Conflict and Emotions
Does Conflict Resolution Lead to Positive Affect from Neutral Stimuli?
Written by Tamara P. Rouwendal Bachelor thesis - Brain and cognition Student number - 0465100
Word count abstract - 124 Word count thesis - 4143 May 2015
Supervisor: Hans Phaf and Alexander Krepel
Faculty of Social and Behavioral Sciences - Brain and Cognition University of Amsterdam
2 Abstract
There are a several studies concerning affect being generated from neutral stimuli. A
fundamental question is how is this affect established. According to the affective monitoring hypothesis conflict takes place in a network which, if unresolved, leads to negative affect but if resolved will lead to positive affect. Using an Eriksen Flanker task with letters, 29
participants were primed with congruent, incongruent and no conflict primers to establish affect on an emotional face judgement task. The results were non-significant but indicates a positive priming effect with the congruent letters and a negative priming effect with the incongruent letters on the faces. There seems to be an indication for the affective monitoring hypothesis being present but further research will be needed to affirm this.
3
Conflict and Emotions: Does Conflict Resolution Lead to Positive Affect from Neutral Stimuli?
Nothing beats a good joke. While the joke is being told the tension builds up as you wait for the punchline to come. Once the punchline is told the tension is released and you burst out laughing. This is known as the tension-release hypothesis. The tension that builds up can be seen as competition in a module between its nodes. The quick release of tension leads to positive affect in the form of for example laughter (Sroufe & Waters, 1976, cited in Phaf et al., 2012). This is closely linked to the affective monitoring hypothesis which can be seen as a mechanism for automatically detecting conflict in a network. The hypothesis states that when competition in a network module is quickly resolved it will lead to positive affect and
unresolved conflict will lead to negative affect (Phaf et al., 2012). Defining affect as a less vague and subjective matter will make it easier to test empirically. So far affect has often been based on self-reports. With the affective monitoring hypothesis affect can be defined in terms of network competition and neural processing, making it a more objective and better
empirically testable definition of affect.
There are two ways in which affect can have an impact. Affect can influence stimuli judgment or affect can be elicited from stimuli (Fazendeiro et al., 2007). This article shall focus on the latter.
The affective monitoring hypothesis seems to be closely linked with the tension-release hypothesis, as mentioned before, but also with other phenomena as processing fluency and the mere exposure effect. Even though these other phenomena are related to the affective monitoring hypothesis they have a different prediction about how positive affect is generated.
Firstly, processing fluency is the ease with which stimuli are processed. High-fluency stimuli are more easily processed and therefore more preferred than low-fluency stimuli.
4
Winkielman et al. (2010) found that positive affect and more liking occur with higher-fluency processing of stimuli. Recurrently exposing the same stimulus increases the processing of fluency of that stimulus leading to positive affect (Zajonc,1968, cited in Cannon et al., 2010). Stimuli that are congruent to each other would lead to high-fluency and therefore positive affect. However unlike the affective monitoring hypothesis this is not generated through solving conflict. While stimuli that are incongruent with each other would accordingly cause low-fluency of stimuli which would lead to negative affect (Phaf et al., 2012).
Another important phenomenon that is closely linked to affective monitoring is the mere exposure effect. It states that stimuli that are seen more often will be more familiar and therefore more positively judged. This means neutral stimuli can be seen as more positive after more exposure, because an increase in perceptual fluency seems to take place. This in turn is linked to feelings of liking and familiarity (Fazendeiro et al., 2007). Cognitive processing and affect are combined in affective monitoring. This is because mere exposure, the feeling of familiarity, takes place at the same time as affective processing in the early stages of processing. (Phaf & Rotteveel, 2005 , cited in Phaf et al., 2012). In this case positive affect is established through the amount of exposure to a particular stimulus. The more
exposure a stimulus has received the more positively it is perceived. Unlike the affective monitoring hypothesis this does not involve the resolution of conflict to establish positive affect.
There have been several studies that focus on neutral stimuli itself producing affect, with their focus directed to studying negative affect. As is the case in the research of Dreisbach and Fischer (2012) and Fritz and Dreisbach (2013). According to Dreisbach and Fischer (2012) conflict is standardly aversive. They tested this by using an affective priming task. The study involved a Stroop task for priming congruent and incongruent colors with the words of different colors. Each Stroop trial was followed by the affective priming task. The
5
affective priming task existed out of two types of affective stimuli, pictures and words, which were either negative or positive. They found that negative targets were reacted to significantly faster on the incongruent than the congruent primes. Establishing that cognitive conflict are aversive signals. This was also confirmed in the research of Fritz & Dreisbach (2013), who tested conflict as aversive signals by conducting two experiments, judging the valence of German words and Chinese characters. Stroop task primes were followed by judging the valence of words or Chinese characters. The primes were either incongruent conflict-primes or congruent non-conflict primes. These neutral target stimuli were judged negatively after conflict instead of non-conflict primes suggesting that conflict is perceived as an aversive signal. In both these research papers the focus is on negative affect but how positive affect is established is not discussed. Conclusions cannot be made about positive affect being
established because no neutral condition was used to compare the conditions to. Only a comparison between the congruent and incongruent primes were looked at.
There has been no research concerning positive affect being induced from neutral stimuli by resolving conflict as stated in the affective monitoring hypothesis. In this study we would therefore like to find out if conflict causes positive affect to be generated from neutral stimuli. According to the affective monitoring hypothesis we would expect positive affect to be produced from neutral stimuli.
In our research we will use a letter Eriksen Flanker task to initiate conflict. There will be three conditions: incongruent, congruent and no conflict. In the no conflict condition there will be no flankers and hence there will be no conflict elicited. The congruent condition will have masked flankers, which will be the same as the target letter causing conflict to be quickly resolved. And an incongruent condition which also has masked flanker letters, but will be different from the target letter and therefore cause conflict. After presenting the letters in the Eriksen Flanker task participants will be exposed to either a happy or angry emotional
6
face. At the start of each trial participants will be asked to either react to the letters or to emotional faces. The reaction times to the emotional faces will be used to measure affect.
We predict that the congruent condition will lead to positive affect after conflict is quickly resolved compared to the no conflict condition which will cause no affect to be established from neutral stimuli because no conflict is established. And the incongruent condition will lead to negative affect because conflict is not resolved compared to no conflict condition where there is no conflict and hence no affect is generated.
7
Method Participants
There are 29 participants partaking in the experiment (17 female and 12 male) with a minimal age criteria of 18 years and a maximum of 42 years (mean age= 24, SD= 4,31). These are mostly psychology students of the university of Amsterdam (UvA) and others outside of the university. Right handed people and people with adequate vision are selected for further analysis. Participants were screened on right-handedness with the Dutch Vragenlijst voor Handvoorkeur (Van Strien, 2003) and were asked about their vision. First-year psychology participants were rewarded with research points.
Design
The experiment has a 3 (conflict conditions: congruent, incongruent and no conflict) x 2 (happy and angry faces) within-subject design. Affect was measured by the reaction times (RT) to the emotional happy and angry faces. The different conditions and photos were counterbalanced. The dependent variables were the reaction times to the emotional faces that were used to measure affect. This was measured with the affect index, which is the calculation of the average reaction times of happy faces subtracted from the average reaction times of angry faces. The independent variables were the target letters in the different conflict conditions; congruent, incongruent and no conflict. The affect indices were calculated for each condition. The percentages of incorrect answers on the face and letter conditions were also calculated. A manipulation check was performed to see if conflict was established in the letter task. This was calculated using the inverse efficiency score (IES). The calculation for IES is the average reaction times divided by the percentage of answers correct on the letter conditions (Townsend & Ashby, 1978).
8
Materials
For the experiment an Eriksen Flanker task (Eriksen & Eriksen, 1974) consisting of letters and a judgement task of photos of emotional faces was used. Before the onset of the Flanker task participants were asked to either respond to the target letter or to the emotional face. Per participant there were 324 trials in total consisting out of 3 blocks with each 108 trials. There were 36 practice trials. Every trial in the Flanker task existed out of a masked flanker letter followed by a target letter and then a judgment task to measure affect. In two-thirds of the trials (72 trials per condition) participants were asked to only respond to the letters in the Flanker task and in one-third of the trials (36 trials per condition) they were asked to respond only to the emotional faces. This was done in order for the tasks to not become too difficult for the participants. Reacting to both tasks could lead to extra interference on the trials. The flanker task is a task where in every trial a masked flanker letter is shown first, either the letter N, H or nothing. The flanker letters are masked by white noise. The Flanker task consisted of three conflict conditions; congruent, incongruent and no conflict condition. The congruent condition contained masked flanker letters which were congruent with the target letter. The incongruent condition contained masked flanker letters which were incongruent with the target letter. The no conflict condition had no flanker letters masked by noise. In all conditions participants only had to respond to the target letters. The flankers were exposed for 30ms and the masks that followed were also exposed for 30ms each. The target letter was exposed for 600ms.
This task is followed by a judgement task which measures affect. In the judgement tasks the participants responded to an emotional face. Participants were shown a black and white photo of an emotional face which was also exposed for 600ms. These faces were selected from the Karolinska Directed Face Database. The selection included ten faces of five
9
models (three females and two males) with either a happy or angry expression. The faces as well as the letters were counterbalanced across the different conditions. Between trials a black screen was exposed between 1000 and 2000ms. The whole trial process is displayed in Figure 1. The letters (white) were a Calibri font with a size of a 100pt and black and white faces (562x762px) were shown on a black background on a 36x14cm HDMI computer screen using
the program Presentation® (Neurobehavioral Systems Inc., USA)1. Participants were able to
respond to the letters and photos using a device with left and right clickable buttons. This was used to measure their reaction times to the different targets. These buttons were always fixed. In the Flanker task the right button corresponded to the letter N and the left button to the letter H. For the judgement task the right button corresponded to a happy expression and the left button to an angry face.
Figure 1. A trial that displays the Eriksen Flanker task and the face judgement task.
1
10
Procedure
All participants signed an informed consent before partaking in the experiment and had the right to stop partaking in the experiment at any time. They filled in the online Dutch
questionnaire Vragenlijst voor Handvoorkeur (Van Strien, 2003) concerning handedness and were also asked about their vision. Before the start of the real experiment participants had 36 practice trials. Instructions were given about the right button corresponding to the letter N and happy face and the letter H and angry face corresponding to the left button. They were
instructed to react only to the letter or face when matching instructions were given on the screen. In two-thirds of the cases they were asked to respond to the letters and the other times they were asked to respond to the faces. After the letters an emotional face (happy or angry) was shown to which participants needed to respond as quickly as possible. Participants were asked to react as fast as possible by clicking the button that matched up with what they saw. In total the whole experiment lasted an hour and was divided into three blocks of 108 trials per participant. In between these blocks participants had a small break. After the experiment there was an exit-interview to check how participants thought they performed and what the experiment was about.
11
Results
This study should be seen as a pilot study for future research. The research planned for an effect size of Cohen’s d = 0,36 based on the research of Dreisbach and Fischer (2012) with a power of 0,95 calculated with the program G Power (Faul et al., 2007) and would have led to a total sample size of n = 85. In the present study we had fewer participants than planned initially. From the 34 participants five participants were excluded, because they had more than 10% errors on the easy letter task, and may not have been sufficiently motivated. If
participants had a more than 10% errors on the letter task we could not be certain that they paid sufficient attention to the letters for conflict to be elicited. Outliers in the reaction time scores (RTs) in the conditions were removed by means of a boxplot method. The boxplot method removed reaction times that had more than an interquartile range of 1.5 from the first and third quartile.
The recalculated power was 0.60 given 29 participants and an effect size of Cohen’s d = 0.36. The power is not sufficient to accurately detect a genuine effect and to draw conclusions with any degree of certainty. We can therefore better see this study as a pilot study informing future research.
Table 1.
The Affect Index of the Inverse Efficiency Scores of the Different Letter Conditions Displayed in Milliseconds.
Letters
Congruent Incongruent No Conflict
Averages Standard Deviations 520 84,83 582 85,95 531 102,18
12 Table 2.
The Error Proportions of Letters in the Different Conditions Expressed in Percentages.
Letters
Congruent SD Incongruent SD No Conflict SD
Averages 2,26 2,50 6,43 5,94 2,79 2,63
The reaction times of the letter conditions have been corrected with the inverse efficiency score (IES) to correct for the speed-accuracy trade-off and to make one dependent variable of the reaction times and percentages correct (Townsend & Ashby, 1978). The IES is calculated by taking the average reaction times and dividing them by the percentage of
answers that were correct. This was done per condition. As shown in Table 2. it was possible for the letter task to use the IES calculation as it met the criteria of having a less than 10% error score per condition.
Table 1. displays the results of the IES reaction times per condition on the letter task. Participants in the congruent letter condition were faster (M= -11.13 ,SE= 9.20) than in the incongruent letter condition (M=50.98, SE= 10.30) compared to the no flanker condition. This difference was significant t(28)=-5.72, p < .001, and represented a large effect, r=0.73. The data seem to suggest there is conflict in the incongruent condition, and the resolution of conflict in the congruent condition, but the latter facilitation is smaller than the former interference. This indicates that the manipulation had worked as it shows conflict is established.
13 Table 3.
The Affect Index of the Different Face Conditions displayed in milliseconds.
Faces
Congruent Incongruent No Conflict
Averages Standard Deviations 8 44,08 - 4 50,24 6 50,78 Table 4.
The Average Error Proportions of Faces in the Different Conditions Expressed in Percentages.
Faces
Congruent SD Incongruent SD No Conflict SD
Happy Angry 10,93 7,29 9,56 6,14 8,83 8,06 8,07 8,21 10,37 6,34 8,49 6,93
14
Figure 2. The affect indices of the different conditions of the face judgement task. The error bars represent the standard deviations.
The IES calculation could not be used for the judgement of emotional faces task, because one of the terms of use was that participants were not allowed to have more than 10% incorrect answers on this task. For this reason the judgement task was converted to affect indices for comparison between the different conditions. The affect index is the score of happy faces subtracted from angry faces. A high affect index means that participants reacted relatively faster on positive faces than negative faces and a low affect index means they reacted relatively slower on positive faces than negative faces. These results are displayed in Table 3. and Figure 2.
Participants reacted to the congruent letter condition had a higher affect index on the faces (M= 8,00 ,SE= 8,18), than they did when they reacted on the no flanker letter condition (M=5,67, SE= 9,43). This difference was however non-significant t(28)=0.22, p=0.83, and represented a small effect size, r=0.04 (Table 3.). This indicates a small positive priming effect. These results are in line with our expectations as we were expecting the affect indices between these conditions to be larger for the congruent flankers than the no flanker
-60 -40 -20 0 20 40 60 80
Congruent No Conflict Incongruent
Af fec t i nd ic es (r ea ct io n tim es in m s)
15
conditions. The effect size is very small. However the difference was non-significant which means that maybe a greater power is needed to see if this is truly an effect.
Participants that reacted to the incongruent letter condition had a lower affect index on the faces (M=-3.85, SE=9.33), than they did when they reacted on the no flanker letter
condition (M=5,67, SE= 9,43). This difference was however non-significant t(28)=-0.68, p=0.50, and represented a small effect size, r=0.13 (Table 3.). This possibly indicates a small negative priming effect. This is in line with our expectations as we expected affect indices to be smaller for the incongruent flankers than the no flanker conditions. The effect size is also fairly small. This was non-significant, but might yield different results in further testing with a greater power.
As shown in Table 4. Participants made a greater proportion of errors in the congruent happy face condition compared to the no conflict condition. They had 10,93% errors on the happy expressions in the congruent condition and10,37% errors on happy expressions in the no conflict condition. The difference between the congruent and no conflict condition is however non-significant, p = .237. The error percentages were also higher in the congruent happy expression condition, which had 10,93% errors, compared to the incongruent happy condition with 8,83% errors. This difference was significant with p < .001. We were
expecting the congruent condition to make less errors on happy faces than the no conflict and incongruent condition, but as shown the opposite result has taken place. We expected
congruent letters to facilitate the choice of a happy expressions as we expected conflict to be resolved. There were also less mistakes made on the incongruent condition, 8,83% than the no conflict condition, 10,37%. The difference was also significant with p < .001.
16
Discussion
The study should be seen as a possible indication for the affective monitoring
hypothesis as the results were non-significant and more power would be needed to see if there was truly an effect. In this study it seems that quickly solving conflict leads to positive affect, as shown with the positive priming effect that was generated from the congruent flankers. It also shows an indication of a negative priming effect of the incongruent flankers, meaning that unsolved conflict leads to negative affect. This is in line with the earlier findings of Dreisbach and Fischer (2012) and Fritz and Dreisbach (2013), who stated that conflict is perceived as aversive signals. However the positive and negative priming effect were very small. These were so small that we cannot clearly state if these results are due to the affective monitoring hypothesis or the fluency hypothesis. The difference between the no conflict and incongruent condition seemed to be greater than the difference between the no conflict and congruent condition. The design of the experiment alternated between two third of the trials where participants had to react on the letters only and one third of the time where they had to react only on the face. This design might have caused the effect sizes to be very small
between the congruent and no conflict letter and face trials. That might be because there was too much time in-between the two types of tasks.
The research gives an indication that positive affect is produced after conflict is resolved and negative affect is generated after conflict stays unresolved from neutral stimuli. This is in line with the affective monitoring hypothesis.
The speed of the tasks made it difficult for participants which may have generated more conflict. This could have led to more interference on the targets. Meaning that conflict may exist or be prolonged in the congruent condition where it should have been resolved when reacting on the target as stated in the affective monitoring hypothesis. Participants also stated that they needed to stay very focused because even a short amount of time of loss of
17
attention led to mistakes being made as they couldn’t remember what they were asked to react to. If participants made too many mistakes this could mean that the task has generated extra conflict because of it being too difficult or too little or no conflict as participants were distracted. Extra conflict would lead to more interference of the flankers on the target. No conflict being generated would lead the manipulation not to work. We would therefore not be able to know if affect is generated through conflict as stated in the affective monitoring hypothesis. Possibly slowing down the offered stimuli may lead to faster reaction times and could lead people to make less mistakes. Also lengthening the practice trials would enable participants to get better used to the speed and instructions of the task, which could also lead to participants producing less errors.
Another possible way to improve upon this research is the letter choice in the Eriksen Flanker task. The letter N and H are very similar in shape and could easily be mistaken for each other. This might have led to extra conflict that would cause interference in the performance of the participants. A better choice would be to take letters that have a distinct shape from each other, as in the letter X and O.
Some participants also had problems identifying the displayed emotion of certain faces, which could have led to more interference and caused them to react more slowly than they would have if the emotion was more clearly stated. This would also make the results unreliable. Changing these faces with clear unambiguous expressions will help solve this problem.
A left-handed button box was used in this experiment as the right-handed button box was broken. This might have been harder for participants to use because participants were right-handed. This could have possibly led to slower reaction times for the right non-dominant middle finger which was used for pressing the positive button, than they would have made with a button box suited for people who are right-handed. This would cause the difference
18
between positive and negative reaction times to become smaller. The difference would then be caused by a difference of finger dominance and induce extra conflict for the right finger compared to the left finger. In other words participants would react slower with the right finger than the left finger. This could be a possible explanation why more errors were made in the congruent condition compared to the incongruent and no conflict condition on happy faces. If this is corrected for with an appropriate button-box the difference between positive and negative reaction times caused by differences in the left and right finger can be lifted.
This pilot study gives an indication of positive affect being generated from neutral stimuli. However as stated before this cannot be clearly found from our results as the differences were very small. In the manipulation check we found more interference than facilitation. As said earlier this could mean that conflict in the congruent condition was
prolonged and did not lead to the resolution of conflict before reacting to the target. Hence we cannot state with certainty if positive affect is established from neutral stimuli by resolving conflict. This by no means means we have to refute our affective monitoring hypothesis. However at this point it is not possible to clearly state if affect is generated because of the fluency hypothesis or the affective monitoring hypothesis. There seems to be a good indication of affective monitoring being present as a way of generating affect through conflict. Further research with a greater power is needed to see if this is the case. This study offers a good opportunity for improving upon the original experiment.
19 References
Beilock S. L. & Holt L. E. (2007). Embodied preference judgments: can likeability be driven by the motor system? Psychological Science, 18, 51–57.
Cannon P. R., Hayes A. E. & Tipper S. P. (2010). Sensorimotor fluency influences affect: Evidence from electromyography. Cognition and Emotion, 24, 681–691.
Chetverikov, A., Jóhannesson, Ó. I. & Kristjánsson, Á. (2014). Blaming the victims of your own mistakes: How visual search accuracy influences evaluation of stimuli. Cognition and Emotion, (ahead-of-print), 1-16.
Dreisbach, G. & Fischer, R. (2012). Conflicts as aversive signals. Brain and Cognition, 78, 94-98.
Eriksen, B.A. & Eriksen, C.W. (1974). Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception & Psychophysics, 16(1), 143-149. Faul, F., Erdfelder, E., Lang, A.G. & Buchner, A. (2007). G*power 3: a flexible statistical
power analysis program for the social, behavioral, and biomedical sciences. Behavioral Res Methods, 39, 175–191.
Fazendeiro, T., Chenier, T. & Winkielman, P. (2007). How Dynamics of Thinking Create Affective and Cognitive Feelings. Psychology and neuroscience of the connection between fluency, linking, and memory.
Fritz, J. & Dreisbach, G. (2013). Conflict as aversive signals: Conflict priming increases judgment for neutral stimuli. Cognitive, Affective, & Behavioral Neuroscience, 13 (2), 311-317.
Heijden, J. van der (2013). Conflicts as positive signals. Unpublished master’s thesis, University of Amsterdam, the Netherlands.
Monahan, J.I., Murphy, S.T. & Zanjonc, R.B. (2000) Subliminal mere exposure: specific, general and diffuse effects. Psychological Science, 11(6), 462-466.
20
Murphy, S.T., Zajonc, R.B. & Tesser, A. (1993) Affect, cognition, and awareness: Affective priming with optimal and suboptimal stimulus exposures. Journal of Personality and Social Psychology, 64 (5), 723-739.
Notebaert, W. & Verguts, T. (2006). Stimulus conflict predicts conflict adaptation in a numerical flanker task. Psychonomic Bulletin & Review, 13(6), 1078-1084. Ogawa, H. & Watanabe, K. (2011). Implicit learning increases preference for predictive
visual display. Attention, Perception, & Psychophysics, 73, 1815-1822.
Phaf, R.H. & Rotteveel, M. (2012). Affective monitoring: A generic mechanism for affect elicitation. Frontiers in Psychology, 3: 47.
Raymond, J. E., Fenske, M. J. & Tavassoli, N. T. (2003). Selective attention determines emotional responses to novel visual stimuli. Psychological Science, 14, 537–542. Schacht, A., Nigbur, R. & Sommer,W. (2009). Emotions in Go/ NoGo conflicts.
Psychological Research, 73, 843–856.
Schacht, A., Dimigen, O. & Sommer, W. (2010). Emotions in Cognitive Conflicts are Not Aversive but are Task Specific. Cognitive, Affective, and Behavioral Neursocience, 10, 349–356.
Strien, J.W. van (2003). Vragenlijst voor Handvoorkeur. Retrieved from http://hdl.handle.net/1765/955 .
Townsend, J.T. & Ashby, F.G. (1978). Methods of modeling capacity in simple
processing systems. In J. Castellan & F. Restle (Eds.), Cognitive theory. 3, 200-239. Westerman, D.L., Lanskam, M. & Olds, J.M. (2014). The Effect of Processing Fluency on
Impressions of Familiarity and Liking. Journal of Experimental Psychology: Learning, Memory, and Cognition.
Willem, S. & Van Der Linden, M. (2006). Mere exposure effect: a consequence of direct and indirect fluency-preference links. Consciousness And Cognition, 15(2), 323-341.
21
Winkielman, G. & Cacioppo, J.T. (2001). Mind at Ease Puts a Smile on the Face:
Psychophysiological Evidence That Processing Facilitation Elicits Positive Affect. Journal of Personality and Social Psychology, 81(6), 989-1000.
