Running head: EXPERIMENTAL INDUCTION OF HABITS
1
Supplemental Materials
Determining the effect of training duration on the behavioral expression of habitual
control in humans: A multi
-
laboratory investigation
by
E. R. Pool
*
, R. Gera, A. Fransen
*
,
O. D. Perez*,
A. Cremer
*
, M. Aleksic
*
,
S.
Tanwisuth
*
, S.
Quail
*
, A
.
O. Ceceli, D
.
A. Manfredi, G
.
Nave, E
.
Tricomi, B
.
Balleine,
T
.
Schonberg, L
.
Schwabe
, and
J
.
P. O'Doherty
Supplemental Analysis of the moderating effects of individual differences relating to
stress, anxiety and impulsivity on habit formation
as a function of training duration
Strategy 2: Extracting non
-
collinear factors
This analytical approach aims at extracting non
-
collinear factors that could be
later entered simultaneously as predictors in the same statistical model testing the effect
o
f training on devaluation sensitivity. This approach has the advantage of testing the
effect of one factor while controlling the variance explained by the other factors but it
might also be prone to inflate significant effects.
Factor Analysis.
We ran an e
xploratory factorial analysis (EFA) using
maximum likelihood estimation on the standardized subscales of the questionnaires (13
subscales in total). We used the package Psych (Revelle
,
2017) with an orthogonal
rotation (varimax). The “Parallel analysis”
method suggested a 4 factors solution to our
data. We derived the factors loadings using a regression method, the validity coefficient
(
R
2
= 0.92, 0.90, 0.88, 0.86) assessing the potential impact of factor sore indeterminacy
(Grice
,
2001) was sufficient fo
r deriving the scores from the EFA.
EXPERIMENTAL INDUCTION OF HABITS
2
For the factor labeling, we labeled the first factor “Stress work”, since the
higher loadings were related to high demands at work and a high workload. We labeled
the second factor “Stress social” since all the higher lo
adings are related to social high
demands (pressure to perform, social tensions, social overload) as well as lack of social
positive events (lack of social recognition). We labeled the third factor “Stress Affect”,
since the higher loadings on this factor
are associated with the presence of negative
affective feelings associated with stress (anxiety, worries, discontent) and the lack of
affective support (social isolation).
Table S1
.
Loading onto Factor 1 “Stress Work”, Factor 2 “Stress Social”, Factor 3
“
Impulsivity” and Factor 4 “Stress Affect”
Stress Work
Stress Social
Impulsivity
Stress Affect
Anxiety composite score
0.23
0.17
0.17
0.65
BIS attentional
0.16
0.09
0.46
0.34
BIS motor
-
0.01
0.08
0.59
0.06
BIS non planning
0.03
-
0.02
0.84
0.07
TICS chronic worrying
0.41
0.31
0.05
0.55
TICS excessive demands at work
0.84
0.25
0.16
0.38
TICS lack of social recognition
0.30
0.58
0.06
0.33
TICS pressure to perform
0.28
0.72
-
0.05
0.32
TICS social isolation
0.08
0.18
0.06
0.79
TICS social overload
0.12
0.84
0.08
0.06
TICS social tensions
0.24
0.56
0.20
0.33
TICS work discontent
0.27
0.26
0.26
0.58
TICS work overload
0.66
0.34
-
0.03
0.15
Note
s
.
The top two scores for each factor are highlighted in bold.
Multi
-
level
Analysis.
We performed a linear mixed effects analysis on the
relationship between the pressing response during the free
-
operant task and the
dimensional factors extracted through the factorial analysis. As fixed effects we entered:
(1) Phase: pre (last tr
aining run) or post (extinction test) devaluation, (2) Cue: valued or
devalued, (3) Training: moderate or extensive, and (4) the factors extracted through the
factorial analysis. As random effects we entered intercepts for Participants as well as by
-
EXPERIMENTAL INDUCTION OF HABITS
3
partic
ipant random slopes for the effect of the interaction between cue and phase. We
entered
Block (repetition per condition)
and the Site of the data collection (Pasadena1,
Pasadena2, Hamburg, Tel
-
Aviv) as control factors. We used the lmer4 package (Bates
et
a
l.,
201
5
) to build the model as follows:
푅푒푠푝표푛푠푒
푟푎푡푒
푝푒푟
푠푒푐표푛푑
～
푃
ℎ
푎푠푒
∗
퐶푢푒
∗
푇푟푎푖푛푖푛푔
∗
퐹푎푐푡표푟
1
+
퐹푎푐푡표푟
2
+
퐹푎푐푡표푟
3
+
퐹푎푐푡표푟
4
+
퐵푙표푐푘
+
푆푖푡푒
+
(
1
+
푃
ℎ
푎푠푒
∗
퐶푢푒
+
퐵푙표푐푘
|
푃푎푟푡푖푐푖푝푎푛푡
)
We report the p
-
values for the model using the lmerTest package (Kuznetsova,
Br
et al.,
2015).
The analysis revealed a significant interaction between Cue, Phase, Training
and the “Stress Affect” factor (
β
=
-
0.25,
SE
= 0.09, 95%CI [
-
0.44,
-
0.06],
p
=
0.010).
Simple slopes follow
-
up tests revealed that the interaction between cue, value, and group
was significant in participants with lower (
-
1 SD) levels of “Stress Affect” (
β
= 0.36,
SE
= 0.13, 95%CI [0.09, 0.63],
p
= 0.010), whereas it was not signific
ant in participants with
a higher (+1 SD) level of “Stress Affect” (
β
=
-
0.13,
SE
= 0.14, 95%CI [
-
0.40, 0.14],
p
=
0.35
; see Figure S1
). We did not find statistical evidence for an interaction between factor
“Stress Work” (
β
=
-
0.01,
SE
= 0.09, 95%CI [
-
0.20, 0.17],
p
= 0.89), “Stress Social” (
β
=
-
0.12,
SE
= 0.09, 95%CI [
-
0.30, 0.06],
p
= 0.21) and “Impulsivity” (
β
=
-
0.10,
SE
= 0.09,
95%CI [
-
0.30, 0.06],
p
= 0.21) and the effect of interest (i.e., the interaction between Cue,
Phase, Tra
ining).
EXPERIMENTAL INDUCTION OF HABITS
4
Figure S1.
A) B
ehavioral adaptation index ([“cue valued pre
-
cue valued post” vs. “cue devalued pre
-
cue devalued post”]
n
= 199) as a function of the level on the “Stress Affect” factor in participants that
received either a moderate or an ex
tensive amount of training. Shaded areas indicate the 95% CI. B) Mean
adjusted
b
ehavioral adaptation index
to moderate vs. extensive training as a function of lower (
−
1 SD) and
higher (+1 SD) level of the
“Stress Affect” factor.
Strategy 3: Directly using
the subscales of Anxiety and Chronic Worries
The interpretation we made of the findings described above is in terms of the
affective component of stress, which is related to worries and anxiety. To provide an
additional confirmation of this hypothesis,
we also entered in two separate models the
subscales corresponding to anxiety and chronic worry as predictors without entering them
into the factor analysis. This analysis has the advantage of testing directly our question on
the role of worries and anxiet
y.
Multi
-
level Analysis.
We performed linear mixed effects analyses on the
relationship between the pressing response during the free
-
operant task and the sub
-
scale
scores. As fixed effects we entered: (1) Phase: pre (last training run) or post (extinction
test) devaluation, (2) Cue: valued or devalued, (3) Training: moderate or extensive, and
(4) the anxiety scale or the chronic worrying subscale. As random effects we entered
intercepts for Participants as well as by
-
participant random slopes for the effec
t of the
interaction between cue and phase. We entered
Block (repetition per condition)
and the
EXPERIMENTAL INDUCTION OF HABITS
5
Site of the data collection (Pasadena1, Pasadena2, Hamburg, Tel
-
Aviv) as control factors.
We used the lmer4 package (Bates 2010) to build the model as follows:
푅푒푠푝표푛푠푒
푟푎푡푒
푝푒푟
푠푒푐표푛푑
～
(
푃
ℎ
푎푠푒
∗
퐶푢푒
)
∗
(
푇푟푎푖푛푖푛푔
∗
퐴푛푥푖푒푡푦
)
+
퐵푙표푐푘
+
푆푖푡푒
+
(
1
+
푃
ℎ
푎푠푒
∗
퐶푢푒
+
퐵푙표푐푘
|
푃푎푟푡푖푐푖푝푎푛푡
)
푅푒푠푝표푛푠푒
푟푎푡푒
푝푒푟
푠푒푐표푛푑
～
(
푃
ℎ
푎푠푒
∗
퐶푢푒
)
∗
(
푇푟푎
푖푛푖푛푔
∗
퐶
ℎ
푟표푛푖푐
푊표푟푟푦푖푛푔
)
+
퐵푙표푐푘
+
푆푖푡푒
+
(
1
+
푃
ℎ
푎푠푒
∗
퐶푢푒
+
퐵푙표푐푘
|
푃푎푟푡푖푐푖푝푎푛푡
)
We reported the p
-
values for the model using lmerTest
package (Kuznetsova,
Brockhoff &
Christensen
,
2015) and corrected it for the
number of tests with a
significance set at
α
= 0.025.
Anxiety.
The analysis revealed a significant interaction between Cue, Phase,
Training and the anxiety composite scale (
β
=
-
0.24,
SE
= 0.09, 95%CI [
-
0.43,
-
0.06],
p
=
0.01). Simple slopes follow
-
up
tests revealed that the interaction between cue, value, and
group was significant in participants with lower (
-
1 SD) levels of “Anxiety” (
β
= 0.36,
SE
= 0.13, 95%CI [0.10, 0.62],
p
= 0.007), whereas it was not significant in participants with
a higher (+1
SD) level of “Anxiety” (
β
=
-
0.13,
SE
= 0.13, 95%CI [
-
0.39, 0.13],
p
= 0.33
;
see Figure S2A
).
Chronic Worrying.
The analysis revealed a significant interaction between
Cue, Phase, Training and the chronic worrying subscale (
β
=
-
0.24,
SE
= 0.09, 95%CI [
-
0.42,
-
0.05],
p
= 0.012). Simple slopes follow
-
up tests revealed that the interaction
between cue, value, and group was significant in participants with lower (
-
1 SD) levels of
“Chronic worrying” (
β
= 0.34,
SE
= 0.13, 95%CI [0.08, 0.60],
p
= 0.010), whereas it was
EXPERIMENTAL INDUCTION OF HABITS
6
not significant in participants with a higher (+1 SD) level of “Chronic worrying” (
β
=
-
0.13,
SE
= 0.13, 95%CI [
-
0.39, 0.13],
p
= 0.33
; see Figure S2B
).
Figure S2.
B
ehavioral adaptation index ["cue valued post
-
cue valued pre"
vs. "cue devalued post
-
cue
devalued pre"] as a function of the level on the composite scale of Anxiety (
n
= 209; A) and of the Chronic
Worries subscale of the Trier Inventory of Chronic Stress (
n
= 207; B) in participants that that received
either a mod
erate or an extensive amount of training. Shaded areas indicate the 95%
Table S2
. Descriptive Statistics of the Non
-
Standardized
Scores of the
Anxiety and Stress Questionnaires
M
(SD)
N
STAI
-
T
34.05
(8.82)
123
STAI
-
S
42.04 (8.99)
86
TICS chronic worrying
5.68 (3.65)
207
TICS excessive demands at work
7.08 (4.74)
208
TICS lack of social recognition
4.20 (3.13)
208
TICS pressure to perform
14.62 (6.75)
205
TICS social isolation
8.22 (5.56)
208
TICS social overload
7.46 (4.91)
208
TICS social tensions
5.00 (4.26)
207
TICS work discontent
10.92 (5.75)
207
TICS work overload
13.10 (6.87)
207
Notes.
M = mean, SD = Standard deviation.
EXPERIMENTAL INDUCTION OF HABITS
7
Supplemental References
Bates D, Mächler M, Bolker B,
Walker S. 2015.
“Fitting Linear Mixed
-
Effects Models
Using lme4.”
J Stat Softw
,
67
:
1
–
48.
Grice J W. 2001. Computing and evaluating factor scores.
Psychol Methods
,
6
:
430.
Kuznetsova
A
,
Brockhoff P B
.
Christensen R H
B. 2015
. lmerTest
: Tests in Linear
Mixed Effects Models.
R package version 2 (0).