative ratings for liking can be
used as a measure of preference.

COPYRIGHT 48 R L Mason and S M Nottingham

Sensory Evaluation

Example of nine-point hedonic category rating


Like extremely
Like very much
Like moderately
Like slightly
Neither like nor dislike
Dislike slightly
Dislike moderately
Dislike very much
Dislike extremely

Example of seven point facial hedonic scale often used for children





Graphic rating scale - the response is recorded by marking a position on a line (also called
visual-analogue scale, line mark scale or unstructured scale) - physical lengths 100-150 mm.
This scale may also use facial expressions for measurement.

• Attribute testing can be used to gain information on the reasons underlying overall
preferences and usually category or line scales are used. These can be hedonic type
attributes or sensory attributes in the form of just right scales as shown below.

not sweet enough just right too sweet

Test Subjects

If information on the acceptance of the product by consumers is required, then it is they who
should do the tasting. However, this is not always practical in preliminary testing of products, so
a compromise can be made by using large numbers staff who assess fairly infrequently.
However, it should always be remembered that this is a compromise, and results are best

COPYRIGHT 49 R L Mason and S M Nottingham

Sensory Evaluation

interpreted only in relative, not absolute, terms. Staff cannot be considered to be a

representative sample of the target population.

When performing consumer testing it is important to include as many panellists as possible.

Personal preferences in foods are being measured which are purely subjective, so the variance
in the data is large. This makes it more difficult to obtain statistically significant results. The
larger the panel, the more chance there is of obtaining a significant result.


The number of consumers to be tested depends on the purpose of the test, the test design and
the precision with which the target population can be identified. In general we require 60 to
120 for most consumer testing. Recruitment and selection of consumers rely on several
criteria or demographics such as:

• Product usage. It is important to determine if you are looking for low, medium or
high users of the product. For speciality products or niche markets, the cost of
consumer testing increases as more people must be contacted before the required
number are found.
• Gender. It is not always necessary to get equal numbers as purchasing or usage
habits vary between products. Researchers should use current market information.
• Age. If a product has broad age appeal then consumers should be selected by age in
proportion to their representation in the user population.
• Nationality. Products targeted towards a specific part of the community or for export
ideally should be tested in that environment. However, it is possible to use foreign
nationals resident here but it depends on how long they have been residing in their
adopted country as they can develop the likes and dislikes characteristic of the
adopted country.
• Social class. This can be based on income or occupation although sometimes it is
difficult to get consumers to reveal such information.
• Others including race, religion, education level, employment, geographic
location, etc. If any of these are important in defining the target audience then the
researcher should consider them.

Source of Consumers

As mentioned it is important to sample properly from the consuming population but because
of cost restraints employees and local residents may be used for things such as product
maintenance. However, for new products or product optimisation or improvement the correct
audience should be selected. These can come from a database of consumers willing to assess
products, telephone survey, leaflet drop, shopping centres, embassies, colleges or door to

Test Location

It is possible to conduct consumer testing in a number of locations depending on the

resources and the results can vary greatly. Locations include:

• Company laboratory facilities, which give good control of the environment and
rapid feedback of results but the sensory booths, are clinical and atypical of a real
COPYRIGHT 50 R L Mason and S M Nottingham
Sensory Evaluation

domestic environment.
• Central location such as school or church halls or shopping centres are convenient as
large numbers can be tested at one time and on a number of products. However the
conditions are artificial compared to normal use at home and the number of questions
that can be asked may be limited.
• Home use tests represent the ultimate in consumer testing as the product is tested
under its normal conditions of use. In addition to the product itself, a check on the
packaging can also be determined. Generally more information can be gathered as the
consumer gets more time and can perform repeated assessments. However it is time
consuming and uses a smaller number of people and the possibility of nonresponse is
great unless consumers are continually reminded.

Test Questionnaire

It is very important that the test questionnaire format is simple, unambiguous, easy to read
and understand. You need to consider the objective of the test and any constraints such as
time, funding etc. In essence you need to be:
• Brief
• Use simple plain English (provide translation for studies involving foreigners)
• Be specific
• Multiple choice questions should be mutually exclusive
• Avoid ambiguity
• Watch the effects of wording
• Don’t ask what they don’t know
• Try and pre-test the questionnaire

For example

How satisfied or dissatisfied were you with the product?

Very satisfied
Slightly satisfied
Neither satisfied nor dissatisfied
Slightly dissatisfied
Very dissatisfied

The product looks like how it is shown on the package.

Agree strongly
Neither agree nor disagree
Disagree strongly

What did you like about the product?

This open-ended question allows for the consumer to add something you may have forgotten
but it is sometimes hard to read the answer (handwriting) and some people do not bother with

COPYRIGHT 51 R L Mason and S M Nottingham

Sensory Evaluation

The question order should go from the more general to the more specific and ask overall
acceptability first before biasing the consumer with more specific issues. Ask the more
sensitive demographic questions last.

Data Analysis

All quantitative data should be subjected to some form of statistical analysis from simple
summary statistics and graphical representation to t-tests and analysis of variance with
pairwise comparisons. Further advanced multivariate methods such as principal components
analysis and cluster analysis along with regression methods to relate consumer data to other
data such as linear regression, partial least squares and preference mapping can also be used.

COPYRIGHT 52 R L Mason and S M Nottingham

Sensory Evaluation


In general, analytical panels are used as “measuring instruments and therefore need to be:

• Valid (able to measure appropriate parameters)

• Reproducible

Panellists can be trained or untrained depending on the degree of difference expected.

Consumers will not detect the small differences that a trained panel would. An untrained
panel would require 20-100 panellists while a trained panel would require 5-20.


Difference tests may be sub-divided into 2 classes:

• Simple difference tests are those that have no direction or characteristic associated
with the difference between the products. Examples of simple difference tests are:

ƒ Triangle test
ƒ Duo Trio test
ƒ Two-out-of-five test
ƒ A not A
ƒ Difference from control

• Directional difference tests are those that have a direction or characteristic associated
with the difference between the products. Examples of directional difference tests

ƒ Paired comparison test

ƒ Ranking
ƒ Rating



(Australian Standard 2542.2.2 - 1983)

Scope and Application

Used to determine whether a perceptible difference exists between two samples. The
difference can involve one or several sensory attributes, but no direction or magnitude of the
difference is measured.

The triangle test is an effective method to determine whether a change in ingredient,

processing, packaging or storage has resulted in product differences. These situations may
arise in product and process development, product matching, in quality control or as a
preliminary test prior to quantitative descriptive testing. A triangle test can also be used for
COPYRIGHT 53 R L Mason and S M Nottingham
Sensory Evaluation
the selection and monitoring of panellists.

With products that produce sensory fatigue, carryover effect or adaptation effects, the triangle
test has limited application.


Three samples, two of which are identical, are presented simultaneously to each panellist for
testing in a predetermined order. The panellist is told that two samples are identical and one
is different (odd). The panellist is required to identify the different sample. The triangle test
is a forced choice test.

Preparation and Procedure

The samples should be representative of the product and all prepared in exactly the same
way. Select four 3-digit random number codes, two for each product. Prepare scoresheets to
provide equal numbers of the following orders:


Make up sets of 3 samples to match the score sheets so that half contain 2 samples of product
A and half contain 2 samples of product B (Total number of sets should be a multiple of 6.).
Make up sets in multiples of the six arrangements as required for the number of panellists.

If total number of panellists or quantity of products available is insufficient to provide equal

numbers of the 6 orders, you still need to make sure there is a balance between sets with 2
‘A’s and 2 ‘B’s. The triangle tests should be presented at random to the panellists. Instruct
each panellist to examine in the specified order (e.g. left to right) and remind them that they
must make a decision.

Count the number of correct responses (those that select the odd sample) and compare the
result with those presented in Table 2.


Specimen answer form for the triangle test

Product Date Time Assessor

One of the three samples presented is different from the other two.

Please examine in the order requested and place a circle around the code of the sample
which is different.

293 594 862

Analysis of results You must make a choice

COPYRIGHT 54 R L Mason and S M Nottingham

Sensory Evaluation

The total number of correct responses is counted as well as the total number of responses and
compared to the statistical tables (Table 2). This is based on the probability that if there is no
real difference the odd sample will be chosen a third of the time.


A company wishes to put a new dessert topping on the market. The product development
section has two different thickening agents available to them, one which is considerably
cheaper. They wish to know if there is any difference in the products made using the 2
different thickeners.

Two batches (A, B) are prepared using the two different thickening agents and samples are
presented to 17 assessors. As each assessor will only make one assessment, it will be
necessary to prepare 27 samples of A and 27 samples of B, and arrange them to provide three
of each of the six possible arrangements as indicated above. One set is discarded and the
remaining 17 sets are randomly distributed between the assessors.

The number of correct responses is 10, ie the number of panelists who correctly selected the
odd sample from the 3 samples presented.

The test organizer will accept a risk of error of 5% (P<0.05), that the test will reveal a
difference when there is none.

Table 2 indicates that for 17 assessors at P<0.05, 10 correct responses are required for

It can be concluded that the product from the two thickening agents are significantly different

What should the test organizer do next???

(Australian Standard 2542.2.4 - 1988)

Scope and Application

Used to determine whether a difference exists between two samples. The difference can
involve one or several sensory attributes, but no direction or magnitude of the difference is
measured. A duo-trio test can be used when one of the products is an existing standard or

A duo-trio test can be applied to determine whether changes in ingredients, processing,

packaging or storage have resulted in differences between products. The duo-trio test finds
application in the selection of panellists, product and process development, product matching,
quality control and as a preliminary test prior to analytical descriptive testing.

Statistically the duo-trio test is less powerful than the triangle test because the chance of
guessing a correct result is one in two. The Duo-Trio test is therefore only used when it is
COPYRIGHT 55 R L Mason and S M Nottingham
Sensory Evaluation

required to form a judgement. This is the case when tasting a product with a lingering after-
taste such as bitterness, spicy or chilli.


Three samples, two of which are identical, are presented to each panellist. One sample is
identified as the reference sample and panellists are instructed to assess the reference sample
first and then identify which of the two samples is the same as the reference. It is a forced
choice test.

Preparation and Procedure

The samples prepared should be representative of the product and prepared in exactly the
same way. If possible, samples are presented simultaneously or if required, sequentially.
There are two forms of this test: balanced reference mode and constant reference mode.

Balanced reference mode

This is used when both the samples are unfamiliar and so both the samples are used as the
reference sample.

Select two 3-digit random codes, one for each product. Prepare scoresheets to provide equal
numbers of the following orders:


Make up sets of 3 samples (reference plus two samples) to match the scoresheets so that half
contain 2 samples of product ‘A’ and half contain 2 samples of product `B'. (Total number of
sets should be a multiple of 4).
If total number of panellists or quantity of products available is insufficient to provide equal
numbers of the 4 orders, then you will still need to check that there is a balance between sets
with 2 ‘A’s and 2 ‘B’s.

The sample sets are allocated at random to the panellists. Instruct each panellist to assess the
reference sample first, followed by the two other samples in order (e.g. left to right). Remind
them that they must make a decision.

Constant reference mode

The constant reference duo-trio test is useful when you have trained panellists. In this test,
one of the samples is a familiar product or designated standard. It is therefore the only one
used as a reference sample. The number of possible presentation orders is thus restricted to:


Select two 3-digit random codes, one for each product and prepare the scoresheets so that
equal numbers of the two orders are presented. (Total sets should be a multiple of 2).
COPYRIGHT 56 R L Mason and S M Nottingham
Sensory Evaluation

Randomly allocate sample sets to panellists and instruct each panellist to assess the reference
sample first, followed by the two other samples in order (e.g. left to right). Remind them that
they must make a decision.

Analysis of results

Count the number of correct responses as well as the total number of responses and use the
statistical Tables 3.



Name: Date: Time: Product: .

You are provided with three samples. The left-hand one is a reference; one of the other two is
the same as the reference.

Taste the samples in the order left to right and circle the number of the sample which is the
same as the reference.

Reference Sample code:. . . . . . . . . Sample code: . . . . . . . . .




A duo-trio test was used to determine if methional could be detected when added to cheddar
cheese in amounts of 0.125 ppm and 0.250 ppm. Each tray had a control sample marked R
and two coded samples, one with methional added and one with no methional. The duo-trio
test was used in preference to the triangle test because less tasting is required to form a
judgment. This fact is important when tasting a substance with a lingering aftertaste, such as

The test was performed on two successive days using eight judges. Each day the judges were
presented with two trays. One tray contained a sample with 0.125 ppm methional and two
control samples and the other contained a sample with 0.250 ppm methional and two control
samples. A total of 16 judgments were made at each level. The results are shown in the
following table.

COPYRIGHT 57 R L Mason and S M Nottingham

Sensory Evaluation

Duo trio test on cheddar cheese containing methional. 0.125 and 0.250 ppm.

Day1 Day 2

JUDGES 0.125 0.250 0.125 0 .250

1 X R R R
2 R R R R
3 X R X R
4 R X X R
5 R R R R
6 X R X X
7 R R R R
8 R R R R
TOTAL 5 7 5 7

X = wrong
R = right
0.125 ppm = 10 out of 16 correct judgments
0.250 ppm = 14 out of 16 correct judgments

Consult Table 3 for 16 judges in a two sample test. This chart shows that 12 correct
judgments are significant at the 5% level.

The conclusion is that methional added to cheddar cheese can be detected at the 0.250 ppm
level but not at the 0.125 ppm level. What would you do next??


ƒ used where a reference standard is available

ƒ less tasting required than triangle test
ƒ can be used with trained or untrained assessors


ƒ No indication of character or degree of any difference

ƒ Statistically less powerful than triangle test


ƒ Quality control — use normal product as control

ƒ Product matching
ƒ Product or process improvement
ƒ Panel selection or training

COPYRIGHT 58 R L Mason and S M Nottingham

Sensory Evaluation


Used to determine whether there is a sensory difference between two samples and to select
and monitor panellists. It is statistically very efficient as the probability of guessing correctly
the different two samples from the five samples presented is low. It can be useful when only
a small number of panellists are available. However, sensory fatigue and memory effects
may affect the test.

As with the triangle and duo-trio tests, assign 3-digit random codes to the samples and then
make up the scoresheets, taking care to prepare the samples in an identical fashion. There
will be 20 possible combinations.

Panellists are instructed to assess each product from left to right and select the two samples
that are different from the other three. Statistical tables exist to determine the significance of
the result.

“A” – “NOT A” TEST

(Australian Standard 2542.2.5 - 1991)

Used to determine whether test samples in a series are the same as or different from the
reference sample. It is an especially useful test where triangle and duo-trio tests cannot be
used. This may be the case where comparisons are required between products that have a
strong or lingering flavour/aftertaste when you will need to control the time between sample
presentation or if there are differences in appearance. It is also useful to determine assessor
sensitivity to a stimulus.

Initially, panellists require familiarisation with the reference or “A” sample. Panellists are
then presented with a series of samples, some of which are the reference sample “A” and
some “not-A”. Generally, the panellist does not have access to the reference “A” while
evaluating the test samples. The panellist must determine whether the sample is the same
(“A”) or different (“not-A”) so it is a forced-choice test. Only one type of “not-A” sample
exists per test series. Panellists may test one, two or up to 10 samples in series (depending on
fatigue factors). The samples are presented randomly with 3-digit codes and one at a time (an
assessment is made and recorded before proceeding to the next sample). All samples are
prepared in an identical way and are representative of the product. The analysis of the data is
quite complex.

Also called the degree of difference (DOD) test.

There is no Australian Standard this test however further information can be obtained in
Meilgaard, Civille and Carr and Aust et al. 1985.

Scope and Application

This test is used to determine whether or not a perceptible overall difference exists between
one or more samples and a control sample and also to give an indication as to the size of any
difference perceived. In quality control situations, trained panellists may also be able to rate
the degree of difference for individual attributes.

COPYRIGHT 59 R L Mason and S M Nottingham

Sensory Evaluation
A difference from control test is a useful test to use when other difference tests, such as
triangle or duo-trio are not suitable because of the normal heterogeneity of the products to be
tested. For example with products such as meats, baked goods and horticultural products it
can be difficult to obtain a homogeneous sample which is necessary for a triangle or duo trio

When used in conjunction with consumer acceptability testing and descriptive testing using a
trained panel, the DFC test is useful for quality control and shelf life testing. In these cases
the relative size of the difference is important for deciding whether the product is an accept or
reject. It can be used to check production samples for the degree of difference from a
recognised control or standard product. In this situation the panellists must be familiar with
the range of differences expected and will require some training with reference samples and
the use of the scale. The test can also be useful in product development situations to
determine which sample is closest to a target product.


Each panellist is presented with an identified control sample plus one or more test samples.
The panellists are asked to rate the size of difference between each test sample and the
identified control sample. Panellists are informed that some of the test samples may be the
same as the control sample. The mean difference from control for each test sample is
compared with the mean difference from control obtained from the blind presentation of the
control sample.

The blind control sample is included as a measure of the placebo effect as it is very rare that
the blind control will actually be rated as absolutely identical to the identified control.


Generally 20-50 people are required. Panellists may be trained or untrained but not a mixture
of the two. For some applications such as in a quality control, the panellists would require
some training. All panellists should be familiar with the test format, how to use the scale and
also be aware that some of the samples will be blind controls.

Preparation and Procedure

All samples should be representative of the product and all prepared in exactly the same way.
Label an identified control sample for each panellist. Label additional blind control samples
as well as the test sample(s) with 3 digit blinding codes. Where possible the control sample
and samples for assessment should be presented simultaneously. Each panellist evaluates the
identified control sample first. The panellists then rate the degree of difference for each test
sample of which some samples will be the blind control. The order of presentation of the test
and blind control samples should be balanced.

For example, half the panellists assess the samples in the order:

1. Identified control vs blind control 2. Identified control vs test sample

While the other half assess the samples as:

COPYRIGHT 60 R L Mason and S M Nottingham

Sensory Evaluation
1. Identified control vs test sample 2. Identified control vs blind control

Examples of scales that may be used for the difference from control test:

Verbal Category Numerical category Scale

No difference 0 = No difference
Very slight difference 1
Slight/moderate difference 2
Moderate difference 3
Moderate/large difference 4
Large difference 5
Very large difference 6
9 = Very large difference

COPYRIGHT 61 R L Mason and S M Nottingham

Sensory Evaluation

Line scale

no difference very large difference

Analysis of Results

Calculate the mean difference from the identified control for each of the test samples and the
blind control samples. If several samples have been evaluated, use a randomised block
analysis of variance using the panellists as blocks. If only one test sample has been evaluated
use a paired t-test to analyse the results.

A company suspects a flavouring ingredient may have been left out of a batch of chunky
vegetable soup. They want to know if this batch of soup is perceived to be different or not
from a control batch of soup. Due to the natural degree of batch to batch variability with the
product, a triangle test or other forced choice difference would be unsuitable due to the risk
of yielding false positives or false negatives.


Name: Date: Time:


Assess the sample marked “control” first.

Assess sample 386 and score the overall sensory difference between the two samples using the
scale below.

not different very different


COPYRIGHT 64 R L Mason and S M Nottingham

Sensory Evaluation



(Australian Standard 2542.2.1 - 1982)

Scope and Application

Used to determine how a specific sensory property differs between two samples. It can be
applied to determine a directional difference (e.g. which sample is sweeter).

A paired comparison test has numerous applications in product or process development, eg

substitution of a new low-calorie sweetener, in quality assurance as well as in storage tests
and in product matching. A paired comparison test can also be used to determine if a more
advanced sensory test should be applied.

The paired comparison test can be used for multiple comparisons, but this results in a large
number of pairs to assess which uses a lot of sample and can cause sensory fatigue. In this
situation it is better to use a rating test.

Before the sensory testing commences, it is necessary to decide whether the results will be
treated as a unilateral or bilateral test. The most common paired comparison tests are two-
sided (bilateral) where there is no prior expectation of the result. Conclusions that can be
drawn are that there is no evidence of a difference or that one sample has a greater intensity
of the chosen attribute or is preferred.

One sided tests (unilateral) also exist when there is prior expectation of the direction of
difference. Conclusions to be drawn include that there is no evidence of a difference or that
the previously declared sample is greater in the attribute intensity or is preferred. The
wording used on bilateral and unilateral score sheets is different.

Test principle

Two coded samples are presented. The panellists complete the scoresheet questions that have
been previously determined by the test objective.


The test is fairly simple requiring minimal training but the panellists must understand the
attribute that is being tested. However, trained panellists may be selected if appropriate.
Twenty is a reasonable number when the panellists have been screened. Statistically,
numbers can be reduced to 7 for a trained panel, but when using completely untrained tasters
such as consumers, then much larger numbers (100+) are needed.

Preparation and Procedure

Two 3-digit randomly coded samples, one of each product, are presented. The sample
presented is representative of the product and all samples are prepared identically. Equal
numbers of AB and BA are randomly allocated to the panellists. Panellists are instructed to

COPYRIGHT 65 R L Mason and S M Nottingham

Sensory Evaluation

assess the samples in a specific order (left to right) and identify which has the higher level of
a particular attribute or is preferred. The test is a forced choice test and ‘no difference’
responses are not allowed.

Analysis of results

Use standard statistical tables for unilateral tests (Table 3) and bilateral tests (Table 4). Count
the number of replies identifying a particular sample most frequently. Compare this value
with the number shown in the statistical table for the number of panellists used.




In front of you are two coded samples of orange juice.

Please assess them in the order shown below from left to right and indicate which sample
is sweetest by circling the appropriate code.
Please cleanse your palate between samples.

Sample code 016 Sample code 983





In front of you are two coded samples of orange juice.

Please assess them in the order shown below from left to right and indicate if sample 016
is sweeter than sample 983. Circle the response below.

Please cleanse your palate between samples.




COPYRIGHT 66 R L Mason and S M Nottingham

Sensory Evaluation


Bilateral test

Two drinks ‘A’ and ‘B’, are offered to a panel of 30 assessors. The two samples are
presented under a random number, eg: ‘789’ and ‘379’. The test supervisor accepts a 5%
level of significance (ie: P< 0.05%). He does not know which of the two samples contains
more sugar.

Question: Which sample is sweeter?

Replies 18 opt for sample ‘A’
12 opt for sample ‘B’
From Table 4 it can be concluded that there is no significant difference in the sweetness of
the two drinks.

Unilateral test

Two drinks, ‘A’ and ‘B’, are offered to a panel of 30 assessors. The two samples are
presented under a random number eg: ‘789’ and ‘379’. The test supervisor accepts a 1% level
of significance (ie: P<0.01%). He knows that drink ‘A’ contains more sugar than drink ‘B’.

Question: Is sample ‘A’ sweeter than sample ‘B’?

Replies 22 yes and 8 No.
From Table 3, it can be concluded that drink ‘A’ is significantly sweeter than drink’B’.


See paired preference.


ƒ Product Development
ƒ Quality Control
ƒ Shelf Life Measurement


(Australian Standard 2542.2.6)

Scope and Application

The ranking test can be considered an extension of the paired comparison test. It is used to
place a series of three or more samples in a rank order to determine whether differences exist
between samples. Samples are ranked for a specified criterion, e.g. an attribute (bitterness,
crunchiness, hardness) or a preference. The criterion needs to be understood by the
panellists. The data obtained is ordinal and therefore provides directional differences between
samples but does not provide information about the degree of difference.

The ranking test is a simple way to compare samples and is useful for reducing the number of

COPYRIGHT 67 R L Mason and S M Nottingham

Sensory Evaluation

test samples prior to performing another test and to evaluate panellist ability. In product
development, a ranking test can be used as a quick method of indicating the effects of
different raw materials, processing, or packaging and storage treatments.

Test principle

Samples are presented to the panellists simultaneously and are placed by the panellists into a
rank order relative to one another according to the specified criterion.


Minimum of 8 but a larger number of panellists is better.

Preparation and Presentation

Three or more 3-digit random coded samples are presented to panellists simultaneously for
assessment in a balanced or random (if more than 4 samples) order. All the samples are
prepared and presented identically. The maximum number of samples will depend on the
type of product. They are instructed to arrange the samples in rank order according to the
level of the specified criterion, and are instructed whether to assign rank 1 for the lowest or
highest level. It is a forced choice test and tied rankings are not permitted. A separate
scoresheet is used and completed separately if the rank order is required for more than one

As a panellist, it is often easier to perform this test by arranging the samples in a provisional
order first and then to re-evaluate them before assigning final ranks.

Analysis of results

Rank totals are calculated for each sample and used to generate test statistics which are
compared to statistical tables. As samples are evaluated only in relation to each other, results
from one test cannot be compared to those from another unless they both tested the same

A cordial manufacturer has been provided with two new samples of lemon flavour that are
cheaper than the existing flavour. The manufacturer wants to know if cordials are made at
the same flavour intensity, would it be cheaper to use either of the two new flavours.
Samples are prepared at the same concentration but in order to test this from a sensory
perspective the 3 samples are presented to 30 assessors who are asked to rank them in order
of flavour intensity. The results are presented below:

COPYRIGHT 68 R L Mason and S M Nottingham

Sensory Evaluation

Assessor Cordial Samples

1 1 2 3
2 1 3 2
3 2 1 3
4 1 3 2
5 3 2 1
6 1 2 3
7 1 3 2
8 3 2 1
9 3 2 1
10 3 2 1
11 2 1 3
12 3 2 1
13 1 2 3
14 3 1 2
15 3 2 1
16 1 2 3
17 2 3 1
18 1 3 2
19 1 2 3
20 2 1 3
21 3 1 2
22 2 1 3
23 1 3 2
24 3 2 1
25 1 3 2
26 1 3 2
27 3 2 1
28 3 2 1
29 1 2 3
30 2 1 3
Rank Sums 58 61 61

F= (58 2 + 612 + 612 ) − 3 × 30(3 + 1)
30 × 3(3 + 1)
= 360.2-360
= 0.2

From Table 7 the critical value for F with 2 degrees of freedom (df = number of samples –1)
is 5.99.

The technician must retain the null hypothesis that there is no difference between the flavour
strength of the three products.


COPYRIGHT 69 R L Mason and S M Nottingham

Sensory Evaluation

(Australian Standard 2542.2.3 1988)

(AS2542.1.3 1995: 7.6 Selection of assessors for rating methods and 7.7 Selection of
assessors for descriptive analysis)

The rating test can be used to measure the perceived intensity of sensory characteristics eg
degree of strawberry flavour in a strawberry milkshake.

For this type of test the basic principles of sensory evaluation should be followed eg coded
samples, controlled test environment, number of samples tested. Panellists should be selected
based on their ability to give consistent ratings to the same sample and to discriminate
between samples checked by statistical analysis. The number of panellists used depends on
the degree of training but generally a minimum of eight highly trained, more if less trained.
Selection and training of panellists will be discussed later in a separate section.

The response scale used for rating may be in the form of a category scale or a line scale. A
category scale is a series of 7 – 15 boxes labelled to identify levels of intensity. With a line
scale, panellists respond by marking a position on a horizontal line labelled with “anchors” at
each end. An advantage of this type of scale is that panellists responses are not limited to a
number of categories on the scale and therefore it may be possible to identify more
differences between samples.

COPYRIGHT 70 R L Mason and S M Nottingham

Sensory Evaluation

Example of a category scale.

Strawberry flavour
Sample number 495 128
Extremely strong

Very strong




An example of a line scale.

Strawberry Flavour

None Very strong

Analysis of results

Ratings must be converted to numerical scores for analysis and interpretation. For category
scales, successive integers are assigned to successive categories and these are used in
analysis, e.g. with a 9-point scale, the integers 1-9 would be used. For graphic scales, the
distance, e.g. in mm, between the response mark and one end of the scale serves as the
response score.

The arithmetic mean and standard deviation, when obtained for each sample, serve as
measures of central tendency and variability, respectively. For statistical analysis, the
analysis of variance technique is appropriate (or a t-test in the case of one or two samples).
Correlation or regression analysis may be used for subjective/objective correlations.


ƒ More than one sensory attribute can be examined.

ƒ Size and direction of differences can be identified.


ƒ Selecting realistic terminology

ƒ Agreement and understanding between assessors in descriptive terms
ƒ Scales are not linear ie: 13 = extremely sweet is not twice as sweet as 7 = moderately

COPYRIGHT 71 R L Mason and S M Nottingham

Sensory Evaluation


ƒ Product Development
ƒ Quality Control
ƒ Storage Trials
ƒ Research


The scoring method was used to determine if there was a difference in bitterness in cheddar
cheese made using three different milk-coagulating enzymes. Samples of cheese from each
treatment were coded and presented to 12 judges for evaluation. The order in which the three
samples were tasted was balanced so that each possible order was used twice: ABC, ACB,

The ratings assigned by the judges were given numerical values, ranging from 0 points for
‘not bitter’ to 5 points for ‘extremely bitter’. The results are shown in the following table.

Judges Samples
A B C Total
1 3 0 1 4
2 2 2 2 6
3 3 1 2 6
4 1 1 0 2
5 3 1 3 7
6 2 1 1 4
7 3 2 2 7
8 2 0 1 3
9 3 1 2 6
10 4 2 3 9
11 1 1 0 2
12 2 2 2 6
Total 29 14 19 62
Mean 2.42a 1.17b 1.58b

The results were submitted to analysis of variance.

Any two values not followed by the same letter are significantly different (P<0.05). Sample
A is significantly (P,0.05) more bitter than sample C and B. There is no significant
difference (P>.05) in bitterness between samples C and B.

COPYRIGHT 72 R L Mason and S M Nottingham

Sensory Evaluation


What kind of data do we have?

When data are graded, as in rating scales then the t-test or other “parametric” statistics such
as analysis of variance are applied. These are discussed in a different section. In other
situations when we categorise performance into right or wrong answers and count numbers of
people who get tests correct or incorrect or those who make one choice over the other we call
this discrete categorical data. For these data we use a special distribution called the binomial
distribution and is useful for tests based on proportions. Some examples of how these tests
are used in sensory tests is given below.

Triangle Test

The triangle test is used when we want to know if there is a detectable difference between
two samples or products. Three samples are presented where two are the same and one is
different. Panellists are asked to pick the odd one out. Purely by luck the panellist has a one
in three chance of getting it right.

This forms the basis of the normal approximation to the binomial test. Lets accept that
p − pexp
z = obs where
pq / N
pobs is the proportion who answered correctly ie X/N
pexp is the proportion of people who we expect by chance ie 1/3
q = 1 - pexp
z is obtained from tables and for a one tailed risk of 5% is equal to 1.65.

By substituting into the equation and solving for X we get

X 1

1.65 = N 3 and
X = 0.778 N + N / 3

Now for a range of N values (ie number who sit the test) we can get a range of X values (ie
the minimum number who must get the test right). These values have conveniently been
calculated and are already tabulated for use (see tables 1, 2).

For example if we have N = 30 panellists we must have at least

0.778√30 + 30/3
or 14.26 correct to achieve significance.

Since we cannot have 0.26 of a person so we round up to 15. Therefore 15 out of 30 people
must get the triangle test right in order to reject the null hypothesis and conclude there is a
difference among the samples.

COPYRIGHT 73 R L Mason and S M Nottingham

Sensory Evaluation

Duo Trio

This is similar to the triangle test except that a standard is presented and two other samples,
one of which is the same as the standard, are also given. The panellist has to pick which of
the two is the same as the standard so has a one in two chance of being correct. Tables for
this are also available but as it is less efficient than the triangle test it is not usually preferred
over the triangle test (see table 3).

Paired Comparison Test

Only two samples are given and panellists are asked to pick which sample is, for example,
sweeter than the other. The same tables as for the duo trio test can be used and a one-tailed
test is used when you expect one sample to be sweeter (for example) than the other. When
there is no preconceived idea of which sample may be sweeter then two-tailed test is
appropriate (see tables 3, 4).

Freidman Test

This is best demonstrated by example. Suppose 18 panellists are asked to rank three orange
juices in order of preference. What we want to know is. Are the ranked values for all
panellists the same? The results are as follows.

Panellist RANK
1 1 3 2
2 2 3 1
3 1 3 2
4 1 2 3
5 3 1 2
6 2 3 1
7 3 2 1
8 1 3 2
9 3 1 2
10 3 1 2
11 2 3 1
12 2 3 1
13 3 2 1
14 2 3 1
15 2 3 1
16 3 2 1
17 3 2 1
18 2 3 1
Sum (column total) 39 43 26

COPYRIGHT 74 R L Mason and S M Nottingham

Sensory Evaluation

The Freidman value F needs to be calculated as follows.

JP ( P + 1)
∑ Tp2 − 3J ( P + 1)

where J - number of judges

P - is the number of samples (products)
T1,T2,T3 - are the rank totals for each sample

For our example we get a F of 8.8. Now when the number of panellists is large or the number
of samples exceeds 5 then F follows the chi-squared distribution with P-1 degrees of
freedom. So looking up the chi-squared table (table 7) gives a critical value of 5.99. Since
our calculated F is greater than this we can reject our question and conclude that there are
significant differences between the samples. Pairwise comparisons can be made using the
formula below.

JP ( P + 1)
1.960 at the 5% level.

Two samples are different if the difference between their rank sums is greater than or equal to

Difference from Control Test

Although the difference from control test is a form of difference testing, the data we collect is
not discreet data so the analysis follows the ‘parametric’ tests that we use on rating scales.
These are discussed in the Statistics for Sensory: Descriptive Testing section.

COPYRIGHT 75 R L Mason and S M Nottingham

Sensory Evaluation


Descriptive testing is used to identify and provide a picture or “profile” of the important
sensory characteristics of a product. With sensory profiling more than two samples can be
assessed simultaneously. This type of test has the advantage of not only being able to tell you
whether or not there is a difference between samples but also the nature and magnitude of
these differences. Appearance, odour, flavour and texture can all be assessed in this way and
the characteristics can be quantified using various techniques and scales as outlined in this


• Tracking changes in the sensory characteristics of a product over time for shelf-life
• Examining the sensory properties of a target product for new product development
• Examining sensory characteristics of different varieties of a product eg to look at
several varieties of apples in order to identify which varieties are sweetest, crunchiest
• Sensory diagnostics of ingredient, process or packaging changes
• Correlations with instrumental methods

The Flavour Profile Method® (Arthur D. Little)

This method was developed by Arthur. D. Little in the late 1940’s early 1950’s. It uses a
panel of 4-6 trained panellists. Panellists are selected by screening for sensory acuity,
interests, attitude and availability. A vocabulary is developed by exposure to a wide range of
products from the product category to be assessed. The list is then reviewed and refined and
reference standards and definitions applied to each term.

The panellists examine the products and the results are reported to the panel leader. Through
discussion in an open session lead by the panel leader, a consensus decision is reached for
each sample. Aroma, flavour and amplitude, which is the balance or blending of the flavour,
is assessed in this way. The scales used with this technique involve the use of numbers and
symbols and therefore cannot be analysed statistically. This method is therefore a qualitative
descriptive test. The main disadvantage with this type of test is that a dominant panel
member or the panel leader could easily influence the panels decision.

COPYRIGHT 76 R L Mason and S M Nottingham

Sensory Evaluation

Questionnaire for flavour profile of beer

Product Name Date

Characteristic Intensity
Amplitude(overall aroma)

Characteristic Intensity
Amplitude(overall flavour)

COPYRIGHT 77 R L Mason and S M Nottingham

Sensory Evaluation

Profile Attribute Analysis®

The Flavour Profile method was renamed the Profile Attribute Analysis with the introduction
of numerical scales. Mean scores could then be calculated and the data statistically analysed.
However consensus methods are still employed by some people. Again, this runs the risk of
a result being skewed by a dominant personality in the group.

The Texture Profile Method®

This method was developed at General Foods in the 1960’s. It was based on the principles of
the Flavour Profile method to assess the textural characteristics of a product. Textural
characteristics are categorised into three groups, mechanical, geometrical and ‘other’

1. Mechanical: relating to the reaction of food to stress eg. hardness, chewiness and
2. Geometrical characteristics: relating to the size, shape and orientation of the particles
within the food eg. grainy, fibrous and aerated
3. Other characteristics: relating to the perception of the moisture and fat contents of the

The order in which the characteristics are assessed is also very important. The order of
assessment is first bite, “chewing” or masticatory second phase and residual or third phase.
Panellists are selected on their ability to discriminate between textural differences in the
product area to be trained. Six to ten panellists are suggested.

Standardised terminology and rating scales are used for the assessments and each scale point
is anchored with a specific food. Initially the technique used an expanded version of the
Flavour Profile scale, however more recently category and line scales have been used.

Panellists each make their own individual judgement and then depending on the type of scale
used, a consensus decision is reached or statistical analysis is performed on the data.

Quantitative Descriptive Analysis (QDA®)

This method of descriptive analysis was developed in the 1970’s. Ten to twelve panellists are
selected by screening for ability to discriminate between products, their ability to verbalise
their perceptions and to work as a group. The first step is to expose the panellists to a wide
range of products from the product category to be assessed. Each panellist individually lists
as many descriptive words possible that describe differences between the products. Hedonic
terms such as nice, good, bad, etc are not allowed. Through a group discussion, the list of
descriptive words is narrowed down to remove duplications and redundant terms until a
standardised vocabulary is reached. This standardised vocabulary then needs to be defined
with verbal definitions or reference standards and anchor points for the scale agreed upon.
The panel also decides the order in which the terms are to be assessed. During this process
the panel leader only acts to facilitate the discussion and provide references but does not
influence or lead the panel. Trial evaluations are then carried out using the agreed vocabulary
and refinements may be made until the panel is happy with the terms used. The panel leader
evaluates the results from these trial sessions and once confident the results are reliable and

COPYRIGHT 78 R L Mason and S M Nottingham

Sensory Evaluation

repeatable the actual assessment can take place. The assessment and trial sessions are
completed in sensory booths following the basic principles of sensory evaluation. An
unstructured 6-inch or 15cm line scale is used to measure the intensities of the agreed
characteristics. Several replicates (3+) are required to validate the data. Data is then
analysed using an analysis of variance. The results are often displayed visually on a spider
web or star diagram.

Results of ANOVA of orange jelly using QDA

Attribute Brand A Brand B SEM Probability

Orange colour 10.2 7.9 0.62 0.011
Orange aroma 7.6 6.9 0.50 0.325
Firmness 9.6 6.6 0.64 0.001
Tartness 8.6 6.9 0.66 0.072
Orange flavour 7.6 6.9 0.72 0.494
Foreign flavour 4.3 4.8 0.48 0.464
Sweetness 7.1 9.6 0.42 <0.001
Rate of 5.1 6.1 0.60 0.242

COPYRIGHT 79 R L Mason and S M Nottingham

Sensory Evaluation

Other Methods
Other methods which you may come across in literature but which will not be discussed in
detail in this workshop are:

Spectrum Method

This is a descriptive analysis technique developed by Civille to cover any or all of

appearance, aroma, flavour, texture or sound characteristics. Panellists use a standardised
lexicon of terms to evaluate the products. This method requires extensive training of the
panel to use standardised scales anchored with multiple reference points and panellists are
trained to use the scale identically. Data is analysed in a similar way to QDA.

Example of intensity scale values (0 to 15) for firmness.

Scale Reference Sample size

3.0 Aerosol whipped cream Redi whip 1oz
5.0 Miracle whip Kraft 1oz
8.0 Cheese whiz Kraft 1oz
11.0 Peanut butter CPC Best Foods 1oz
14.0 Cream cheese Kraft/Philadelphia 1oz

Time Intensity

This is used to track the changes in perception of a particular attribute of a product over time.
For example you might rate the intensity of mint flavour perceived in chewing gum over a 3
minute period. This can be measured using pencil and paper or using one of the sensory
software packages with time intensity facilities.

Free Choice Profiling

Unlike other descriptive testing techniques this method does not use an agreed vocabulary to
assess the samples. Each panellist generates their own list of terms and scales, although they
must use these consistently for all samples. The data from this type of assessment is then
analysed using Generalised Procrustes analysis. The main advantage of this technique is the
time saved on training a panel, however interpretation of individual attributes can be
subjective as the terms are not defined as with other descriptive testing methods.

COPYRIGHT 80 R L Mason and S M Nottingham

Sensory Evaluation


We mentioned earlier about different types of data and how they are analysed using different
statistical methods. In this section we will look at the most common form of statistical
analysis for rating or preference type data, the paired t-test and Analysis of Variance (one
way and two way also known as repeated measures analysis of variance). We will also
introduce some advanced methods for separating data into logical groups using Principal
Component Analysis.

The paired t test

A common question we have in sensory evaluation is when we are comparing two products
or samples and we want to know if they are the same or different. We can use statistics and
in particular the paired t test to determine statistical difference. We calculate a t value from
the formula below and compare it to some tabled values for probabilities less than our
accepted risk, usually a probability of 0.05.

t= mean of difference scores

standard deviation of difference scores/ N

Here is an example taken from O’Mahony. Intensity scores for two products are measured by
10 panellists on a 25 point scale.

Panellist Product A Product B Difference - d d2
1 20 22 2 4
2 18 19 1 1
3 19 17 -2 4
4 22 18 -4 16
5 17 21 4 16
6 20 23 3 9
7 19 19 0 0
8 16 20 4 16
9 21 22 1 1
10 19 20 1 1
N = 10 ∑ d = 10 ∑d 2
= 68
d =1

Now for some calculations.

mean of difference scores, d, = 1

the standard deviation of d is computationally =

∑d 2
− (( ∑ d ) 2 / N )
= 58 / 9 = 2.538
N −1
so t = = 1.25
2.538 10
The degrees of freedom term

COPYRIGHT 81 R L Mason and S M Nottingham

Sensory Evaluation

If we had four flavour scores, and knew three of them plus the variance or standard deviation
then we could calculate the value of the fourth unknown data point. In general, degrees of
freedom are equal to the sample size, minus the number of parameters we are estimating. We
need this value when looking up statistical tables.

The tabulated t value for df = 9, two-tailed, p=0.05 is 2.262 (see table 6). A word on two-
tailed and one-tailed test. If we simply wish to test whether a mean is different from the
population then we use a two-tailed test. However, if it is directional ie. greater than or less
than then we need a one-tail test. Our value of 1.25 is less than this so we do not reject our
notion that product A is the same as product B. Our data indicates that the difference scores
are not significantly different from 0.

Analysis of Variance

If we have only two samples we want to compare then we can use the paired t-test as
described earlier and establish a difference if it is there. If we have four samples then we
could do six paired t-tests and cover all possible pairings of the four treatments. This
however becomes very inefficient and unreliable as the number of samples increase. An
alternative to this is to use a technique known as analysis of variance to compare several
samples at the same time.

Analysis of variance looks at the amount of variance attributed to the samples or treatments
and also estimates the error variance or natural variation. By then doing a ratio of these
variances (ie signal to noise) we can then compare this to tabulated values. The distribution
is known as the F distribution and we calculate a F ratio or F test.

Most computer packages now do analysis of variance, sometimes described as one-way

analysis of variance and two-way analysis of variance. A two-way analysis of variance is
used when the same judges or panellists rate the same samples (sometimes called repeated
measures). Lets look at an example.

Suppose we have 10 panellists rating three samples of mango for mango flavour intensity on
a nine point scale. Their results are entered into a computer that then completes a two-way
analysis of variance and gives the following table.

COPYRIGHT 82 R L Mason and S M Nottingham

Sensory Evaluation

Source of variation d.f Mean square F

Samples 2 23.456 11.80 **

Panellists 9 3.667 1.84 ns
Error 18 1.987

The F ratio for samples is 11.80 which is greater than the tabulated value of 6.01 at P=0.01
with 2 and 18 degrees of freedom (often ** indicates significance at P=0.01, * for P=0.05).
This means that the variability due to samples is greater than that occurring naturally so there
are differences between the mango flavour intensity of the three mango samples.

Now to further test which pairs are significantly different we have a number of options. The
most common test would be the least significant difference (lsd) test which is based on the t-
test. The lsd tells us what the minimum difference between two means must be for there to
 2
be a significant difference. The formula is t α ,df ems  where ems is the error mean
 n
square and n is the number of observations per sample and t come from tables of calculated

In this example the lsd (P=0.01) is 1.32, so any two means with a difference greater than this
lsd are significantly different.

Other pairwise comparison tests are Duncan’s multiple range and Tukeys honestly significant
difference (HSD). Formulas for these tests can be found in most statistical textbooks or in
some cases the computer package may do the test for you.

You will also note that the panellist’s F ratio is not significant (ns), indicating the average
score given by any one judge is not that different to another judge’s score. Quite often the
panellists F ratio is significant, indicating that they are using different parts of the scale. With
highly trained panels, panellists tend to agree on the use of the scales.

An extension to the two-way analysis of variance is the three-way analysis of variance where
we add replicates to the AOV table to provide a complete analysis of the experimental data.

Difference from Control Test

The analysis of the data from this test can take a number of forms but I will outline the most
common and simplest to use. If you have a blind control and one test sample then you can
perform a paired t-test. If you have more than two samples then you can use analysis of
variance techniques and pairwise comparisons to determine differences.

An example taken from Meilgaard et al is given below. Forty-two judges are asked to
measure the perceived overall sensory difference between two prototypes (samples F and N)
and the regular analgesic cream (control). A category scale is used where 0 is no difference
and 10 is extreme difference.

COPYRIGHT 83 R L Mason and S M Nottingham

Sensory Evaluation

Difference from control test – Analgesic cream

Judge Blind Product Product Judge Blind Product Product

control F N control F N
1 1 4 5 22 3 6 7
2 4 6 6 23 3 5 6
3 1 4 6 24 4 6 6
4 4 8 7 25 0 3 3
5 2 4 3 26 2 5 1
6 1 4 5 27 2 5 5
7 3 3 6 28 2 6 4
8 0 2 4 29 3 5 6
9 6 8 9 30 1 4 7
10 7 7 9 31 4 6 7
11 0 1 2 32 1 4 5
12 1 5 6 33 3 5 5
13 4 5 7 34 1 4 4
14 1 6 5 35 4 6 5
15 4 7 6 36 2 3 6
16 2 2 5 37 3 4 6
17 2 6 7 38 0 4 4
18 4 5 7 39 4 8 7
19 0 3 4 40 0 5 6
20 5 4 5 41 1 5 5
21 2 3 3 42 3 4 4
Mean 2.4 4.8 5.4

Source of variation d.f Mean square F

Judge 41 6.183 6.04 **

Product 2 105.365 102.93 **
Error 82 1.024

Sample means with Fishers LSD0.05 = 0.44

Sample Blind control Product F

Mean response 2.4a 4.8b

Sample Blind control Product N

Mean response 2.4a 5.4b

Within a row, means not followed by a similar letter are significantly different at the 95%
confidence level.

It is concluded that both samples F and N were found to be significantly different from the
control. Descriptive testing is recommended to determine the nature of these differences.

COPYRIGHT 84 R L Mason and S M Nottingham

Sensory Evaluation

Advanced analysis using principal components

When we have lots of data and have asked our panel a number of questions we need a
technique for reducing the data down to a manageable size. The aim of the technique is to
reduce the number of variables that describe a sample to that of fewer dimensions. If we
reduce to two then we can plot the results onto a graph. These two dimensions are called
principal components and are a linear combination of the original variables.

It is useful for classifying a number of products by grouping them according to the variables
that describe them. Below is an example of the separation of nine coffees using principal



Wetwood Roast

Vittoria Australian Blend Robert Timms Mocha Kenya

Douwe Egbert premium Melita Colombian Premium Style
Andronicus Mocha Kenya Robert Timms New Guinea Gold
Harris Gold Label

COPYRIGHT 85 R L Mason and S M Nottingham

Sensory Evaluation


When developing a sensory panel, there are several areas that need to be addressed that include:

ƒ The need for a panel in the organisation (R&D, QA/QC)

ƒ Organisation and management support and commitment (time and money)
ƒ Resources required
o Sensory staff
o Interest and availability of potential panellists
o Samples and references for screening and training
o Availability of a panel room and booths
o Facilities for data collection and statistical analysis

Establishment of a trained sensory panel can be divided into 2 steps:

ƒ Selection
ƒ Training

Selection for Descriptive Testing (Australian Standard 2542.1.3 - 1995)


Panel members are usually recruited from staff in laboratories, offices and the plant of a
company. Some companies test their products at a different company facility. External
panellists may also be recruited from the community nearby if the sensory panel work is
going to be very time consuming.

Talks, circulars, noticeboards or personal invitations may be used to recruit potential

panellists. Information should be provided to the prospective panellists concerning the
application of sensory evaluation, what will be involved for the panellists and the envisaged
work program.

Pre-screening questionnaire

Potential panellists need to complete a pre-screening questionnaire to obtain background

information on their:

ƒ interest in participating in the screening and training program as well as ongoing work
ƒ availability
ƒ general good health (note any illnesses or allergies and permanent impairment to the
ƒ any food idiosyncrasies (strong food dislikes or reactions to foods)
ƒ other information that might be relevant (age, sex, nationality, cultural and religious
background, previous sensory experience, smoking habits)

Panellists should not be asked to assess a food that they dislike.

In a company situation, distribute questionnaires for employees to fill in, detailing the above

COPYRIGHT 86 R L Mason and S M Nottingham

Sensory Evaluation

criteria. If you make all the questions optional you will find that the majority of people
respond truthfully. Pre-screening questionnaires can also be used to select individuals who
can describe sensory concepts. Record all the information you receive in some form of
database. Based on the above criteria, decide which prospective panellists are to proceed in
the screening process.


Individual interviews are required to determine whether prospective panellists will work well in
a group situation as well as for the analytical approach required in descriptive testing. An
interview is also used to confirm interest and availability.

For a descriptive sensory panel, there is a large investment involved in terms of both time and
money. It is best to complete a thorough screening process rather than training unsuitable
subjects. During the selection process, it is important to make note of both attendance and
personalities of panellists. A panellist who is repeatedly late or unavailable can be more trouble
than they are worth. Someone who distracts other panellists by talking or making comments,
despite repeated requests to remain silent while testing, is a liability, not an asset.

However, it is recognised that the best panellists available may need to be used although they
may not necessarily meet all the requirements.

Sensory screening tests

Screening is completed to obtain information on prospective panellists who need to be able

ƒ Detect differences in attributes present and their intensities
ƒ Describe the attributes using verbal descriptors and scaling methods for the different
ƒ Be able to recall and apply attribute references when required

Prior to the first screening test, a preliminary session is a good idea to set the rules that may need
to be enforced politely but firmly.

Instructions for panellists

ƒ Avoid eating, drinking, smoking or chewing gum for 30 minutes before testing.
ƒ Do not talk or distract other panellists while testing.
ƒ Read any instructions on the scoresheet before starting to evaluate samples.
ƒ Make sure you evaluate the samples in the required order.
ƒ Don't forget to fill in your name and the date.
ƒ Do not discuss samples with other panellists until after they have evaluated the samples.
ƒ Have confidence in your own judgement.
ƒ Ignore your personal likes and dislikes.

Sensory screening tests also give the prospective panellists an indication of the methods used in
sensory analysis. The screening tests used should be chosen with the envisaged sensory
program in mind. Basic tastes and odours are commonly used for screening tests as well as
materials that illustrate the attributes that may be included in the sensory program. Samples of

COPYRIGHT 87 R L Mason and S M Nottingham

Sensory Evaluation

the actual food products may also be used.

A series of triangle or duo-trio tests may be completed to assess the ability of the potential
panellists to detect small differences between stimuli at supra-threshold levels. Preferably,
potential panellists should respond correctly 100% of the time.

Matching tests may be used to evaluate the ability of a prospective panellist to distinguish
between different sensory stimuli.

In order to evaluate the ability of the panellists to describe sensory responses, a series of
products can be presented and potential panellists asked to describe the sensory impression. The
products used should be related to those that will be used in the envisaged sensory testing.
For example, a range of odours may be presented:

Chemical name Name most commonly associated with the odour

Benzaldehyde Bitter almonds, ...
Octene-3-ol Mushroom, ...
Phenyl-2 ethyl acetate Floral, ...
Diallyl sulfide Garlic, ...
Camphor Camphor, ...
Menthol Peppermint, ...
Eugenol Clove, ...
Anethol Aniseed, ...
Vanillin Vanilla, ...
Geosmin Musty/mouldy, ...
Beta-ionone Violets, raspberries, ...
Butyric acid Rancid butter, ...
Acetic acid Vinegar, ...
Isoamyl acetate Fruit, acid drops, ...
Dimethylthiophene Grilled onions, ...

Panellists are given these samples to assess one at a time and asked to describe the odour using
his/her own words. A system of marking can be devised e.g. 4 points for absolutely correct, 3
points for correct in general terms, 2 points for a vague association, 1 point for a wrong
association and 0 points for no response. A satisfactory level for selection of panellists needs to
be specified in relation to the materials used. Similar techniques can be applied for taste and

The potential panellists may be screened for their ability to rank or rate products for selected
attributes using the same technique as the final panel will use. All potential panellists are
presented with the samples in the same order. Panellists are chosen if a satisfactory level is
attained which will depend on the intensities of the samples used. Also check that they have
used most of the scale.

COPYRIGHT 88 R L Mason and S M Nottingham

Sensory Evaluation


In this phase, it is important that the panellists develop confidence as well as the skills for
product assessment. Panellists must be taught the correct procedures for evaluating samples
and ways to reduce or eliminate sensory adaptation. They must also learn to disregard their
personal preferences. Between 40 and 120h of training are required for a descriptive sensory
panel which will depend on the product, the number of attributes as well as the validity and
reliability required. A trained panel usually consists of 10-20 panellists.

The initial stage of training involves vocabulary development. The entire range of products
is presented to the panellists. They are instructed to individually assess the sensory
differences between the samples and record any differences as descriptive words. On
completion of this task, the panellists each list the attributes used to describe each sample. At
this time, it is very important that the panel leader does not lead or judge the descriptive
words generated by the panellist although they can ask for clarification. The panellists
themselves will usually start to move towards a general consensus once the total attribute list
has been generated.

It is then the role of the panel leader to provide reference standards for the attributes that have
been previously selected by general panel consensus. The references can be used to help the
panellists to identify and remember a sensory attribute found in the sample. The references
may be chemicals, ingredients or products. The panellists then assess the samples alongside
the references until a consensus is reached regarding the sensory attributes, reference
standards and definitions. This process should continue until the panellists are all happy and
understand the terms used.

Towards the end of training, a scoresheet is created by the panellists. The panellists decide
on the order in which the attributes are to be assessed. Generally the panel leader decides on
the type of scale used, although the panel decides on the verbal anchors to be used.

Once the panellists have become familiar with the samples, references and definitions, panel
evaluation sessions are completed that should be similar to the final testing situation. The
panellists are presented with coded samples in triplicate and asked to rate them using the
scoresheets and attribute scales they have trained with. By statistically analysing the data, the
panel leader will be able to determine if further training is required or if the evaluation phase can

Like any instrument, the performance of individual panellists as well as the panel as a whole
needs to be monitored to check they are producing reliable results. Reliability is checked by
completing test replications and the descriptive data obtained is analysed statistically using an
analysis of variance.

Motivation of panellists is one of the most important factors in maintaining an efficient trained
sensory panel. If panellists are motivated and interested they will perform well. For panellists, a
sense of completing meaningful work is an important source of motivation. When appropriate
on completion of a project, feedback should be given to the panel as to the project objectives and
outcomes and the contribution of the sensory results. Individual panellist feedback is also
important. They should be made to feel that attendance at sensory evaluation sessions is

COPYRIGHT 89 R L Mason and S M Nottingham

Sensory Evaluation

important. This can be reinforced by running sessions strictly and efficiently to keep their time
input to a minimum.

Throughout training as well as during ongoing sensory evaluation sessions, it is important to

keep the channels of communications open through panel discussion at the completion of a
training session or a sensory testing session.

Ongoing records of panellists' training and experience are invaluable. In some instances training
can occupy more time than the actual experimental testing sessions, especially when you first
start. However, if the job is done correctly right from the start, your trained panel will be one of
the most valuable resources in the company. Make sure you look after them.

An aside: Expert panels

Panellists who have a great deal of experience in assessing a particular product are often referred
to as "Expert tasters". Commodities that utilise expert tasters include the tea, coffee, wine and
dairy industries. These panels usually include only 2 or 3 highly trained tasters. These tasters
are particularly sensitive to the nuances of a specific product. They also have the ability to carry
the characteristics of standard samples in their sensory memory. It takes a great deal of practice
to develop the skill and requires continued tasting to stay "tuned". They are usually responsible
for arranging the tasting conditions and samples themselves, in addition to actually tasting and
making a final report. This type of panel is most frequently used to assign a quality grade to a
finished product, as in butter and cheese grading. In the wine and coffee industries one expert
may use these skills to blend individual components to produce a final product with the desired

COPYRIGHT 90 R L Mason and S M Nottingham

Sensory Evaluation


As with any other scientific experiment your sensory testing needs to be reported in a clear
and concise manner. The Australian standards for each test type details what should be
included in the report.

The results obtained should be interpreted and conclusions drawn using all the information
gathered in the experiment. Recommendations may also need to be included depending on
the nature of the work.

Remember that it is much easier to write the report if you keep a record as you go along!

COPYRIGHT 91 R L Mason and S M Nottingham

Sensory Evaluation


American Meat Science Association, “Guidelines for Cookery and Sensory Evaluation of
Meat”, AMS, USA, 1978.

Amerine, M A, Pangborn, R M and Roessler, E B, “Principles of Sensory Evaluation of

Food”, New York: Academic Press, 1965.

ASTM, “Manual on Sensory Testing Methods”, STP 434, Am. Soc. Test. Makr.,
Philadelphia, Pennsylvania, 1968.

Aust, L B, Gacula, M C, Beard,S A and Washam, R W. “Degree of Difference Test Method

in Sensory Evaluation of Heterogeneous Product Types. Journal of Food Science, 50: 511 –
513, 1985

Bartoshuk, L, “Separate worlds of taste” Psychology Today 14 (9): 48-57, 1980.

Bartoshuk, L M, “The biological basis of food perception and acceptance” Food Quality &
Preference 4: 21-32, 1993.

Bourne, M C, “Food Texture and Viscosity: Concept and Measurement”, Academic Press
Inc., California, 1982.

Chi-Tang Ho, Manley, C H, “Flavor Measurement”, Marcel Dekker, Inc. 1993.

Gacula M C., “Design and analysis of Sensory Optimization”, Food & Nutrition Press. 1993.

Gacula, M C and Singh, J, “Statistical Methods in Food and Consumer Research, New York:
Academic Press, 1984.

Jellinek, G, “Sensory Evaluation of Food: Theory and Practice”, Chichester: Ellis

Horwood; 1985.

Lawless, H T, “Pepper potency and the forgotten flavour sense” Food Technology 43 (11):
52, 57-58, 1989.

Lawless, H T & Heymann, H, “Sensory Evaluation of Food: Principles and Practices”,

Chapman & Hall, New York, 1998.

Lyman, B, “A Psychology of Food”, Van Nostrand Reinhold Co. Inc., New York, USA,

Lyon, D H, Francombe, M A, Hasdell, T A and Lawson, K, (editors) “Guidelines for

Sensory Analysis in Food Product Development and Quality Control”. Chapmann and Hall,
London, UK, 1992.

McBride, R L, “The Bliss Point Factor”, Sun Books, Australia, 1990.

COPYRIGHT 92 R L Mason and S M Nottingham

Sensory Evaluation

McBride, R L, (editor), “Psychological Basis of Sensory Evaluation”, Elsevier Applied

Science, London, UK, 1990.

McRae, R, Robinson, R K & Sadler, M J (eds) “Encyclopedia of Food Science, Food

Technology and Nutrition”, Volume 6, Academic Press, London, 1993.

Meilgaard, M, Civille, G V and Carr, B T, “Sensory Evaluation Techniques: Boca Raton,

Fla: CRC Press, 1999. (3rd Edition)

Miflora Minoza-Gatchalian, “Sensory Evaluation Methods with Statistical Analysis (for

Research Product Development and Quality Control)”. 1981.

Moskowitz, H R, “New Directions for Product Testing and Sensory Analysis of Foods”, Food
& Nutrition Press, Inc. 1985.

Moskowitz, H, “Applied Sensory Analysis of Food”, Volumes 1 and 2, CRC Press, Florida,
USA, 1988.

O’Mahoney, M, “Sensory Evaluation of Food: Statistical Methods and Procedures”, New

York: Marcel Dekker, Inc, 1986.

O’Mahony, M & Ishii, I “Do you have an umami tooth?” Nutrtion Today May/June, 1985.

Piggott, J R, “Sensory Analysis of Food”, London: Elsevier Applied Science, 1988 (2nd
edition now available).

Piggott, J R, “Statistical Procedures in Food Research”, London: Elsevier Applied Science,


Piggott, J R, Paterson, A “Understanding Natural Flavors”. Blackie Academic &

Professional. 1994.

Poste, L M, Mackie, D A, Butter, G and Larmond, E, “Laboratory Methods for Sensory

Analysis of Food”, Agriculture Canada Publication 1864/E, 1991.

Rutledge, K P and Hudson, J M, “Sensory Evaluation: Method for Establishing and Training
a Descriptive Flavour Panel, Food Technology 44 (12): 78-84, 1990.

Stone, H and Sidel, J L, “Sensory Evaluation Practices”, 2nd edition, New York: Academic
Press, 1992.

Thomson, D M H, “Food Acceptability”, Elsevier Applied Science, London, UK, 1988.

COPYRIGHT 93 R L Mason and S M Nottingham

Sensory Evaluation


Gacula, M C, “Journal of Sensory Studies”. Food & Nutrition Press, Inc.

MacFie, H J., Meiselman, H L., “Food Quality and Preference”. Elsevier Applied Scien

COPYRIGHT 94 R L Mason and S M Nottingham


Table 1: Probability of X or More Correct Judgments in n Trials (one-tailed, p = 1/3)a

n\x 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
5 868 539 210 045 004
6 912 649 320 100 018 001
7 941 737 429 173 045 007
8 961 805 532 259 088 020 003
9 974 857 623 350 145 042 008 001
10 983 896 701 441 213 077 020 003
11 988 925 766 527 289 122 039 009 001
12 992 946 819 607 368 178 066 019 004 001
13 995 961 861 678 448 241 104 035 009 002
14 997 973 895 739 524 310 149 058 017 004 001
15 998 981 921 791 596 382 203 088 031 008 002
16 998 986 941 834 661 453 263 126 050 016 004 001
17 999 990 956 890 719 522 326 172 075 027 008 002
18 999 993 967 898 769 588 391 223 108 043 014 004 001
19 995 976 921 812 648 457 279 146 065 024 007 002
20 997 982 940 848 703 521 339 191 092 038 013 004 001
21 998 987 954 879 751 581 399 240 125 056 021 007 002
22 998 991 965 904 794 638 460 293 163 079 033 012 003 001
23 999 993 974 924 831 690 519 349 206 107 048 019 006 002
24 999 995 980 941 862 737 576 406 254 140 068 028 010 003 001
25 999 996 985 954 888 778 630 462 304 178 092 042 016 006 002
26 997 989 964 910 815 679 518 357 220 121 058 025 009 003 001
27 998 992 972 928 847 725 572 411 266 154 079 036 014 055 002
28 999 994 979 943 874 765 623 464 314 191 104 050 022 008 003 001
29 999 996 984 955 897 801 670 517 364 232 133 068 031 013 005 001
30 999 997 988 965 916 833 714 568 415 276 166 090 043 019 007 002 001
31 998 991 972 932 861 754 617 466 322 203 115 059 027 011 004 001
32 998 993 978 946 885 789 662 516 370 243 144 078 038 016 066 002 001
33 999 995 983 957 905 821 705 565 419 285 177 100 051 023 010 004 001
34 999 996 987 965 922 849 744 612 468 330 213 126 067 033 014 006 002 001
35 999 997 990 973 937 873 779 656 516 376 252 155 087 044 020 009 003 001
36 998 992 978 949 895 810 697 562 422 293 187 109 058 028 012 005 002 001
47 998 994 963 959 913 838 735 607 469 336 223 135 075 038 018 007 003 001
38 999 996 987 967 928 863 769 650 515 381 261 164 095 051 025 011 004 002 001
39 999 997 990 973 941 885 800 689 560 425 301 196 118 066 033 016 007 003 001
40 999 997 992 979 952 903 829 726 603 470 32 231 144 083 044 021 010 004 001
41 998 994 983 961 920 854 761 644 515 385 268 173 104 057 029 014 006 002 001
42 999 995 987 968 933 876 791 683 558 428 307 205 127 073 038 019 008 003 001
43 999 996 990 974 945 895 820 719 600 471 347 239 153 091 050 025 012 005 002 001
44 999 997 992 980 955 912 845 753 639 514 389 275 182 111 063 033 016 007 003 001
45 999 998 994 984 963 926 867 783 677 556 430 313 213 135 079 043 022 010 004 002 001
46 998 995 987 970 938 887 811 713 596 472 352 246 161 098 055 029 014 006 003 001
47 999 996 990 976 949 904 836 745 635 514 392 282 189 119 070 038 019 009 004 002 001
48 999 997 992 980 958 919 859 776 672 554 433 318 220 142 086 048 025 012 006 002 001
49 999 998 994 984 965 932 879 803 706 593 473 356 253 168 105 061 033 017 008 003 001
50 999 998 995 987 972 943 896 829 739 631 513 395 287 196 126 076 042 022 011 005 002 001

COPYRIGHT R L Mason and S M Nottingham

Table 2: Minimum Numbers of Correct Judgments to Establish Significance at Various Probability
Levels for the Triangle tests (one tailed, p = 1/3)
Probability Levels
No. of trials (n) 0.05 0.04 0.03 0.02 0.01 0.005 0.001
5 4 5 5 5 5 5
6 5 5 5 5 6 6
7 5 6 6 6 6 7 7
8 6 6 6 6 7 7 8
9 6 7 7 7 7 8 8
10 7 7 7 7 8 8 9
11 7 7 8 8 8 9 10
12 8 8 8 8 9 9 10
13 8 8 9 9 9 10 11
14 9 9 9 9 10 10 11
15 9 9 10 10 10 11 12
16 9 10 10 10 11 11 12
17 10 10 10 11 11 12 13
18 10 11 11 11 12 12 13
19 11 11 11 12 12 13 14
20 11 11 12 12 13 13 14
21 12 12 12 13 13 14 15
22 12 12 13 13 14 14 15
23 12 13 13 13 14 15 16
24 13 13 13 14 15 15 16
25 13 14 14 14 15 16 17
26 14 14 14 15 15 16 17
27 14 14 15 15 16 17 18
28 15 15 15 16 16 17 18
29 15 15 16 16 17 17 19
30 15 16 16 16 17 18 19
31 16 16 16 17 18 18 20
32 16 16 17 17 18 19 20
33 17 17 17 18 18 19 21
34 17 17 18 18 19 20 21
35 17 18 18 19 19 20 22
36 18 18 18 19 20 20 22
37 18 18 19 19 20 21 22
38 19 19 19 20 21 21 23
39 19 19 20 20 21 22 23
40 19 20 20 21 21 22 24
41 20 20 20 21 22 23 24
42 20 20 21 21 22 23 25
43 20 21 21 22 23 24 25
44 21 21 22 22 23 24 26
45 21 22 22 23 24 24 26
46 22 22 22 23 24 25 27
47 22 22 23 23 24 25 27
48 22 23 23 24 25 26 27
49 23 23 24 24 25 26 28
50 23 24 24 25 26 26 28
60 27 27 28 29 30 31 33
70 31 31 32 33 34 35 37
80 35 35 36 36 38 39 41
90 38 39 40 40 42 43 45

COPYRIGHT 96 R L Mason and S M Nottingham

Table 3: Minimum Numbers of Correct Judgments to Establish Significance at Various Probability
Levels for Paired - Comparison and Duo-Trio Tests (one-tailed, p=1/2)
Probability levels
No of trials (N) 0.05 0.04 0.03 0.02 0.01 0.005 0.001
7 7 7 7 7 7
8 7 7 8 8 8 8
9 8 8 8 8 9 9
10 9 9 9 9 10 10 10
11 9 9 10 10 10 11 11
12 10 10 10 10 11 11 12
13 10 11 11 11 12 12 13
14 11 11 11 12 12 13 13
15 12 12 12 12 13 13 14
16 12 12 13 13 14 14 15
17 13 13 13 14 14 15 16
18 13 14 14 14 15 15 16
19 14 14 15 15 15 16 17
20 15 15 15 16 16 17 18
21 15 15 16 16 17 17 18
22 16 16 16 17 17 15 19
23 16 17 17 17 18 19 20
24 17 17 18 18 19 19 20
25 18 15 18 19 19 20 21
26 18 18 19 19 20 20 22
27 19 19 19 20 20 21 22
28 19 20 20 20 21 22 23
29 20 20 21 21 22 22 24
30 20 21 21 22 22 23 24
31 21 21 22 22 23 24 25
32 22 22 22 23 24 24 26
33 22 23 23 23 24 25 26
34 23 23 23 24 25 25 27
35 23 24 24 25 25 26 27
36 24 24 25 25 26 27 28
37 24 25 25 26 26 27 29
38 25 25 26 26 27 28 29
39 26 26 26 27 28 28 30
40 26 27 27 27 28 29 30
41 27 27 27 28 29 30 31
42 27 28 28 29 29 30 32
43 28 28 29 29 30 31 32
44 28 29 29 30 31 31 33
45 29 29 30 30 31 32 34
46 30 30 30 31 32 33 34
47 30 30 31 31 32 33 35
48 31 31 31 32 33 34 36
49 31 32 32 33 34 34 36
50 32 32 33 33 34 35 37
60 37 38 38 39 40 41 43
70 43 43 44 45 46 47 49
80 48 49 49 50 51 52 55
90 54 54 55 56 57 58 61
100 59 60 60 61 63 64 66

Source : .E.B .Roessler et al.. Journal of Food Science, 1978, 43, 940-947
COPYRIGHT 97 R L Mason and S M Nottingham
Table 4: Minimum Numbers of Agreeing Judgements Necessary to Establish Significance at Various
Probability Levels for the Paired – Preference Tests and Difference (two tailed, p=1/2).

Probability Levels
No. of trials (n) 0.05 0.04 0.03 0.02 0.01 0.005 0.001
7 7 7 7 7
8 8 8 8 8 8
9 8 8 9 9 9 9
10 9 9 9 10 10 10
11 10 10 10 10 11 11 11
12 10 10 11 11 11 12 12
13 11 11 11 12 12 12 13
14 12 12 12 12 13 13 14
15 12 12 13 13 13 14 14
16 13 13 13 14 14 14 15
17 13 14 14 14 15 15 16
18 14 14 15 15 15 16 17
19 15 15 15 15 16 16 17
20 15 16 16 16 17 17 18
21 16 16 16 17 17 18 19
22 17 17 17 17 18 18 19
23 17 17 18 18 19 19 20
24 18 18 18 19 19 20 21
25 18 19 19 19 20 20 21
26 19 19 19 20 20 21 22
27 20 20 20 20 21 22 23
28 20 20 21 21 22 22 23
29 21 21 21 22 22 23 24
30 21 22 22 22 23 24 25
31 22 22 22 23 24 24 25
32 23 23 23 23 24 25 26
33 23 23 24 24 25 25 27
34 24 24 24 25 25 26 27
35 24 25 25 25 26 27 28
36 25 25 25 26 27 27 29
37 25 26 26 26 27 28 29
38 26 26 27 27 28 29 30
39 27 27 27 28 28 29 31
40 27 27 28 28 29 30 31
41 28 28 28 29 30 30 32
42 28 29 29 29 30 31 32
43 29 29 30 30 31 32 33
44 29 30 30 30 31 32 34
45 30 30 31 31 32 33 34
46 31 31 31 32 33 33 35
47 31 31 32 32 33 34 36
48 32 32 32 33 34 35 36
49 32 33 33 34 34 35 37
50 33 33 34 34 35 36 37
60 39 39 39 40 41 42 44
70 44 45 45 46 47 48 50
80 50 50 51 51 52 53 56
90 55 56 56 57 58 59 61
100 61 61 62 63 64 65 67

COPYRIGHT 98 R L Mason and S M Nottingham

Table 5a ¨ 5 % Points for the Distribution of F

n2\n1 1 2 3 4 5 6 8 12 24 8
1 161.40 199.50 215.70 224.60 230.20 234.00 238.90 243.90 249.00 254.30
2 18.51 19.00 19.16 19.25 19.30 19.33 19.37 19.41 19.45 19.50
3 10.13 9.55 9.28 9.12 9.01 8.94 8.84 8.74 8.64 8.53
4 7.71 6.94 6.59 6.39 6.26 6.16 6.04 5.91 5.77 5.63
5 6.61 5.79 5.41 5.19 5.05 4.95 4.82 4.68 4.53 4.36
6 5.99 5.14 4.76 4.53 4.39 4.28 4.15 4.00 3.84 3.67
7 5.59 4.74 4.35 4.12 3.97 3.87 3.73 3.57 3.41 3.23
8 5.32 4.46 4.07 3.84 3.69 3.58 3.44 3.28 3.12 2.93
9 5.12 4.26 3.86 3.63 3.48 3.37 3.23 3.07 2.90 2.71
10 4.96 4.10 3.71 3.48 3.33 3.22 3.07 2.91 2.74 2.54
11 4.84 3.98 3.59 3.36 3.20 3.09 2.95 2.79 2.61 2.40
12 4.75 3.88 3.49 3.26 3.11 3.00 2.85 2.69 2.50 2.30
13 4.67 3.80 3.41 3.18 3.02 2.92 2.77 2.60 2.42 2.21
14 4.60 3.74 3.34 3.11 2.96 2.85 2.70 2.53 2.35 2.13
15 4.54 3.68 3.29 3.06 2.90 2.79 2.64 2.48 2.29 2.07
16 4.49 3.63 3.24 3.01 2.85 2.74 2.59 2.42 2.24 2.01
17 4.45 3.59 3.20 2.96 2.81 2.70 2.55 2.38 2.19 1.96
18 4.41 3.55 3.16 2.93 2.77 2.66 2.51 2.34 2.15 1.92
19 4.38 3.52 3.13 2.90 2.74 2.63 2.48 2.31 2.11 1.88
20 4.35 3.49 3.10 2.87 2.71 2.60 2.45 2.28 2.08 1.84
21 4.32 3.47 3.07 2.84 2.68 2.57 2.42 2.25 2.05 1.81
22 4.30 3.44 3.05 2.82 2.66 2.55 2.40 2.23 2.06 1.78
23 4.28 3.42 3.03 2.80 2.64 2.53 2.38 2.20 2.00 1.76
24 4.26 3.40 3.01 2.78 2.62 2.51 2.36 2.18 1.98 1.73
25 4.24 3.38 2.99 2.76 2.60 2.49 2.34 2.16 1.96 1.71
26 4.22 3.37 2.98 2.74 2.59 2.47 2.32 2.15 1.95 1.69
27 4.21 3.35 2.96 2.73 2.57 2.46 2.30 2.13 1.93 1.67
28 4.20 3.34 2.95 2.71 2.56 2.44 2.29 2.12 1.91 1.65
29 4.18 3.33 2.93 2.70 2.54 2.43 2.28 2.40 1.90 1.64
30 4.17 3.32 2.92 2.69 2.53 2.42 2.27 2.09 1.89 1.62
40 4.08 3.23 2.84 2.61 2.45 2.34 2.18 2.00 1.79 1.51
60 4.00 3.15 2.76 2.52 2.37 2.25 2.10 1.92 1.70 1.39
120 3.92 3.07 2.68 2.45 2.29 2.17 2.02 1.83 1.61 1.25
8 3.84 2.99 2.60 2.37 2.21 2.09 1.94 1.75 1.52 1.00

Source : Table 9 is taken from Table V of Fisher and Yates : 1974 Statistical Tables for Biological, Agricultural and Medical
Research published by Longman Group UK Ltd. London (previously published by Oliver and Boyd Ltd. Edinburgh) and by
permission of the authors and publishers.

COPYRIGHT 99 R L Mason and S M Nottingham

Table 5b ¨ 1 % Points for the Distribution of F

n2\n1 1 2 3 4 5 6 8 12 24 8
1 4052 4999 5403 5625 5764 5859 5981 6106 6234 6366
2 98.49 99.00 99.17 99.25 99.30 99.33 99.36 99.42 99.46 99.50
3 34.12 30.81 29.46 28.71 28.24 27.91 27.49 27.05 26.60 26.12
4 21.20 18.00 16.69 15.98 15.52 15.21 14.80 14.37 13.93 13.46
5 16.46 13.27 12.06 11.39 10.97 10.67 10.29 9.89 9.47 9.02
6 13.74 10.92 9.78 9.15 8.75 8.47 8.10 7.72 7.31 6.88
7 12.25 9.55 8.45 7.85 7.46 7.19 6.84 6.47 6.07 5.65
8 11.26 8.65 7.59 7.01 6.63 6.37 6.03 5.67 5.28 4.86
9 10.56 8.02 6.99 6.42 6.06 5.80 4.47 5.11 4.73 4.31
10 10.04 7.56 6.55 5.99 5.64 5.39 5.06 4.71 4.33 3.91
11 9.65 7.20 6.22 5.67 5.32 5.07 4.74 4.40 4.02 3.60
12 9.33 6.93 5.95 5.41 5.06 4.82 4.50 4.16 3.78 3.36
13 9.07 6.70 5.74 5.20 4.86 4.62 4.30 3.96 3.59 3.16
14 8.86 6.51 5.56 5.03 4.69 4.46 4.14 3.80 3.43 3.00
15 8.68 6.36 5.42 4.89 4.56 4.32 4.00 3.67 3.29 2.87
16 8.53 6.23 5.29 4.77 4.44 4.20 3.89 3.55 3.18 2.75
17 8.40 6.11 5.18 4.67 4.34 4.10 3.79 3.45 3.08 2.65
18 8.28 6.01 5.09 4.58 4.25 4.01 3.71 3.37 3.00 2.57
19 8.18 5.93 5.01 4.50 4.17 3.94 3.63 3.30 2.92 2.49
20 8.10 5.85 4.94 4.43 4.10 3.87 3.56 3.23 2.86 2.42
21 8.02 5.78 4.87 4.37 4.04 3.81 3.51 3.17 2.80 2.36
22 7.94 5.72 4.82 4.31 3.99 3.76 3.45 3.12 2.75 2.31
23 7.88 5.66 4.76 4.26 3.94 3.71 3.41 3.07 2.70 2.26
24 7.82 5.61 4.72 4.22 3.90 3.67 3.36 3.03 2.66 2.21
25 7.77 5.57 4.68 4.18 3.86 3.63 3.32 2.99 2.62 2.17
26 7.72 5.53 4.64 4.14 3.82 3.59 3.29 2.96 2.58 2.13
27 7.68 5.49 4.60 4.11 3.78 3.56 3.26 2.93 2.55 2.10
28 7.64 5.45 4.57 4.07 3.75 3.53 3.23 2.90 2.52 2.06
29 7.60 5.42 4.54 4.04 3.73 3.50 3.20 2.87 2.49 2.06
30 7.56 5.39 4.51 4.02 3.70 3.47 3.17 2.84 2.47 2.01
40 7.31 5.18 4.31 3.83 3.51 3.29 2.99 2.66 2.29 1.80
60 7.08 4.98 4.13 3.65 3.34 3.12 2.82 2.50 2.12 1.60
120 6.85 4.79 3.95 3.48 3.17 2.96 52.66 2.34 1.95 1.38
8 6.64 4.60 3.78 3.32 3.02 2.80 2.51 2.18 1.79 1.00

Source : Table 9 is taken from Table V of Fisher and Yates : 1974 Statistical Tables for Biological, Agricultural and Medical
Research published by Longman Group UK Ltd. London (previously published by Oliver and Boyd Ltd. Edinburgh) and by
permission of the authors and publishers.

COPYRIGHT 100 R L Mason and S M Nottingham

Table 6: Critical Value of ta

Level of significance for one-tailed test

0.1 0.05 0.025 0.01 0.005 0.0005
Level of significance for two-tailed test
df 0.2 0.1 0.05 0.02 0.01 0.001
1 3.078 6.314 12.706 31.821 63.657 636.619
2 1.886 2.92 4.303 6.965 9.925 31.598
3 1.638 2.353 3.182 4.541 5.841 12.941
4 1.533 2.132 2.776 3.747 4.604 8.61
5 1.476 2.015 2.571 3.365 4.032 6.859
6 1.44 1.943 2.447 3.143 3.707 5.959
7 1.415 1.895 2.365 2.998 3.499 5.405
8 1.397 1.86 2.306 2.896 3.355 5.041
9 1.383 1.833 2.262 2.821 3.25 4.781
10 1.372 1.812 2.228 2.764 3.169 4.587
11 1.363 1.796 2.201 2.718 3.106 4.437
12 1.356 1.782 2.179 2.681 3.055 4.318
13 1.35 1.771 2.16 2.63 3.012 4.221
14 1.345 1.761 2.145 2.624 2.977 4.14
15 1.341 1.753 2.131 2.602 2.947 4.073
16 1.337 1.746 2.12 2.583 2.921 4.015
17 1.333 1.74 2.11 2.567 2.898 3.965
18 1.33 1.734 2.101 2.552 2.878 3.922
19 1.328 1.729 2.093 2.539 2.861 3.883
20 1.325 1.725 2.086 2.528 2.845 3.85
21 1.323 1.721 2.08 2.518 2.831 3.819
22 1.321 1.717 2.074 2.508 2.819 3.792
23 1.319 1.714 2.069 2.5 2.807 3.767
24 1.318 1.711 2.064 2.492 2.797 3.745
25 1.316 1.708 2.06 2.485 2.787 3.725
26 1.315 1.706 2.056 2.479 2.779 3.707
27 1.314 1.703 2.052 2.473 2.771 3.69
28 1.313 1.701 2.048 2.467 2.763 3.674
29 1.311 1.699 2.045 2.462 2.756 3.659
30 1.31 1.697 2.042 2.457 2.75 3.646
40 1.303 1.684 2.021 2.423 2.704 3.551
60 1.296 1.671 2 2.39 2.66 3.46
120 1.289 1.658 1.98 2.358 2.617 3.373
00 1.282 1.645 1.96 2.326 2.576 3.2

The value listed in the table is the critical value of t for the number of degrees of freedom listed in the left
column for a one- or two-tailed test at the significance level indicated at the top of each column. If the
observed t is greater than or equal to the tables value, reject Ho.

Source: Table III of Fisher and Yates, Statistical Tables for Biological, Agricultural and Medical
Research, published by Longman Group Ltd, London (previously published by Oliver and Boyd Ltd,
Edinburgh) and by permission of the authors and publishers.

COPYRIGHT 101 R L Mason and S M Nottingham

Table 7: Critical Values of Chi-Squarea

Level of significance for one-tailed test

0.10 0.05 0.025 0.01 0.005 0.0005
Level of significance for two-tailed test
df 0.2 0.1 0.05 0.02 0.01 0.001
1 1.64 2.71 3.84 5.41 6.64 10.83
2 3.22 4.6 5.99 7.82 9.21 13.82
3 4.64 6.25 7.82 9.84 11.34 16.27
4 5.99 7.78 9.49 11.67 13.28 18.46
5 7.29 9.24 11.07 13.39 15.09 20.52
6 8.56 10.64 12.59 15.03 16.81 22.46
7 9.8 12.02 14.07 16.62 18.48 24.32
8 11.03 13.36 15.51 18.17 20.09 26.12
9 12.24 14.68 16.92 19.68 21.67 27.88
10 13.44 15.99 18.31 21.16 23.21 29.59
11 14.63 17.28 19.68 22.62 24.72 31.26
12 15.81 18.55 21.03 24.05 26.22 32.91
13 16.98 19.81 22.36 25.47 27.69 34.53
14 18.15 21.06 23.68 26.87 29.14 36.12
15 19.31 22.31 25 28.26 30.58 37.7
16 20.46 23.54 26.3 29.63 32 39.29
17 21.62 24.77 27.59 31 33.41 40.75
18 22.76 25.99 28.87 32.35 34.8 42.31
19 23.9 27.2 30.14 33.69 36.19 43.82
20 25.04 28.41 31.41 35.02 37.57 45.32
21 26.17 29.62 32.67 36.34 38.93 46.8
22 27.3 30.81 33.92 37.66 40.29 48.27
23 28.43 32.01 35.17 38.97 41.64 49.73
24 29.55 33.2 36.42 40.27 42.98 51.18
25 30.68 34.38 37.65 41.57 44.31 62.62
26 31.8 35.56 38.88 42.86 45.64 54.05
27 32.91 36.74 40.11 44.14 46.96 55.48
28 34.03 37.92 41.34 45.42 48.28 56.89
29 35.14 39.09 42.69 46.69 49.59 58.3
30 36.25 40.26 43.77 47.96 50.89 59.7
32 38.47 42.59 46.19 50.49 53.49 62.49
34 40.68 44.9 48.6 53 56.06 65.25
36 42.88 47.21 51 55.49 58.62 67.99
38 45.08 49.51 53.38 57.97 61.16 70.7
40 47.27 51.81 55.76 60.44 63.69 73.4
44 51.64 56.37 60.48 65.34 68.71 78.75
48 55.99 60.91 65.17 70.2 73.68 84.04
52 60.33 65.42 69.83 75.02 78.62 89.27
56 64.66 69.92 74.47 79.82 83.51 94.46
60 68.97 74.4 79.08 84.58 88.38 99.61
The table lists the critical values of chi square for the degrees of freedom shown at the left for tests
corresponding to those significance levels heading each column. If the observed value of xobs2 is greater
than or equal to the tabled value, reject Ho.

Source: Table IV of Fisher and Yates, Statistical Tables for Biological, Agricultural and Medical
Research, published by Longman Group Ltd, London (previously published by Oliver and Boyd Ltd,
Edinburgh) and by permission of the authors and publishers.

COPYRIGHT 102 R L Mason and S M Nottingham