Why a standard IAT effect cannot provide evidence for ...€¦ · Web viewFurthermore, the...

Why a standard IAT effect cannot provide evidence for association formation: The role of

similarity construction

Karoline Bading1, Christoph Stahl1, & Klaus Rothermund2

1 University of Cologne

2 Friedrich Schiller University Jena

Author note

Correspondence concerning this article should be addressed to Karoline Bading, Herbert-

Lewin-Straße 2, Raum 3.33, 50931 Köln. E-mail: [email protected]

mailto:[email protected]

Why a standard IAT effect cannot provide evidence for association formation: The role of

similarity construction

Appendix A

Experiment 1

Figure 1. Experiment 1: Mean evaluative ratings as a function of US Valence, CS-US Relation,

and IAT Type. Error bars represent 95% within-subjects confidence intervals.

The evaluative ratings of the four families were submitted to a 2 (US Valence) × 2 (CS-

US Relation) × 2 (IAT Type) mixed-design ANOVA. The between-subjects factor IAT Type

yielded a marginally significant main effect, F (1,62)=3.63, MSE=3.50, p=.061, η¿

G2 =.015, but

did not enter any significant interaction, all ps ≥ .166. The main effect reflected higher evaluative

ratings in the “valence of action” group (M = 4.01, SD = 2.41) than in the “valence of sound”

Evaluative Ratings

group (M = 3.56, SD = 2.48). Furthermore, we found a replication of the results by Moran and

Bar-Anan (2013), i.e. significant main effects of US Valence, F (1,62)=66.44, MSE=2.24,

p<.001, η¿

G2 =.150, and of CS-US Relation, F (1,62)=14.08, MSE=2.72, p<.001, η

¿

G2 =.043, as

well as a significant interaction of the two factors, F (1,62)=93.79, MSE=5.13, p<.001,

η¿

G2 =.364. This interaction reflected the fact that the direction of the simple main effect of US

Valence depended on the level of CS-US Relation. Among “starting” families, the “melody”

family (M = 5.53, SD = 1.77) was preferred over the “scream” family (M = 1.27, SD = 1.78),

F (1,63)=161.42, MSE=3.61, p<.001, η¿

G2 =.594. Among “ending” families, the “scream” family

(M = 4.78, SD = 1.99) was preferred over the “melody” family (M = 3.56, SD = 1.84),

F (1,63)=12.99, MSE=3.66, p=.001, η¿

G2 =.093. Furthermore, the “starting melody” family was

preferred over the “ending melody” family, F (1,63)=31.90, MSE=3.89, p<.001, η¿

G2 =.232,

while the “ending scream” family was preferred over the “starting scream” family,

F (1,63)=100.47, MSE=3.94, p<.001, η¿

G2 =.468 (see Figure 1).

For participants in the “VoS” group, we computed Do scores in such a way that positive

values reflected faster stimulus classification when the “melody” family shared a key with

positive attributes, and the “scream” family shared a key with negative attributes (i.e., a

preference for the “melody” family over the “scream” family).

The mean Do score for the “starting” IAT was significantly greater than zero, t (31)=4.10,

p<.001 (one-tailed), indicating a preference for the “starting melody” family over the “starting

Overall IAT effects“VoS”-group

scream” family (M = 0.19, SD = 0.27). The same was true for the “ending” IAT: the mean Do

score was also significantly greater than zero, t (31)=1.98, p=.028 (one-tailed), indicating a

preference for the “ending melody” family over the “ending scream” family (M = 0.15, SD =

0.43). In a paired-samples t-test no difference between the mean Do scores of the two IATs was

found, M d=0.04, 95% CI ¿, 0.24 ¿, t (31)=0.45, p=.658.

For participants in the “VoA” group, we computed Do

scores in such

a way that positive values reflected faster stimulus classification when the family performing the

positive action (“starting melody” family in the “melody” IAT, “ending scream” family in the

“scream” IAT) shared a key with positive attributes, and the family performing the negative

action (“ending melody” family in the “melody” IAT, “starting scream” family in the “scream

IAT”) shared a key with negative attributes (i.e., a preference for the “positive action” family

over the “negative action” family). The mean Do score for the “melody” IAT as well as the mean

Do score for the “scream” IAT were significantly greater than zero. The positive mean Do score

indicates a preference for the “starting melody” family over the “ending melody” family in the

“melody” IAT (M = 0.24, SD = 0.51), t (31)=2.63, p=.007 (one-tailed), and a preference for the

“ending scream” family over the “starting scream” family in the “scream” IAT (M = 0.16, SD =

0.41), t (31)=2.29, p=.015 (one-tailed). In a paired-samples t-test no difference between the

mean Do scores of the two IATs was found, M d=0.07, 95% CI ¿, 0.25¿, t (31)=0.83, p=.411.

“VoA”-group

Experiment 2

Figure 2. Experiment 2: Mean evaluative ratings as a function of US Valence and CS-US

Relation, and IAT Type. Error bars represent 95% within-subjects confidence intervals.

The evaluative ratings of the four creature families were submitted to a 2 (US Valence) ×

2 (CS-US Relation) × 2 (IAT Type) mixed-design ANOVA. Once again, significant main effects

of US Valence, F (1,117)=71.92, MSE=3.19, p<.001, η¿

G2 =.096, and of CS-US Relation,

F (1,117)=5.44 , MSE=4.47, p=.021, η¿

G2 =.011, as well as a significant interaction of the two

factors, F (1,62)=93.79, MSE=5.13, p<.001, η¿

G2 =.364, were found. The factor IAT Type did not

yield a main effect or enter any significant two- or three-way interactions, all ps ≥ .150. The

interaction of US Valence and CS-US Relation followed the same pattern as in experiment 1.

Among “starting” families, the “melody” family (M = 6.46, SD = 1.83) was preferred over the

“scream” family (M = 1.92, SD = 2.32), F (1,119)=248.40, MSE=4.98, p<.001, η¿

G2 =.543.

Among “ending” families, the “scream” family (M = 5.53, SD = 2.22) was preferred over the

Evaluative Ratings

“melody” family (M = 3.75, SD = 2.21), F (1,119)=36.20, MSE=5.22, p<.001, η¿

G2 =.139.

Furthermore, the “starting melody” family was preferred over the “ending melody” family,

F (1,119)=96.20, MSE=4.57, p<.001, η¿

G2 =.309, while the “ending scream” family was

preferred over the “starting scream” family, F (1,119)=112.92, MSE=6.92, p<.001, η¿

G2 =.389

(see Figure 2).

For participants in the “VoS” group, we computed Do scores in such a way that positive

values reflected faster stimulus classification when the “melody” families shared a key with

positive attributes, and the “scream” families shared a key with negative attributes (i.e., a

preference for the “melody” families over the “scream” families).

The mean Do score in the VoS group was significantly greater than zero, t (39)=4.47,

p<.001 (one-tailed), indicating a preference for the “melody” families over the “scream” families

(M = 0.38, SD = 0.54).

For participants in the “VoA” group, we computed Do

scores in such

a way that positive values reflected faster stimulus classification when the families performing

the positive actions (“starting melody” and“ending scream”) shared a key with positive attributes,

and the families performing the negative actions (“ending melody” and“starting scream”) shared

a key with negative attributes (i.e., a preference for the “positive action” families over the

“negative action” families).

Overall IAT effects

“VoS”-group

“VoA”-group

The mean Do score in the VoA group was significantly greater than zero, t (39)=3.50,

p=.001 (one-tailed), indicating a preference for the “positive action” families over the “negative

action” families (M = 0.24, SD = 0.44).

For participants in the “VoR” group, we computed Do

scores in such

a way that positive values reflected faster stimulus classification when the “starting” families

shared a key with positive attributes, and the “ending” families shared a key with negative

attributes (i.e., a preference for the “starting” over the “ending” families).

The mean Do score in the VoR group was significantly greater than zero, t (39)=1.97,

p=.028 (one-tailed), indicating a preference for the “starting” families over the “ending” families

(M = 0.13, SD = 0.43).

Appendix B

“VoR”-group

Order effects in Experiment 1

Figure 3. Experiment 1: Family-specific D scores in the ‘VoS’ group as a function of US

Valence, CS-US Relation, and IAT Order. Error bars represent 95% confidence intervals.

A 2 (US Valence) × 2 (CS-US Relation) × 2 (IAT Order: starting vs. ending IAT first)

mixed-design ANOVA on the Df scores in the “VoS” group revealed a significant main effect of

US Valence, F (1,30)=13.22, MSE=0.32, p=.001, η¿

G2 =.124. All other main effects and

interactions did not reach significance, all ps ≥ .095. This suggests that the preference for the

“melody” over the “scream” family (i.e., the simple main effects of US Valence) was comparable

across both levels of CS-US Relation (i.e., the “starting” vs. “ending” IATs), and furthermore,

that the order in which the IATs were performed did not affect the size of this preference.

Descriptively, the simple main effect of US Valence in the “starting” IAT was larger when this

was the second one to be performed. For the “ending” IAT the pattern was reversed: the simple

main effect of US Valence was larger when the “ending” IAT was performed first (see Figure 1).

Order Effects in the “VoS” group (Exp. 1)

Figure 4. Experiment 1: Family-specific D scores in the ‘VoA’ group as a function of US

Valence, CS-US Relation, and IAT Order. Error bars represent 95% confidence intervals.

A 2 (US Valence) × 2 (CS-US Relation) × 2 (IAT Order: melody vs. scream IAT first)

mixed-design ANOVA on the Df scores in the “VoA” group revealed a significant interaction of

US Valence and CS-US Relation, F (1,30)=4.73, MSE=0.69, p=.038, η¿

G2 =.071. All other main

effects and interactions did not reach significance, all ps ≥ .107. This suggests that the preference

for the “starting” (“ending”) over the “ending” (“starting”) family in the “melody” (“scream”)

IAT was not affected by the order in which these two IATs were performed. Descriptively, the

preference for the “starting” over “ending” family in the “melody” IAT was larger when the

“melody” IAT was performed first. The same was true for the “scream” IAT: the preference for

the “ending” over the “starting” family was larger when the “scream” IAT was performed first

(see Figure 2).

Order Effects in the “VoA” group (Exp. 1)

Appendix C

Exemplary IAT instructions

You will now be presented with images of the yellow and green creatures.

Your task is to classify these images according to their family membership.

Press space bar to continue.

For classifying the creatures, please use the ‘c’ and ‘m’ key on the computer keyboard.

From now on, we will refer to the ‘c’ key as the ‘left key’ and to the ‘m’ key as the ‘right

key’.

Press the left key when an image of a yellow creature appears.

Press the right key when an image of a green creature appears.


When classifying the creatures, please react as fast as possible!

At the same time, please try making as few mistakes as possible!


Experiment 1Screen 1

Screen 2

Screen 3

When you make a mistake, a red ‘x’ will appear below the image of the creature.

When you see this red ‘x’, press the correct key as fast as possible in order to continue the

task.


The correct key assignment will be shown to you throughout the entire task.

In the upper left corner you will see for which creatures you have to press the left key

(“yellow creatures”).

In the upper right corner you will see for which creatures you have to press the right key

(“green creatures”).


Please now place your two index fingers on the two keys (‘c’ and ‘m’).

In order to react faster, leave your index fingers on the keys throughout the entire task.

Press space bar to start the task.

Screen 4

Screen 5

Screen 6 (followed by target practice block)

Screen 7 (followed by attribute practice block)

Your next task is to classify words as ‘positive’ or ‘negative’.

When a negative word appears, press the left key.

When a positive word appears, press the right key.

When you make a mistake, a red ‘x’ will appear.

When you see this red ‘x’, press the correct key as fast as possible.


Your next task is to perform both tasks “simultaneously”.

On every trial, you will see an image of a creature or a word.

Press the left key when an image of a yellow creature or a negative word appears.

Press the right key when an image of green creature or a positive word appears.



In order to practice this “double task” you will receive a short practice block.

Press space bar to begin.

You have completed the practice block!

Screen 8 (followed by combined practice block)

Screen 9 (followed by main block)

You will now receive a longer block in which you will perform the previously practiced

“double task”.


You have completed the first “double task”!

Your next task is to again classify the yellow and green creatures according to their family

membership.

However, the correct key assignment is now reversed!

Press the left key when an image of a green creature appears.

Press the right key when an image of a yellow creature appears.


You will now have to perform another “double task”!

Press the left key when the image of green creature or a negative word appears.

Press the right key when the image of a yellow creature or a positive word appears.



Screen 10 (followed by 2nd target practice block)

Screen 11 (followed by 2nd combined practice block)





“double task”.


You will now be presented with images of the yellow, green, pink and brown creatures.

Your task is to classify these images according to their family membership.


For classifying the creatures, please use the ‘c’ and ‘m’ key on the computer keyboard.

From now on, we will refer to the ‘c’ key as the ‘left key’ and to the ‘m’ key as the ‘right

key’.

Press the left key when an image of a pink or a yellow creature appears.

Screen 12 (followed by 2nd main block)

Experiment 2Screen 1

Screen 2

Press the right key when an image of a green or a brown creature appears.


When classifying the creatures, please react as fast as possible!

At the same time, please try making as few mistakes as possible!


When you make a mistake, a red ‘x’ will appear below the image of the creature.

When you see this red ‘x’, press the correct key as fast as possible in order to continue the

task.


The correct key assignment will be shown to you throughout the entire task.

In the upper left corner you will see for which creatures you have to press the left key

(“pink and yellow creatures”).

In the upper right corner you will see for which creatures you have to press the right key

(“green and brown creatures”).

Screen 3

Screen 4

Screen 5


Please now place your two index fingers on the two keys (‘c’ and ‘m’).

In order to react faster, leave your index fingers on the keys throughout the entire task.


Your next task is to classify words as ‘positive’ or ‘negative’.

When a negative word appears, press the left key.

When a positive word appears, press the right key.




Your next task is to perform both tasks “simultaneously’.

On every trial, you will see an image of a creature or a word.

Press the left key when an image of pink or a yellow creature or a negative word appears.

Screen 6 (followed by target practice block)

Screen 7 (followed by attribute practice block)

Screen 8 (followed by combined practice block)

Press the right key when an image of green or a brown creature or a positive word

appears.







“double task”.


You have completed the first “double task”!

Your next task is to again classify the yellow, green, pink and brown creatures according

to their family membership.

However, the correct key assignment is now reversed!

Press the left key when an image of a green or a brown creature appears.

Screen 9 (followed by main block)

Screen 10 (followed by 2nd target practice block)

Press the right key when an image of a pink or a yellow creature appears.


You will now have to perform another “double task”!

Press the left key when the image of green or a brown creature or a negative word appears.

Press the right key when the image of pink or a yellow creature or a positive word

appears.







“double task”.


Screen 11 (followed by 2nd combined practice block)

Screen 12 (followed by 2nd main block)

Why a standard IAT effect cannot provide evidence for ...€¦ · Web viewFurthermore, the...

Documents

Transcript of Why a standard IAT effect cannot provide evidence for ...€¦ · Web viewFurthermore, the...