Measuring emotions in
commercial products
May Ng
Emotional measures
If liking is the first and only thing
GOtobeyond
liking
that came
your mind?
WHAT

is emotion?


Brief
Intense
Often focused on a referent
Example:
King & Meiselman
The chocolate made her happy
(2010)
Multidimensional
EMOTION
MODEL
Engagement
High Activation
Activated Unpleasant Activated Pleasant 450
Larsen & Diener
(1992)
450
Pleasant Unpleasant Unactivated Unpleasant Unactivated Pleasant Low Activation
WHY
measure
emotion?

Liking data is not enough

Emotional quality of products is
important for:
–
–
Differential advantage
Purchase decision
HOW
to measure
emotion?
Examples:


EsSense Profile
Consumer defined
Check-All-That-Apply
(CATA)
EsSense Profile (King & Meiselman., 2010)

Overall acceptability (9 point scale)

Emotion terms (5 point scale)
–
–
39 terms; most positive
Selected from published psychological emotion list
Reference:
King, S. C., & Meiselman, H. L. (2010). Development of a method to
measure consumer emotions associated with foods. Food Quality and
Preference, 21 (2), 168-177
OBJECTIVES
COMPARE
EsSense
Profile
v.s. Consumer defined
CATA
EMOTION data
GO BEYOND LIKING???
METHODS &
MATERIALS
Samples
11 UK Commercial
Blackcurrant Squashes
Labeled as P1- 11
6
Added
Sugar
2
Niche
Added
Sugar
3
No
Added
Sugar
Subjects
Group 1
n=100
Group 2
n=100
Methodologies
EsSense
Profile
Consumer defined
CATA
METHODS &
MATERIALS
Methodologies
EsSense Profile
Consumer defined CATA
Predetermined
psychological
list
Consumer self
generated & defined
list
5 point scale
9–
(Quantitative data)
(1:1 Rep Grid Interview; n=29)
point scale Check-All-That-Apply
to
measure liking (Qualitative data)
EMOTION LEXICONS
EsSense Profile (39) C-defined CATA (36)
25
POSITIVE
(+ve)
3
NEGATIVE
(-ve)
Adventurous
Bored
Active
Disgusted
Affectionate
Worried
Calm
Energetic
Enthusiastic
Free
Friendly
Glad
Good
Good-natured
Happy
Interested
Pleased
Joyful
Satisfied
Loving
Secure
Merry
Tender
Nostalgic
Warm
Peaceful
Whole
Pleasant
11
UNCLEAR
(+/-ve)
16
POSITIVE
(+ve)
1
19
NEGATIVE UNCLEAR
(+/-ve)
(-ve)
Aggressive
Daring
Eager
Guilty
Mild
Polite
Steady
Tame
Understanding
Wild
Approval
At ease
Attentive
Comforted
Curious
Desire
Good
Happy
Interested
Pleasant
Pleased
Refreshed
Reminiscence
Satisfaction
Trust
Warm
Guilty pleasure
Angry
Annoyed
Bored
Cautious
Confused
Disappointment
Discontented
Disgusted
Displeasure
Not refreshed
Regret
Resentment
Sceptical
Shocked
Sickly
Uncomfortable
Unhappy
Unpleasant
Worried
EMOTION LEXICONS
EsSense Profile (39) C-defined CATA (36)
25
POSITIVE
(+ve)
Adventurous
Active
Affectionate
Calm
Energetic
Enthusiastic
Free
Friendly
Glad
3
NEGATIVE
(-ve)
Bored
Disgust
Worried
Good
Good-natured
Happy
Interested
Joyful
Loving
Merry
Pleased
Nostalgic Satisfied
Peaceful
Secure
Pleasant Tender
Warm
Whole
1
16
19
POSITIVE NEGATIVE UNCLEAR
(+/-ve)
(+ve)
(-ve)
11
UNCLEAR
(-/+ve)
Aggressive
Daring
Eager
Guilty
Mild
Polite
Steady
Tame
Understanding
Wild
Approval
At ease
Attentive
Comforted
Curious
Desire
Angry
Annoyed
Bored
Cautious
Confused
Disappointment
Discontented
Good
Happy
Disgust
Interested Displeasure
Not refreshed
Regret
Resentment
Pleasant
Sceptical
Shocked
Pleased
Sickly
Refreshed
Uncomfortable
Unhappy
Unpleasant
Reminiscence
Reminiscence
Satisfied
Trust
Warm
Worried
Guilty
pleasure
DATA ANALYSIS
EsSense Profile
Quantitative
data
C-defined CATA
Qualitative
data
ANOVA
ANOVA and Tukeys Multiple
Chi square test
& to determine which products differed from
Comparison test
Multiple Comparisoneach other
to determine which Tukeys
products differed
from each other
liking scores
Multiple Correspondence
Analysis
Principal Component Analysis
(MCA)
(PCA)
Emotion mean data
to determine product positioning
Consumer liking as supplementary variable.
Individual responses to CATA question
Products as supplementary variable
to determine product positioning
Results:
EsSense Profile
Liking data
Product
Similar
low liking
scores
11 UK commercial
blackcurrant juice
squashes
(Labelled as P1 – P11)
Added Sugar
Niche Added Sugar
No Added Sugar
Similar
high liking
scores
4
3
Liking
Group
scores
4.1
A
4.3
AB
10
5.0
BC
8
5.3
CD
6
5
5.5
6.0
CDE
DEF
2
6.0
DEF
1
6.3
EFG
9
6.5
FG
7
6.7
FG
11
6.8
G
Do products with similar liking scores induce
different emotional responses??
Increased in Liking
RESULTS:
EsSense Profile
RESULTS:
EsSense Profile
Correlation
29/ 39 emotion circle:
terms were found significant in
Variables (axes F1 and F2: 89.28 %)
‘Tame’
blind assessment
1
0.75
0.5
Bored
Tame
Energetic
Good‐ naturedPeaceful
Interested Calm
JoyfulMerry
Good
Pleasant
Secure PoliteTender
SatisfiedSteady
Loving
22 +ve
& Warm
3+/- ve
Friendly
Understanding
WholeHappy Eager
Glad
FreeEnthusiastic Liking
Active
Affectionate
Pleased
Adventurous
Daring
F2 (5.58 %)
0.25
3 –ve
0
‐0.25
Disgusted
Worried
‐0.5
‐0.75
‐1
‐1
‐0.75
‐0.5
‐0.25
0
F1 (83.70 %)
0.25
0.5
0.75
1
RESULTS:
EsSense Profile
PCA:
Biplot (axes F1 and F2: 89.28 %)
4
10
NEGATIVE
Low liking
‐80
‐64
‐48
2
5
POSITIVE
High liking
6
‐32
‐16
0
7
16
9
1
3
8
‐5
32
11
48
Liking
64
RESULTS:
EsSense Profile
Q: Do products with similar liking scores induce
different emotional responses?
A: Yes. There is a significant difference for some emotions.
The difference in mean data is ≤ 0.5 point over a 5 point scale.
Low liking scores:
3
4
10
More ‘Tame’
High liking scores:
1
Added
Sugar
2
5
Niche
Added
Sugar
7
No
Added
Sugar
9
4
11
2
5
Less 'Adventurous’
& Less ‘Daring’
Results:
C-Defined CATA
RESULTS:
C-Defined CATA
Similar
low liking
scores
11 UK commercial
blackcurrant juice
squashes
(Labelled as P1 – P11)
Added Sugar
Niche Added Sugar
No Added Sugar
Similar
high liking
scores
Product
Liking
scores
3
10
4
8
6
2
5
9
11
1
7
4.06
4.26
4.62
4.81
4.83
5.45
6.0
6.36
6.57
6.61
6.62
Group
A
A
AB
AB
AB
BC
CD
D
D
D
D
Do products with similar liking scores induce
different emotional responses??
Increased in Liking
Liking data
RESULTS:
C-Defined CATA
PLEASANT
High liking
Added
Sugar
Niche
Added
Sugar
No
Added
Sugar
LOW
ACTIVATION
HIGH
MCA:
UNPLEASANT
Low liking
Multidimensional
EMOTION
MODEL
Engagement
High Activation
Activated Unpleasant Activated Pleasant 450
Larsen & Diener
(1992)
450
Pleasant Unpleasant Unactivated Unpleasant Unactivated Pleasant Low Activation
RESULTS:
C-Defined CATA
MCA:
Desire
Higher
activated
pleasant
emotion
Interested
Added
Sugar
Niche
Added
Sugar
No
Added
Sugar
Higher
activated
unpleasant
emotion
Angry
Discontent
RESULTS:
C-Defined CATA
Q: Do products with similar liking scores induce
different emotional responses?
A: Yes. Key examples:
Low liking scores:
3
4
10
4
3
More
‘Refreshed’
More
‘Attentive’
Frequency count for ‘Refreshed’
High liking scores:
1
5
7
33 consumers
18
9 consumers
11
5
4
3 More
‘Bored’
9
Less ‘Reminiscence’
Less ‘Warm’
Furthermore, consumers with same liking have different
emotional
responses
.
Added
Niche
No
Sugar
Added
Sugar
Added
Sugar
RESULTS:
C-Defined CATA
MCA: Product 7 (Highest Liking)
ID16
Same liking scores of 7
but
Different emotions
More consumers
have selected
pleasant/ positive
emotions
ID33
Key Findings
KEY
FINDINGS
EsSense Profile



C-Defined CATA
Liking data is not enough
Similar product/ emotion plot
Emotions can be explained using degree of
pleasantness and activation as illustrated by
Larsen and Diener (1992)
PROS AND CONS
EsSense Profile
C-Defined CATA

Easy and
straightforward to use


Limited number of
negative emotion
terms


Reported more
negative emotions
Better discrimination
Time consuming and
limited statistical
analysis
OTHER WORK...


Can sensory properties evoke emotions?? YES!!
Emotion data were also collected under different
conditions:
Condition 1:
Product
Condition 2:
Packaging
Condition 3:
Informed
tasting
Take home message
Exploiting
emotional
responses
Products with similar liking scores
is crucial to connect people to brand
can inducevia
different
responses
sensoryemotional
experience
REMINDER: POSTER Measuring emotional response
to food products
Eaton C1, Bealin-Kelly F2 & Hort J1
1Sensory
Science Centre, University of Nottingham, UK.
2SABMiller, Woking, UK.
Poster Board 1, Training room 1.
ACKNOWLEDGEMENT
Supervisors:
Dr Joanne Hort (UoN)
Dr Carolina Chaya (UPM)
Acknowledge:
Dr Ben Lawlor (GSK)
Dr Tracey Hollowood(Sensory Dimensions)
Prof Hal MacFie (Consultant)
& all the volunteers who have taken part in my study!
Sponsor:
SPECIAL THANKS TO:
The PFSG 2011 Student Travel Award
Thank you for your attention
Any Questions?
Contacts
May Ng :
Dr. Joanne Hort :
[email protected]
[email protected]
•
•
Reference:
Larsen, R. J., & Diener, E. (1992). Promises and problems with the ircumplex
model of emotion. Review of Personality and Social Psychology, 13, 25-59.
•
•
Reference:
King, S. C., & Meiselman, H. L. (2010). Development of a method to measure
consumer emotions associated with foods. Food Quality and Preference, 21 (2),
168-177
Multiple Correspondence Analysis


Popular technique to explore the relationships among
multiple categorical variable
Data input: Individual CATA questions and 11 products
Principal Component Analysis
Methods & Materials
C-defined CATA
Process of terms generation & selection
 Stage1: Terms elicitation
- 1:1 Rep Grid Interview (n=29)
- Subjects presented with triads of products

StageRefreshing,
2: Defining constructs
Please indicate
how two products
differ in the same
way from the third
constructs
-Everypremium,
subject
has his/ her own list of
guilty
pleasure
etc
-Product
assessment
using individual constructs
etc
289

365
972
Selection of terms based on word frequency table
109