FIRST EXPERIMENT

FIRST EXPERIMENT 3.1 Overview The purpose of this first study was to determine if explanatory style as measured by the ASQ could explain systematic differences in appraisal and emotions. In addition, we tested the feasibility of the experimental setting for the research questions to insure that 1.) the overall cover story and the nature of the social intelligence test was believable and motivated participants to really get involved; 2.) the social intelligence test situation was stressful enough to stimulate causal search; and 3.) anger emotions would be reported. Besides testing the basic parameters of the induction method, we wanted to create a relationship between the two participants. Indeed, authors such as Averill (1982) have suggested that individuals are more likely to become angry with someone when they have a relationship. However, because we selected participants based upon extreme scores on the ASQ, finding individuals with a real relationship would have been difficult. We were therefore inspired by a procedure to induce interpersonal closeness experimentally (Aron, Melinat, Aron, Vallone, & Bator, 1997). Specifically, Aron et al. (1997) devised a getting to know each other procedure in which two persons engage in a 45-minute structured interaction based on a series of questions. The questions start out quite general (e.g., What is your name? Where do you come from?) and become ever more intimate (e.g., What do you regret most in your life? What does friendship mean to you?). Aron et al. (1997) found that this procedure could create enough closeness between persons to study some variables concerning interpersonal relationships. We therefore conducted a pilot study to test the procedure and adapted it for our study. Finally, we were interested in studying the personality – appraisal – emotion relationship and to test the partially contradictory predictions made by appraisal / attribution theories and optimism literature concerning how attribution style and anger might be related in social situations. 60 3.2 Method The overall method and procedure of Study 1 was similar to the one described in Chapter 2. Only those elements unique to this study will therefore be mentioned in the following section. 3.2.1 Recruitment and participants 109 students filled out the ASQ during class time. The mean score on the ASQ (CoNeg) was 69.9 with a standard deviation of 11.29 and participants who had scores of 80 or above (Internals) and 60 and below (Externals) were retained. The study was introduced during class time two weeks later, and the codes and dates of birth of all potential participants was passed around the lecture room to give those students who were on the list the possibility to sign up. Thirty-seven students originally signed up for the study. However, in order to have enough participants, we also recruited two students with a score of 61 and one with a score of 63 on the ASQ (CoNeg). In total, therefore, 40 participants (6 males and 34 females) took part in the study; 19 Internals (Range of scores = 80-103; M = 87, SD= 6.2) and 21 Externals (Range of scores = 43-63; M = 56.9, SD 4.85). The students ranged from 19 to 36 years of age (M = 22.3, SD = 4.28). 3.2.2 Procedure Questionnaire phase: The participants came to one of two sessions organized in a seminar room at the university to fill out the acquaintanceship and personality questionnaires already listed in Chapter 2. During this session, participants were also randomly paired with a same sex participant whom they did not know, and were asked to sign up for the experimental session at a time that suited both of their schedules. Experimental phase: As explained in Chapter 2, each pair of students came separately to the laboratory to take part in the social intelligence test. A unique aspect of this first study, however, was the relationship-building procedure. 61 Relationship-building procedure. After the general context of the research had been described and the participants had signed the consent form, the second part of the barrier was put up in order to obstruct all eye contact between the participants. The experimenter (“E”) then gave the students the relationship-building questions (see Appendix 14 for the questions) and read them the following instructions: “When we work with other persons by telephone, by fax or by e-mail – even across large distances – we often wish to know who we are dealing with. This is particularly true when we work on complex problems or when we frequently work with the same person. Consequently, professionals also pose each other questions of a more private nature during their telephone conversations so as to get to know each other and sometime in the hope of being able to work together more efficiently and effectively. Some individuals even find that it is much easier to communicate with someone through these new forms of communication (telephone, fax, e-mail) than discussing with someone face-to-face.” The E explained that in preparation for the social intelligence test, the participants would have 20 minutes to interact as though they were on the telephone. To make the interaction more efficient, however, they were requested to rely on series of questions designed to help strangers get to know each other in a relatively short time. In addition, they were asked to take turns asking each other the questions in the order they were presented on the instruction sheet. The E made sure that the participants understood the procedure and then gave them a short questionnaire to measure first impressions of the interaction partner and emotions (Appendix 15). The participants were then left alone in the experiment room and interacted for twenty minutes while the E sat in the second room and filmed and observed the participants (to make sure the instructions were being followed and that no problems arose). After twenty minutes, the E interrupted the conversation and gave the participants a second partner perception questionnaire (Appendix 16). Social intelligence test: The social intelligence test was introduced and carried out as described in Chapter 2. 62 3.3 Overview of the measures and predictions The participants filled out questionnaires at four different moments during the experiment: T0: before the relationship-building procedure; T1: after the relationship-building paradigm and in anticipation of the social intelligence test; T2: after the social intelligence test; and T3: after receiving the poor score on the test. These questionnaires included manipulation check and implication questions, as well as the 17 emotions on a nine-point rating scale already described in Chapter 2. The following section will describe the specific predictions for this study and the items used to measure them. Appendix 17 also contains a list of variables included for exploratory purposes. 3.3.1 Relationship-building questions In order to determine if the relationship-building procedure had been successful, participants received a series of questions at T0 and T1. These questions included the items from the Rubin Liking Scale (Rubin, 1970) as well as six additional items we especially created for the experiment. The Rubin Liking Scale is one of the few validated relationship quality questionnaires (Sternberg, 1997) and is comprised of 13 questions, which participants rate on scales of 1 (not at all true; disagree completely) to 9 (definitely true; agree completely). The questionnaire was translated into French by a native French-speaking research assistant with very good knowledge of English and then back translated into English by a native English-speaking research assistant with very good knowledge of French. The following two items from the original scale were not included: 1.) When I am with___, we are almost always in the same mood: and 2.) I think that ____ is unusually well adjusted. The first question did not make sense in the present context because more than one meeting would have been necessary to answer the question. The concept of “being well adjusted” is very particular to American culture and has no equivalent in French. Finally, because the Rubin scale was relatively old, and we did not know if the French translation would adequately capture partner perception and liking in the present cultural context, we created six new questions for exploratory purposes. 63 3.3.2 Hypotheses 1.) According to many emotion theorists (e.g., Folkman & Lazarus, 1985), appraisal and resulting emotions are processes that will vary with both the situation and the individual. Indeed, similar situations – such as a frustrating achievement context, or receiving a low score on an important test – probably would be evaluated in similar ways by many types of individuals and thus engender rather similar emotions. We therefore expected Internals and Externals to have rather similar emotion patterns across measurement points. In other words, because the experiment had four measurement points concerning different and distinct situations, we predicted a main effect of measurement point on emotions. Personality, on the other hand, should partially account for systematic appraisal and emotion differences within a particular measurement point. In particular, Internals and Externals were not expected to report significantly different emotions at T1 because these measures served primarily as baselines. We did expect significant differences at T2, after the social intelligence test, and at T3, after the reception of the low score, because these situations were created to stimulate causal search and negative emotions. 2.) Explanatory style has been defined as a generalized appraisal tendency across situations. Based upon the literature review, we therefore predicted that Externals would be significantly more likely to make external attributions than Internals at T2 and T3, and that Internals would make significantly more internal attributions than Externals at T2 and T3. We also wished to explore if external attributions would be made towards the other person or towards other things. The Table 3.1 below summarizes the questions and predictions. In addition to the questions already described in Chapter 2, we added a three-part exploratory question (Responsible: self, Responsible: partner, Responsible: external). This exploratory question asked participants to distribute 100 points between the partner, the self, and other external forces concerning their degree of responsibility: “If you could distribute 100 points between your partner, yourself, and other external factors (which you need to specify), to indicate the responsibility of each party for the results on this test, how would you do this?” 64 Table 3.1: Personality and causal attribution predictions Variable Name ¹ Prediction ² Variables for Internal Attribution Cause: self (2,3)* Internals ↑ Cause: self intentional (2,3)* Internals ↑ Responsible: self (2,3) **** Internals ↑ Variables for External « Partner » Attributions Cause: partner (2,3)* Externals ↑ Cause: partner intentional (2,3)* Externals ↑ Responsibility: self vs. partner (2, 3)** Externals ↑ Better performance with another (2,3)*** Externals ↑ Responsible: partner (2,3) **** Externals ↑ Variables for External « Other » Attributions Cause: chance (2,3)* Externals ↑ Responsible: external (2,3) **** Externals ↑ Note. ¹ The actual questions corresponding to the variable names can be found in Appendix 12. The number in parentheses indicates the measurement points during which the questions were posed (0 = T0, 1= T1, 2=T2, 3= T3). ² This column indicates whether Internals are expected to have a significantly higher score than Externals on a particular question (or vice versa) * Single item questions with a 9-point scale ranging from “not at all” to “very much”. The higher the score, the more this item is being endorsed. ** Nine-point variable where a score of 1= the self and a score of 9 = the partner. The higher the score on these questions, the more responsibility is being attributed to the partner. *** The higher the score on this 9-point scale, the more the respondents believe that they could have performed better with another person (i.e., an indirect measure of blaming the partner) **** These exploratory questions asked participants to distribute 100 points between the partner, the self, and other external forces concerning each party’s degree of responsibility. 3.) Anger, irritation, frustration, and contempt are generally considered to be other-directed emotions (e.g., Averill, 1982; Scherer, 2001) and shame, guilt, and embarrassment self-directed or self-conscious emotions (e.g., Lewis & Haviland, 1993; Haidt, 2003; Roseman, 2001; Scherer, 2001; Tangney & Fischer, 1995). Based on our literature review, we predicted that Externals would report more other-directed emotions than Internals and that Internals would report more self-directed or self-conscious emotions than Externals at T2 and T3. 4.) We predicted that the relationships between attribution style and particular emotions would be mediated by specific causal attributions. In particular, we predicted that Externals making external attributions would report more other-directed emotions and that Internals making internal attributions would report more self-directed emotions at T2 and T3. 65 3.4 Results 3.4.1 Overview of analyses The analyses will be presented in a series of steps. First, the accuracy and normality of the dependent variables will be examined. Second, the manipulation check and implication questions will be analyzed to inspect if the experimental procedure was taken seriously by the participants and considered to be stressful. The success of the relationship-building procedure will also be determined. Third, we will check if the experimental procedure was able to induce emotions, by examining if measurement point had a main effect on emotion. Fourth, we will examine systematic differences between Internals and Externals concerning the predicted emotions and attributions at T2 and T3. 3.4.2 Preliminary analyses Dependent variables were examined through various SPSS programs in order to check for accuracy of data and missing values. We searched for univariate outliers by examining z-scores and identified variables with scores over 3.29 (Tabachnick & Fidell, 2001). Interestingly, 37 negative emotions out of 64 possible emotions across the four measurement points had scores that fell into this range. We therefore examined the emotion variables carefully, by visually inspecting distributions, and analyzing normality using both skewness and Kurtosis indicators. Appendix 18 lists all emotion variables ordered in terms of skewness and Kurtois. These results show that positive emotions were relatively normal in their distribution or slightly positively skewed. The more negative the emotion terms, however, the more these variables were positively skewed. This could be due partially to social desirability and the unwillingness of participants to report strong, negative emotions. More likely, however, this type of variable distribution is normal for an experimental situation in which very strong, negative emotions would rarely be experienced or reported5. We then looked for consistent high or low respondents by examining boxplots of the variables. Although this inspection showed that some participants 6 reported more negative emotions than others, we could find no indication that these were unrealistic scores. On the 5 Although emotion theorist would like to have normally distributed negative emotions, there are many ethical ramifications if the induction of emotions such as anger of fear is too powerful. 6 In particular, one External (Number 14, a male) and two Internals (Number 22 and 40, both females) were responsible for several outlying variables within negative emotions. 66 contrary, it seemed very reasonable to expect some participants to experience more negative emotions or stronger negative emotions than others in this type of situation. Because we were interested in understanding individual differences, we therefore decided not to eliminate these participants. 3.4.3 Manipulation check questions and implication The second step was to insure that individuals had been sufficiently motivated and implicated in the task and did not guess the real purpose of the experiment. The responses to the manipulation check questions, in which we asked participants to describe in their own words the purpose of the study, confirmed that the students generally believed the cover story and did not guess the real aims of the research. In addition, participants thought it was important to do well on the task and reported giving their best effort, because all means on these questions were at least above 5 on a 9-point scale and many were much higher (Table 3.2 below shows the means and standard deviations of all implication and manipulation check question). For example, before the task, 90% of participants gave a rating of 5 or above (on a scale ranging from 1 = not at all to 9 = very important) on the question “How important is it for you to receive a good result on this test”). After the task, 77.5% of participants gave a rating of 5 or above on this same question. In addition, 97.5 % of participants gave a rating of 5 or above (scale ranging from 1 = no effort to 9 = very best effort) on the question “Do you feel you gave your best effort on this task” which was asked after the task. Finally, most participants had perceived the social intelligence test to be stressful, with 80 % of the participants rating the stress level at 5 or above. Table 3.2: Means and standard deviations of all manipulation check and implication variables Overall Internals Externals Variable Name M SD M SD M SD Importance: self (1) 6.00 1.49 6.00 1.52 6.00 1.49 Importance: partner (1) 5.80 1.24 5.71 1.10 5.89 1.41 Importance: self (2) 5.48 1.97 5.81 1.54 5.11 2.35 Effort: self (2) 7.98 1.39 8.05 1.02 7.89 1.73 Effort: partner (2) 8.13 1.20 8.19 0.93 8.05 1.47 Effort: self (3) 7.43 2.17 7.10 2.14 7.79 2.20 Effort: partner (3) 7.70 1.70 7.43 1.50 8.00 1.89 Test perceived as stressful (2) 5.60 1.93 5.48 1.50 5.74 2.35 Note. All means and standard deviations are based upon raw scores. All variables were rated on a nine-point scale. 67 Independent T-tests were conducted to determine if there were any significant differences between Internals and Externals or between male and female participants in the level of implication, effort, or perception of stress. We found no such differences. We interpreted these combined results to mean that our experimental situation was successful in inducing the most important prerequisites for engendering emotions (goal relevance and implication) and for allowing individual differences to emerge (stress). 3.4.4 Relationship-building paradigm The next step was to determine if the relationship-building paradigm had been successful. We first conducted standardized alpha analyses on the exploratory relationship quality questions (6 items) and the Rubin liking scale (11 items) separately. Because both alphas were very high, we decided to put together the questions and to create a total relationship quality questionnaire (TRQ, 17 items; see Table 3.3 for all standard deviations and standardized alpha values). We then conducted a repeated-measures ANOVA with two between-subject factors (attribution style and gender) to test differences in the TRQ scores between T0 and T1. Overall, we found that the relationship-building paradigm had been successful (F(1, 38) = 72.069, p < .000.7, η² = .334). There was no interaction effect with attribution style. However, the six males participants were generally less influenced by the relationship building than the female participants (F(1, 36) = 7.951, p < .01, η² = .181). There were no three-way interaction effects. These results indicate that the relationship-building was equally successful in inducing some relationship-like qualities in both Internals and Externals, albeit somewhat more successful for females than for males. Table 3.3: Means, standard deviations, and standardized alphas of the relationship questions Variables Mean SD Standardized alphas Exploratory questions (6 items) Pre-test 5.77 1.23 .8495 Post-test 6.90 1.31 . 8925 Rubin liking scale (11 items) Pre-test 4.98 1.26 .9245 Post-test 5.95 1.40 .9326 Total Relationship Quality Questionnaire Pre-test 5.26 1.17 .9384 (TRQ, 17 items) Post-test 6.28 1.33 .9555 Note. All analyses based on raw scores 7 Unless stated otherwise, we will always report Pillai’s trace result, as recommended by Olson (1976). 68 3.4.5 Testing the hypotheses Emotions As a next step we examined if the experimental situation was able to induce emotions. First, we checked if the procedure had induced different emotions across the four measurement points. Second, we checked if Externals or Internals were generally prone to report particular emotions across the situations. Finally, we were interested in interaction effects. To examine these questions, we conducted separate within-subject repeated-measure ANOVAs for each emotion8 across the four measurement points with attribution style as a between-subject factor9. Table 3.4 gives the means and standard deviations of all emotions at each measurement point. Table 3.5 presents the within and between-subject effects for each emotion, reporting the F value and the η² effect size. Table 3.6 presents the within-subject contrasts. Finally, Table 3.7 gives the means of the emotions at each measurement point for the two attribution style groups. Based on the results presented in these four Tables, we will first interpret the within-subject main effects, then the between-subject main effects, and finally the within-subject interaction effects. Table 3.4: Means and standard deviations of all emotions at each measurement point Measurement point T0 T1 T2 T3 Emotion M SD M SD M SD M SD Anger 1.23 0.73 1.15 0.80 1.65 1.25 1.78 1.42 Irritation 1.65 1.53 1.18 0.71 2.58 2.25 2.10 1.88 Frustration 1.25 0.84 1.30 1.18 3.43 2.82 2.93 2.16 Contempt 1.08 0.35 1.00 0.00 1.05 0.32 1.05 0.32 Shame 1.20 0.61 1.18 0.59 1.43 1.47 1.58 1.08 Guilt 1.23 0.58 1.15 0.53 1.45 0.99 1.58 1.11 Embarrassment 2.63 1.94 1.70 1.71 2.18 1.78 1.83 1.15 Sadness 1.58 1.26 1.18 0.50 1.28 0.64 2.13 1.79 Joy 4.55 1.80 4.15 1.96 3.43 1.88 2.80 2.16 Satisfaction 3.73 2.30 3.20 1.95 2.68 1.73 1.80 1.24 Hope 4.25 2.77 3.75 2.52 3.53 2.20 2.48 2.35 Aroused 3.50 2.20 3.58 2.24 3.50 2.21 1.45 0.99 Nervous 3.15 2.24 3.28 2.43 3.53 2.39 1.50 1.22 Anxiety 2.63 1.93 2.95 2.28 2.30 2.05 1.45 1.43 Fear 2.35 1.75 2.38 1.81 1.93 1.53 1.25 0.93 Surprise 2.10 1.45 2.58 1.89 2.30 1.86 2.10 1.84 Disgust 1.08 0.35 1.00 0.00 1.10 0.38 1.58 1.17 Note. All means and standard deviations are based on raw scores. 1 = did not experience the emotion; 9 = experienced the emotion very strongly 8 We eliminated contempt from the analyses, because only two persons reported this emotion. 9 Because only six males participated in the study, we were not able to include gender as a between-subject factor. 69 Table 3.5: Repeated-measure ANOVAs for each emotion Within-subject effects Between-subject effects Measurement point Measurement point x Attgroup¹ Attgroup¹ Emotion Df (Huynh-Feldt)² F η² F η² F (df=1) η² Anger 2.89 3.796** .091 3.415* .082 2.545 .063 Irritation 2.538 7.537*** .166 2.058 .051 10.135** .211 Frustration 2.202 22.287*** .370 1.457 .037 4.519* .106 Shame 2.070 1.76 .044 .463 .012 3.991 .095 Guilt 2.454 2.697 .066 .245 .006 1.54 .039 Embarrass. 2.79 4.583** .108 .496 .013 2.185 .054 Sadness 1.79 6.025** .137 1.206 .031 2.773 .068 Joy 2.945 14.734*** .279 2.051 .051 .099 .003 Satisfaction 2.734 12.682*** .250 .088 .002 .293 .008 Hope 2.661 9.158*** .194 1.262 .032 2.614 .064 Aroused 2.849 18.204*** .324 .491 .013 .525 .014 Nervous 3 16.344*** .301 2.37 .059 5.687* .130 Anxiety 2.808 11.631*** .234 1.691 .043 3.587 .086 Fear 2.588 7.987*** .174 .377 .010 .646 .017 Surprise 3 1.041 .027 4.742** .111 .713 .018 Disgust 1.356 7.318** .161 .334 .009 .357 .009 Note. ¹ Attgroup = Attribution style group (Binary variable, 1= Externals; 2= Internals) ²Because the majority of emotions did not meet the sphericity assumption10, we report the Huynh-Feldt11 correction for all