Page 1
Using Implicit Measures in Attitude and Personality Research
Wilhelm Hofmann
University of Chicago
Booth School of Business
1
SPSP 2012 GSC and Training Committee Innovative Methods pre‐conference
Overview
I. What are implicit measures useful for?
II. Conceptual approaches to the use of implicit measures
III. The Implicit Measurement Zoo: Which procedure to pick?
IV. Resources
2
Page 2
Mental operations
I) What are implicit measures useful for?
Aspects of attitudes and traits that are difficult to assess via self‐report due to
• Introspective limits
• Self‐presentational concerns
3
Mental experience
≠
The Idea behind Implicit Measurement
Mental Associations
Thinking, Feeling, Behavior
4
e.g., implicit stereotypes,Implicit prejudiceImplicit self‐esteem…
Page 3
Implicit Measurement
Measurement Procedure
Mental Associations
Task Responses
Measurement Outcome
Inference
Automatic Process
Person
5
De Houwer & Moors, 2010
Link to Dual‐System Theories
6
System 1Associative Processing
System 2 Propositional Reasoning
• Fast, intuitive, habitual response generation
• Schemas and scripts• Basic affective and motivational
orientations
• Slow, deliberate, rule‐based• Judgments, decisions, intentions• Reasoned, planned behavior
More on dual‐systems/processes: Evans, 2008, ARP; Strack & Deutsch, 2004, PSPR; Gawronski & Bodenhausen, 2006, PB; Smith & DeCoster, 2000, PSPR; Sherman, Gawronski, & Trope, in prep., Guilford Press
Page 4
Example: Implicit Gender Stereotypes
• Are hiring decisions and salary offers influenced by implicit gender stereotypes?
• Procedure: Implicit Association Test (IAT)(Greenwald, McGhee, & Schwartz, 1998, JPSP)
• Demonstration (not included in this slideshow): Clear the desk in front of you and prepare to tap on your desk
7
The Traditional IAT(Greenwald et al., 1998)
Block N trials Task Left keyassignment
Right keyassignment
1 20 Target discrimination FEMALE MALE
2 20 Attribute discrimination Career Family
3 20 Initial combined block (p) FEMALE, Career MALE, family
4 40 Initial combined block (t) FEMALE, Career MALE, family
5 20 Reversed target discrimination MALE FEMALE
6 20 Reversed combined block (p) MALE, Career FEMALE, family
7 40 Reversed combined block (t) MALE, Career FEMALE, family
8
Note. p = originally denoted “practice” block; t = “test” block
Page 5
Career Family
Left Right
Career Family
Left Right
FEMALE MALE
MALE FEMALE
9
„Incompatible“ Block
„Compatible“ Block
Difference → IAT Effect
difficult & slow
easy & fast
IAT Effect
0
100
200
300
400
500
600
700
800
Female + Career& Male + Family
Male + Career& Female + Family
Average Block Reaction Tim
es
10
‐200
‐100
0
100
200
IAT Effect
Relative Preference
Female+Career
Male+Career
For a more sophisticated scoring algorithm (D‐Score), see Greenwald et al., 2003, JPSP
Page 6
Some Variations of Implicit Constructs (as measured with the IAT procedure)
11
Implicit Stereotypes
MALECareer
FEMALEFamily
Implicit Prejudice
YOUNGgood
ELDERLYbad
Brand Attitudes
COKEgood
PEPSIbad
Implicit Self‐Esteem
MEgood
OTHERSbad
Implicit Self‐Concept
MEangry
OTHERScalm
Political Attitudes
Al Goregood
Bushbad
Social Attitudes, Group Research
Consumer, Health, Self‐Regulation, Political etc.
Self‐Esteem, Personality Self‐Concept
II. Conceptual Approaches
• Implicit Measure as Outcome (DV)
– Universal attitudes
– Known groups approach
– Experimental manipulations
– Method‐specific effects
• Implicit Measure as Predictor (IV)
– Relation between Implicit and Explicit Cognition
– Behavior Prediction
12
Page 7
Conceptual Approaches
Implicit Measure as DV
– Universal attitudes
– Known groups approach
– Experimental manipulations
– Method‐specific effects
13
Research Example
• Greenwald et al., 1998, JPSP
ImplicitAttitude
‐200
‐100
0
100
200
IAT Effect
Implicit Attitude
Flower+ good
Insect+ good
mean IAT effects should not be over‐interpreted!
Conceptual Approaches
Implicit Measure as DV
– Universal attitudes
– Known groups approach
– Experimental manipulations
– Method‐specific effects
14
Research Example
• Snowden et al., 2003, Archives of Sexual Behavior
ImplicitAttitude
‐200
‐100
0
100
200
HeterosexualMen
HomosexualMen
Implicit Attitude
Women+ attractive
Men+ attractive
Page 8
Conceptual Approaches
Implicit Measure as DV
– Universal attitudes
– Known groups approach
– Experimental manipulations/Interventions
– Method‐specific effects
15
Research Examples
• Olson & Fazio, 2001, Psych Science
• Hollands et al., 2001, Health Psych
• Wittenbrink et al., 2001, JPSP
ImplicitAttitude
‐0.5
0
0.5
CS1+posUSCS2+negUS
CS1+negUSCS2+posUS
Implicit Attitude
CS1+ positive
CS2+ positive
‐Evaluative Conditioning
‐Context Effects Olson & Fazio, 2001
Conceptual Approaches
Implicit Measure as DV
– Universal attitudes
– Known groups approach
– Experimental manipulations
– Method‐specific effects
16
Research Examples
• Mierke & Klauer, 2003, JPSP
• Rothermund & Wentura, 2004, JEP:G
• Bluemke & Friese, 2006, JESP
• Steffens, 2004, Exp Psych
ImplicitConstruct
construct‐ related variance
method‐specific variance
error variance
Variance Decomposition
Cognitive abilitiesStimulus SalienceStimulus SelectionStrategies (e.g., faking)...
Measurement Process
Page 9
Conceptual ApproachesImplicit Measure as “IV”
– Relation between Implicit and Explicit Cognition
– Behavior Prediction
17
Research Examples
• Brown & Ryan, 2003, JPSP
• Ranganath et al., 2005, JESP
• Hofmann et al., 2005, PSPB
• Gawronski et al., 2007, JESP
• Koole et al., 2001, JPSP
ExplicitConstruct
ImplicitConstruct
Accessibility/StrengthIntuition/Affective focusValidation/EndorsementSpontaneity/Time pressure
Implicit‐Explicit Correlation
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low implicitself‐esteem
High implicitself‐esteem
Explicit Self‐Esteem
No time pressure
Time pressure
Koole et al., 2001
Explicit report
Conceptual ApproachesImplicit Measure as “IV”
– Relation between Implicit and Explicit Cognition
– Behavior Prediction
18
Research Examples
• Dunton & Fazio, 1997, PSPB
• Nier, 2005, GPIR
• Payne et al., 2005, JPSP
• …
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Low AMP score‐
High AMP score+
Explicit Attitude
High concern with acting prejudiced
Low concern with acting prejudiced
+
‐
Payne et al., 2005
ExplicitConstruct
ImplicitConstruct
Accessibility/StrengthIntuition/Affective focusValidation/EndorsementSpontaneity/Time pressure
Implicit‐Explicit Correlation
Self‐Presentation
Page 10
Conceptual ApproachesMore complex cases of
Implicit & Explicit Attitude Change
19
• Gawronski & Bodenhausen, 2006, Psych Bull (review)
ExplicitAttitude
ImplicitAttitude
Influence
→APE‐Model
Conceptual ApproachesImplicit Measure as IV
– Relation between Implicit and Explicit Cognition
– Behavior Prediction
20
Research Examples
• Egloff & Schmukle, 2002, JPSP
• Payne et al., 2008, Cog. & Emo.
• Back et al., 2009, JPSP
• Greenwald et al., 2009, JPSP (meta‐analysis)Incremental Validity Approach
ExplicitConstruct
ImplicitConstruct
Behavioral Outcome
Page 11
Conceptual ApproachesImplicit Measure as IV
– Relation between Implicit and Explicit Cognition
– Behavior Prediction
21
Research Examples
• Egloff & Schmukle, 2002, JPSP
• Payne et al., 2008, Cog. & Emo.
• Back et al., 2009, JPSP
• Greenwald et al., 2009, JPSP (meta‐analysis)
Incremental Validity
Conceptual ApproachesImplicit Measure as IV
– Relation between Implicit and Explicit Cognition
– Behavior Prediction
22
Research Examples
• Dovidio et al., 1997, JPSP
• Asendorpf et al., 2001, JPSP
ControlledBehavior
AutomaticBehavior
ExplicitConstruct
ImplicitConstruct
Asendorpf et al., 2005
Double‐Dissociation Model
Page 12
Conceptual ApproachesImplicit Measure as IV
– Relation between Implicit and Explicit Cognition
– Behavior Prediction
23
Research Examples
• Hofmann et al., 2007, JESP
• Conner et al., 2007, PSPB
• Friese et al., 2008, ERSP (review)
ExplicitConstruct
ImplicitConstruct
Moderated Predictive Validity
Behavioral Outcome
Control ResourcesControl Motivation
Process Reliance...
Conceptual ApproachesImplicit Measure as IV
– Relation between Implicit and Explicit Cognition
– Behavior Prediction
24
Research Examples
• Hofmann et al., 2007, JESP
• Conner et al., 2007, PSPB
• Friese et al., 2008, ERSP (review)
Moderated Predictive Validity
-1.5
-1
-0.5
0
0.5
1
1.5
low (-1 SD) high (+1 SD)
Implicit Attitude Measure
Can
dy C
onsu
mpt
ion
Depletion No Depletion
-1.5
-1
-0.5
0
0.5
1
1.5
low (-1 SD) high (+1 SD)
Restraint Standards
Can
dy C
onsu
mpt
ion
Depletion No Depletion
*
*
Page 13
Conceptual ApproachesMore complex case:
Implicit‐Explicit Consistency as IV
25
Research Examples
Attitudes:
• Brinol et al., 2006, JPSP
Self‐esteem:
• Jordan et al., 2003, JPSP
• Zeigler‐Hill, 2006, JPSP
Intelligence self‐concept:
• Dislich et al., 2012, EJPExplicit
Construct
ImplicitConstruct
DV
Inter‐action
Hot and under‐researched avenues
– Correspondence between different implicit procedures (e.g., Bosson et al., 2000, JPSP; Gawronski & Bodenhausen, 2005, JPSP)
– Development and long‐term change of implicit constructs (e.g., Dunham, Baron, & Banaji, 2008, TiCS)
– Neural and physiological correlates (e.g., Cunningham et al.,
2003, JPSP)
– Process‐dissociation approaches (Conrey et al., 2005, JPSP)
– Extensions to new applied areas such as law, politics, clinical & health, etc. Era of Application
26
Page 14
III. The Implicit Measurement Zoo27
• Priming and related measures– Sequential Conceptual Priming (LDT)
– Sequential Evaluative Priming (EP)
– Affect Misattribution Procedure (AMP)
• IAT and variants– Implicit Association Test (IAT)
– Single‐Category IAT (SC‐IAT)
– Brief Implicit Association Test (BIAT)
– Recoding Free IAT (IAT‐RF)
– Single‐Block IAT (SB‐IAT)
– Go/No‐Go Association Task (GNAT)
• Extrinsic Affective Simon Task (EAST)
• Approach‐Avoidance Measures– Approach‐Avoidance Task (AAT)
– Implicit Association Procedure (IAP)
– Evaluative Movement Assessment (EMA)
– Stimulus Response Compatibility Task (SRCT); “Manikin‐Task”
• Paper and Pencil Measures– Name‐Letter Task (NLT)
– Linguistic Intergroup Bias (LIB)
– Breadth‐based Adjective Rating Task (BART)
– Stereotypic Explanatory Bias (SEB)
– Paper & Pencil IAT
...
The Implicit Measurement Zoo:Which procedure to pick?
• Depends a lot on what you want to measure & on your constraints
– Does it make theoretical sense to include an implicit measure in your research?
– Is focus on experimental manipulations/mean differences (“I as DV”) or on correlational/predictive research (“I as IV”)?
• issue of reliability
– Does the procedure allow you to appropriately represent the construct you are interested in?
• e.g., absolute vs. relative comparison
– Computer‐based or paper & pencil?
– Time constraints?• are briefer options available?
28
Page 15
Selected Procedures: Pros and Cons
• Priming
– Concept Priming and Evaluative Priming
– Affect Misattribution Procedure (AMP)
• Traditional Implicit Association Test
– Some problems with the traditional IAT and their suggested solutions
• Single‐Category IAT
• Personalized IAT
• Single‐Block IAT
29
Career
30
Word Nonword
75 ms
125 ms
response
Sequential Concept Priming(e.g., Banaji & Hardin, 1996; Wittenbrink et al., 1997)
Prime
Target
Page 16
31
Word Nonword
Career
Word Nonword
Career
Word Nonword
Family
Word Nonword
Family
Supposed mechanism: spreading activation in associative semantic network
Sequential Concept Priming
32
Word Nonword
Career
Word Nonword
Career
Word Nonword
Family
Word Nonword
Family
Stereotyping index = (RT(malefamily) – RT(male career) + RT(femalecareer) – RT(femalefamily))/2
(For further details on different scoring indices, see Wittenbrink, 2007)
Sequential Concept Priming
slow slowfast fast
Page 17
Positive Negative
Peace
Evaluative Priming(e.g., Fazio et al., 1986; 1995)
33
Positive Negative
Peace
Main difference between concept and evaluative priming: Affective decision instead of lexical decision (→response competition)
(For further details on priming measures, see Wentura & Degner, 2010; Wittenbrink, 2007)
Positive Negative
Illness
Positive Negative
Illness
Sequential Priming: Pros and Cons
Pros
• very unobtrusive
• optional subliminal prime presentation
• allows both absolute and relative comparisons
34
Cons
• time‐intensive
• complicated (especially with regard to indices)
• relatively small effects
• very low reliability → not very suitable for correlational research (implicit as IV)
Good alternative to evaluative priming: AMP
Page 18
Affect Misattribution Procedure (AMP) (Payne, Cheng, Govorun, Stewart, 2005, JPSP)
35
unpleasant pleasant
75 ms125 ms
100 ms
response
NegativePrime
Target
Affect Misattribution Procedure36
unpleasant pleasant
75 ms125 ms
100 msresponse
PositivePrime
Page 19
Affect Misattribution Procedure37
unpleasant pleasant
75 ms125 ms
100 msresponse
Supposed Mechanism: misattribution of activated affect to judgment of ambivalent target
DV = percentage positive responses to target stimuli when preceded by prime of interest
Prime of interest
AMP: Pros and Cons
38
Pros
• easy to implement
• allows both absolute and relative comparisons
• good reliability. Suitable for correlational research (“implicit as IV”)
• promising findings regarding incremental validity
Cons
• mechanism not well understood yet
• sometimes very large overlap with explicit measures (possibly more “explicit” than other implicit measures)
Page 20
The Traditional IAT(Greenwald et al., 1998)
Block N trials Task Left keyassignment
Right keyassignment
1 20 Target discrimination FEMALE MALE
2 20 Attribute discrimination Career Family
3 20 Initial combined block (p) FEMALE, Career MALE, family
4 40 Initial combined block (t) FEMALE, Career MALE, family
5 20 Reversed target discrimination MALE FEMALE
6 20 Reversed combined block (p) MALE, Career FEMALE, family
7 40 Reversed combined block (t) MALE, Career FEMALE, family
39
Note. p = originally denoted “practice” block; t = “test” block
The traditional IAT: Pros and Cons
Pros
• high reliability (both internal consistency and retest)
• ease of administration
• very well‐researched
• clear indication of incremental validity
40
Cons
• relative comparison measure
• various sources of unwanted method‐specific variance identified
Page 21
Some IAT Problems and Suggested Remedies
Problem D‐Score Algorithm
Single CategoryIAT
Personalized IAT
Single Block IAT
Cognitive Skill Confounds
x x
Order Effects x x
Relative Comparison
x
Extra‐personal Associations
x
RecodingStrategies
x
41
D‐Score Algorithm(Greenwald et al., 2003, JPSP)
42
Substantially reduces:• Cognitive ability confounds (e.g., task‐switching; Klauer & Mierke, 2003)
• Compatibility order effects (compatible block first produces larger IAT scores than vice versa)
SPSS and SAS scripts available at: http://faculty.washington.edu/agg/iat_materials.htm
Page 22
Single‐Category IAT(Karpinski & Steinman, 2006, JPSP)
Addressed Problem: relative nature of IAT– Hard to know what specific association drives an IAT effect
– Sometimes a suitable comparison part is lacking
Solution: only one Target Category, balancing of number of trials
43
SPIDERnegative positive negative
SPIDERpositive
“compatible” block “incompatible” block
Personalized IAT(Olson & Fazio, 2004, JPSP)
Addressed Problem: is IAT influenced by extra‐personal associations?
Solution (Olson & Fazio, 2004): – exchange “positive” and “negative” attribute category labels with
more personalized ones (“I like”; “I dislike”)
44
Candy-bars“I like”
Apples“I dislike”
Apples“I like”
Candy-bars“I dislike”
“compatible” block “incompatible” block
For further discussions, see Nosek & Hansen, 2008, EJPA; Gawronski et al., 2008, SPPC
Page 23
Single‐Block IAT(Teige-Mocigemba et al., 2008, EJPA)
Eliminates block structure of the IAT:• thus, no compatibility order effects• reduces cognitive skill confounds• disables participant recoding strategies
45
Football
negative
Baseball
Baseball
positive
Football
Stimulus presentation
Outlook:Process Dissociation Approaches
• No implicit measure is process‐pure
• Process Dissociation (PD) approaches separate multiple processes that determine responses (e.g., error rates in the IAT)– Simple PD Approach (Jacoby, 1991; Payne, 2001):
automatic and controlled process
– QUAD model (Conrey et al., 2005; Sherman et al., 2008) ‐ 4 parameters:Automatic Activation, Stimulus Discrimination, Overriding Bias, Guessing
• For more information, seeSherman, J. W., Klauer, K. C., & Allen, T. J. (2010). Mathematical modeling of implicit social cognition: The machine in the ghost. In B. Gawronski & B. K. Payne (Eds.), Handbook of implicit social cognition (pp. 156‐175). New York: Guilford.
Tutorial etc.: http://psychology.ucdavis.edu/labs/sherman/site/research.html
46
Page 24
IV. Resources
• Hardware
• Programming Software
– Inquisit
– DirectRT
– Eprime
• Project Implicit®
• Recommended Books and Hands‐On Chapters
47
Hardware and Software48
• Hardware: any modern computer will do
• Software– Inquisit (Millisecond): http://www.millisecond.com
– DirectRT (Empirisoft): http://www.empirisoft.com
– Eprime (Psychology Software Tools, Inc.): http://www.pstnet.com/eprime
– FreeIAT: http://www4.ncsu.edu/~awmeade/FreeIAT/FreeIAT.htm
Page 25
Project Implicit®(www. https://implicit.harvard.edu)
• IAT demonstrations
• Background information
49
Resources: Books on Implicit Measures
Handbook of implicit social cognition: Measurement, theory, and applications
Bertram Gawronski, & Keith Payne (Eds.)2010. Guilford Press.
50
Implicit Measures of Attitudes
Bernd Wittenbrink & Norbert Schwarz (Eds.)2007. Guilford Press.
Attitudes: Insights from the new implicit measures
Richard Petty, Russell Fazio, & Pablo Briñol (Eds.)2008. Psychology Press.
Page 26
Practical Hands‐On Chapters
IAT (and its variants)• Teige‐Mocigemba, S., Klauer, K. C., & Sherman, J. W. (2010). Practical guide to
Implicit Association Test and related tasks. In B. Gawronski & B. K. Payne (Eds.), Handbook of Implicit Social Cognition (S. 117‐139). New York: Guilford.
• Lane, K. A., Banaji. M. R., Nosek. B. A., & Greenwald, A. G. (2007). Understanding and using the Implicit Association Test: IV: What we know (so far). In B. Wittenbrink & N. S. Schwarz (eds.) Implicit Measures of Attitudes: Procedures and Controversies (59‐102). New York: Guilford.
Priming and AMP• Wentura, D. & Degner, J. (2010). A Practical Guide to Sequential Priming and
Related Tasks. In B. Gawronski, & B. K. Payne (Eds.), Handbook of implicit social cognition (pp. 95‐116). New York: Guilford.
Paper and Pencil• Sekaquaptewa, D., Vargas, P., & von Hippel, W. (2010). A practical guide to
paper and pencil implicit measures of attitudes. In B. Gawronski & B. K. Payne (Eds.), Handbook of Implicit Social Cognition (pp. 140‐155). New York: Guilford.
51