Experiment 2

The aim of Experiment 2 was to further investigate the processing of the power check and its efferent effects on facial expressions. Findings from Experiment 1 suggest that low power appraisal check results differentially affect the response patterning of the goal conduciveness check over the cheek region compared with high power appraisal check results. Furthermore, findings from Experiment 1 indicate that the upper face is not affected by efferent effects of the power check.

In Experiment 1, the absence of the predicted power check effects in the upper face (over the frontalis and the corrugator region) might have resulted because the manipulation of the power check was not strong enough to drive the predicted efferent effects, since in the concurrently recorded brain activity, significant effects of the power check were found.

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 128 Furthermore, the Component Process Model predicts that when control is possible, Coping depends on the power check. This proposition indicates that the evaluation of power is strengthened by antecedently processed control checks. Therefore, in Experiment 2, the manipulation of the power check was increased by implementing a manipulation of the control check (by changing the frequency of having high and low power during the gambling task). Thus, the gambling task in Experiment 2 had blocks of gambling where participants had frequent high power feedback (i.e., high control block) and blocks where they had infrequent high power feedback (i.e., low control block). Previous findings on the

phenomenon of illusion of control showed that changes of the frequency of action-outcome contingencies in pure chance tasks changed participants’ perception about the degree to control the task (Biner et al., 1998; Langer, 1975; Langer & Roth, 1975; Wohl & Enzle, 2002). In these tasks, participants that frequently perceived that their action resulted in a desired outcome reported higher degrees of control over the task. Accordingly and based on the findings of Experiment 1, we expected that frequent high power feedback would increase participants’ perceived degree of control and consequently, attenuate the efferent effects of high power check results. In contrast, infrequent high power feedback was predicted to decrease the perceived degree of control and consequently, magnify the efferent effects of the low power check. Thus, effects of the power check were expected to be magnified in blocks of infrequent high power feedback (low control) in comparison to blocks of frequent high power feedback (high control). Apart from that, for Experiment 2, the predictions about the efferent effects of the manipulated appraisal checks were the same as in Experiment 1.

3.3.4.1. Materials and Methods 3.3.4.1.1. Participants

Thirty-four female students of the University of Geneva took part in Experiment 2.

Due to technical problems, stimulus trigger information of six participants was inadequately

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 129 recorded, and thus, these participants were removed from the analysis. The remaining 28 participants ranged in years of age from 18 to 30 (M = 21.21, SD = 3.06). They were all healthy, right handed (mean of Edinburgh Handedness Inventory = 88.93, SD = 12.27), and had normal or corrected to normal vision. Given the changes of the task (see below), the guaranteed amount of participation fee was 25 CHF.

3.3.4.1.2. Experimental paradigm and procedure

The gambling task of Experiment 2 followed Experiment 1 (Figure 3B) except for: (a) the amount of circles at “choice of circles”, which was adjusted to the two levels of the goal conduciveness check (i.e., winning and losing money), (b) the increase of the number of trials (from 384 to 864 trials), due to the power-frequency manipulation. Consequently, the

duration of the gambling task would have increased to more than one hour. To limit the duration of the gambling task to 45 min, the presentation times were reduced for the fixation cross (randomized duration between 300 and 700 ms), response feedback of selected circle (300 ms), and response feedback of choice about the outcome (300 ms). Moreover, (c) as a result of the increased number of trials, the amount of guaranteed play money was set to 25 CHF, and (d) the monetary magnitude of wins was 0.05 CHF and of losses was -0.05 CHF.

Resulting in 26 CHF of maximal additional amount of play money participants could win.

To implement the additional manipulation of the control check, in the form of gambling blocks of frequent high power feedback and blocks of infrequent high power feedback, each block consisted of 144 gambling trials. The high control block (i.e., frequent high power feedback) gave high power feedback in 75% of the trials and low power feedback in 25% of the trials. In contrast, in low control blocks (i.e., infrequent high power feedback), the percentage of high and low power feedback was reversed. Importantly, in order to control for frequency effects other than those of the power-frequency manipulation, the amount of wins and losses was equal in each block (50% wins, 50% losses). Further, participants were

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 130 told in the beginning of each block how much control over the upcoming gambling block they could expect. In the beginning of high control blocks, participants read: “In most of the trials it will be you who decides about the outcome of a gambling trial”; whereas before a low control block started, they read: “In most of the trials it will be the computer who decides about the outcome of a gambling trial”.

Given the manipulation of the control check, the preserved power check manipulation (high and low power), and the reduction of the level of the goal conduciveness manipulation (wins and losses), the eight feedback-stimuli conditions of Experiment 2 were: “high control, high power loss”, “high control, high power win”, “high control, low power loss”, “high control, low power win”, “low control, high power loss”, “low control, high power win”,

“low control, low power loss”, and “low control, low power win”.

3.3.4.1.3. Data acquisition and preprocessing

The same set-up of data acquisition, preprocessing steps, artifact correction procedures, and percentage score calculation (of subsequent 100-ms time intervals after feedback-stimulus onset) were applied in Experiment 2 as in Experiment 1. In Experiment 2, the amount of rejected trials of the entire EMG data due to artifacts and outliers was 2.17% in total.

3.3.4.1.4. Statistical analyses of temporal profiles of appraisal check effects

Based on the results of Experiment 1 and previous studies (Delplanque et al., 2009;

Lanctot & Hess, 2007), the differential unfolding of efferent appraisal check effects including Control as a factor was tested from 400 ms to 1,400 ms after feedback onset. Therefore, the temporal profile analysis included eleven 100-ms time intervals. The EMG data of each facial region was submitted to a 2 (Power: high vs. low) × 2 (Goal conduciveness: loss vs. win) × 2 (Control: high vs. low) × 11 (Time: 100-ms time-intervals starting from 400 ms) repeated measures ANOVA with Greenhouse-Geisser correction. In the results the epsilon-value of the

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 131 Greenhouse-Geisser correction are reported as ε, and the degrees of freedom are reported as uncorrected values. All tests were performed at an alpha level of 5%. All reported effect sizes are partial ŋ² (in the results simply noted as ŋ²). All statistical analyses were carried out with IBM SPSS Statistics 19.

3.3.4.2. Results

According to the three tested predictions, a priori specified contrasts were calculated, if the four-factorial repeated measures ANOVA revealed (1) significant main effects of goal conduciveness over the corrugator and the cheek region, or main effects of power or control over the frontalis, the corrugator, and the cheek regions; and (2) significant interaction effects of Goal Conduciveness × Time over the corrugator and cheek region, or of Power × Time, Control ×Time or Power × Control × Time over the frontalis, the corrugator, and the cheek regions. Cumulative effects were predicted between the goal conduciveness check and power check, given the results of Experiment 1. Nevertheless, no specific a priori hypotheses were formulated about the interaction effects between the manipulated appraisal checks. The results of the repeated measures ANOVAs for the frontalis, the corrugator, and the cheek region are summarized in Table 3.

3.3.4.2.1. Activity over the frontalis region

The four-factorial ANOVA yielded no significant main or interaction effects of power, control, goal conduciveness, and time, (F-values < 2.57, pGG –values > .120). These results are similar to the ones of Experiment 1, suggesting that the muscle activity over the frontalis region was not significantly affected by efferent effects of the appraisal check manipulation, contrary to the predictions.

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 132 Table 3

Results of the Repeated Measures ANOVA for each Facial Region of Experiment 2

Frontalis region Corrugator region Cheek region Note. N = 28. For each facial muscle region a Greenhouse-Geisser adjusted repeated measures ANOVA was calculated with the within-subject factors of goal conduciveness (GC: win vs. losses), power (free vs.

no choice about GC), control (frequent vs. infrequent high power), and time (11 post-stimulus 100-ms time intervals).

†p < .10. *p < .05.

3.3.4.2.2. Activity over the corrugator region

The ANOVA revealed only one significant effect in the form of an interaction effect between Goal Conduciveness × Power, F(1, 27) = 4.20, p = .050, ŋ² = .14. None of the predicted main effects of Power, Control, or Goal conduciveness, or any respective

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 133 interaction effects were significant (F-values < 1.68, pGG -values > .206). These ANOVA results indicate that efferent effects of the goal conduciveness check and the power check cumulatively affected the response patterning over the corrugator region. Furthermore, the fact that there were no significant control effects is in line with a previous study showing that corrugator activity is unaffected by stimulus frequency (Schacht, Nigbur, & Sommer, 2009).

The pattern of the interaction effect between Goal conduciveness × Power is

illustrated in Figure 6. Post hoc tests revealed that as a function of the degree of power, wins differently affected muscle activity over the corrugator region, t(27) = 2.04, p = .051.

Corrugator activity was increased in response to “low power wins” compared with “high power wins”. No other comparisons were significant (t-values < 1.56, p-values > .131). This response patterning indicates that appraisal results of low or high power (i.e., bounded or free choice about the win, respectively) cumulatively differentiated the efferent effects of

appraised goal conducive events (wins) over the corrugator region. In comparison to previous studies, it appears unusual that corrugator activity was increased in response to wins (positive events) and unaffected by losses (negative events), and further that corrugator activity varied as a function of the power check results (cf. Cacioppo et al., 1992) . It is possible that the interaction effect constitutes an appraisal check effect of task/goal relevance and goal conduciveness. In fact, the ERP results of the concurrently recorded brain activity suggest that the power check information has been evaluated in terms of goal conduciveness because feedback-related negativity main effects of the power check manipulation were found. Thus, these results on feedback-related negativity indicate that having bounded choice about the gambling outcome (low power) was negatively evaluated compared with having free choice (high power). Consistent with prediction (1) and findings in previous studies, the interaction effect over the corrugator region advocates that wins were generally evaluated as task/goal relevant, and “high power” was additionally evaluated as goal conducive, which lead to

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 134 muscle tone relaxation, in comparison to “low power”, which was evaluated as goal

obstructive and increased corrugator activity.

Figure 6. EMG amplitude variation over the corrugator region (Experiment 2). The standard errors of the mean are also shown. *p = .051.

Activity over the cheek region

The ANOVA yielded a significant four-fold interaction effect between Goal

Conduciveness × Power × Control × Time, F(10, 270) = 3.64, p_GG = .037, ŋ² = .12, ε = .185.

Further, the interaction effect between Power × Time was marginally significant,

F(10, 270) = 2.45, pGG = .076, ŋ² = .08, ε = .270. The only significant main effect was Time, F(10, 270) = 7.49, p_GG = .005, ŋ² = .22, ε = .141. Neither main effects of goal conduciveness, power, and control nor respective interaction effects were found (F-values < 2.60, pGG -values > .118). This four-fold interaction effect suggests that efferent effects of the goal conduciveness check and the power check differentially unfolded over the cheek region as a function of the appraised degree of control (see Figure 7).

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 135 Post hoc tests examined the temporal profile of the Goal Conduciveness × Power × Control × Time interaction effect. In total, for each of the eleven time intervals (from 400–

1,400 ms after feedback-stimulus onset) a three-factorial repeated measure ANOVA was calculated. Given the prediction that high control check results attenuate and low control check results magnify the effects of the power check, the unfolding of effects of the goal conduciveness check and the power check effects were followed-up for both degrees of control.

At 400 ms after feedback-stimulus onset no predicted effect of the goal conduciveness check was found (F-values < 2.54, p-values > .123), contrasting with prediction (2). The first significant effect of appraisal checks was found at 1300 ms in the form of an interaction effect between Goal Conduciveness × Power × Control , F(1, 27) = 4.82, p = .037, ŋ² = .15.

At 1,300 ms, post hoc tests followed-up the three-fold interaction effect. High and low degrees of control differentially affected the cumulative effects of the goal conduciveness check and the power check over the cheek region. As predicted, when control was high, no differentiated response patterning of the goal conduciveness check and the power check was found (F-values < 0.54, p-values > .469). In contrast, when control was low, the response patterning was differentiated by both the goal conduciveness check and the power check, F(1, 27) = 4.99, p = .034, ŋ² = .16. Follow-up tests showed that muscle activity over the cheek region was increased following “low power losses” compared with “low power wins”, t(27) = 2.18, p = .038. Whereas, muscle activity was only marginally increased following

“high power wins” compared with “high power losses”, t(27) = 1.87, p = .073. This cumulative response patterning indicates that muscle activity over the cheek region was similarly increased following very negative feedback-stimuli (“low power losses”, i.e., nothing can be done about the loss) and very positive feedback-stimuli (i.e., “high power wins”, i.e., guaranteed win). This interpretation is in line with the proposition that the cheek

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 136 region shows a curvilinear response patterning of valence (cf. Lang, Bradley, & Cuthbert, 1998; Lang et al., 1993; Ravaja et al., 2008).

At 1,400 ms, similar to the antecedent time interval, a significant interaction effect between Goal Conduciveness × Power × Control was found, F(1, 27) = 4.78, p = .038, ŋ² = .15, and additionally a significant main effect of Power was obtained, F(1, 27) = 6.38, p = .018, ŋ² = .19. Post hoc tests on the three-fold interaction effect yielded a similar response patterning as at 1,300 ms: When control was high, no differentiated response pattern was revealed (F-values < 1.05, p-values > .314), whereas, when control was low, both the goal conduciveness check and the power check affected the muscle activity over the cheek region, F(1, 27) = 4.42, p = .045, ŋ² = .14. Specifically, when control was low, muscle activity over the cheek region increased in response to “low power losses” compared with “low power wins”, t(27) = 2.25, p = .033; no other significant effects were found (t-values < 1.47, p-values > .154). Additionally, when control was low, over the cheek region a power check main effect was obtained, F(1, 27) = 7.30, p = .012, ŋ² = .21, indicating that muscle activity was increased following low power feedback in comparison to high power feedback. These results are also compatible with the proposition the cheek region shows a u-shaped quadratic pattern of valence—increased muscle activity in response to either very negative events (i.e.,

“low power losses”) or very positive events (i.e., “high power wins”). Moreover, these results support prediction (1) that appraisal check results of low power trigger increased muscle activity over the cheek region in comparison to appraisal check results of high power.

To sum up, for the first time cumulative effects involving appraisal checks of Coping have been found over the cheek region. These cumulated effects of appraisal check results are in line with the prediction of the component patterning that muscle activity over the cheek region is increased following low power compared with high power check results. As predicted, these cumulative effects were increased in response to low power feedback and

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 137 low control. Specifically, the theory predicts mouth stretch and lip corner retraction following low power and low control checks. Thus, these cumulated effects varied as a function of the appraised degree of control. In line with previous studies on the response specificity of the cheek region’s zygomaticus muscle, the present results indicate that participants evaluated

“low power losses” as very negative and “high power wins” as very positive feedback-stimuli in the gambling task.

3.3.4.3. Discussion

The aim of Experiment 2 was to further investigate the efferent effects of the goal conduciveness check and the power check when the control check is additionally

manipulated. In the gambling task, the control check was manipulated by implementing two blocks of different frequencies of high power feedback. Concurrently to the recorded brain activity, facial EMG was recorded over the frontalis, the corrugator, and the cheek region to examine facial expressions in response to the stimuli. For each trial, feedback-stimuli simultaneously presented information about the goal conduciveness check and the power check conveying at the same time whether participants will win or lose money (i.e., the goal conduciveness check), and whether participants can freely decide about the outcome of that trial (i.e., two choice options [high power] or single choice option [low power]).

Additionally, prior to each block of gambling, participants were informed about the upcoming degree of control over the gambling task. The results of the facial EMG showed

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 138

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400

High win High loss Low win Low loss (B) High control

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400

Time (ms)

Figure 7. Temporal profiles of EMG amplitude variations over the cheek region (Experiment 2). For (A) low control and (B) high control gambling blocks the unfolding of the goal conduciveness and the power check effects is illustrated. Provided are also the standard errors of the mean. *p < .05.

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 139 differentiated effects of the goal conduciveness check over the corrugator and the cheek region as a function of the two appraisal checks of Coping–the control check and the power check. Moreover, efferent effects of the power check unfolded dynamically over the cheek region, indicating that the additional manipulation of the control check indeed magnified these efferent effects.

In the concurrently recorded brain activity, results showed sequential effects of the centrally processed appraisal checks (see chapter 3.2). In contrast to the sequential effects in Experiment 1, in Experiment 2, the feedback-related negativity was sensitive to the control check and the power check. The result of the power check suggests that low power feedback was evaluated as goal obstructive, whereas high power feedback was evaluated as goal conducive. Similarly, the results of the control check showed highest negative deflections of the feedback-related negativity to the infrequent low power feedback, which is consistent with effects of changed frequency of the feedback-stimuli (cf. Holroyd et al., 2006). Contrary to our prediction, the feedback-related negativity was insensitive to the goal conduciveness check. On the subsequent component, the P300, the results confirmed those of Experiment 1.

On the P300, multiple main effects occurred of the goal conduciveness check, the monetary magnitude, and the power check. In addition, the P300 yielded a main effect of the control check, which is in line with previous findings of changed stimulus-frequency in a gambling task (cf. Hajcak et al., 2005), and an interaction between the goal conduciveness check and the power check, which is partly comparable to the effects in the facial EMG data over the corrugator region (“high power wins” had larger P300s than “low power wins”) and the cheek region (“high power losses” had larger P300s than “low power losses"). This

interaction indicates an integration of these appraisal check results. At first sight, these results in the feedback-related negativity seem incompatible with the sequence hypothesis of the Component Process Model, predicting that the processing of the goal conduciveness check

CHAPTER 3.3:APPRAISAL PATTERNS IN FACIAL EXPRESSIONS 140 precedes the processing of the control check. Further, these results on the feedback-related negativity contrast with previous findings that this component is solely sensitive for favorable reward feedback (e.g., Hajcak, Moser, Holroyd, & Simons, 2006; Holroyd et al., 2006;

Pfabigan et al., 2011). However, the fact that the results on the P300 were similar in both experiments (larger P300s following high power than following low power feedback) suggests that the changed findings on the feedback-related negativity in Experiment 2

Pfabigan et al., 2011). However, the fact that the results on the P300 were similar in both experiments (larger P300s following high power than following low power feedback) suggests that the changed findings on the feedback-related negativity in Experiment 2