APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
Journal of Experimental Psychology:
Human Perception and Performance
2015, Vol. 41, No. 5, 000
© 2015 American Psychological Association
0096-1523/15/$12.00 http://dx.doi.org/10.1037/xhp0000130
It Depends: Approach and Avoidance Reactions to Emotional Expressions
are Influenced by the Contrast Emotions Presented in the Task
AQ: au
Andrea Paulus and Dirk Wentura
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Saarland University
Studies examining approach and avoidance reactions to emotional expressions have yielded conflicting
results. For example, expressions of anger have been reported to elicit approach reactions in some studies
but avoidance reactions in others. Nonetheless, the results were often explained by the same general
underlying process, namely the influence that the social message signaled by the expression has on
motivational responses. It is therefore unclear which reaction is triggered by which emotional expression,
and which underlying process is responsible for these reactions. In order to address this issue, we
examined the role of a potential moderator on approach and avoidance reactions to emotional expressions, namely the contrast emotion used in the task. We believe that different approach and avoidance
reactions occur depending on the congruency or incongruency of the evaluation of the 2 emotions
presented in the task. The results from a series of experiments supported these assumptions: Negative
emotional expressions (anger, fear, sadness) elicited avoidance reactions if contrasted with expressions
of happiness. However, if contrasted with a different negative emotional expression, anger and sadness
triggered approach reactions and fear activated avoidance reactions. Importantly, these results also
emerged if the emotional expression was not task-relevant. We propose that approach and avoidance
reactions to emotional expressions are triggered by their evaluation if the 2 emotions presented in a task
differ in evaluative connotation. If they have the same evaluative connotation, however, reactions are
determined by their social message.
Keywords: emotional expression, approach/avoidance, behavioral reaction, contrast, intention
diate behavioral reactions (e.g., Marsh et al., 2005; Roelofs et
al., 2010; Seidel et al., 2010; Wilkowski & Meier, 2010). At
first sight, the results from the individual studies seem to
support the notion that approach and avoidance reactions are
triggered by the social message signaled by the emotional
expression. However, across studies the results were often
inconsistent: While some studies observed faster avoidance
reactions to expressions of anger (Marsh et al., 2005; Rotteveel
& Phaf, 2004; Seidel et al., 2010; Stins et al., 2011), others
found faster approach reactions (Krieglmeyer & Deutsch, 2013;
Wilkowski & Meier, 2010). Similar inconsistent results were
reported for fear expressions (Marsh et al., 2005; Wilkowski &
Meier, 2010). Interestingly, these inconsistent results were often explained by the same general underlying process, namely
that the social message signaled by the emotional expression triggers
behavioral tendencies: Faster avoidance reactions to anger expressions, for example, were explained as adaptive reactions triggered by
the direct threat signaled by the expression (Marsh et al., 2005). Faster
approach reactions to the same expression, on the other hand, were
attributed to the social challenge signaled by the expression and
the adaptive value of approaching the opponent in order to overcome
the challenge (see, e.g., Wilkowski & Meier, 2010). This example
demonstrates that it is not clear which reaction is triggered by
which expression, and which underlying process is responsible for
the observed reactions. The aim of our study was to address these
questions by examining the role of a possible moderator: The
contrast emotion with which the emotion in question is paired in an
experiment.
Emotional expressions are powerful social signals—they convey information about the affective state of the expresser but also
communicate a social message (e.g., Horstmann, 2003; Paulus,
Musial, & Renn, 2014; Paulus & Wentura, 2014). This social
message informs the interaction partner about the expresser’s
behavioral intentions and requests. A happy face, for example,
usually communicates the expresser’s wish for affiliation, whereas
an angry expression can signal the expresser’s intention to attack.
This information is of direct relevance for the perceiver. Accordingly, it has been suggested that an emotional expression triggers
immediate behavioral responses like approach and avoidance reactions (Marsh, Ambady, & Kleck, 2005; Paulus & Wentura,
2014; Seidel, Habel, Kirschner, Gur, & Derntl, 2010; Stins et al.,
2011; Wilkowski & Meier, 2010). This assumption is in line with
theories stating that approach and avoidance reactions are motivational reactions that are functional for the organism’s current needs
(e.g., Lang & Bradley, 2013).
In order to examine these assumptions empirically, a number
of studies recently examined if emotional faces trigger imme-
Andrea Paulus and Dirk Wentura, Faculty of Behavioral Sciences,
Department of Psychology, Saarland University.
Correspondence concerning this article should be addressed to Andrea
Paulus, Faculty of Behavioral Sciences, Department of Psychology, Saarland University, Building A2 4, D-66123 Saarbrücken, Germany. E-mail:
[email protected]
1
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
PAULUS AND WENTURA
2
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
The Contrast Emotion Matters
T1
To the best of our knowledge, all published experiments examining approach and avoidance reactions to emotional expressions
employed two different emotional expressions together. Typically,
participants’ task was to classify the expressions by executing
approach or avoidance movements, for example by moving a
virtual figure toward or away from an emotional face appearing on
the screen depending on the expression (with a reversal of instructions after half of the trials; see, e.g., Krieglmeyer & Deutsch,
2013). The results were then interpreted separately for each expression (i.e., anger and fear in our example) and theoretically
generalized to other contexts. However, we believe that the respective contrast emotion codetermines the character of the task.
Therefore, the reactions to one expression will be influenced by the
other emotion presented in the task.
Why do we predict that the contrast emotion influences approach and avoidance reactions to emotional expressions? We
believe that the answer lies in the congruency or incongruency of
the evaluative component of the two emotions (i.e., whether the
two emotions are both negative or both positive, or whether one is
positive, the other negative) presented in the task: Depending on
this congruency, the behavioral reactions will be triggered by
different processes, thereby causing different results. If the two
concurrent emotions differ on the evaluative dimension, this is the
most salient and simple feature that distinguishes the two emotions. Accordingly, approach and avoidance reactions will be
triggered by the evaluation of the respective emotions in such a
situation (i.e., positive stimuli typically elicit approach tendencies
whereas negative stimuli typically elicit avoidance tendencies;
e.g., Chen & Bargh, 1999; Solarz, 1960). If both expressions have
the same evaluative connotation (e.g., anger and fear), however,
this dimension becomes much less salient. In this case, other
features of the expressions have to be used to differentiate the two
expressions. Accordingly, approach and avoidance reactions will
then be influenced by this dimension. This might result in an
approach related reaction to specific negative expressions. Our
approach therefore suggests that the same emotional expression
can trigger different reactions depending on the contrast emotion it
is paired with.
The question whether the reaction to one emotion is influenced
by the choice of the contrast emotion has not been empirically
addressed (but see Krieglmeyer & Deutsch, 2013, for a similar
argument). This is particularly surprising given the two expressions used in approach and avoidance experiments often differed:
Anger, for example, has been paired with happiness (e.g., Roelofs
et al., 2010; Rotteveel & Phaf, 2004; Stins et al., 2011), fear
(Krieglmeyer & Deutsch, 2013; Marsh et al., 2005; Wilkowski &
Meier, 2010), and sadness (Seidel et al., 2010).
We assume that this difference between studies might help to
explain some of the inconsistent results regarding approach and
avoidance reactions to emotional expressions (see Table 1 for an
overview of the relevant studies). According to our rationale, a
negative emotion should yield avoidance behavior if it is paired
with happy expressions (which should be associated with approach). Indeed, Rotteveel and Phaf (2004), Roelofs et al. (2010),
as well as Stins et al. (2011) paired anger with happiness and
reported anger-related avoidance behavior; Seidel, Habel, Kirschner, Gur, and Derntl (2010) paired disgust and happiness and
Table 1
Overview of Approach/Avoidance Studies, Sorted by Congruency
and Incongruency of Evaluative Contrast
Study
Positive/negative contrast
Rotteveel & Pfaf
(2004); Exp. 1
Roelofs et al. (2010)
Stins et al. (2011)
Seidel et al. (2010)
Negative/negative contrast
Bossuyt et al. (2014))a
Krieglmeyer & Deutsch
(2013); Exp. 2a
Seidel et al. (2010)
Wilkowski & Meier
(2010); Exp. 2 & 4
Bossuyt et al. (2014)b
Krieglmeyer & Deutsch
(2013); Exp. 2b
Marsh et al. (2005)
Happy
AS
AS
AS
AS
Anger
⬎
⬎
⬎
⬎
Fear Sad Disgust
AS
AS
AS
AS
AS
⬎
AS
AS
AS
⬎
⬍
AS
AS
AS
⬎
⬍
AS
AS
AS
AS
⬍
⬍
AS
AS
AS
Note. AS ⫽ approach score (i.e., RTs of approach movements minus RTs
of avoidance movements).
This result was observed if the approach reaction was conceptualized as
an aggressive/dominant approach (see General Discussion section for a
detailed description). b This result was observed if the approach reaction
was conceptualized as an affiliative/submissive approach (see General
Discussion section for a detailed description).
a
reported disgust-related avoidance behavior. Neither fearful nor
sad faces have hitherto been contrasted with happy faces.
We furthermore assume that if two negative emotions are
paired, behavioral reactions are triggered by a different feature
than the evaluative connotation of the expressions; this might
result in approach related reactions to negative expressions. There
is abundant evidence for behavioral tendencies associated with
negative emotions that are in contrast to avoidance behavior. For
example, Wilkowski and Meier (2010) paired anger with fear
expressions and reported anger-related approach behavior and
fear-related avoidance behavior. This also holds for the studies by
Krieglmeyer and Deutsch (2013) and Bossuyt, Moors, and De
Houwer (2014).1 Using the same pairings, Marsh, Ambady, and
Kleck (2005) found approach-related behavior to expressions of
fear and avoidance-related behavior to expressions of anger (see
General Discussion section for a discussion of the inconsistencies
between studies). Seidel et al. (2010) found sadness to trigger
approach behavior if paired with expressions of anger.
Which feature of the emotional expression might be responsible
for the reactions if two expressions with the same evaluative
connotation are contrasted? We propose that such a feature might
be the social message signaled by the emotional expressions. We
believe the results regarding approach reactions to angry faces and
avoidance reactions to fearful ones are in line with such an explanation: It is sensible to approach a social opponent to overcome the
challenge signaled by their angry face, and it is also adaptive to
1
As can be seen in Table 1, Krieglmeyer and Deutsch (2013) and
Bossuyt et al. (2014) varied the meaning of the behavioral responses. We
will elaborate on this issue in the General Discussion section.
Fn1
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
APPROACH/AVOIDANCE REACTIONS TO EMOTIONAL EXPRESSIONS
activate an avoidance reaction in the presence of a fearful face
signaling a threat in the environment (but see Marsh et al., 2005 for
a different argument, see General Discussion section).The pattern
found for sad faces fits a social meaning account as well: Even
though sadness has a negative evaluative connotation, it signals a
request for help. Therefore, this message should elicit an approach
reaction (if it is not contrasted with happy faces).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Empirical and Theoretical Support for the “Contrast
Emotion” View
Even though there is no direct empirical support for our assumption that the contrast emotion influences approach and avoidance
reactions (i.e., evidence from studies directly examining approach
and avoidance), there exists empirical and theoretical support for
our hypothesis in the broader social cognition literature: First,
Bijlstra, Holland, and Wigboldus (2010), examined the role of the
contrast emotion in a series of experiments exploring the impact of
the expresser’s social category on emotion recognition. Earlier
research in this field had already shown that emotional expressions
were recognized faster if the evaluation of the emotion and the
evaluation of the expresser’s social category matched (e.g., Hugenberg, 2005; Hugenberg & Sczesny, 2006). Bijlstra and colleagues,
however, were able to demonstrate that depending on the evaluative congruency of the two emotions presented in the task, different patterns emerged: If the two emotions differed in evaluation
(e.g., anger/sadness vs. happiness), emotion recognition was influenced by the evaluation of the expresser’s social category (e.g.,
White Dutch vs. Moroccan Dutch), replicating the known pattern.
If the two emotions shared evaluation, however, emotion recognition was influenced by the semantic match between the expresser’s
social category and the expressed emotion, leading to different
results for the same expression (e.g., faster reactions to Dutch
Moroccans expressing anger than to Dutch Moroccans expressing
sadness). This study showed that the same expression can have
different effects depending on the contrast emotion used, and thus
provides support for our notion that the contrast emotion can
determine which feature of the emotion drives the reaction.
Second, effects of contrast emotion have been found in studies
examining emotion recognition with faces that seem to approach
or withdraw from the perceiver. One study found that withdrawing
angry and approaching happy faces were recognized faster than the
opposite composition (van Peer, Rotteveel, Spinhoven, Tollenaar,
& Roelofs, 2010); another study, however, observed the reverse
result for angry faces if they were paired with fearful expressions
(Nelson, Adams, Stevenson, Weisbuch, & Norton, 2013). This
pattern further supports the notion that the contrast emotion influences reactions to emotional expressions.
From a theoretical perspective, support for our assumptions
comes from the evaluative coding account of approach and avoidance reactions (Eder & Rothermund, 2008): This theory explains
the approach/avoidance effects found with standard positive and
negative stimuli with reference to the match between the stimuli’s
evaluation and the evaluation of the respective movement. More
precisely, it states that the mental representations of approach and
avoidance behavior include a positive and a negative connotation.
If in a given trial the evaluation of the stimuli matches the evaluation of the executed behavior’s mental representation, reaction
times (RTs) are faster compared with trials with a mismatch (but
3
see Krieglmeyer, Deutsch, De Houwer, & De Raedt, 2010). This
theory therefore meshes well with our assumptions, in that it posits
that approach and avoidance movements in response to positive
and negative stimuli are elicited by the coding of the stimuli’s
evaluation and not necessarily their motivational meaning.
Taken together, there is both (indirect) empirical as well as
theoretical support for our assumption that approach and avoidance movements are determined by the evaluative component if
the two presented emotions differ in evaluative connotation and by
a different feature of the expression if they do not. Therefore, we
believe it would be fruitful to systematically examine the role of
the contrast emotion in approach and avoidance experiments. That
was the main goal of our study. The second goal was to examine
the causal mechanism underlying approach and avoidance reactions if the two emotions differ in evaluation. We predict that in
this situation reactions are triggered by the social message signaled
by the emotional expressions.
Overview
We present six experiments, all of which employed the same
basic procedure: Participants reacted to emotional expressions
displayed on a computer screen by moving a small virtual figure
toward or away from the expression (e.g., De Houwer, Crombez,
Baeyens, & Hermans, 2001; Krieglmeyer & Deutsch, 2010). The
time to initiate the movement served as the dependent variable.
The main goal of the experiments was to examine how the evaluation of one expression (the contrast expression) influences reactions to the other expression (the target expression), when evaluative congruency is manipulated.
In Experiments 1a and 1b, participants worked through one
block of approach and avoidance tasks pairing a negative target
expression (i.e., anger, fear) with a positive contrast expression
(i.e., happiness). This was done to establish the basic finding that
anger and fear are both associated with avoidance reactions if
paired with happiness. Participants in the remaining experiments
always completed two blocks of approach and avoidance tasks. In
one of the blocks, a negative target expression (i.e., anger, fear, or
sadness) was paired with a positive contrast expression (happiness), whereas in the other block, the same target expression was
paired with a different negative contrast expression (i.e., anger or
fear). The order of the two blocks was counterbalanced between
participants.
We chose anger, fear, and sadness as target expressions for
several reasons. First of all, studies examining approach and avoidance reactions to anger and fear have yielded conflicting results:
Both expressions have been reported to elicit approach as well as
avoidance reactions. In most of the studies, the two emotions were
either contrasted with each other (Krieglmeyer & Deutsch, 2013;
Marsh et al., 2005; Wilkowski & Meier, 2010) or with happiness
(Rotteveel & Phaf, 2004; Seidel et al., 2010; Stins et al., 2011).
Therefore, we selected these expressions to shed some light on the
inconsistent findings.
Second, we selected those emotions as target expressions because they have the same evaluative connotation but signal different social meanings, which according to our assumptions should
elicit different reactions: Whereas the social challenge signaled by
expressions of anger and the request for help signaled by expressions of sadness should elicit an approach reaction, the warning of
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
PAULUS AND WENTURA
4
T2
a common threat signaled by expressions of fear should elicit an
avoidance reaction. Thus, while we expected all three emotions to
elicit an avoidance reaction if contrasted with expressions of
happiness, we expected anger and sadness to elicit an approach
reaction and fear to elicit avoidance if contrasted with a different
negative expression (see Table 2). Therefore, reactions to anger
and sadness should be influenced by the evaluative connotation of
the contrast expression, whereas reactions to fear should not differ.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Experiments 1a and 1b
Experiments 1a and 1b were designed to examine the influence
of a positive contrast emotion on approach and avoidance reactions
to negative emotional expressions. We used happiness as the
contrast emotion, and measured approach and avoidance reactions
to angry expressions (Experiment 1a) and fearful expressions
(Experiment 1b). We hypothesized that fearful as well as angry
expressions would elicit faster avoidance than approach reactions
if contrasted with happiness. In both experiments, participants’
task was to categorize the emotional expression by moving a small
virtual figure toward or away from the expression by pressing
assigned keys on a computer keyboard. Participants were instructed to imagine that the figure represented themselves (see,
e.g., De Houwer et al., 2001; Krieglmeyer & Deutsch, 2010 for a
similar procedure).
Method
Participants. In Experiment 1a, 56 nonpsychology undergraduate students (27 women, 29 men) from Saarland University
participated. The age range was 19 to 34 years with a median of 23
years. In Experiment 1b, 41 nonpsychology undergraduate students (25 women, 16 men) from Saarland University participated.
The age range was 17 to 26 years with a median of 22 years. Both
experiments lasted approximately 20 minutes. Participants were
paid 4 Euros.
Design. Both experiments followed a 2 (movement: approach,
avoidance) ⫻ 2 (emotional expression: target expression [fear/
anger], contrast expression [happiness]) design. In Experiment 1a,
expressions of anger were presented together with expressions of
happiness, and in Experiment 1b, expressions of fear were presented together with expressions of happiness. Both factors were
varied within participants.
The differential influence of emotional expression on the reaction time (RT) of approach and avoidance reactions was relevant
for our power considerations. It is associated with dfN ⫽ 1 and its
effect size can therefore be expressed as dZ (Cohen, 1988). Examining approach and avoidance reactions to positive and negative
words with a task similar to the one used in our experiments (the
Table 2
Expected Behavioral Reactions to the Target Emotions as a
Function of the Contrast Emotions
Target emotion
Positive contrast
emotion
Negative contrast
emotion
Anger
Sadness
Fear
avoidance
avoidance
avoidance
approach
approach
avoidance
so-called manikin task), Krieglmeyer and Deutsch (2010, Exp. 1)
found dZ in the range of 0.50 to 0.79. Looking at approach and
avoidance reactions to angry and happy faces, Rotteveel and Phaf
(2004; Exp. 1) reported an effect of dZ ⫽ 0.54. Thus, a conservative assumption is to assume dZ ⫽ 0.50 (i.e., a “medium” effect as
defined by Cohen, 1988). To detect an effect of this magnitude
with a probability of 1 ⫺ ␤ ⫽ .80, an ␣-value of .05, a minimum
sample size of 34 participants was required (actual power with
N ⫽ 41 [56] was 1 ⫺ ␤ ⫽ .88 [.96]; calculations were done using
G.Power 3.1.3; Faul, Erdfelder, Lang, & Buchner, 2007).
Material. Stimuli consisted of emotional expressions shown
by men and women. Pictures were taken from the Radboud Faces
Database (Langner et al., 2010), the Amsterdam Dynamic Facial
Expression Set (van der Schalk, Hawk, Fischer, & Doosje, 2001),
the Warsaw set of emotional facial expression pictures (Olszanowski, Pochwatko, Kukliński,Ścibor-Rylski, & Ohme, 2008), and
our own collection (Paulus, Rohr, Neuschwander, Seewald, &
Wentura, 2012). For Experiment 1a, we created two stimulus sets,
each consisting of the images of five men and five women expressing anger and happiness. Use of stimulus sets was counterbalanced between participants. This was done in order to enhance
the generalizability of the results. For Experiment 1b, we selected
fear and happiness expressions shown by four women and four
men.2 All images depicted headshots with a straight orientation
and gaze directed at the viewer. The face and the top of the neck
were shown on a white background. The images measured 240 ⫻
240 pixels and were presented on a CRT display set to a resolution
of 1,024 ⫻ 786 pixels.
Procedure. The procedure closely followed the one described
in Krieglmeyer and Deutsch (2010): Each trial started with a
fixation cross presented in the middle of the screen. Participants
were instructed to press and hold the “5” key on the number pad
as soon as the fixation cross appeared. Triggered by the key press,
a virtual stick figure appeared either below or above the fixation
cross. After 750 ms, the fixation cross was replaced by an emotional expression (positioning of the figure was counterbalanced
such that the figure appeared above and below each expression
equally often). Participants were instructed to move the figure
toward or away from the expression by pressing either the “2” or
the “8” key on the number pad three times. Each press of “2”
moved the figure up 20 pixels and each press of “8” moved it down
20 pixels. Key presses temporarily shortened one of the figure’s
legs in an alternate fashion, creating the impression that it walked
toward or away from the emotional expression. After three keypresses, the screen turned blank. Half of the participants first
categorized the expressions by moving the figure toward happy
expressions and away from angry (Experiment 1a) and fearful
(Experiment 1b) expressions. For the other half of participants, the
instructions were reversed. After half of the trials, the assignment
of movement direction to emotional expression was reversed.
Participants were instructed to react as quickly and accurately as
possible. They received an error message if their reaction was
incorrect.
2
These small differences in number of stimuli between experiments
were caused by the fact that they were not planned and conducted in strictly
parallel order. However, we do not believe these differences to be a
problem since each main result is replicated by the result of a further
experiment in this article.
Fn2
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
APPROACH/AVOIDANCE REACTIONS TO EMOTIONAL EXPRESSIONS
Experiment 1a comprised a total of 160 trials; Experiment 1b
featured 128 trials. There were 16 practice trials in both experiments, using emotional expressions from two men and two women
not presented in the main experiment.
The time lapse between the release of “5”and the first press of
“2” or “8” was defined as the RT.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Results
Fn3
Fn4
F1, F4
T3
Fn5
Trials with an incorrect response (Experiment 1a: 4.3%; Experiment 1b: 3.8%) and trials with outlying RTs (i.e., RTs three
interquartile ranges above the third or below the first quartile;
Tukey, 1977) with respect to the individual distribution (Experiment 1a: 1.4%; Experiment 1b: 1.0%) were discarded from the
analyses.3
For the sake of brevity and comprehensibility, we calculated
difference scores in both experiments. That is, we subtracted RTs
of approach movements from RTs of avoidance movements
(termed approach score for the remainder of the article; see Seidel
et al., 2010; Krieglmeyer & Deutsch, 2013 for a similar approach).4 These approach scores indicate the relation between
approach and avoidance movements: A higher (lower) difference
score indicates relatively more (less) approach compared to avoidance related behavior. Please note that in all experiments, approach
reactions were generally faster than avoidance reactions (i.e., a
main effect of movement). This effect has repeatedly been reported
in the literature (e.g., Krieglmeyer & Deutsch, 2013; Rinck &
Becker, 2007; Wilkowski & Meier, 2010). Thus, our focus will be
on tests of relative differences, indicating whether mean approach
scores differ significantly for the two emotions, and whether each
emotion score indicates relatively more approach or avoidance.
Because of the main effect of movement, it is not meaningful to
test whether the mean approach score for a given emotion deviates
from zero (i.e., testing whether the RTs for approach and avoidance reactions differ for a given emotion). However, to anticipate,
in order to obtain an estimate of the main effect that is independent
of emotion-induced facilitations or reductions, we included baseline trials in Experiments 3 and 4. In these trials, participants
responded to arrows pointing upward or downward by moving the
virtual figure in the respective direction (see Krieglmeyer &
Deutsch, 2013). These trials provided an estimate of the approach
bias of c ⫽ 20 ms (see the Appendix for details). We added this
constant as a baseline in Figures 1– 4 to give an indication of
whether results for a given emotion reflect approach/avoidancerelatedness (only) in relative terms or in absolute terms. Mean RTs
for Experiment 1a and 1b are presented in Table 3.
Experiment 1a. We compared the difference scores for expressions of anger and happiness with an ANOVA for repeated
measures. The results showed that the relation between approach
and avoidance movements differed for the two expressions, F(1,
55) ⫽ 10.81, p ⫽ .002, ␩p2 ⫽ .164 (dZ ⫽ 0.445): Whereas approach
reactions were faster than avoidance reactions for expressions of
happiness, the pattern was reversed for expressions of anger (see
left panel of Figure 1). As can be seen in Figure 1, the constant c
(as an estimate of the approach-bias of the task; see above) is not
included in the 95% CI for happiness (thereby yielding evidence
for approach tendencies in an absolute sense). For fear, the constant c was (just) included in the confidence interval (i.e., the CI
ranges from ⫺55 ms to 22 ms).
5
Experiment 1b. A repeated measures ANOVA comparing the
approach scores for expressions of fear and happiness yielded the
expected significant effect, F(1, 40) ⫽ 9.98, p ⫽ .003, ␩p2 ⫽ .200
(dZ ⫽ 0.49). The means show that participants were faster when
approaching happy expressions than when avoiding them, whereas
the opposite was true for fear (see right panel of Figure 1). As can
be seen in Figure 1, the constant c (as an estimate of the approachbias of the task) again is not included in the 95% CI for happiness
but included in the confidence interval for fear (i.e., the CI ranges
from ⫺38 ms to 25 ms). However, collapsed across anger (Exp.
1a) and fear (Exp. 1b), the CI ranged from ⫺38 ms to 13 ms,
indicating that evaluation-based avoidance tendencies were found
not only in relative but also in absolute terms.
Discussion
The results of Experiments 1a and 1b support our hypotheses:
When contrasted with expressions of happiness, angry as well as
fearful expressions were avoidance related. This is an important
finding because both expressions have been linked with approach
and avoidance reactions (e.g., Marsh et al., 2005; Wilkowski &
Meier, 2010). Our experiments provide evidence that both expressions elicit the same reaction if they are contrasted with oppositely
evaluated expressions of happiness. These experiments are therefore the first step in showing that behavioral reactions to anger and
fear are not only determined by the emotions themselves but also
by the contrast emotion used.
The result that anger as well as fear expressions elicit avoidance
reactions when paired with expressions of happiness can also help
to understand the process underlying approach and avoidance
reactions to emotional expressions: It seems likely that the reactions in this set-up were triggered by the evaluation of the emotional expressions and not by their respective social messages
because those should generate opposing results: The social message signaled by an anger expression should elicit an approach
reaction and the social message signaled by a fearful expression an
avoidant one. This is an important finding, because several studies
examining approach and avoidance reactions to emotional expressions contrasted a negative expression with a positive one and
argued that effects were caused by the behavioral intention and not
the evaluation of the emotional expressions (e.g., Roelofs et al.,
2010; Seidel et al., 2010). We suggest that these conclusions have
to be reconsidered.
However, to shed light on the bigger picture it is important to
examine if anger and fear expressions are also avoidance related if
they are contrasted with a negative instead of a positive emotion.
This was the aim of Experiments 2a and 2b.
3
Different outlier criteria (i.e., lower and higher general cutoffs, individual cutoffs) essentially led to the same results. This holds for all
experiments throughout the article.
4
Note that all tests using the approach score as the DV are equivalent
to tests involving a within-participants factor of movement (with the
conditions approach vs. avoidance); for example, the test for a main effect
of emotion with regard to the DV approach score is equivalent to the test
for an emotion ⫻ movement interaction with RTs as the DV.
5
We only report dz for the results which were critical for the power
calculations for the following experiments.
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
PAULUS AND WENTURA
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
6
Figure 1. Mean approach scores relating to anger and happiness expressions (Experiment 1a) and fear and
happiness expressions (Experiment 1b). The dotted line indicates the constant c.
Experiments 2a and 2b
Participants in Experiments 2a and 2b received two blocks of
approach and avoidance tasks: In one of the blocks, they performed approach and avoidance reactions to expressions of happiness and anger (Experiment 2a) or expressions of happiness and
fear (Experiment 2b). In the other block, they reacted to expressions of anger and fear (Experiments 2a and 2b). Anger was thus
the target expression in Experiment 2a, and fear was the target
expression in Experiment 2b. Happiness and anger/fear, respectively, served as contrast expressions.
Regarding Experiment 2a, we expected a difference in approach
and avoidance reactions to anger expressions depending on
whether the contrast emotion was happiness or fear: Reactions to
anger expressions should be determined by anger’s negative evaluation if the contrast emotion is positively evaluated (thus making
the evaluative connotation salient); they should be determined by
anger’s social message if the contrast emotion also had a negative
evaluative connotation (as the evaluative congruency should make
the difference between the specific emotions on the social dimension more salient). Therefore, we expected anger to elicit avoidance if contrasted with expressions of happiness (thereby replicating Experiment 1a) but approach if contrasted with expressions of
fear. The latter prediction was derived from the fact that a series of
recently published studies has indicated that expressions of anger
elicit approach rather than avoidance reactions, in particular if the
Table 3
Mean Reaction Times for Approach and Avoidance Movements
(Standard Deviations in Parentheses; Experiment 1a, 1b)
Movement
Experiment
Emotion
Approach
Avoidance
Experiment 1a
Happiness
Anger
Happiness
Fear
729 (202)
792 (222)
712 (144)
765 (141)
817 (255)
776 (234)
799 (141)
758 (151)
Experiment 1b
approach reactions do not serve an affiliative goal (Bossuyt,
Moors, & De Houwer, 2014; Krieglmeyer & Deutsch, 2013;
Wilkowski & Meier, 2010). We therefore predicted a significant
interaction of type of emotional expression (target expression,
contrast expression) and block. Regarding responses to fearful
expressions in Experiment 2b, we expected a different pattern:
Because both the social message communicated by a fearful expression and its evaluation should elicit an avoidance reaction, we
expected fearful expressions to trigger an avoidance reaction both
when contrasted with expressions of anger and when contrasted
with happiness. Therefore, we expected to find a significant main
effect of type of emotional expression (target expression, contrast
expression) with the approach score as the DV which should not be
moderated by block.
Method
Participants. A total of 62 nonpsychology undergraduate students (30 women, 32 men) from Saarland University participated
in Experiment 2a. The age range was 18 to 33 years with a median
of 24 years. In Experiment 2b, 61 nonpsychology undergraduate
students (30 women, 31 men) from Saarland University participated. The age range was 18 to 30 years with a median of 22 years.
Both experiment lasted approximately 40 minutes. Participants
were paid 6 Euros.
We recruited a somewhat greater number of participants for
Experiments 2a and 2b than for Experiment 1 for two reasons.
First, the effects found in Experiments 1a and 1b were a bit smaller
than those found in the reference studies. Second, in Experiment
2a we aimed at detecting a two-way interaction between emotional
expression (target expression vs. contrast expression) and evaluation of contrast expression (positive vs. negative) with the approach score as DV instead of a main effect. Given a sample size
of N ⫽ 61 and an ␣-value of .05, effects of size dZ ⫽ 0.36 (i.e.,
“small” to “medium” effects as defined by Cohen, 1988) could be
detected with a probability of 1 ⫺ ␤ ⫽ .80 (Faul et al., 2007).
Design. Both experiments followed a 2 (movement: approach,
avoidance) ⫻ 2 (emotional expression: target expression, contrast
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
APPROACH/AVOIDANCE REACTIONS TO EMOTIONAL EXPRESSIONS
T4
F2
expression) ⫻ 2 (evaluation of contrast expression: positive, negative) design. Anger was the target expression in Experiment 2a,
whereas fear was the target expression in Experiment 2b. In both
experiments, happiness was the positive contrast expression. The
negative contrast expression was fear in Experiment 2a and anger
in Experiment 2b. In both experiments, all factors were varied
within participants.
Material. The emotional expressions presented in Experiments 2a and 2b were taken from the same pool as those used in
Experiments 1a and 1b and were edited the same way. We selected
fearful, angry, and happy expressions of 10 men and 10 women.
We created two stimulus-sets from these 60 images, each comprising the three expressions of five men and five women. The two
sets were used for the two different contrast emotion conditions.
Procedure. The procedure was identical to Experiment 1 with
the exception that participants worked through two blocks of
approach and avoidance tasks. The two different contrast emotion
conditions were blocked, such that the target expression was
contrasted with a positive expression in one block (Block A), and
contrasted with a negative expression in the other block (Block B).
The sequence of the two blocks was counterbalanced across participants. As mentioned earlier, different stimulus-sets were presented in the two blocks, such that all images were presented
equally often. The assignment of stimulus-set to block/condition
was counterbalanced across participants. The two blocks were
separated by a 5-min filler task, the ZVT (Zahlenverbindungstest
[Number Connection Test]; Oswald & Roth, 1987). This basic
intelligence task was chosen because it was unrelated to the task at
hand and easy to administer.
Results
Trials with incorrect responses (Experiment 2a: 5.5%, Experiment 2b: 4.4%) were discarded from analysis, as were trials with
RT outliers (i.e., RTs more than three interquartile ranges above
the third quartile or below the first quartile; Tukey, 1977) with
respect to the individual distribution, assessed separately for the
two blocks (Experiment 2a, Block 1: 1.2%, Block 2: 1.3%; Experiment 2b, Block 1: 1.3%, Block2: 1.4%). Again, we computed
approach scores by subtracting RTs of approach movements from
RTs of avoidance movements. Mean RTs for Experiment 2 are
presented in Table 4.
Experiment 2a. We conducted a 2 (emotional expression:
anger, contrast expression) ⫻ 2 (contrast expression: happiness,
fear) analysis of variance with approach scores as the dependent
variable. We found a main effect of emotional expression, F(1,
61) ⫽ 4.64, p ⫽ .035, ␩p2 ⫽ .071 and a main effect of contrast
expression, F(1, 61) ⫽ 4.89, p ⬍ .031, ␩p2 ⫽ .074. These main
effects were qualified by the expected interaction between emotional expression and contrast expression, F(1, 61) ⫽ 18.88, p ⬍
.001, ␩p2 ⫽ .236 (dZ ⫽ 0.55; see left panel of Figure 2).
In order to test our hypothesis that reactions to anger would be
influenced by the contrast emotion, we compared the approach
scores of anger contrasted with happiness and anger contrasted
with fear. This comparison reached significance, F(1, 61) ⫽ 27.79,
p ⬍ .001, ␩p2 ⫽ .289. As expected, avoidance reactions were
executed faster than approach reactions if anger was contrasted
with happiness, but approach reactions were executed faster than
avoidance reactions if anger was contrasted with fear (see left
7
Table 4
Mean Reaction Times for Approach and Avoidance Movements
(Standard Deviations in Parentheses; Experiment 2a, 2b)
Movement
Experiment
Experiment 2a
Block 1
Block 2
Experiment 2b
Block 1
Block 2
Emotion
Approach
Avoidance
Happiness
Anger
Fear
Anger
669 (198)
745 (239)
822 (268)
780 (214)
745 (239)
703 (177)
833 (232)
843 (225)
Happiness
Fear
Anger
Fear
694 (157)
769 (177)
817 (231)
885 (250)
771 (149)
733 (139)
898 (240)
873 (231)
panel of Figure 2). As can be seen in the left panel of Figure 2, the
constant c (as an estimate of the approach-bias of the task) is not
included in either of the 95% CIs (thereby providing evidence for
approach and avoidance, respectively, in absolute terms).
The corresponding analysis for the contrast emotions also
yielded a significant difference between the mean approach score
for expressions of happiness and the mean approach score for
expressions of fear, F(1, 61) ⫽ 8.87, p ⫽ .004, ␩p2 ⫽ .201. As can
be seen in the left panel of Figure 2, the constant c (as an estimate
of the approach-bias of the task) was not included in the 95% CI
of happiness but included in the 95% CI of fear.
We furthermore ran separate analyses for each block, comparing
the mean approach scores between anger and the respective contrast expression. The analyses reached significance for both
blocks, F(1, 61) ⫽ 21.96, p ⬍ .001, ␩p2 ⫽ .265, for the contrast
with happiness, and F(1, 61) ⫽ 4.61, p ⫽ .036, ␩p2 ⫽ .070, for the
contrast with fear.
Experiment 2b. In accordance with Experiment 2a, we calculated the same 2 (emotional expression: fear, contrast expression) ⫻ 2 (contrast expression: happy vs. anger) analysis of variance on the approach scores (see right panel of Figure 2).
However, in this experiment we expected to find a significant main
effect of emotional expression that should not differ between the
two contrast emotions (i.e., the relation between approach and
avoidance reactions to fear should not depend on whether fear was
contrasted with happiness or anger). As expected, the main effect
of emotional expression reached significance, F(1, 60) ⫽ 19.79,
p ⬍ .001, ␩p2 ⫽ .248 (dZ ⫽ 0.57). The right panel of Figure 2 shows
that fear avoidance reactions were faster than fear approach reactions both when fear was paired with happiness and when it was
paired with anger. Both contrast expressions (happiness, anger)
elicited faster approach than avoidance reactions. This effect was
not moderated by contrast expression, F(1, 60) ⫽ 0.23, p ⫽ .632,
␩p2 ⫽ .004, and the main effect of contrast expression was not
significant, F(1, 60) ⫽ 2.68, p ⫽ .107, ␩p2 ⫽ .043.
Even though we did not expect a significant difference, we also
compared the approach scores of fear contrasted with happiness
with the approach scores of fear contrasted with anger. As expected, the difference was not significant F(1, 60) ⫽ 1.74, p ⫽
.283, ␩p2 ⫽ .019. The right panel of Figure 2 shows that fearful
expressions elicited faster avoidance than approach reactions irrespective of the control expression. The constant c was not included
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
8
PAULUS AND WENTURA
Figure 2. Interaction between the target expression (Experiment 2a: anger; Experiment 2b: fear) and the
contrast expression observed in Experiment 2a and Experiment 2b. The dotted line indicates the constant c.
in the 95% CI if fear was contrasted with happiness but c was
included if fear was contrasted with anger. However, collapsing
across blocks yielded a CI ranging from ⫺51 ms to 4 ms, indicating that fear-based avoidance tendencies were found not only in
relative but also in absolute terms.
The corresponding analysis for the control expression likewise
found no significant difference between the approach scores for
happiness and anger, F ⬍ 1. The right panel of Figure 2 shows that
both happy and angry expressions elicited faster approach than
avoidance reactions if they were contrasted with fearful expressions. The constant c was included in neither of the 95% CIs.
For the sake of completeness, we also ran a separate analysis for
each block, comparing the approach score between fear and the
respective contrast expression. Both analyses showed a significant
difference in the approach score between the two expressions, F(1,
60) ⫽ 21.95, p ⬍ .001, ␩p2 ⫽ .268, for the contrast with happiness,
and F(1, 60) ⫽ 6.42, p ⫽ .014, ␩p2 ⫽ .097, for the contrast with
anger.
Discussion
The results from Experiments 2a and 2b supported our assumptions: We expected that approach and avoidance reactions to
expressions of anger should be moderated by the respective contrast expression, whereas reactions to expressions of fear should
not be influenced by the contrast expression. Accordingly, avoidance reactions should be faster than approach reactions if anger
was contrasted with expressions of happiness, and approach reactions should be faster than avoidance reactions if anger was contrasted with expressions of fear. Expressions of fear, on the other
hand, should elicit avoidance reactions irrespective of the contrast
expression. The results of Experiment 2a (anger) and 2b (fear)
showed exactly this pattern, supporting our assumption that the
contrast emotional matters.
Furthermore, we believe that these results also help to understand the process underlying approach and avoidance reactions to
emotional expressions. We assume that behavioral reactions are
triggered by the evaluation of the emotional target expression if the
two emotions employed differ in evaluative connotation. By contrast, reactions should be triggered by the social meaning of the
emotional target expression if the two emotions have the same
evaluative connotation. The results are in line with these assumptions: Both anger and fear expressions elicited an avoidance reaction if paired with expressions of happiness. However, if paired
with each other, anger elicited an approach reaction and fear
elicited an avoidance reaction. Exactly this pattern was expected
based on the literature: It has been shown that fear has a warning
function (Paulus & Wentura, 2014; Whalen, 1998; Whalen et al.,
2001) that should activate an avoidance reaction. Expressions of
anger, however, should elicit an approach reaction which is adaptive in overcoming the social challenge signaled by the expression
(Krieglmeyer & Deutsch, 2013; Wilkowski & Meier, 2010).
However, there is an alternative explanation for the pattern
regarding anger and fear expressions, which we call the judgmental tendency hypothesis: It is conceivable that expressions of anger
elicit approach and expressions of fear elicit avoidance if contrasted with each other because of structural features of the experimental task. To spell out this argument, we need to preface it with
a more general note on binary decision tasks.
It can be argued that binary decision tasks are not always solved
in a symmetrical way (i.e., “if the stimulus is of Type A, respond
with X”; “if the stimulus is of Type B, respond with Y”), but
asymmetrically: Participants might internally code stimuli of one
type as the target and stimuli of the other type as the nontarget.
Subtle differences between the two category types (e.g., salience of
stimuli) can thereby determine which stimulus type is coded as the
target and which one as the nontarget. Then, to the extent that
response types can be considered “affirmative” versus “negating,”
responding with the more affirmative response to the targets and
with the negating response to the nontargets will yield faster
responses than the reverse assignment (see Klauer & Stern, 1992;
Rothermund & Wentura, 2004; Wentura, 2000).
In regard to our experiments, we can potentially see the approach response as an affirmative response, whereas the avoidance
response might be considered a negating response. Moreover, the
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
APPROACH/AVOIDANCE REACTIONS TO EMOTIONAL EXPRESSIONS
likelihood of being coded as the target might be higher for expressions of happiness than expressions of anger, and higher for
expressions of anger than other negative facial expressions. As a
result, approach reactions to happy faces (irrespective of contrast
category) and to angry faces (if not contrasted with happy faces)
would be executed faster than avoidance reactions.
There are two possible ways to examine if such a process is
driving our results. First, we could introduce a different negative
emotion with a social signal that is plausibly approach-related. If
our results are caused by the judgmental tendency approach described above, this expression should trigger an avoidance behavior if it is contrasted with happiness and if it contrasted with anger
because happiness and anger would be coded as the target expression. If our results are, however, caused by the evaluation or the
social meaning of the emotional expression, respectively, then this
expression should trigger an avoidance reaction if paired with
happiness but an approach reaction if paired with anger. Expressions of sadness fulfill the required criteria: Sadness has a negative
evaluative connotation but signals a request for help, which should
elicit an approach reaction, as has been demonstrated in approach
and avoidance experiments before (Seidel et al., 2010). Thus, in
Experiment 3 we essentially replicated Experiment 2b, replacing
fear expressions with sadness expressions. We expected sadness
expressions to be (relatively) avoidance-related if contrasted with
happy expressions (because of evaluation-triggered responses) but
(relatively) approach-related if contrasted with angry expressions
(because of responses based on social message).
The second possibility to rule out the judgmental tendency
hypothesis is the use of a binary decision task that is orthogonal to
the emotion categories. Experiment 4 used such a task. If participants are not instructed to categorize the emotional expressions it
seems unlikely that one type of expression will be coded as the
target and the other one as the nontarget. Therefore, if a task in
which emotional expression is not task-relevant produces the same
result pattern as the emotion-categorization task, the caveat is
refuted. As a second benefit, such an approach would also allow us
to rule out that the results from our experiments contrasting two
negative emotions were simply caused by a semantic match between the emotional expression and the movement (we will elaborate more on this argument below).
Experiment 3
In Experiment 3, sadness was the target expression, and it was
either contrasted with expressions of happiness or expressions of
anger. If approach and avoidance reactions are triggered by the
evaluation in case of evaluative incongruency and by the social
message in case of evaluative congruency, expressions of sadness
should elicit an avoidance reaction if contrasted with happiness
and an approach reaction if contrasted with anger. Therefore, we
expected sadness to show a pattern similar to anger expressions in
Experiment 2a.
Method
Participants. In Experiment 3, 40 nonpsychology undergraduate students (20 women, 20 men) from Saarland University
participated. The age range was 18 to 33 years with a median of 21
years. The experiment lasted approximately 40 min. Participants
were paid 6 Euros.
9
We expected a two-way interaction in Experiment 3; the significant corresponding interaction in 2a was associated with dZ ⫽
0.55. Thus, a conservative assumption was to assume dZ ⫽ 0.50
(i.e., a “medium” effect as defined by Cohen, 1988). To detect an
effect of this magnitude with a probability of 1 ⫺ ␤ ⫽ .80 and an
␣-value of .05, a sample size of 34 participants was required
(actual power with N ⫽ 40 was 1 ⫺ ␤ ⫽ .87; Faul et al., 2007).
Design. The experiment followed a 2 (movement: approach,
avoidance) ⫻ 2 (emotional expression: sadness, contrast expression) ⫻ 2 (evaluation of contrast expression: positive, negative)
design with all factors varied within participants.
Material. The emotional expressions presented in this experiment were taken from the same sets as those employed in the
previous experiments and were edited the same way. We selected
sad, angry, and happy expressions of eight men and eight women.
Two stimulus-sets were created from these 60 images, each comprising the three expressions from four men and four women. The
two sets were used for the two different contrast emotion conditions.
Procedure. The procedure was identical to Experiment 2b
with the following two exceptions: First, expressions of sadness
replaced the expressions of fear. Second, as already mentioned
above (see Results section of Experiment 1), in addition to the
emotional face trials, there were baseline trials requiring participants to respond to arrows pointing upward or downward. In each
block, 32 of these trials were randomly intermixed with face trials.
There were 16 additional practice trials using arrows.
Results
Trials with incorrect responses (3.8%) were discarded from
analysis, as were trials with RT outliers (i.e., RTs three interquartile ranges above the third quartile or below the first quartile;
Tukey, 1977) with respect to the individual distribution, assessed
separately for the two blocks (Block 1: 1.0%, Block 2: 1.0%). The
results from the baseline trials (of both Experiment 3 and 4) are
reported in the Appendix. As in the previous experiments, we
calculated approach scores by subtracting the RT for approach
movements from the RT for avoidance movements. Mean RTs for
Experiment 3 are presented in Table 5.
As a first step, we calculated a 2 (emotional expression: sad,
contrast) ⫻ 2 (contrast expression: happy, anger) ANOVA for
Table 5
Mean Reaction Times for Approach and Avoidance Movements
(Standard Deviations in Parentheses; Experiment 3, 4)
Movement
Experiment
Experiment 3
Block 1
Block 2
Experiment 4
Block 1
Block 2
Emotion
Approach
Avoidance
Happiness
Sadness
Anger
Sadness
683 (122)
724 (150)
822 (182)
832 (135)
741 (136)
723 (109)
838 (133)
868 (140)
Happiness
Sadness
Anger
Sadness
924 (321)
924 (323)
910 (278)
899 (296)
1004 (303)
949 (247)
991 (272)
974 (273)
T5
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
10
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
F3
Fn6
PAULUS AND WENTURA
repeated measures with approach scores as the dependent variable.
Importantly, the expected interaction between emotional expression and contrast expression was significant, F(1, 39) ⫽ 4.86, p ⫽
.033, ␩p2 ⫽ .11 (dZ ⫽ 0.32; see Figure 3). No main effect of
emotional expression (F ⬍ 1) or contrast expression, F(1, 39) ⫽
1.06, p ⫽ .306, ␩p2 ⫽ .027, emerged.
Because we hypothesized that expressions of sadness would
activate avoidance when contrasted with expressions of happiness
but approach when contrasted with expressions of anger, we tested
whether the mean approach score for sadness with contrast happiness was significantly smaller than the approach score for sadness with contrast anger; this was the case, F(1, 39) ⫽ 4.12, p ⫽
.049, ␩p2 ⫽ .10.
We also ran the corresponding analysis for the contrast expressions, that is, we tested whether the mean approach score for
happiness was significantly larger than the one for anger; surprisingly, this was the case, F(1, 39) ⫽ 4.12, p ⫽ .049, ␩p2 ⫽ .10.6
Finally, we compared the mean approach scores of the two
emotions separately for each block. For the block contrasting
sadness with happiness, a significant difference emerged, F(1,
39) ⫽ 7.89, p ⫽ .008, ␩p2 ⫽ .168. For the block contrasting sadness
with anger, as expected the difference was not significant (F ⬍ 1).
In this experiment, behavioral tendencies were a bit weaker (in
absolute terms) as in the preceding ones: The constant c was
included in the 95% CI of all mean approach scores except the
happiness score (see Figure 3).
Discussion
As expected, sadness evoked a different pattern of response
tendencies depending on whether it was contrasted with happiness
or anger expressions. Most importantly, both anger and sadness
did not differ with regard to response tendencies, thereby providing first evidence against a simple judgmental tendency hypothesis.
In Experiment 4, we aimed to further corroborate this claim by
employing a binary task that was orthogonal to emotion. This
Figure 4. Interaction between the target expression sadness and the
contrast expression observed in Experiment 4. The dotted line indicates the
constant c.
approach also allowed us to rule out that reactions to target
emotions paired with emotions of the same evaluative connotation
were caused by simple judgmental tendencies. It would furthermore allow us to rule out the possibility that the results of the
experiments contrasting two negative emotions arose from the
semantic match between the emotional expression and the movement. The concept of anger, for example, includes approach (e.g.,
Anderson & Bushman, 2002; Carver & Harmon-Jones, 2009),
whereas the concept of fear includes avoidance (e.g., Lerner &
Keltner, 2001). Because the task of the previous experiments was
to categorize the emotional expression, it is likely that the broader
semantic concept of the emotion also became activated, making
the match or mismatch between emotion concept and movement
evident. Such a process, however, seems much less likely if
participants are instructed to categorize stimuli depending on a
feature independent of emotion, as the semantic concept of the
emotion is less likely to be activated.
Finally, a demonstration of emotion-related behavioral tendencies in a situation where emotion is not task-relevant would carry
significance beyond our specific hypotheses, as emotion was task
relevant in all experiments (we know of) examining behavioral
reactions to emotional expressions and there is an ongoing debate
in the literature concerning the existence of such effects all together (e.g., Phaf, Mohr, Rotteveel, & Wicherts, 2014).
Experiment 4
Experiment 4 used the same emotions as Experiment 3. However, in Experiment 4, emotion was no longer relevant to participants’ task. We expected results to mimic the results of Experiment 3, supporting our notion that approach and avoidance
reactions to emotional faces are triggered by evaluative connotation and social meaning, respectively.
Figure 3. Interaction between the target expression sadness and the
contrast expression observed in Experiment 3. The dotted line indicates the
constant c.
6
Coincidentally, this F test result is indeed identical (up to 2 decimals)
to the preceding one.
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
APPROACH/AVOIDANCE REACTIONS TO EMOTIONAL EXPRESSIONS
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Method
Participants. In Experiment 4, 40 nonpsychology undergraduate students (18 women, 22 men) from Saarland University
participated. The age range was 18 to 31 years with a median of 23
years. The experiment lasted approximately 40 minutes. Participants were paid 6 Euros.
We expected a two-way interaction as in Experiment 2a and 3;
these interactions were associated with dZ ⫽ 0.55 and dZ ⫽ 0.32
in, respectively. Given a sample size of N ⫽ 40 and an ␣-value of
.05, effects of size dZ ⫽ 0.44 (i.e., the average of dZ ⫽ 0.55 and
dZ ⫽ 0.32) could be detected with a probability of 1 ⫺ ␤ ⫽ .77
(Faul et al., 2007).
Design, material, and procedure. Experiment 4 was identical to Experiment 3 with the following two exceptions: First, we
slightly blurred the left or the right side of each of the emotional
expressions. Based on previous studies and our subjective impression while viewing the stimuli, the blurring likely had a negligible
effect (see also Paulus & Wentura, 2014). Second, participants
were instructed to categorize the images based on which side of the
face was blurred, by moving the virtual figure toward or away
from the face. The blurring was strong enough to ensure easy
classification. The assignment of right and left side to approach
and avoidance movements, respectively, was counterbalanced
across participants and was constant for each participant.
Results
Trials with incorrect responses (5.8%) were discarded from
analysis, as were trials with RT outliers (i.e., RTs three interquartile ranges above the third quartile or below the first quartile;
Tukey, 1977) with respect to the individual distribution, assessed
separately for the two blocks (Block 1: 0.7%, Block 2: 1.0%).
Mean RTs for Experiment 4 are presented in Table 5. The results
from the baseline trials are reported in the Appendix.
As in the previous experiments, we calculated approach scores
by subtracting the RT for approach movements from the RT for
avoidance movements. As a first step, we ran a 2 (emotional
expression: sad, contrast) ⫻ 2 (contrast expression: happiness,
anger) ANOVA for repeated measures with approach scores as the
dependent variable. A main effect of emotional expression
emerged, F(1, 39) ⫽ 4.63, p ⫽ .038, ␩p2 ⫽ .106. Importantly, the
expected interaction between emotional expression and contrast
expression was significant as well, F(1, 39) ⫽ 5.76, p ⫽ .021, ␩p2 ⫽
.129 (dZ ⫽ .38; see Figure 4).
Because we hypothesized that expressions of sadness would
activate avoidance when contrasted with expressions of happiness
but approach when contrasted with expressions of anger, we tested
whether the mean approach score for sadness with contrast anger
was significantly larger than the approach score for sadness with
contrast happiness; this was the case, F(1, 39) ⫽ 9.03, p ⫽ .005,
␩p2 ⫽ .19.
We also ran the corresponding analysis for the contrast expressions, that is, we tested whether the mean approach score for
happiness was significantly larger than the one for anger; this was
not the case (F ⬍ 1).
Finally, we compared the mean approach scores within the
blocks separately. For the block contrasting sadness with happiness, a significant difference emerged, F(1, 39) ⫽ 10.20, p ⫽ .003,
␩p2 ⫽ .21 For the block contrasting sadness with anger, the differ-
11
ence was not significant (F ⬍ 1). Behavioral tendencies in this
experiment were stronger than in Experiment 3: The constant c
was only included in the 95% CI for sadness if it was paired with
happiness. It was not included in all other cases (see Figure 4).
Discussion
Experiment 4 complements the results observed in Experiment
3 and supports our notion that approach and avoidance reactions to
emotional expressions are triggered by the evaluative connotation
and the social message signaled by the expression: Again, expressions of sadness elicited an avoidance reaction if contrasted with
expressions of happiness, but an approach reaction if contrasted
with expressions of fear. Importantly, these behavioral tendencies
emerged even though the emotional expression was not taskrelevant. These results are clearly in line with our main assumptions; they cannot be reconciled with the judgmental tendency
hypothesis.
Apart from ruling out the judgmental tendency approach, these
results show that it is very unlikely that approach and avoidance
reactions to emotional expressions are (mainly) triggered by a
semantic match between the concept of the expression and the
movement: If this semantic match was responsible for the results,
approach reactions to angry and sad faces should be stronger if the
emotion was task relevant, because the emotion category would be
more strongly activated in this situation. The results of Experiment
4, however, replicate the results of Experiment 3—if anything,
they were even stronger. Therefore, we think it is rather unlikely
that our results were caused by the semantic match between the
emotion category and the movement.
General Discussion
Approach and avoidance reactions to emotional expressions are
influenced by the contrast category used; in a number of experiments we were able to show that the same emotional expression
can elicit an avoidance reaction if contrasted with one emotion but
an approach reaction if contrasted with a different emotion: Anger,
fear, and sadness expressions elicited (relative) avoidance reactions if they were contrasted with expressions of happiness. If
contrasted with another negative emotion, however, anger and
sadness elicited (relative) approach reactions. Fear, on the other
hand, still triggered (relative) avoidance reactions. The results
therefore provide strong support for the main hypothesis of our
study, namely that the contrast emotion matters.
Furthermore, we believe these results are in line with our second
hypothesis, namely that approach and avoidance reactions are
elicited by the evaluative connotation of the emotion if two emotions of different evaluative connotation are presented together in
a task, but by the social meaning if the two emotions have the same
evaluative connotation. Sadness, for example, has a negative evaluative connotation but signals a request for help. Accordingly, it
activated avoidance reactions if presented with expressions of
happiness but approach reactions if presented with anger expressions. Fear, on the other hand, also has a negative evaluative
connotation but signals threat in the environment. Accordingly, it
elicited an avoidance reaction both when presented with expressions of happiness and when presented with an emotion of negative evaluative connotation.
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
12
Fn7
PAULUS AND WENTURA
Our findings are important with regard to several points. First of
all, they help to explain many of the inconsistent results regarding
approach and avoidance reactions to emotional expressions reported in the literature. In our view, the most striking inconsistencies can be found between different studies examining approach
and avoidance reactions to expressions of anger: A number of
studies reported faster avoidance reactions to anger expressions
(e.g., Marsh et al., 2005; Roelofs et al., 2010; Rotteveel & Phaf,
2004; Stins et al., 2011), whereas others reported faster approach
reactions (Krieglmeyer & Deutsch, 2013; Wilkowski & Meier,
2010). Most of the studies reporting avoidance reactions, however,
contrasted anger with happiness (Roelofs et al., 2010; Rotteveel &
Phaf, 2004; Stins et al., 2011), whereas those finding approach
reactions paired anger with fear.7 Our results are exactly in line
with these findings: When contrasted with happiness, anger elicited avoidance, and when contrasted with fear or sadness, anger
elicited approach in our experiments.
As a second implication, we believe our results provide
evidence regarding the process underlying approach and avoidance reactions to emotional expression: Depending on the contrast emotion, these behavioral tendencies are either triggered
by the evaluative connotation or the social message of the
emotional expression. We believe that it is the salience of the
respective expression features that determines which feature
most affects the responses. If two emotional expressions of
opposing evaluative connotation are presented in an experiment, the evaluative connotation of the respective emotions is a
very salient feature of the expressions. Accordingly, it will
influence approach and avoidance reactions. If, however, two
emotional expressions of the same evaluative connotation are
presented together, evaluative connotation is no longer a differentiating feature. We believe that in this situation the social
meaning associated with the emotional expressions becomes
more salient, triggering the response.
Another inference that can be drawn from our results regards
the automaticity of the underlying process. In Experiment 4,
participants categorized the faces along a feature dimension
orthogonal to the emotional expression. The results observed in
this experiment were identical to those found in Experiment 3,
in which the same emotional expressions were presented but
participants’ task was to categorize the expressions. In both
experiments, a negative emotional expression (i.e., sadness)
elicited avoidance if contrasted with a positive expression but
approach if contrasted with a negative expression (i.e., anger).
These findings inform the ongoing debate on whether approach
and avoidance reactions are elicited automatically, that is, irrespective of conscious intentions to evaluate the valence of the
expression (e.g., Phaf et al., 2014).
However, one might ask how the postulate of an automatic
reaction goes along with an influence of contextual factors as
observed in our experiments. We believe that this is not a conflict:
In our view, the same emotional expression is linked to several
behavioral reactions. Which reaction is triggered in a given moment is influenced by the context. If the context asks the perceiver
to attend to the evaluative connotation of the expression, the link
between evaluative connotation and reaction becomes active in an
automatic manner. If, however, the context asks the perceiver to
attend to the social message, the link between this social message
and the reaction is activated. This view is in line with findings
from different fields of psychology reporting an influence of
context, goal, or instruction on automatic reactions (e.g., Bermeitinger, Wentura, & Frings, 2011; Casper, Rothermund, & Wentura,
2010; Folk, Remington, & Johnston, 1992; Kiefer & Martens,
2010; Wittenbrink, Judd, & Park, 2001).
At a last point, the finding that approach and avoidance reactions to emotional expressions are not only triggered by the expressions themselves but are also influenced by the other expressions presented in the experiment is important for the
interpretation of many studies examining approach and avoidance
reactions to emotional expressions: Results from these studies
have often only been interpreted with respect to the emotional
expression of interest, disregarding the influence of the contrast
emotion (e.g., Seidel et al., 2010; Stins et al., 2011; Wilkowski &
Meier, 2010; see also van Peer et al., 2010, for a similar argument
in a different paradigm). Our results show that the reaction to one
emotion should not be considered in isolation but in the context of
its presentation. Depending on the context, the same emotional
expression can elicit divergent reactions. We believe that the
results from several studies might have to be reevaluated in light of
these findings.
Our study also relates to two other studies that recently attempted to reconcile the inconsistent results obtained with anger
expressions (Bossuyt et al., 2014; Krieglmeyer & Deutsch, 2013).
Both studies focused on the meaning of approach behavior and
hypothesized that this behavior can serve either an aggressive or an
affiliative goal. To test their prediction, the approach reaction in
both studies was either conceptualized as an aggressive or an
affiliative/submissive response. The results showed that if the
approach reaction was conceptualized as an aggressive approach rather than an affiliative/submissive approach, anger
faces induced stronger facilitation of approach (relative to
avoidance). This pattern was reversed for fearful faces. We
think that our theoretical approach complements rather than
contradicts the one proposed by Krieglmeyer and Deutsch
(2013) and Bossuyt et al. (2014), as it specifies which reaction
is triggered by an emotional face if no explicit meaning of the
respective movements is provided. Therefore, we believe that
our study adds to the previous studies by providing an additional piece of evidence to answer the question of which reaction is triggered by which emotional expression.
Despite providing many answers, our results also raise some
questions. One concerns the role of neutral expressions: If a
negative emotional expression is paired with a neutral expression,
7
In a single experiment, Marsh et al. (2005) contrasted anger and fear
and found anger to be associated with avoidance and fear to be associated
with approach. This result is in contrast to what others and we have found.
However, there is one remarkable feature of this experiment that we would
like to highlight: The two blocks of trials that assessed push and pull
response times to anger and fear were randomly intermixed with four
control blocks. Two of those blocks assessed push and pull response times
to admired (e.g., Gandhi) versus hated (e.g., Hitler) individuals; the two
remaining blocks assessed push and pull response times to women who
were either highly attractive or disfigured. Thus, for the majority of
participants the two fear/anger blocks were embedded within blocks of
comparisons different from emotions. It would be a research endeavor on
its own to explore possible carry-over effects (e.g., that pushing acquires a
connotation of other-relevant negativity, Peeters, 1983, during the blocks
presenting hated/admired persons).
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
APPROACH/AVOIDANCE REACTIONS TO EMOTIONAL EXPRESSIONS
will the response be triggered by the negative emotion’s evaluation
or its social meaning? Looking at our findings, one might predict
that the response will be determined by the emotion’s evaluation
since negative and neutral expressions can be distinguished along
this dimension. However, other studies examining approach and
avoidance reactions to emotional expressions suggest that it is the
social meaning that determines the response in this situation.
Wilkowski and Meier (2010), for example, contrasted expressions
of anger with neutral expressions and observed faster approach
than avoidance reactions to angry faces. Similar results were found
by Krieglmeyer and Deutsch (2013). We can only speculate about
the reason for these findings. In our view, one potential answer
might lie in the nature of neutral expressions: Neutral expressions
are ambiguous with regard to their evaluative connotation (Shah &
Lewis, 2003), and it has been shown that they can elicit responses
similar to negative emotional expressions (Rohr, Degner, & Wentura, 2015). We believe that because of this ambiguity, factors like
the social meaning may be activated when the expressions are
processed in order to encode it. However, this assumption needs
empirical testing.
A second point that should be addressed is the nature of avoidance reactions: Do actual avoidance reactions—that is, avoidance
in absolute terms— occur? In our experiments, the constant c
(which served as an estimate for the approach bias) was not
included in the majority of the approach score CIs of the emotions
(and contexts) for which we expected an approach reaction. This
finding indicates that “true” approach reactions were shown. By
contrast, for the emotions (and contexts) for which we expected an
avoidance reaction, the constant c was only excluded from about
half of the relevant approach score CIs. Therefore, the results are
less clear regarding the avoidance reactions. We believe one reason for this might be our assessment of the approach bias. We
followed the procedure described by Krieglmeyer and Deutsch
(2013) and included baseline trials in the main procedure, in which
participants responded to arrows pointing upward and downward,
in order to measure the approach bias. However, in hindsight, this
method might not be perfectly suitable for an assessment of the
size of the approach bias for several reasons. First of all, when
responding to the arrows, participants are asked to move the virtual
figure “up” or “down” instead of “towards” and “away.” Therefore, the task participants are asked to carry out in these trials is
quite different from the task in the critical trials. Second, in our
experiments, the two types of trials were intermixed but there were
fewer baseline trials than critical trials. It is therefore reasonable to
assume that responses in the critical trials were more fluent, and
that participants might have experienced substantial switch costs
when switching from the main task to the baseline task. These
switch costs might make the estimate of the constant c quite noisy.
Therefore, we believe that a reliable answer to the question of
whether actual avoidance reactions occur will have to await the
establishment of a different procedure to assess the magnitude of
the approach bias. However, we do not believe that the inconclusive answer regarding avoidance behavior poses a problem to the
main point of the present study. We set out to test if the presented
contrast emotion influences the behavioral reactions to emotional
expressions. This question is clearly answered: Behavioral reactions to the same emotion differed depending on the contrast
expression.
13
Another point concerns the process underlying approach and
avoidance reactions to emotional expressions. We propose that
approach and avoidance reactions are triggered by the social
message if the two emotions presented in the task have the same
evaluative connotation. Might a simpler approach explain our
results? Assume for a moment that emotional expressions (a) can
be described along a (continuous) feature termed “approachability;” and (b) emotions can be ordered with regard to this feature
such that happy expressions are more approachable than sad ones,
sad ones more approachable than angry ones and angry ones more
approachable than fearful ones. This rank order then predicts
which emotion produces (relative) approach and which emotion
produces (relative) avoidance if two of these emotions are paired
in an approach avoidance task.8 Such an assumption would then
imply that it is inherent to the approach-avoidance task to turn
even small quantitative differences in “approachability” into qualitative differences: The more approachable expression should elicit
an approach reaction and the less approachable expression an
avoidance reaction, even if the difference in approachability between the emotions is only marginal. However, we believe there
are several problems with this suggestion. First, the rank order is
established ad hoc on the basis of empirical results, it is not
theory-based. This disadvantage can only (partly) be compensated by clarity of results. Second, if a quantitative difference
like approachability was driving the approach and avoidance
reactions in our experiments, one would expect to always find
a difference in reactions if two different emotions are presented
in an approach and avoidance task. However, in Experiment 3
and Experiment 4 we observed the exact same pattern of results
for angry and sad faces. This would imply that angry and sad
faces do not differ at all in their approachability. We believe
that this is rather unlikely.
However, even if the “approachability” of the emotional expressions cannot explain our results, one might object that our interpretation that approach and avoidance reactions are triggered by the
social message if the two emotions presented in the task have the
same evaluative connotation is nevertheless a bit premature: The
results only show that the reaction in this situation is triggered by
a feature other than evaluative connotation. Direct evidence that it
is indeed the social message triggering the reaction is not provided.
In general, we agree with this claim. However, we believe there are
(at least) three strong arguments supporting our conclusion. First
of all, our results regarding four different emotional expressions
(happiness, anger, fear, sadness) are perfectly in line with predictions derived from a social message perspective. Second, we
cannot think of a different feature other than the social message
which triggers, for example, approach reactions in the case of
angry and sad facial expressions. These expressions differ in many
aspects but the approach-related social message (see also our
argument above). Last, our results nicely match to those observed
in studies in which the social message was more directly manipulated (e.g., Paulus & Wentura, 2014; Wilkowski & Meier, 2010).
Therefore we believe we can be rather confident in our assertion
that the social message triggers approach and avoidance reactions
if the two expressions presented in the task have the same evalu8
We thank an anonymous reviewer for this suggestion.
Fn8
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
14
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Fn9
PAULUS AND WENTURA
ative connotation.9 However, and most importantly, even if we
agree that future studies should provide more direct evidence for
this claim, we do not think that this diminishes the main conclusion
of our results: Depending on the contrast emotion, the same
emotional expression can elicit divergent reactions.
A related consideration concerns the question of whether the
rationale that behavioral tendencies are evaluation-based in case of
emotional expressions with two contrasting evaluative connotations and based on the social signal in case of matching evaluative
connotations always hold. Recent work of ours indicates that this
is not the case. In a couple of experiments, we examined the
influence of the expresser’s group membership on approach and
avoidance reactions (Paulus & Wentura, 2014). The results
showed that depending on group membership, the same expression
can elicit diverging reactions: While in-group happiness elicited
relatively more avoidance than in-group fear, the opposite pattern
emerged for expressions shown by out-group members. We argued
that the relationship between expresser and perceiver influenced
the social message signaled by the emotional expression. For
example, a smile expressed by a positively evaluated in-group
member should signal a desire to affiliate, and a fearful expression
should act as a warning. In contrast, the same expressions shown
by negatively evaluated out-group member could signal dominance or submission, respectively. At first sight, the interpretation
that approach and avoidance reactions to emotional expressions
can be determined by the social message even if the two contrasted
expressions differ in evaluative connotation (i.e., happiness and
fear) seems to conflict with the rationale outlined in the present
paper. However, we believe the findings of the two studies complement rather than contradict each other. Both provide evidence
that in certain contexts, the social message signaled by the expression determines the reactions. In Paulus and Wentura (2014), we
were able to show that group membership can provide such a
context, which can lead to the overriding of simple evaluationbased automatisms. The present set of experiments demonstrates
that such a context can also be established by the contrast emotion
used, in the absence of additional expresser-related categorization
cues (e.g., indicating group membership as in Paulus & Wentura,
2014). Future studies should examine other potential contextual
features, such as for example nature of the relationship between
expresser and perceiver.
A last point is the relevance of our results for emotion
processing in real life settings. Obviously, the results presented
here were found in experimentally designed experiments in
very controlled settings. However, we believe that the findings
might still be applied to the “real life:” If a person is exhibiting
a negative facial expression in an overall positive situation
(e.g., showing anger all of a sudden while at a joyful meeting
with friends), his or her interaction partner would most likely
first react to this negativity (e.g., draw back and check what‘s
going on) and then analyze its social meaning. If, however, the
situation is already negative when the negative facial expression is shown (e.g., showing an anger face during a fight with
his or her spouse), the social meaning would be most relevant
and influence reactions accordingly.
Taken together, our results provide evidence that the contrast
expression presented in an approach and avoidance task influences behavioral reactions to the emotional target expression.
This finding helps to explain many of the inconsistent results
reported in the literature. We believe it also sheds light on the
process underlying approach and avoidance reactions to emotional expressions.
9
An anonymous reviewer suggested that our results might be explained
by differences in evaluative connotation between the negative emotional
expressions. We agree that theoretically such differences are able to explain our data. However, there are at least two arguments pointing against
such an explanation: First, an analysis of the evaluative ratings of part of
the stimuli employed in the experiment shows that expressions of anger
(M ⫽ 1.99; SD ⫽ 0.20), sadness (M ⫽ 2.25; SD ⫽ 0.75), and fear (M ⫽
2.13; SD ⫽ 0.17) were rated as equally negative: F(2, 36) ⫽ 1.30, p ⫽
.285. (Please note that this analysis is only based on approximately half of
the negative stimuli employed in the experiment [54%] because valence
ratings of the images were only available for the Radboud dataset.) Second,
this explanation would imply that expressions of anger and sadness are
perceived as equally negative (both elicit an approach reaction) and both
expressions are perceived as less negative than fear (which elicits an
avoidance reaction). We believe that such rather complex differences are
rather unlikely.
References
Anderson, C. A., & Bushman, B. J. (2002). Human aggression. Annual
Review of Psychology, 53, 27–51. http://dx.doi.org/10.1146/annurev
.psych.53.100901.135231
Bermeitinger, C., Wentura, D., & Frings, C. (2011). How to switch on and
switch off semantic priming effects for natural and artifactual categories:
Activation processes in category memory depend on focusing specific
feature dimensions. Psychonomic Bulletin & Review, 18, 579 –585.
http://dx.doi.org/10.3758/s13423-011-0067-z
Bijlstra, G., Holland, R. W., & Wigboldus, D. J. (2010). The social face of
emotion recognition: Evaluations versus stereotypes. Journal of Experimental Social Psychology, 46, 657– 663. http://dx.doi.org/10.1016/j
.jesp.2010.03.006
Bossuyt, E., Moors, A., & De Houwer, J. (2014). On angry approach and
fearful avoidance: The goal-dependent nature of emotional approach and
avoidance tendencies. Journal of Experimental Social Psychology, 50,
118 –124. http://dx.doi.org/10.1016/j.jesp.2013.09.009
Carver, C. S., & Harmon-Jones, E. (2009). Anger and approach: Reply to
Watson (2009) and to Tomarken and Zald (2009). Psychological Bulletin, 135, 215–217. http://dx.doi.org/10.1037/a0015026
Casper, C., Rothermund, K., & Wentura, D. (2010). Automatic stereotype
activation is context dependent. Social Psychology, 41, 131–136. http://
dx.doi.org/10.1027/1864-9335/a000019
Chen, M., & Bargh, J. A. (1999). Consequences of automatic evaluation:
Immediate behavioral predispositions to approach or avoid the stimulus.
Personality and Social Psychology Bulletin, 25, 215–224. http://dx.doi
.org/10.1177/0146167299025002007
Cohen, J. (1988). Statistical power analysis for the behavioral sciences
(2nd ed.). Hillsdale, NJ: Erlbaum.
De Houwer, J., Crombez, G., Baeyens, F., & Hermans, D. (2001). On the
generality of the affective Simon effect. Cognition and Emotion, 15,
189 –206. http://dx.doi.org/10.1080/02699930125883
Eder, A. B., & Rothermund, K. (2008). When do motor behaviors (mis)match
affective stimuli? An evaluative coding view of approach and avoidance
reactions. Journal of Experimental Psychology: General, 137, 262–281.
http://dx.doi.org/10.1037/0096-3445.137.2.262
Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). GⴱPower 3: A
flexible statistical power analysis program for the social, behavioral, and
biomedical sciences. Behavior Research Methods, 39, 175–191. http://
dx.doi.org/10.3758/BF03193146
Folk, C. L., Remington, R. W., & Johnston, J. C. (1992). Involuntary
covert orienting is contingent on attentional control settings. Journal of
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
APPROACH/AVOIDANCE REACTIONS TO EMOTIONAL EXPRESSIONS
AQ: 1
Experimental Psychology: Human Perception and Performance, 18,
1030 –1044. http://dx.doi.org/10.1037/0096-1523.18.4.1030
Horstmann, G. (2003). What do facial expressions convey: Feeling states,
behavioral intentions, or action requests? Emotion, 3, 150 –166. http://
dx.doi.org/10.1037/1528-3542.3.2.150
Hugenberg, K. (2005). Social categorization and the perception of facial
affect: Target race moderates the response latency advantage for happy
faces. Emotion, 5, 267–276. http://dx.doi.org/10.1037/1528-3542.5.3
.267
Hugenberg, K., & Sczesny, S. (2006). On wonderful women and seeing
smiles: Social categorization moderates the happy face response latency
advantage. Social Cognition, 24, 516 –539. http://dx.doi.org/10.1521/
soco.2006.24.5.516
Kiefer, M., & Martens, U. (2010). Attentional sensitization of unconscious
cognition: Task sets modulate subsequent masked semantic priming.
Journal of Experimental Psychology: General, 139, 464 – 489. http://dx
.doi.org/10.1037/a0019561
Klauer, K. C., & Stern, E. (1992). How attitudes guide memory-based
judgments: A two-process model. Journal of Experimental Social Psychology, 28, 186 –206. http://dx.doi.org/10.1016/0022-1031(92)90038-L
Krieglmeyer, R., & Deutsch, R. (2010). Comparing measures of approachavoidance behavior: The manikin task vs. two versions of the joystick
task. Cognition and Emotion, 24, 810 – 828. http://dx.doi.org/10.1080/
02699930903047298
Krieglmeyer, R., & Deutsch, R. (2013). Approach does not equal approach:
Angry facial expressions evoke approach only when it serves aggression.
Social Psychological & Personality Science, 4, 607– 614. http://dx.doi
.org/10.1177/1948550612471060
Krieglmeyer, R., Deutsch, R., De Houwer, J., & De Raedt, R. (2010).
Being moved: Valence activates approach-avoidance behavior independently of evaluation and approach-avoidance intentions. Psychological
Science, 21, 607– 613. http://dx.doi.org/10.1177/0956797610365131
Lang, P. J., & Bradley, M. M. (2013). Appetitive and defensive motivation:
Goal-directed or goal-determined? Emotion Review, 5, 230 –234. http://
dx.doi.org/10.1177/1754073913477511
Langner, O., Dotsch, R., Bijlstra, G., Wigboldus, D. J., Hawk, S. T., & van
Knippenberg, A. (2010). Presentation and validation of the Radboud
faces database. Cognition and Emotion, 24, 1377–1388. http://dx.doi
.org/10.1080/02699930903485076
Lerner, J. S., & Keltner, D. (2001). Fear, anger, and risk. Journal of
Personality and Social Psychology, 81, 146 –159. http://dx.doi.org/10
.1037/0022-3514.81.1.146
Marsh, A. A., Ambady, N., & Kleck, R. E. (2005). The effects of fear and
anger facial expressions on approach- and avoidance-related behaviors.
Emotion, 5, 119 –124. http://dx.doi.org/10.1037/1528-3542.5.1.119
Nelson, A. J., Adams, R. R., Jr., Stevenson, M. T., Weisbuch, M., &
Norton, M. I. (2013). Approach-avoidance movement influences the
decoding of anger and fear expressions. Social Cognition, 31, 745–757.
http://dx.doi.org/10.1521/soco.2013.31.6.745
Olszanowski, M., Pochwatko, G., Kukliński, K., Ścibor-Rylski, M., &
Ohme, R. (2008, June). Warsaw set of emotional facial expression
pictures—Validation study. Poster presented at the EAESP General
Meeting, Opatija, Croatia.
Oswald, W. D., & Roth, E. (1987). Der Zahlen-Verbindungs-Test (ZVT).
Ein sprachfreier Intelligenz-Test zur Messung der “kognitiven Leistungsgeschwindigkeit” (Handanweisung; 2, überarbeitete und erweiterte
Auflage) [●●●]. Göttingen, Germany: Hogrefe.
Paulus, A., Musial, E., & Renn, K. (2014). Gender of the expresser
moderates the effect of emotional faces on the startle reflex. Cognition
and Emotion, 28, 1493–1501. http://dx.doi.org/10.1080/02699931.2014
.886557
Paulus, A., Rohr, M., Neuschwander, M., Seewald, M., & Wentura, D.
(2012). Validation data of the Saarbrücken emotion database. (Unpublished data). Saarland University, Germany.
15
Paulus, A., & Wentura, D. (2014). Threatening joy: Approach and avoidance reactions to emotions are influenced by the group membership of
the expresser. Cognition and Emotion, 28, 656 – 677. http://dx.doi.org/
10.1080/02699931.2013.849659
Peeters, G. (1983). Relational and informational patterns in social cognition. In W. Doise & S. Moscovici (Eds.), Current issues in European
social psychology (Vol. 1, pp. 201–237). New York, NY: Cambridge
University Press.
Phaf, R. H., Mohr, S. E., Rotteveel, M., & Wicherts, J. M. (2014).
Approach, avoidance, and affect: A meta-analysis of approachavoidance tendencies in manual reaction time tasks. Frontiers in Psychology, 5, 378. http://dx.doi.org/10.3389/fpsyg.2014.00378
Rinck, M., & Becker, E. S. (2007). Approach and avoidance in fear of
spiders. Journal of Behavior Therapy and Experimental Psychiatry, 38,
105–120. http://dx.doi.org/10.1016/j.jbtep.2006.10.001
Roelofs, K., Putman, P., Schouten, S., Lange, W. G., Volman, I., & Rinck,
M. (2010). Gaze direction differentially affects avoidance tendencies to
happy and angry faces in socially anxious individuals. Behaviour Research and Therapy, 48, 290 –294. http://dx.doi.org/10.1016/j.brat.2009
.11.008
Rohr, M., Degner, J., & Wentura, D. (2015). The “emotion misattribution”
procedure: Processing beyond good and bad under masked and unmasked presentation conditions. Cognition and Emotion, 29, 196 –219.
http://dx.doi.org/10.1080/02699931.2014.898613
Rothermund, K., & Wentura, D. (2004). Underlying processes in the
implicit association test: Dissociating salience from associations. Journal of Experimental Psychology: General, 133, 139 –165. http://dx.doi
.org/10.1037/0096-3445.133.2.139
Rotteveel, M., & Phaf, R. H. (2004). Automatic affective evaluation does
not automatically predispose for arm flexion and extension. Emotion, 4,
156 –172. http://dx.doi.org/10.1037/1528-3542.4.2.156
Seidel, E. M., Habel, U., Kirschner, M., Gur, R. C., & Derntl, B. (2010).
The impact of facial emotional expressions on behavioral tendencies in
women and men. Journal of Experimental Psychology: Human Perception and Performance, 36, 500 –507. http://dx.doi.org/10.1037/a0018169
Shah, R., & Lewis, M. B. (2003). Locating the neutral expression in the
facial emotion space. Visual Cognition, 10, 549 –566. http://dx.doi.org/
10.1080/13506280244000203a
Solarz, A. K. (1960). Latency of instrumental responses as a function of
compatibility with the meaning of eliciting verbal signs. Journal of
Experimental Psychology, 59, 239 –245. http://dx.doi.org/10.1037/
h0047274
Stins, J. F., Roelofs, K., Villan, J., Kooijman, K., Hagenaars, M. A., &
Beek, P. J. (2011). Walk to me when I smile, step back when I’m angry:
Emotional faces modulate whole-body approach-avoidance behaviors.
Experimental Brain Research, 212, 603– 611. http://dx.doi.org/10.1007/
s00221-011-2767-z
Tukey, J. W. (1977). Exploratory data analysis. Reading, PA: Addison
Wesley.
van der Schalk, J., Hawk, S. T., Fischer, A. H., & Doosje, B. (2011).
Moving faces, looking places: Validation of the Amsterdam Dynamic
Facial Expression Set (ADFES). Emotion, 11, 907–920. http://dx.doi
.org/10.1037/a0023853
van Peer, J. M., Rotteveel, M., Spinhoven, P., Tollenaar, M. S., & Roelofs,
K. (2010). Affect-congruent approach and withdrawal movements of
happy and angry faces facilitate affective categorization. Cognition and
Emotion, 24, 863– 875. http://dx.doi.org/10.1080/02699930902935485
Wentura, D. (2000). Dissociative affective and associative priming effects
in the lexical decision task: Yes versus no responses to word targets
reveal evaluative judgment tendencies. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26, 456 – 469. http://dx.doi
.org/10.1037/0278-7393.26.2.456
APA NLM
tapraid5/zfn-xhp/zfn-xhp/zfn00515/zfn3313d15z xppws S⫽1 8/4/15 15:44 Art: 2014-0213
PAULUS AND WENTURA
16
Whalen, P. J. (1998). Fear, vigilance and ambiguity: Initial neuroimaging
studies of the human amygdala. Current Directions in Psychological
Science, 7, 177–188. http://dx.doi.org/10.1111/1467-8721.ep10836912
Whalen, P. J., Shin, L. M., McInerney, S. C., Fischer, H., Wright, C. I., &
Rauch, S. L. (2001). A functional MRI study of human amygdala
responses to facial expressions of fear versus anger. Emotion, 1, 70 – 83.
http://dx.doi.org/10.1037/1528-3542.1.1.70
Wilkowski, B. M., & Meier, B. P. (2010). Bring it on: Angry facial
expressions potentiate approach-motivated motor behavior. Journal of
Personality and Social Psychology, 98, 201–210. http://dx.doi.org/10
.1037/a0017992
Wittenbrink, B., Judd, C. M., & Park, B. (2001). Spontaneous prejudice in
context: Variability in automatically activated attitudes. Journal of Personality and Social Psychology, 81, 815– 827.
Appendix
●●●
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
AQ: 2
TA1
In Experiments 3 and 4, we included baseline trials to obtain an
estimate of approach bias in our experimental task (i.e., an estimate
of how much faster approach reactions are generally executed
relative to avoidance reactions). In the baseline trials, participants
responded to arrows pointing upward or downward by moving the
virtual figure in the respective direction (see Krieglmeyer &
Deutsch, 2013). The trial sequence included 64 of these trials in
both experiments.
Results
Mean RTs and bias scores are presented in Table A1.
A 2 (movement: toward, away) ⫻ 2 (experiment) ANOVA with
movement as a within-participants factor and experiment as a
between-participants factor yielded only a significant main effect
of movement, F(1, 78) ⫽ 52.09, p ⬍ .001, ␩p2 ⫽ .400 (F ⬍ 1 for
Table A1
Mean RTs for the Baseline Trials (Experiments 3 and 4)
Movement
Experiment 3
Experiment 4
Overall
Towards
Away
Approach biasa
531
546
16 (4)
536
560
24 (4)
533
553
20 (3)
a
Difference between RT for “away” and “towards” movements; standard
errors in parentheses.
the main effect of experiment; F(1, 78) ⫽ 2.35, p ⫽ .130, ␩p2 ⫽
.029 for the interaction).
Received October 10, 2014
Revision received July 20, 2015
Accepted July 22, 2015 䡲