Measuring Citizens’ Implicit and Explicit Attitudes towards the European Union
Michaela Maier, Stefan Krause, Jürgen Maier, Nico Jahn & Silke Adam
Abstract
The assessment of citizens’ attitudes towards Europe is highly relevant for political research
and most often explicit measures are used. However, during the last 20 years research in
psychology has led to the notion that human behavior is not only determined by controlled,
“explicit“ attitudes but also by so-called “implicit“ attitudes. That political objects cause such
implicit reactions has been proven (Lodge & Taber, 2005) as well as that implicit attitude
measures are relevant for the explanation of political behavior beyond explicit tools (Choma
& Hafer, 2009; Galdi et al. 2008; Roccato et al., 2010). It seems especially reasonable to
implement implicit in addition to explicit EU-attitude measures in EU research, assuming that
a) European issues in general are less salient for citizens, b) in pro-European countries (like
Germany) it still seems socially less desirable to express individual EU-skeptical positions,
and c) that current issues like the EURO-crisis boost affective reactions which might be more
accessible through implicit measures. In our empirical study, we therefore combine a standard
explicit questionnaire with an affective misattribution procedure (AMP) to assess the
correlation between explicit and implicit EU attitudes as well as their combined effects on
citizens’ openness to engage in EU-skeptical information behavior or with an EU-skeptical
party. Based on a survey and test with a sample representative for the German population
(n=920) we find a good reliability for the developed AMP and stable significant correlations
between the implicit and corresponding explicit EU-attitude measures. In addition, our
findings show that implicit reactions towards Europe have a rather small but stable impact
when explaining citizens’ information seeking about Europe and their openness to engage
with an EU-skeptical party in addition to explicit measures.
Introduction
The assessment of citizens’ attitudes towards the European Union is highly relevant for
political research and most often explicit measures are used (e.g., Boomgaarden et al., 2011;
Hooghe & Marks, 2005). However, during the last 20 years, researchers in social, personality
and cognitive psychology have intensively worked on the notion that human behavior is not
only determined by controlled, “explicit” thoughts and attitudes, but also significantly by socalled “implicit” reactions and attitudes (for an overview see Friese, Hofmann, & Schmitt
2008). This differentiation has led to the development of the so-called “two-process-models”
(for overviews see Chaiken & Trope, 1999; Evans, 2008; Smith & DeCoster, 2000) and
implicit attitude measures. The advantage of implicit tests is that they allow measuring
attitudes before they can by actively edited by their owners, respectively the assessment of
attitudes which citizens might not have conscious access to under certain circumstances.
That most objects in Political Science cause such implicit, automatic, affective reactions has
been proven (Lodge & Taber, 2005), as well as the finding that implicit attitude measures are
relevant for the explanation of voting behavior (Choma & Hafer, 2009; Roccato et al., 2010).
An additive model combining explicit and implicit measures outplays conventional models of
voting behavior which only specify explicit attitudes (Roccato et al., 2010), this is especially
true for undecided voters (Galdi, Arcuri, & Gawronski, 2008).
Beyond this evidence for the general relevance of implicit measures for political sociology, it
seems especially reasonable to implement explicit and implicit measures with regards to EU
attitudes for the following reasons: First, EU-topics very often are not salient (“top of head”)
for the citizens. Second, current issues (e.g. economically the depths- and EURO-crisis;
culturally the possible membership of Turkey) might foster the formation of implicit attitudes,
affective reactions and fears more than other issues. Third, political ingroup-/outgroup-effects
may be assessed very well with the help of implicit measures (vgl. Burdein et al., 2004; 2006;
Lodge & Taber, 2005; Nosek et al., 2009; Wittenbrink et al., 1997). And last but not least,
answers to explicit questions are typically distorted by their perceived social desirability
(Leary & Kowalski, 1990; Paulhus, 1984). Taking into account the broad consensus among
political elites in Germany that European integration is a desirable political goal and the fact
that citizens still broadly support the EU (vgl. Eurobarometer 77.3; May, 2012) effects of
social desirability might be quite strong when asking explicit questions about Europe. The use
of an implicit measure avoids this problem (Kim, 2003; Steffens, 2004).
Due to these aspects, to us it seems especially reasonable to try to use implicit together with
explicit measures when assessing EU attitudes. We therefore combine a standard explicit
questionnaire with an affective misattribution procedure (AMP) as described by Payne et al.
(2005) to assess the correlation between implicit and explicit EU attitudes as well as their
combined effects on citizens’ openness to engage in EU-skeptical information seeking or with
an EU-skeptical party. Based on a survey and test with a sample representative for the
German population (n=920) we also analyze possible moderation effects by political interest.
2
In the following, we will first outline the theoretical foundations of implicit and explicit
attitude measures based on Strack and Deutsch’s (2004)’ Reflective-Impulsive Model (RIM),
summarize the empirical findings at hand regarding the relevance of implicit and explicit
attitudes in political sociology and derive our research questions and hypotheses. We will then
describe the implicit and explicit measures we’ve used in our empirical study in detail, before
we’ll present and discuss the findings.
Theoretical foundations of implicit and explicit attitudes (measures)
Typically, questionnaires assessing EU attitudes are based on traditional reflective self-report
items and participants’ answers are usually accepted as truthful responses to these items.
However, people might be unable (e.g., as they might not hold deliberate attitudes toward the
given object due to low personal relevance) and/or unwilling (e.g. due to perceived social
desirability) to report their “real” opinions about the EU, and therefore self-report
measurements might be biased. These clear limitations of self-report measurements, which
only capture deliberate attitudes, were mainly recognized by social, personality and cognitive
psychologists who subsequently developed alternative measurement instruments. These new
measurement tools reduce people’s ability to control their responses and also do not require
introspection for the assessment of attitudes. Therefore these tools are capable of further
describing the mental processes behind attitudes and are difficult to fake (Greenwald &
Banaji, 1995). Such measurement instruments are usually labeled as implicit measurements,
whereas traditional self-report measurements are often described as explicit measurements.
One important line within this theoretical implicit-explicit-approach is Strack and Deutsch’s
(2004) Reflective-Impulsive Model (RIM), which describes (implicit) associative/impulsive
and (explicit) propositional/reflective processes as distinct systems. Explicit measurements
gather knowledge or deliberate beliefs in the so-called reflective system, whereas implicit
measurements tap into the automatized associative structure of the impulsive system. The
central core of the impulsive system are associations which are activated automatically or
spontaneously by corresponding stimuli in the environment. The impulsive system is
conceived as a network or pattern of connected associative elements. These connections are
formed if stimuli are approached in close temporal or spatial proximity. Every time a stimulus
activates an element, close elements are also activated and the connections between these
elements are strengthened. Consequently, related associations are easier accessible depending
on the frequency of recent co-activation. An associative network in the impulsive system is
3
activated rapidly and automatically without the use of cognitive resources. Moreover
associative structures reflect basic aspects of the environment including “cognitive, affective,
or motor reactions, without representing the causes of such multimodal correlations” (Strack
& Deutsch, 2004, p. 223). Associated elements activate each other, but this co-activation does
not include an evaluation. Associated elements are only connected based on contiguity and
similarity, and no deliberate processes are involved in this system (Deutsch & Strack, 2010).
For example, a person might associate the EU with the current debt crisis and the fear of
personal financial insecurity, because media exposure has created such links. New findings
show that implicit associations are rather sensitive towards such context cues (Gawronski.
2012) which is in line with Zaller’s (1992) conceptualization of opinion formation.
Nevertheless, this person may not deliberately believe in this negative connection. He or she
could be even a strong EU supporter who positively regards the EU's efforts to solve the
European debt crisis.
Explicit deliberate believes or attitudes are represented as propositions and are part of the
RIM’s other system − the so-called reflective system. Unlike its impulsive counterpart, the
reflective system works non-automatically, not stimulus-driven and therefore it demands high
cognitive capacity. Unlike the impulsive system, representations in the reflective system are
flexibly generated and changed. Propositions in the reflective system are represented in a
language-based format and therefore, people are able to express them (e.g. via self-report
measures). They are also considered as connections between different elements, but these
connections are based on different relational schemata. Therefore, attitudes and knowledge
are much more organized (in a semantic fashion) compared to simple associative connections
of the impulsive system (Strack & Deutsch, 2004). As can be seen from the EU example, the
central difference between the two systems is the role of subjective truth or accuracy. In the
impulsive system the activation of connected associated elements (e.g. EU, debt crisis and
fear) occurs independently from whether these linked associations are considered as correct or
incorrect. By contrast, the reflective system is concerned with the validation of different
activated propositions. This validation depends on reasoning, logical consistency and
avoiding cognitive dissonance between considered propositions (Gawronski & Strack, 2004;
Hofmann, Gschwendner, & Schmitt, 2005; Strack & Deutsch, 2004).
Under certain conditions, implicit associations activated in the impulsive system can be
translated into an explicit proposition-based format in order to enter the reflective system.
Attitudes towards objects of great personal interest are considered to be often recalled,
4
applied, experienced and also involve a great deal of “basic social and personal values”
(Boninger, Krosnick, & Berent, 1995, p. 63). If people intensely devote themselves to an
attitude object (e.g., the EU), the likelihood increases that implicit associations are translated
into a proportional format (also see Zaller, 1992). Subsequently, the former association
becomes easier available for the reflective system as an additional proposition to consider.
Since the reflective system is driven by cognitive consistency between different propositions,
a stronger correlation between implicit and explicit measures is expected in such cases of high
involvement (Deutsch & Strack, 2010; Hofmann, Gschwendner, Nosek, & Schmitt, 2005;
Nosek, 2005, Nosek, 2007; Strack & Deutsch, 2004).
Implicit and explicit attitudes (measures) in Political Research
The relevance of the individual involvement (e.g. political interest, personal relevance of
political attitude objects) has already gained significant interest in the emerging research line
on the interactions between implicit and explicit political attitudes. For example, Karpinski
(2005) showed that the relationship between implicit and explicit attitudes toward U.S.presidential candidates is moderated by attitude importance (perceived importance of and
interest in U.S.-politics). Friese et al. (2007) could replicate these findings and showed that
higher general interest in politics strengthened the consistency between explicitly and
implicitly measured attitudes towards the five German major parties in the 2002 German
parliamentary elections. Correspondingly, Choma and Hafer (2009) investigated the
relationship between explicit and implicit political orientation moderated by political
sophistication. For people with higher political knowledge, which the authors link to attitude
importance and interest in politics, the relationship between implicit and explicit political
orientations was also much stronger than for citizens with lower levels of sophistication.
But what about political topics which are not intensively elaborated about? Indeed, EU-topics
are not very often salient (“top of head”) for the citizens. In that case of a relatively low
involvement in the topic, citizens might stronger rely on their “gut feeling”, i.e. implicit
attitudes, than more strongly interested citizens. In line with this assumption, Galdi, Arcuri, &
Gawronski (2008) showed that especially for undecided voters implicit measures are better
predictors for voting behavior then traditional self-reports. In other words, the lesser a
political topic is elaborated on and not considered as import, the higher should be the
influence of implicit attitudes on subsequential political behavior (Friese et al., 2012).
5
Not only voting behavior of undecided voters is influenced by their implicit attitudes. Even
information processing could be biased by implicit attitudes. Since especially undecided (or
uninterested) voters do not hold an elaborated opinion, they often rely on the only available
information- their implicit “gut feeling”- when they have to interpret ambiguous political
material. Moreover, they also selective expose themselves to confirmatory political
information which is in line with their implicit attitudes (Gawronski, Galdi & Arcuri, 2013).
These theoretical considerations were empirical tested by Galdi, Gawronski and Friese
(2012), who conducted a study concerning the influence of implicit attitudes on information
search using the example of Turkey’s possible inclusion into the European Union. They
showed that participants who stated being undecided about Turkey’s inclusion selectively
exposed themselves to newspaper articles which was consistent with their implicit attitude on
the issue. Whereas selective exposure for decided participants was predicted by explicit
beliefs and not by implicit associations. So, undecided or less interested voters may have a
positive or negative implicit associations about a political issue which may lead them to
selectively search for information that supports the preference implied by implicit
associations.
According to the theoretical framework and the empirical research findings presented, this
study aims at investigating the interplay of implicit and explicit attitudes towards the
European Union in Germany. We assume that
H1 implicit and explicit EU attitudes will show a significant correlation.
However, this interrelation should be significantly weaker than between explicit EU attitude
measures as the implicit measure according to its theoretical foundation should tackle rather
loose associations with the EU evoked by contextual cues (Zaller, 1992), while explicit
attitudes should show a strong consistency.
Taking into account the broad consensus among political elites in Germany that European
integration is a desirable political goal and the fact that German citizens in general still
broadly support the EU (vgl. Eurobarometer 77.3; May 2012), we expect effects of social
desirability on the explicit questions about Europe which should be avoided by the implicit
measures, and we would assume that
6
H2 implicit EU attitudes in general will be less positive than explicit EU-attitudes.
Beyond the correlational connection between implicit and explicit EU attitude measures, we
also assume that
H3 the implicit attitudes will have a predictive power regarding EU-relevant political
behavior beyond the explicit measures.
According to the RIM, we expect a moderating effect of political interest on the relevance of
the cueing-effects of the context and the subsequent relevance of implicit measures for the
prediction of political behavior. Taking into account Zaller’s (1992) observations regarding
the nonmonotonic effects of political interest,
H4 we expect the strongest impact of the implicit measure in the group of citizens with
moderate political interest.
In the group of citizens with high political involvement, we expect the greatest impact of
explicit EU attitudes on political behavior due to high levels of deliberation. In the group of
citizens with low political involvement however, we expect rather random answers towards
the explicit questions (including the dependent variables) and therefore the worst fit of the
models
In the following, we will describe our empirical study which included standardized explicit
EU-attitude measures, as well as an Affect Misattribution Procedure (AMP) which we’ve
developed for the assessment of implicit EU-attitudes.
Methods and Operationalizations
Method and Fieldwork
The data for this study was collected in representative online-survey that also included an
implicit attitude test. The fieldwork for this study was done in May and June 2013 by the
market research institute gfk, Nuremberg, using their online access panel. Initially, 1.758
people were invited to participate in the study per email. From the 1.752 persons who
responded, a quota sample of 1.072 persons was drawn taking into account the variables age,
education, gender and geographic region. The quotas were chosen in accordance to the
distribution of the criteria in the total German population. The selected participants received a
7
link to an online-questionnaire which included the explicit EU attitude-measures as well as an
Affective Misattribution Procedure (AMP) as implicit measure, one reminder was sent out.
The quality of the interviews was ensured by using the GfK-tool TIGO. This tool a) detects
specific patterns of participants’ answers to standardized questions (e.g. straightlining); b)
evaluates the answers to open questions; and c) takes into account the time that a participant
spent in filling in the questionnaire (min. 11 minutes, max. 60 minutes). Interviews which
were evaluated as “not satisfactory” according to these criteria were removed from the sample
(n=44). In addition, 125 incomplete interviews were dismissed, so that 17 additional
participants had to be recruited in order to reach the aspired number of cases. In total, 920
interviews were rated as good quality and used for the analysis of the explicit questionnaire
items. For the implicit measure, additional criteria were used to ensure good data quality (see
below). 47 percent of the participants were female; the average age was 44.9 years (SD =
13.7); 26.2 percent had not passed secondary school, 36.4 had passed secondary school, 17.8
percent had qualified for university, and 19.6 percent had obtained a university degree.
The Affect Misattribution Procedure
To assess citizens‘ implicit reactions towards the Europe Union we’ve developed a so-called
”Affect Misattribution Procedure“ (AMP) as described by Payne et al. (2005; also see
Hofmann & Baumert, 2010). AMPs are not based on reaction-times and therefore better
suited in the context of online-studies than other implicit measures which rely on a proper
assessment of reaction-time, e.g. the Implicit Association Test by Greenwald et al. (1998).
Moreover, the IAT has been recently criticized for the mandatory assessment two different
attitude objects, which make it impossible to assess implicit attitudes towards a single object
(Gawronski & Houwer, 2012; Teige-Mocigemba, Klauer, & Sherman, 2010). The AMP on
the other side can be used both as a relative implicit preference measure for one out of two or
more objects (Payne, Govorun, & Arbuckle, 2008) but also as an absolute implicit measure
for one single category (Bar-Anan & Nosek, 2012; Payne et al., 2008; Payne, McClernon, &
Dobbins, 2007). In addition, the AMP is a rather economic implicit measurement tool
because a typical session is quite short (usually less than five minutes), and AMPs in general
show good internal consistency with a range between Cronbach’s α = .70 and .90 (Gawronski
& Houwer, 2012; Payne et al. 2005; Payne, Govorun, & Arbuckle, 2008).
8
The AMP follows the general idea of projective tests: Participants in the test shall come to
evaluations in ambivalent situations and transfer their reactions originally directed towards an
object (“prime”) to an ambivalent target. In our case participants are exposed to ”primes“
representing the European Union (i.e. pictures closely related to the European Union), primes
representing Germany (.i.e. pictures closely related to Germany) or neutral prime pictures
taken from the standardized International Affective Picture System (Lang, Bradley, &
Cuthbert, 2008) for milliseconds (see Figure 1). Then an ambivalent target follows that shall
be evaluated as positive or negative. As target pictures we adopted the Chinese pictographs
from the original AMP (Payne et al. 2005). For each of the three prime picture categories,
twelve different pictures were used twice and every single prime picture was randomly paired
with a different Chinese pictograph (out of a pool of 82). Participants are explicitly asked to
monitor themselves so that the prime cannot influence them (consciously) and to direct the
evaluation only to the target instead. Many studies have proven (Gawronski & Walther, 2008;
Hofmann, Friese, & Roefs, 2009; Miller, Brewer, & Arbuckle, 2009; Oikawa, Aarts, &
Oikawa, 2011; Payne et al., 2010; Rydell, McConnell, & Mackie, 2008), that the affective
reaction toward the prime (in our case: EU and Germany) systematically affects the
evaluation of the target (Chinese letter) and that it reflects the implicit attitude towards the
prime (Payne, 2005).
The AMP was programmed with Javascript and embedded in the online-survey. Participants
were instructed to concentrate only on the Chinese pictographs and indicate as quickly as
possible if they pleased them or not (by pressing “Q“ for “bad” or “P” for “good“). After ten
test ratings, all prime pictures were presented twice, resulting in 72 trials. Every trial started
with a fixation point (50 ms), followed by a randomly chosen prime picture (either a EU,
Germany or neutral picture; 100 ms), a blank screen (125 ms), a randomly chosen Chinese
pictograph as target (300ms), and a grey noise picture (see Hofmann & Baumert, 2010, for
similar presentation times). After the participant’s evaluation, the next trial started
automatically.
The AMP requires participants to respond spontaneously in order to measure implicit
affective associations towards a prime picture; therefore all trial responses exceeding a
threshold of 1500 ms were excluded from further analysis (16.98% of all EU-related trials,
16.30% of all Germany-related trials and 16.73% of all neutral prime trials). In order to
guarantee data quality, participants who had less than 50 % valid trials in one prime category
(based on reaction time threshold) were not included in the data analysis. Likewise 40
9
participants who were able to read Chinese pictographs and 93 participants who did not show
any variation in their key response (so-called straight-lining) were also removed from the
final set. Based on all these criteria the resulting n for the EU-related trials was 732, for the
Germany-related trials 723 and for the neutral trials 719.
Fixation
point
!
Prime
Blank
screen
Target
Grey noise
Figure 1. Example for a single AMP trial (here with an EU prime picture).
In order to obtain a metric measure, the proportion of prime pictures rated as “good” in each
!
prime condition (EU, Germany or neutral) was calculated. Since good ratings were coded as 1
and bad ratings were coded as 0, the proportion values range between 0 and 1 (Hofmann &
Baumert, 2010; Payne et al., 2005). Besides the proportion of good ratings for the European
Union, four additional measures were calculated; the first takes into account the responses on
the neutral pictures as a baseline in order to control for interindividual response tendency
(Bar-Anan & Nosek, 2012; Payne et al., 2008; Payne et al., 2007). Therefore, the difference
score between the valid proportion of good ratings after EU prime pictures minus the valid
proportion of good ratings after neutral prime pictures (as baseline) was calculated. The same
procedure was applied for Germany picture (good rating after Germany picture minus valid
proportion of good ratings after neutral prime pictures). In order to be able to compare
affective reactions to the European Union and Germany, additional difference scores between
the valid proportion of good ratings after EU primes and the valid proportion of good ratings
after Germany primes were calculated. Both difference scores range between -1 and 1. For the
EU-Germany-score higher numbers reflect a relative implicit preference for the EU as
opposed to Germany; for the Germany-EU-score higher numbers reflect a relative implicit
preference for Germany (Payne et al. 2008).
10
Explicit EU-attitude measures
Participants’ explicit EU-attitudes were assessed with two measures: An index of four
standardized questions commonly used to estimate EU attitudes as well as an evaluative space
grid. The index was composed from four original questions (see also Hooghe & Marks, 2005,
p. 427; Maier, Adam, & Maier, 2012) assessing a) the European Union’s overall image; b) the
evaluation of the Germany’s EU membership; c) the evaluation of advantages and
disadvantages Germany had from EU membership; d) the evaluation of the EU’s general
performance, each measured on a 5-point Likert-scale (see appendix). The index, which also
ranges from 1 (‘very negative attitudes’) to 5 (‘very positive attitudes’), shows a Cronbach’s
alpha of .88.
The evaluative space grid was based on the suggestions of Larsen et al. (2009; see also
Retzbach, Maier & Jahn, 2013) and consisted of 5x5 cells (see Figure 3). On the x-axis
positive attitudes toward the European Union were assessed on a scale from 1 (‘not at all
positive’) to 5 (‘very positive’), while on the y-axis negative attitudes toward the EU were
represented also on a scale from 1 (‘not at all negative’) to 5 (‘very negative’). Participants
were asked to assess their positive and negative attitudes towards the EU simultaneously. The
difference between the ratings is interpreted as bipolar valence rating (Larsen et al., 2009;
Retzbach, Maier & Jahn, 2013) and operationalized in a difference score ranging from 4
(‘fully positive evaluation of the EU’) to -4 (‘fully negative evaluation of the EU’). This
difference score will be used as second explicit EU attitude measure in the following.
very
negative
☐
☐
☐
☐
☐
negative
☐
☐
☐
☐
☐
kind of
negative
☐
☐
☐
☐
☐
☐
☐
☐
☐
☐
☐
☐
☐
☐
☐
not at all
positive
a little
positive
kind of
positive
positive
very
positive
a little
negative
not at all
negative
Figure 2: The evaluative EU-space grid
11
Dependent variables
Two dependent variables were chosen in order to assess the predictive power of the explicit
and implicit attitude measures: 1) Interest in information critically analyzing the EU and 2)
the intention to vote for a party perceived as EU-skeptical. Both measures shall represent
citizens’ intention to engage in EU-skeptical information seeking (e.g. during an election
campaign) as well as in EU-skeptical political behaviour on election day. In the questionnaire,
information behaviour was assessed using two items “How strong is your interest in
information regarding the advantages of the European Union?” and “How strong is your
interest in information critically analysing the EU?”. Both items were rated on a scale from 1
‘not interested in all’ to 5 ‘very strong interest’. For the first dependent variable a difference
score was computed, subtracting the interest for positive information from the interest in
critical information on the EU. The score consequently takes values from 4 (‘only interested
in critical information’) to -4 (‘only interested in positive information’).
For the second dependent variable the information which party the participant would vote for
if European elections would be held on the next Sunday (“If elections to the European
parliament were held next Sunday, which party would you vote for?”) was combined with the
rating how this party stands regarding EU integration (“How would you assess the attitude of
Party X regarding the integration of the member countries within the European Union? 1 ‘The
party supports the further integration’, 7 ‘The party is of the opinion that the integration has
already gone much too far’). The combination variable takes values from 1 ‘voting intention
for a pro-European party’ to 7 ‘voting intention for an EU-skeptical party’.
Logic of analysis
The goal of this paper is to evaluate possible benefits of measuring citizens’ attitudes towards
the European Union with explicit and implicit measures simultaneously. As the explicit
measurement is standard in the research field, the more specific question must be whether
implicit measures can assess EU attitudes reliably, validly and whether they can add to the
predictive power of the explicit measures at hand. Therefore, after a first description of the
measures, the focus of the analyses will lay on the assessment of the reliability, validity and
predictive potential of the suggested implicit measure (AMP). The reliability of the AMP will
be assessed based on a retest, its construct validity will be estimated by a correlation with
similar explicit concepts, and its predictive power will be assessed by adding it to explicit
12
measures for the explanation of interest in EU-skeptical information as well as of the intention
to vote for a EU-skeptical party in an OLS regression analysis. As a last step, possible
moderation effects from political interest will be analyzed. Along the way, the predictive
power of the two explicit measures suggested for this paper (standard index of questionnaire
items vs. evaluative space grid) will be compared.
Results
Descriptives
Table 1 provides the descriptive data for the AMP and the explicit measures included in the
study. The first interesting point is that the share of positive ratings for neutral pictures is
higher the share of negative ratings (M=.57). Given the neutral valence of the pictures
documented in the International Affective Picture System (Lang et al., 2008), positive and
negative ratings in this category should be more balanced and based on the distribution found
in our study, participants’ tendency for positive evaluation may be assumed. Therefore it
seems advisable to control for this response set either by working with AMP difference scores
or by including the AMP evaluation of neutral pictures as control variable when working with
the share of positive EU- and Germany-ratings. Taking this into account, the rather positive
evaluations of the EU and Germany should not be overinterpreted at this point. The AMPdifference scores reveal a very slightly more positive evaluation of Germany than the EU on
the aggregate level.
Table 1: Descriptives of the implicit and explicit measures included
N
Min.
Max.
M
SD
Implicit measures
AMP Share positive EU-ratings*
AMP Share positive DE-ratings
AMP Share positive ratings neutral pictures
AMP Difference pos. EU-pos. neutral
AMP Difference pos. DE-pos. neutral
AMP Difference pos. EU-pos. DE
AMP Difference pos. DE-pos. EU
732
723
719
709
706
713
713
0
0
0
-.94
-.1.00
-1.00
-.96
1
1
1
.90
.96
.96
1.00
.59
.63
.57
.02
.06
-.04
.04
.27
.27
.26
.23
.25
.23
.23
920
920
920
1
1
1
5
5
5
2.99
3.43
3.09
1.05
1.23
1.24
920
920
1
1
5
5
2.45
2.99
1.04
.98
Evaluative space grid
920
* Based on min. 50% good trials per AMP. α=.88.
-4
4
-.01
2.03
Explicit measures
Item General EU image
Item EU-Membership is a good thing
Item EU more advantages than
disadvantages
Item EU works well
Index EU-attitude**
13
Both explicit measures also show rather balanced global EU attitudes (Mindex = 2.99; Mevaluative
grid
= -.01), while differences become visible between the single items constituting the index.
Especially the item “EU-membership is a good thing for Germany” (M = 3.43) receives
significantly more support than the other items, while the item “The EU works well as it is”
(M = 2.45) receives lowest support. All in all, the implicit as well as the explicit EU-attitudes
seem balanced on the aggregate level, however, the mean values of the explicit items vary,
which indicates that EU-support is rather ambivalent than clearly positive. When interpreting
this finding it should be kept in mind that this study was conducted in midst of the 2012/13
debt-crisis.
Reliability of the implicit measure
Turning to the reliability (rtt) of the AMP first, we split this measure into two similar halves
for each participant which is possible as all pictures were presented twice. We then calculated
the proportion of “good” ratings for each category in both samples, and correlated these
proportions corrected with the Spearman-Brown prediction formula (for details see,
Moosbrugger & Kelava, 2012). The two proportions of “good” ratings for EU primes
correlated with rtt = .86; for the Germany primes they also correlated with rtt = .86 and for the
neutral primes they correlated with rtt = .84. The difference scores between “good” ratings of
the EU and the neutral pictures correlated with rtt = .63. Both difference scores EU-Germany
and Germany-neutral correlated respectively with rtt = .67 (for more information on possible
AMP reliability measures see Appendix). These stable correlations between the proportions of
good ratings per category as well as between the difference scores support the retest reliability
of our AMP. Their strength is comparable to implicit measures in other fields (for a
metaanalysis see Hofmann, 2005; Retzbach, Maier & Jahn, 2013) and higher than the values
we found in other domains ourselves (Retzbach, Maier & Jahn, 2013). In sum, the retest
reliability leaves us with a first positive impression.
Construct Validity of the implicit measure
In order to assess the construct validity of the AMP, we correlated the ratings with three
corresponding explicit measures (see Table 2): the index from four EU-attitude items which is
standard in EU-research, the evaluative space grid assessing positive and negative EU
attitudes simultaneously, and the measure for exclusive European identity (for all three
measures, see methods section).
14
First, we find highly significant positive correlations between all AMP measures for implicit
reactions to the European Union with the explicit EU-attitude measures. The strongest
correlations with the explicit measures are found between the share of positive EU-ratings
(r=.23 to r=.31) as well as the difference score between positive EU and Germany ratings
(r=.22 to r=.27). However, the share of positive EU-ratings also shows a strong positive
correlation with the share of positive ratings for the neutral pictures (r=.64). As mentioned
before, this high correlation might indicate participants’ general tendency for positive
reactions in the AMP. The difference score between positive EU and positive neutral picture
ratings controls for this tendency. Consequently, the correlations with the explicit measures
turn out lower for this measure (r=.13 to r=.22), but they still reach the level described as
typical for correlations between implicit and explicit measures (Hofmann et al., 2005; see also
Gawronski & Payne, 2010). The difference score between positive EU and positive Germany
ratings also controls for this response set, however it shows higher correlations (r=.22 to
r=.27), as this measure represents a preference for the EU over Germany. It is not astonishing
that participants who like the EU even better than their home country also show positive
explicit EU-attitudes. It is interesting to see that for all three implicit EU measures the
correlations are lowest with the evaluative grid (r=.14 to r=.23). It seems that this tasks
requires even more cognitive resources than the standard questionnaire items and therefore
shows lowest correlations with the implicit measures.
Table 2: Correlations of the AMP-measures with related explicit constructs
Implicit EU-attitudes
AMP Share positive EU-ratings
AMP Difference pos. EU-pos. neutral
AMP Difference pos. EU-pos. DE
Implicit BRD-attitudes
AMP Share positive DE-ratings
AMP Difference pos. DE-pos. neutral
AMP Difference pos. DE-pos. EU
EU-Index
EU-Grid
European
identity
AMP Share
positive
ratings
neutral
pictures
.31***
(N=732)
.22***
(N=709)
.27***
(N=713)
.23***
(N=732)
.14***
(N=709)
.22***
(N=713)
.25***
(N=732)
.13**
(N=709)
.26***
(N=713)
.64***
(N=709)
-
.08*
(N=723)
-.02
(706)
-.27***
(713) s.a.
.04
(723)
-.05
(706)
-.22***
(N=713) s.a.
.03
(N=723)
-.10*
(706)
-.26***
(713) s.a.
.58***
(N=706)
-
Levels of significance: * p<0.05; ** p<0.01; *** p<0.001.
15
-
-
In contrast to the measures for implicit EU-attitudes, the implicit measures for attitudes
towards Germany in general do not show significant relationships with the explicit EUmeasures. This is an interesting finding, as it indicates that attitudes towards the EU and
towards Germany are not antagonists but independent concepts. This has to be taken into
account when further applying the AMP-measures: Using the difference score between the
good EU and the good Germany ratings is only adequate for the operationalization of
exclusive territorial identities but not for EU-attitudes as such. Such an exclusive preference
for Germany is measured by the difference score between positive Germany and positive EU
ratings. This measure shows highly significant negative relations with all three explicit EU
attitude measures which is completely in line with research that has shown exclusive national
identity to be one of the strongest predictor for EU-skeptical attitudes (e.g., Hooghe & Marks,
2005; Maier, Adam & Maier, 2012).
In sum, the theoretically assumed correlations between the AMP-measures and the
corresponding explicit concepts are highly significant and show a strength which has been
reported for the correlations of implicit and explicit attitude measures in general (Hofmann et
al., 2005). In line with Dual-process theories like the RIM (Strack & Deutsch, 2004) these
moderate correlations suggest that implicit and explicit attitudes share common elements,
while at the same time contain exclusive factors as well.
Predictive Potential of the implicit and explicit measures
In the third step, we analyse whether the AMP can add to the power of explicit measures
when predicting interest in information critically analysing the EU and the intention to vote
for a party which is perceived as EU-skeptical. Both dependent variables were chosen due to
their high relevance for current research on European elections and campaigns. Finally, we
control for moderation effects of political interest.
Interest in critical information about Europe is measured on a 7-point scale from +4 (only
interested in EU-critical information) to -4 (only interested in positive EU-information). The
mean value is M = 0.18 (SD = 0.84), meaning that participants of the study had a slightly
higher interest in EU-skeptical than in pro-EU information. The first model assesses the
explanatory power of the two explicit measures, while models 2 and 3 evaluate the
explanatory power of the isolated AMP-measures (see Table 3). It becomes clear that the
explicit measures have a much higher predictive potential (R2model1 = .19) than the implicit
16
measures (R2model2 = .01; R2model3 = .03). Model 4 combines the explicit with the EU-AMP
measure, and in this step the EU-AMP loses statistical significance. However, when including
the implicit measure for positive reactions towards Germany in model 5, this implicit measure
shows an statistically significant impact, even though the explanatory power of the overall
model increases only slightly (R2model5 = .20). In order to control for the stability of the
implicit effect, the two standard predictors of EU-attitudes, territorial identity and economic
considerations (Hooghe & Marks, 2005; Maier, Adam & Maier, 2012) are included in model
6. Even in this strict test, the AMP-measure for positive affect for Germany stays statistically
significant. However, the explicit variables do not show the expected impacts anymore.
Assuming that this is an artefact due to multicollinearity between the explicit concepts, in the
7th model, the EU-index is not included anymore. As a results, model 7 shows significant
effects of the EU-grid as well as of European identity and economic considerations which are
absolutely in line with literature and theoretical expectations (Hooghe & Marks, 2005; Maier,
Adam, Maier, 2012). At the same time, the implicit measure for positive reactions to
Germany prevails its statistical significance. In sum, models 5 and 7 include the same implicit
and similar explicit concepts and show a predictive power about R2 = .20 which can be
considered as good. The implicit affect towards Germany shows a stable positive effect for
interest in EU-skeptical information, even though the relative predictive power of this
measure is very small compared with the explicit measures. It can be summarized that not the
measure operationalizing the affective reaction towards Europe, but the affective reaction
toward the own country has a stable effect on interest in EU-skeptical information seeking.
In order to check for the stability of our findings, we replicate the regression analyses for a
second dependent variable, the intention to vote for a party which the participant perceived as
EU-skeptical (see methods section). On the scale from 1 ‘The party supports further
integration’ to 7 ‘The party is of the opinion that integration has already gone much too far’,
the mean value for the preferred parties is M = 3.13 (SD = 1.75), meaning that most
participants intend to vote for parties which they perceive as slightly pro-European. The
regression models presented in Table 4 fully support our findings from the regression
analyses described before: Not the measure assessing implicit reactions toward Europe but the
implicit reactions toward the own country (Germany) yield a stable significant effect on EUrelated voting intentions: participants who show a strong positive affect for Germany, have a
significantly stronger intention to vote for a party which they themselves perceive as EUskeptical than people who do not have such a strong implicit attachment to Germany. Again,
17
this finding is completely in line with the literature on explicit EU-attitudes which has shown
that an exclusive national identity in addition to economic considerations is the best predictor
for EU-skepticism. This effect of the implicit measures even holds if the corresponding
explicit measure (territorial identity) is included in the model. However, the impact of the
implicit measure in the model is again very small compared to the explicit measure.
As final step of the analysis, we control for moderation effects of political interest. As
specified in hypothesis 4, we expect the strongest effects from contextual cues and therefore
on implicit attitudes in the group of respondents with medium political interest (see e.g.,
Zaller, 1992). Table 5 shows the results for the regression model No. 7 (control model 2)
again for both dependent variables – EU-skeptical information seeking and EU-skeptical
voting intention – for three groups with low, medium and high political interest. For EUskeptical information seeking, our findings are fully in line with the expectations: The
explanatory power is lowest in the group with low political interest (R2 = .11). Only explicit
pro-European attitudes measured with the space grid show a highly significant negative effect
on EU-skeptical information seeking. The predictive power significantly increases in the
group with medium political interest (R2 = .19). Here, in the implicit positive affect for
Germany becomes the strongest predictor for interest in anti-EU information, its impact goes
even beyond the explicit control variables. While the predictive power of the overall model
further increases in the group with high political interest (R2 = .25), the implicit measure
looses its relevance in this group. As suggested by our hypothesis based on Strack and
Deutsch’s (2004) Reflective-Impulsive Model but also on the standard literature in the field
(e.g., Zaller, 1992), the impact of implicit attitudes on EU-skeptical information seeking
proves to be strongest in the group with medium political interest, fully supporting hypothesis
4.
Unfortunately, these results can’t be replicated for the second dependent variable – voting
intention for a EU-skeptical party – in the same way: Even if we were to disregard the model
for the group with low political interest due to the small number of participants (n = 55), the
fit of the overall model is rather low for the groups with medium (R2 = .10) and high (R2 =
.15) political interest. In addition, the picture we get about the effects from the implicit and
explicit measures is by far not as clear here as for the first independent variable. At this point
we might state that the moderation-analysis does not seem to work as well for this second
dependent variable and further analyses will be necessary here.
18
SE
.123
.040
.019
B
.663
-.160***
-.113***
-.317*
-.033
.153
.155
2
Implicit model
only AMP-EU
B
SE
.377
.082
-.681***
-.189
.678***
.175
.161
.161
3
Implicit model
AMPs EU/DE
B
SE
.257
.087
Explicit Controls
Territorial identity
Econ. considerations
2
Adjusted R
.19
.01
.03
F
105.14
4.14
8.73
N
919
708
699
Note: Cell entries are unstandardized B coefficients and standard errors.
***: p<.001, **: p<.01, *: p<.05.
Implicit measures
AMP share pos. EU
AMP share pos. neut.
AMP share pos. DE
Intercept
Explicit measures
EU-index
EU-grid
1
Explicit model
Table 3: Explicit and implicit predictors of EU-skeptical information seeking
.19
24.53
708
.143
-.175
-.168***
-.117***
.145
.141
.048
.022
4
Combined
model EU
B
SE
.691
.153
.20
35.22
699
-.130
-.261
.453**
-.160**
-.114***
.167
.147
.148
.048
.022
5
Combined
model EU/DE
B
SE
.596
.156
.167
.147
.149
.054
.023
SE
.180
-.073*
.028
-.055
.035
.21
26.80
699
-.097
-.237
.427**
-.087
-.107***
B
.447
6
Control model 1
.165
.146
.149
n.i.
.019
SE
.134
-.084**
.028
-.078*
.032
.20
30.77
699
-.143
-.215
.435**
n.i.
-.123***
B
.254
7
Control model 2
SE
.322
.106
.051
B
4.574
-.459***
-.166**
-.851*
-.262
.366
.374
2
Implicit model
only AMP-EU
B
SE
3.698
.200
-1.545***
-.657
1.515***
.407
.382
3.78
3
Implicit model
AMPs EU/DE
B
SE
3.378
.212
Explicit Controls
Territoral identity
Econ. considerations
2
Adjusted R
.18
.02
.05
F
69.43
6.450
9.405
N
632
478
472
Note: Cell entries are unstandardized B coefficients and standard errors.
***: p<.001, **: p<.01, *: p<.05.
Implicit measures
AMP share pos. EU
AMP share pos. neut.
AMP share pos. DE
Intercept
Explicit measures
EU-index
EU-grid
1
Explicit model
Table 4: Explicit and implicit predictors of EU-skeptical voting intention
20
.17
25.84
478
-.122
-.464
-.525***
-.098
.349
.347
.124
.058
4
Combined
model EU
B
SE
5.024
.396
.18
21.74
472
-.653
-.716*
1.057**
-.492***
-.092
.395
.358
.355
.124
.058
5
Combined
model EU/DE
B
SE
4.711
.406
.399
.360
.361
.145
.058
SE
.477
-.059
.071
.005
.087
.18
15.58
472
-.614
-.688
1.011**
-.459**
-.088
B
4.515
6
Control model 1
.403
.367
.370
n.i.
n.i.
SE
.334
-.219**
.066
-.277***
.071
.13
15.28
472
-.888*
-.582
1.154*
n.i.
n.i.
B
3.824
7
Control model 2
Conclusion
In sum, our findings support the idea of using implicit in addition to explicit measures in
research on attitudes towards the European Union. The measures gained from the Affect
Misattribution Procedure (AMP) showed highly significant positive correlations with the
explicit standard measures (hypothesis 1). The moderate strength of these correlations
supports the theoretical assumption that implicit and explicit measures tackle different attitude
formation processes as described by dual-process theories like the RIM and that the one can’t
replace the other type of measurement.
Regarding the question whether implicit measures can avoid response behavior perceived as
socially desirable (hypothesis 2), we can’t provide a clear answer. The implicit as well as the
global explicit measures showed rather balanced EU attitudes. The strongest differences
became visible between the single items constituting the explicit attitude-index. Especially the
item “EU-membership is a good thing for Germany” received very strong support, however,
other explicit items were rather denied, e.g. while the item “The EU works well as it is” (M =
2.45). In sum, implicit as well as the global explicit EU-attitudes were balanced, however, the
variance between the single explicit items indicates that EU-support in Germany seems rather
ambivalent than clearly positive these days.
The test whether the implicit measures have predictive power for the explanation of political
behavior beyond the corresponding explicit measures (hypothesis 3) brought a positive result:
Even when including very powerful explicit control variables into the regression models, it
was an implicit measure that had the strongest effect on EU-skeptical information seeking and
voting behavior. However, this measure was not EU-related but referred to a strong positive
affect for the own nation state. This means that the participants of the study who showed very
strong positive reactions towards Germany in the implicit test, were explicitly strongly
interested in EU-skeptical information and EU-skeptical parties. However, this finding is
completely in line with the research on explicit EU-attitudes which has shown that an
exclusive national identity together with economic considerations has the strongest effect on
EU-attitudes.
Hypothesis 4 claimed that implicit attitudes should have the strongest effect on political
behavior in the group of participants with medium political interest. This hypothesis was
strongly supported by the analysis of information seeking as dependent variable. In addition,
in the group with low political interest, the model fit was the lowest – es expected based on
the observations by Zaller (1992). At the same time, in the group of participants with high
political interest, the explicit measures showed the strongest impact as suggested by the RIM.
However, this analysis could only partially be replicated for the second dependent variable
(voting intention) and additional analyses are necessary here.
22
SE
.270
.038
.448
.380
.365
n.i.
-.100**
Implicit measures
AMP share pos. EU
-.150
AMP share pos. neut.
-.147
AMP share pos. DE
.428
.026
-.485*
.557*
n.i.
-.076**
B
.406
.227
.201
.216
.028
SE
.209
-.460
.152
.270
n.i.
-.199***
B
-.071
.288
.257
.252
.033
SE
.235
-3.417**
-1.634
3.712**
n.i.
-.305*
1.227
1.137
1.084
.125
-.255
-.198
.328
n.i.
*
-.146
.544
.507
.554
.068
-.816
-.735
d
1.035
n.i.
-.237***
.661
.575
.551
.059
Dependent: EU-skeptical voting intention
Group 1:
Group 2:
Group 3:
low political
medium
high political
interest
political
interest
interest
B
SE
B
SE
B
SE
4.143
.583
2.923
.345
3.071
.339
Explicit Controls
b
Territoral identity
-.018
.065
-.108**
.041
-.082
.046
.181
.197
-.285** .105
-.079
.099
a
z
z
z
Econ. considerations
-.125
.071
-.149**
.049
.056
.057
n.i.
n.i.
n.i.
2
Adjusted R
.11
.19
.25
.33
.10
.15
F
3.79
13.19
14.98
6.350
5.53
8.37
N
134
309
254
55
205
210
Note: Cell entries are unstandardized B coefficients and standard errors.
a
b
c
d
***: p<.001, **: p<.01, *: p<.05, : p=0.08, : p=0.08, : p=0.07, : p=0.06.
z
In these models, economic considerations were not included due to multicollinearity with the space grid. The explanatory power of the models is not negatively affected by this
exclusion.
Intercept
Explicit measures
EU-index
EU-grid
B
.470
Dependent: EU-skeptical information seeking
Group 1:
Group 2: medium
Group 3:
low political
political interest
high political
interest
interest
Table 5: Moderation-effects of political interest
Acknowledgements
Our work on this paper was supported by a grant from The Netherlands Institute for
Advanced Study in the Humanities and Social Sciences (NIAS) as well as by the Research
Network Communication, Media and Politics at the University of Koblenz-Landau. We thank
Andrea Retzbach, Anna Baumert, Tobias Rothmund and Manfred Schmitt for their
extraordinary helpful comments on our project and Maxim Egorov, Julian Erben, and Frank
Schneider for their work on the development of the instruments.
References
Arcuri, L., Castelli, L., Galdi, S., Zogmaister, C., & Amadori, A. (2008). Predicting the Vote: Implicit Attitudes
as Predictors of the Future Behavior of Decided and Undecided Voters. Political Psychology, 29(3), 369–
387. doi:10.1111/j.1467-9221.2008.00635.x
Bar-Anan, Y., & Nosek, B. A. (2012). Reporting Intentional Rating of the Primes Predicts Priming Effects in the
Affective Misattribution Procedure. Personality and Social Psychology Bulletin, 38(9), 1194–1208.
doi:10.1177/0146167212446835
Boninger, D. S., Krosnick, J. A., & Berent, M. K. (1995). Origins of attitude importance: Self-interest, social
identification, and value relevance. Journal of Personality and Social Psychology, 68(1), 61–80.
doi:10.1037/0022-3514.68.1.61
Boomgaarden, H. G., Schuck, A. R. T., Elenbaas, M., & Vreese, C. H. de. (2011). Mapping EU attitudes:
Conceptual and empirical dimensions of Euroscepticism and EU support. European Union Politics, 12(2),
241–266. doi:10.1177/1465116510395411
Bühner, M. (2011). Einführung in die Test- und Fragebogenkonstruktion (3., aktualisierte und erw. Aufl). PS
Psychologie. München [u.a.]: Pearson Studium.
Burdein, I., Lodge, M., & Taber, C. (2004). Implicit identifications in political information processing: An
experimental test of the hot identification hypothesis. Unpublished manuscript, Stony Brook,
Burdein, I., Lodge, M., & Taber, C. (2006). Experiments on the Automaticity of Political Beliefs and Attitudes.
Political Psychology, 27(3), 359–371. doi:10.1111/j.1467-9221.2006.00504.x
Chaiken, S., & Chen, S. (1999). The Heuristic-Systematic Model in Its Broader Context. In S. Chaiken & Y.
Trope (Eds.), Dual-process Theories in Social Psychology (pp. 73–96). New York, NY, USA: Guilford
Press.
Chaiken, S., & Trope, Y. (Eds.). (1999). Dual-process Theories in Social Psychology. New York, NY, USA:
Guilford Press.
Choma, B. L., & Hafer, C. L. (2009). Understanding the relation between explicitly and implicitly measured
political orientation: The moderating role of political sophistication. Personality and Individual Differences,
47(8), 964–967. doi:10.1016/j.paid.2009.07.024
Choma, B. L., & Hafer, C. L. (2009). Understanding the relation between explicitly and implicitly measured
political orientation: The moderating role of political sophistication. Personality and Individual Differences,
47(8), 964–967. doi:10.1016/j.paid.2009.07.024
Deutsch, R., & Strack, F. (2010). Building blocks of social behavior: Reflective and impulsive processes. In B.
Gawronski & B. K. Payne (Eds.), Handbook of Implicit Social Cognition: Measurement, Theory, and
Applications. Measurement, Theory, and Applications (pp. 62–79). Guilford Publications.
Evans,Jonathan St. B. T. (2008). Dual-Processing Accounts of Reasoning, Judgment, and Social Cognition.
Annual Review of Psychology, 59(1), 255–278. doi:10.1146/annurev.psych.59.103006.093629
Friese, M., Bluemke, M., & Wänke, M. (2007). Predicting Voting Behavior with Implicit Attitude Measures.
Experimental Psychology (formerly "Zeitschrift für Experimentelle Psychologie"), 54(4), 247–255.
doi:10.1027/1618-3169.54.4.247
Friese, M., Hofmann, W., & Schmitt, M. (2008). When and why do implicit measures predict behaviour?
Empirical evidence for the moderating role of opportunity, motivation, and process reliance. European
Review of Social Psychology, 19(1), 285–338. doi:10.1080/10463280802556958
Friese, M., Smith, C. T., Plischke, T., Bluemke, M., Nosek, B. A., & Krueger, F. (2012). Do Implicit Attitudes
Predict Actual Voting Behavior Particularly for Undecided Voters? PLoS ONE, 7(8), e44130.
doi:10.1371/journal.pone.0044130
Galdi, S., Arcuri, L., & Gawronski, B. (2008). Automatic Mental Associations Predict Future Choices of
Undecided Decision-Makers. Science, 321(5892), 1100–1102. doi:10.1126/science.1160769
Galdi, S., Gawronski, B., Arcuri, L., & Friese, M. (2012). Selective Exposure in Decided and Undecided
Individuals: Differential Relations to Automatic Associations and Conscious Beliefs. Personality and Social
Psychology Bulletin, 38(5), 559–569. doi:10.1177/0146167211435981
Gawronski, B., Galdi, S., & Arcuri, L. (in press). What Can Political Psychology Learn from Implicit Measures?
Empirical Evidence and New Directions. Political Psychology,
Gawronski, B., & Walther, E. (2008). The TAR Effect: When the Ones Who Dislike Become the Ones Who Are
Disliked. Personality and Social Psychology Bulletin, 34(9), 1276–1289. doi:10.1177/0146167208318952
Gawronski, B., & Houwer, J. de (Eds.). (2012). Implicit measures in social and personality psychology (2nd
ed.). New York: Cambridge University Press. Retrieved from https://biblio.ugent.be/publication/2959505
Gawronski, B., & Houwer, J. de. (2012). Implicit Measures in Social and Personality Psychology. In B.
Gawronski & J. de Houwer (Eds.), Implicit measures in social and personality psychology (2nd ed.). New
York: Cambridge University Press.
Gawronski, B., & Payne, B. K. (Eds.). (2010). Handbook of Implicit Social Cognition: Measurement, Theory,
and Applications: Measurement, Theory, and Applications: Guilford Publications. Retrieved from
http://www.lob.de/cgi-bin/work/suche2?titnr=257171065&flag=citavi
Gawronski, B., & Strack, F. (2004). On the propositional nature of cognitive consistency: Dissonance changes
explicit, but not implicit attitudes. Journal of Experimental Social Psychology, 40(4), 535–542.
doi:10.1016/j.jesp.2003.10.005
Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes.
Psychological Review, 102(1), 4–27. doi:10.1037/0033-295X.102.1.4
Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998). Measuring individual differences in implicit
cognition: The implicit association test. Journal of Personality and Social Psychology, 74(6), 1464–1480.
doi:10.1037//0022-3514.74.6.1464
Hofmann, W. (2005). A Meta-Analysis on the Correlation Between the Implicit Association Test and Explicit
Self-Report Measures. Personality and Social Psychology Bulletin, 31(10), 1369–1385.
doi:10.1177/0146167205275613
Hofmann, W., & Baumert, A. (2010). Immediate affect as a basis for intuitive moral judgement: An adaptation
of the affect misattribution procedure. Cognition & Emotion, 24(3), 522–535.
doi:10.1080/02699930902847193
Hofmann, W., Friese, M., & Roefs, A. (2009). Three ways to resist temptation: The independent contributions of
executive attention, inhibitory control, and affect regulation to the impulse control of eating behavior.
Journal of Experimental Social Psychology, 45(2), 431–435. doi:10.1016/j.jesp.2008.09.013
Hofmann, W., Gschwendner, T., Nosek, B. A., & Schmitt, M. (2005). What moderates implicit—explicit
consistency? European Review of Social Psychology, 16(1), 335–390. doi:10.1080/10463280500443228
Hooghe, L., & Marks, G. (2005). Calculation, Community and Cues: Public Opinion on European Integration.
European Union Politics, 6(4), 419–443. doi:10.1177/1465116505057816
Jost, J. T., Kay, A. C., & Thorisdottir, H. (Eds.). (2009). Series in political psychology. Social and psychological
bases of ideology and system justification. Oxford, New York: Oxford University Press.
Karpinski, A. (2005). Attitude Importance as a Moderator of the Relationship Between Implicit and Explicit
Attitude Measures. Personality and Social Psychology Bulletin, 31(7), 949–962.
doi:10.1177/0146167204273007
Kim, D.-Y. (2003). Voluntary Controllability of the Implicit Association Test (IAT). Social Psychology
Quarterly, 66(1), 83–96.
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (2008). International affective picture system (IAPS): Affective
ratings of pictures and instruction manual. Gainesville, FL: University of Florida.
Larsen, J. T., Norris, C. J., McGraw, A. P., Hawkley, L. C., & Cacioppo, J. T. (2009). The evaluative space grid:
A single-item measure of positivity and negativity. Cognition & Emotion, 23(3), 453–480.
doi:10.1080/02699930801994054
Leary, M. R., & Kowalski, R. M. (1990). Impression management: A literature review and two-component
model. Psychological bulletin, 107(1), 34–47.
Leary, M. R., & Kowalski, R. M. (1990). Impression management: A literature review and two-component
model. Psychological bulletin, 107(1), 34–47.
25
Lodge, M., & Taber, C. S. (2005). The Automaticity of Affect for Political Leaders, Groups, and Issues: An
Experimental Test of the Hot Cognition Hypothesis. Political Psychology, 26(3), 455–482.
doi:10.1111/j.1467-9221.2005.00426.x
Maier, M., Adam, S., & Maier, J. (2012). The impact of identity and economic cues on citizens' EU support: An
experimental study on the effects of party communication in the run-up to the 2009 European Parliament
elections. European Union Politics, 13(4), 580–603. doi:10.1177/1465116512453957
Miller, K. P., Brewer, M. B., & Arbuckle, N. L. (2009). Social Identity Complexity: Its Correlates and
Antecedents. Group Processes & Intergroup Relations, 12(1), 79–94. doi:10.1177/1368430208098778
Moosbrugger, H., & Kelava, A. (2012). Testtheorie und Fragebogenkonstruktion (2., aktualisierte und
überarbeitete Auflage). Springer-Lehrbuch. Berlin, Heidelberg: Springer-Verlag Berlin Heidelberg.
Nosek, B. A., Banaji, M. R., & Jost, J. T. (2009). The Social and Psychological Bases of Ideology and System
Justification. In J. T. Jost, A. C. Kay, & H. Thorisdottir (Eds.), Series in political psychology. Social and
psychological bases of ideology and system justification (pp. 480–506). Oxford, New York: Oxford
University Press.
Nosek, B. A. (2005). Moderators of the Relationship Between Implicit and Explicit Evaluation. Journal of
Experimental Psychology: General, 134(4), 565–584. doi:10.1037/0096-3445.134.4.565
Nosek, B. A. (2007). Implicit-Explicit Relations. Current Directions in Psychological Science, 16(2), 65–69.
doi:10.1111/j.1467-8721.2007.00477.x
Oikawa, M., Aarts, H., & Oikawa, H. (2011). There is a fire burning in my heart: The role of causal attribution in
affect transfer. Cognition & Emotion, 25(1), 156–163. doi:10.1080/02699931003680061
Paulhus, D. L. (1984). Two-component models of socially desirable responding. Journal of Personality and
Social Psychology, 46(3), 598–609. doi:10.1037/0022-3514.46.3.598
Payne, B. K., Cheng, C. M., Govorun, O., & Stewart, B. D. (2005). An inkblot for attitudes: Affect
misattribution as implicit measurement. Journal of Personality and Social Psychology, 89(3), 277–293.
doi:10.1037/0022-3514.89.3.277
Payne, B. K., Govorun, O., & Arbuckle, N. L. (2008). Automatic attitudes and alcohol: Does implicit liking
predict drinking? Cognition & Emotion, 22(2), 238–271. doi:10.1080/02699930701357394
Payne, B. K., Krosnick, J. A., Pasek, J., Lelkes, Y., Akhtar, O., & Tompson, T. (2010). Implicit and explicit
prejudice in the 2008 American presidential election. Journal of Experimental Social Psychology, 46(2),
367–374. doi:10.1016/j.jesp.2009.11.001
Payne, B. K., McClernon, F. J., & Dobbins, I. G. (2007). Automatic affective responses to smoking cues.
Experimental and Clinical Psychopharmacology, 15(4), 400–409. doi:10.1037/1064-1297.15.4.400
Retzbach, A., Maier, M., & Jahn, N. (under review). The formation of explicit and autimatic affective
evaluations of nanotechnology and the influence on risk judgements and consumer behavior.
Roccato, M., & Zogmaister, C. (2010). Predicting the Vote through Implicit and Explicit Attitudes: A Field
Research. Political Psychology, 31(2), 249–274. doi:10.1111/j.1467-9221.2009.00751.x
Rydell, R. J., McConnell, A. R., & Mackie, D. M. (2008). Consequences of discrepant explicit and implicit
attitudes: Cognitive dissonance and increased information processing. Journal of Experimental Social
Psychology, 44(6), 1526–1532. doi:10.1016/j.jesp.2008.07.006
Smith, E. R., & DeCoster, J. (2000). Dual-Process Models in Social and Cognitive Psychology: Conceptual
Integration and Links to Underlying Memory Systems. Personality and Social Psychology Review, 4(2),
108–131. doi:10.1207/S15327957PSPR0402_01
Steffens, M. C. (2004). Is the Implicit Association Test Immune to Faking? Experimental Psychology (formerly
"Zeitschrift für Experimentelle Psychologie"), 51(3), 165–179. doi:10.1027/1618-3169.51.3.165
Strack, F., & Deutsch, R. (2004). Reflective and Impulsive Determinants of Social Behavior. Personality and
Social Psychology Review, 8(3), 220–247. doi:10.1207/s15327957pspr0803_1
Teige-Mocigemba, S., Klauer, K. C., & Sherman, J. W. (2010). A practical guide to the Implicit Association
Test and related tasks. In B. Gawronski & B. K. Payne (Eds.), Handbook of Implicit Social Cognition:
Measurement, Theory, and Applications. Measurement, Theory, and Applications (pp. 117–139). Guilford
Publications.
Wittenbrink, B., Gist, P. L., & Hilton, J. L. (1997). Structural properties of stereotypic knowledge and their
influences on the construal of social situations. Journal of Personality and Social Psychology, 72(3), 526–
543. doi:10.1037/0022-3514.72.3.526
Zaller, J. (1992). The nature and origins of mass opinion. Cambridge [England], New York, NY, USA:
Cambridge University Press.
26
APPENDIX
I.
II.
Additional Information on the AMP
Information on the explicit attitudes measures
I. Additional Information on the AMP
Payne et al. (2005) estimation of reliability is slightly different to our approach. They used
Cronbach’s alpha in order to estimate the reliability of their AMP difference scores. But a
Cronbach’s alpha reliability estimation needs a fuel data set, otherwise the entire case is
removed from the analysis. Since single AMP trials were excluded from our data (e.g. due to
slow reaction times), Cronbach’s alpha is not a robust estimator of reliability in our case and
therefore split-half reliability estimations were reported in the method section above (Bühner,
2011). Nevertheless we adapted Payne et al. (2005) Cronbach’s alpha estimation for our data
as well. So, for each of the three difference scores (EU-neutral; Germany-neutral and EUGermany) a set of 24 single difference scores was created and treated as individual items.
First, each prime was scored as +1 for a “good” evaluation or 0 for a “bad” evaluation. Then a
score on each randomly selected EU prime trial was subtracted from a randomly selected
neutral prime trial. The same procedure was repeated for every single Germany prime minus a
randomly selected neutral prime and for every single EU prime minus a randomly selected
Germany prime. Each trial was used in only one pair. This created three sets of 24 difference
scores which could each range between -1, 0, and +1. For each set the 24 difference scores
were used for a Cronbach’s alpha reliability estimation. The analysis for the difference score
EU-neutral revealed a value of α = .55 (n=71); for Germany-neutral a value of α = .60 (n=71)
and for EU-Germany a value of α = .58 (n=64).
Neutral Reference Prime Pictures
27
Picture are used based on the IAPS (Lang et al., 2008). IDs from the top left to the right
bottom: 7000, 7004, 7006, 7009, 7010, 7080, 7150, 7705, 7002, 7175, 7235, and 7186.
The neutral images were selected based on mean valence (target value = 5; with SDs as small
as possible and mean arousal ratings (M and SD as small as possible).
Description
ID
Valence mean
Valence SD
Arousal mean
Arousal SD
Rolling Pin
7000
5.00
0.84
2.42
1.79
Spoon
7004
5.04
0.60
2.00
1.66
Bowl
7006
4.88
0.99
2.33
1.67
Mug
7009
4.93
1.00
3.01
1.97
Basket
7010
4.94
1.07
1.76
1.48
Fork
7080
5.27
1.09
2.32
1.84
Umbrella
7150
4.72
1.00
2.61
1.76
Cabinet
7705
4.77
1.02
2.65
1.88
Towel
7002
4.97
0.97
3.16
2.00
Lamp
7175
4.87
1.00
1.72
1.26
Chair
7235
4.96
1.18
2.83
2.00
Abstract Art
7186
4.63
1.6
3.6
2.36
II. Information on the explicit attitude measures
Explicit measures for EU attitudes (combined as index):
a) All in all, how positive or negative is the image you have from the European Union?
5 = very positive
4 = quite positive
3 = partly positive, partly negative
2 = quite negative
1 = very negative
28
b) Do you think that Germany’s membership in the European Union is a good thing?
5 = completely agree
4 = rather agree
3 = partly agree, partly disagree
2 = rather disagree
1 = completely disagree
c) In general, do you think that Germany has gained more advantages than disadvantages
from membership in the European Union?
5 = completely agree
4 = rather agree
3 = partly agree, partly disagree
2 = rather disagree
1 = completely disagree
d) The European Union works well as it is.
5 = completely agree
4 = rather agree
3 = partly agree, partly disagree
2 = rather disagree
1 = completely disagree
29