1/16
Looks like no tags are added yet.
Discuss uses for difference tests
Used for discriminating between stimuli with small differences (confusable)
Formulation changes
Ingredient supplier change
Copy-cat product
Describe the different varieties of difference tests
General Difference
Duo-Trio
3 samples (1 reference, 2 blinded with one like the reference)
4 possible orders
Which is the same as the reference?
Triangle
3 samples (randomized with 1 different)
6 possible orders
Which is the odd one out?
Dual Standard
4 samples (2 reference, 2 blinded)
4 possible orders
Match the unknown to the 2 standards?
Tetrad unspecified
4 samples
6 possible orders
Sort into groups of 2?
Specific (name the attribute) Difference
Paired Comparison 2-AFC
2 samples
Which has more attribute intensity? (sweetness, saltiness)
3-AFC
3 samples (randomized with 1 different)
3 possible orders (A being stronger and B being weaker)
One has less (or more) of a specific attribute, find the weaker (or stronger) one?
n-AFC
Tetrad specified
4 samples
6 possible orders
Sort into groups of 2 and indicate the stronger or weaker pair in a specific attribute?
Analyze difference test data by conducting binomial statistics
General Difference
Duo-Trio
½ probability
1-tailed
Use table G.4.a
Triangle
1/3 probability
1-tailed
Dual Standard
½ probability
1-tailed
Tetrad unspecified
1/3 probability
First pick p = 1
Second pick p = 1/3
1-tailed
Specific (name the attribute) Difference
2-AFC
½ probability
1-tailed (greater or lesser known, difference test)
2-tailed (unknown change, preference test)
3-AFC
1/3 probability
1-tailed
n-AFC
Tetrad specified
1/6 probability
First pick (AABB) pick A: p = ½
Second pick (ABB or AAB) pick the matching one: p = 1/3
1-tailed
Introduce the paradox of discriminatory non-discriminators and its explanation using Thurstonian modeling
Judges perform better on 3-AFC tests (Which sample is sweeter?) than Triangle tests (Identify the odd sample) even though same # of samples per trial and binomial statistics. Thurstone’s Law of Comparative Judgment notes that discrimination between stimuli on a psychological continuum form normal distribution (perception is not always the same). Confusable samples overlap in their distributions where a lower intensity sample can be perceived as more intense than a higher intensity sample, and vise versa. When S.D. is ~3 with moderate degree of difference, samples are confusable. When S.D. is ~9 with large degree of difference, samples are distinguishable. During a triangle test, judges find the “odd 1 out” by a comparison of distance strategy, i.e. which one is most different. During a 3-AFC test, judges scan from perceptual extremes to find the “strongest” or “weakest.” Pic 3 row down: Triangle test picks the dot on the left due to distance, but 3-AFC picks the dot on the right due to most intense perception.

Analyze the paradox of discriminatory non-discriminators for various difference tests
As the S.D. or d’ between samples increases, the proportion of correct responses increases. General difference tests are less correct compared to specific difference tests, i.e. triangle vs 3-AFC and duo-trio vs 2-AFC.

Discuss the power of tests and its importance in test selection
Power
Probability of finding a difference if the difference actually exists (1- Beta, where Beta = type 2 error)
Type 2 error is failing to reject the null Ho when the Ho isn’t true, a false negative. You conclude no effect exists when it actually does.
Power depends on d’, alpha, # of judges
Higher proportion correct for d’ → more powerful test
Need larger # of tests for less powerful tests to demonstrate significant differences
2 and 3 AFC tests are more powerful than duo-trio and triangle tests. But we don’t always know the attribute, so Tetrad tests are a happy medium where not as many participants are needed.
Estimate proportion of population who are true discriminators
P = 1 - ((1-(D/N))/(1-p))
Define descriptive analysis (DA)
Detection and description of qualitative, quantitative and temporal aspects of a product by a trained panel of subjects.
Qualitative: complete sensory attributes
Lexicon development and defining attributes
Quantitative: measurement of intensities
Rating intensity on scales or rankings
Temporal: change in attribute intensity with time
Change in attributes overtime (first bite to residue)
Understand when to use DA
Discrimination/difference tests only identify if there is a difference(s), but DA gives the complete description or profile of a product.
Define the sensory properties of a target product for new product development
Define the characteristics for a control or a standard for QA/QC
Monitor product quality over time (shelf life)
Correlate with instrumental parameters
Match with another brand
Understand the sensory landscape for a product category/identify “white space”
Discuss the general process of DA, including panel selection, training, and panelist monitoring
Relatively few judges (10-20)
Panel needs to be screened, trained and monitored
Selection:
Sensory acuity/sensitivity
Availability
Verbal ability
Personality
Training:
Introduction to method
Terminology development
Introduction to scaling
Scaling standards/references to anchor scale
Practice with products
Practice in a group
Practice in an individual booth setting
Monitoring
Variability of each panelist & group over time
Outliers are eliminated prior to actual sample testing
Retraining to address disagreements (cross-over)
Panelists are used as analytical instruments
Laboratory testing condition (in Booths)
Standards/references are typically used
Discuss the history and basic components of legacy methods Flavor Profile & Texture Profile
Flavor Profile (FP)
Arthur D. Little & Co. in 1940-1950s by Lören Sjostrom, Stanley
Cairncross, Jean Caul
First used to describe MSG in complex flavor systems
Consensus (group score) technique, which is no longer commonly used
Amplitude: Definition: Degree of Blending
If one note stands out, low amplitude
“In a sense, the amplitude is not to be understood, just to be experienced” (Lawless & Heyman 2010)
Amplitude of whipped cream < whipped cream + sugar < whipped cream + sugar + vanilla
Ordinal Scales (0-3)
Profile Attribute Analysis (PAA)
Evolved from flavor profile and is the current Arthur D. Little technique, where panel training similar (4-6 trained panelists)
Intensity measurements done on 7-pt numerical scales
Give anchors like Spectrum method
Data analysis: Means, t-tests, ANOVA, PCA, etc
Texture Profile (TP)
General Foods (became Kraft) in 1960s by scientists
Builds on FP approach
Based on engineering principles and uses standardized terminology & rating scales
Mechanical - reaction of material to applied stress
Kinesthetic, hardness, chewiness
Rheological (viscosity)
Sensory (thickness)
Geometrical - highly organized structure of arrangements of particles
particle size and shape perceived by tactile sensations or visually observed
Other characteristics: Fat/moisture - Mouthfeel
Perception of oiliness or juiciness
Discuss the advantages and disadvantages of these methods
FP
Advantages
Can store record of flavor and compare scores over time (same panelists -> consistency)
Makes allowances for panelists who lack sensitivity: consensus
If references are used and training is good, can be accurate
Reproducible?? (claimed by ADL)
Disadvantages
Expensive
Long training
Subjective: Consensus, Panel leader could sway judgment
Suggestion effects: Panelists interact and induce suggestion
Often used without reference standards: not accurate
Single standards: concepts not completely aligned
“True FP resists most attempts for mathematical characterization of the data” (Lawless & Heyman 2010)
TP
Advantages
Extensive training ʹ select appropriate descriptive terms
Defining all terms and all procedures - reduce some of the variability
Reference scales - further reduce panel variability
Can conduct statistical analyses
Disadvantages
Use of products which can change to anchor the scales
Separation of texture from other sensory properties
a-priori decision on attributes used (defined lexicon (vs consumers creating terms)
Panel leader is a subject
Long training
Expensive
Discuss the history and basic components of the Quantitative Descriptive Analysis (QDA) and Sensory Spectrum methods
QDA
Tragon Corp. in 1970s by Herbert Stone and Joel Sidel
Be responsive to all the sensory properties of a product
80-100 words developed in the first 2 sessions (by consumers)
Rely on limited # of subjects for each test (10-12)
Use subjects qualified before participation
Verbal ability, availability, work as a group, perceive differences
Users & likers (unique to this methodology
Be able to evaluate multiple products in individual booths
Use a language development process free from leader influence
Consensus based (can take 7-10 hours)
Be quantitative and use a repeated trial design
Bother individuals and panels are monitored for performance
2-4 sample replications used
Relative scale
7. Have a useful data analysis system
ANOVA, LSD, Duncan, Tuckey, Newman-Keuls’, Scheffé
Spider web plot (Radar plot) to visualize the data
Sensory Spectrum (evolved TP)
Gail Civille (worked at general foods on the TP)
Complete and detailed profile of a product
Intensity of attributes based on absolute or universal scales with references
Generally includes all modalities of product
Applicable to many product categories, not just food
Discuss the advantages and disadvantages of these methods
QDA
Advantages
Disadvantages
Lack of formal instruction could lead to erroneous term development (non-standard lexicon)
The “free” approach to scaling may lead to inconsistency of results
The lack of immediate feedback to panelists on a regular basis reduces the opportunity for learning and expansion of terminology
Connecting “spokes” of the “spider web” can be misleading
Discuss pro’s and con’s of rapid descriptive methods compared to traditional descriptive analysis
To understand key elements of rapid descriptive methods, including check-all-that-apply (CATA), rate-all-that-apply (RATA), and nappingultra flash profile (Napping-UFP)
Time Intensity (one attribute over time)
Temporal Dominance of Sensation
Temporal Check-All-That-Apply (TCATA)
To learn the correct method of data analysis for CATA, RATA, and Napping-UFP