d. The Present Experiment

The present study tested if high incentive can compensate the effort-mobilization deficit of individuals processing sadness primes during the performance of a difficult task. Evidence for this effect would support the idea that affect primes’ influence on cardiovascular response is moderated by motivational context variables, rather than being stable. Moreover, we sought for further evidence for the idea that affect-primes’ impact on effort-related cardiovascular response is emotion-category specific rather than valence specific. Therefore we investigated primes of two negative emotions—sadness and anger—which were expected to have different motivational effects, and promised participants either low or high monetary reward for success on a difficult short-term memory task.

Our predictions were based on the theorizing outlined in the IAPE model (Gendolla, 2012) and its integration with the principles of motivational intensity theory (Brehm & Self, 1989). The predicted combined effects of sadness vs. anger primes, low vs. high incentive, and varying levels of objective task difficulty are depicted in Figure 9.

Predictions for the prime x objective task difficulty interaction with relatively low incentive depicted in Panel A of Figure 9 were already tested and supported in the study by Freydefont et al. (2012). Consequently, we focused on the objectively difficult task condition to test if high incentive can compensate the effort mobilization deficit of people primed with sadness when objective difficulty is high. The relevant conditions are marked with the grey boxes in Figure 9.

97 Figure 9. Theoretical predictions for the combined effect of sadness (grey line) vs. anger (black line) affect primes and success incentive on effort mobilization on different objective task difficulty levels. Panel A shows predictions for low incentive; Panel B shows predictions for high incentive. The predicted effort levels for objectively highly difficult tasks, which are relevant for the present experiment, are marked by the grey boxes.

As explained above, our main effort-related dependent variable was PEP reactivity. For an objectively difficult task, we expected the weakest PEP response in the sadness-prime/low-incentive condition. Here the high effort that was necessary due to the high level of subjectively experienced demand was not justified by low incentive, which should lead to disengagement (Freydefont et al., 2012; Silvestrini & Gendolla, 2011c). By contrast, high incentive should boost effort in the sadness-prime condition by justifying the high necessary effort, leading to very strong PEP reactivity. Responses in the two anger-prime conditions were anticipated to fall in between the two sadness-prime cells, because anger is associated with ease, and thus reduces subjective demand (Freydefont et al., 2012; Gendolla & Silvestrini, 2011). Given that high effort was thus not necessary, justifying high effort should have no effect.

98 II.4.3. Method

II.4.3.a. Participants and Design

Sixty-nine healthy undergraduates (43 women, mean age 22 years) participated voluntarily and were randomly assigned to a 2 (Prime: anger vs. sadness) x 2 (Incentive:

low vs. high) between-persons design. Participants received 10 Swiss Francs (approximately USD 11) for their participation. Seven participants had incomplete cardiovascular data due to measurement problems. In order to provide the same statistical power for all inferential tests, those datasets were entirely excluded from the analysis, leaving a final sample of N = 62. The women/men distribution was balanced across the experimental conditions.

II.4.3.b. Affect Primes

We used low contrast, low resolution, grey scale, front perspective pictures of averaged human facial expressions from The Averaged Karolinska Directed Emotional Faces (AKDEF; Lundqvist & Litton, 1998) inventory as primes. Half of these averaged faces were female (AKDEF pictures FNES, FSAS, FANS) and half were male (AKDEF pictures MNES, MSAS, MANS).

II.4.3.c. Apparatus and Physiological Measures

PEP (in milliseconds [ms]) and HR (in beats per minute [bpm]) were continuously assessed with a Cardioscreen 1000 system (medis, Ilmenau, Germany; see Scherhag, Kaden, Kentschke, Sueselbeck, & Borggrefe, 2005, for a validation study) that sampled electrocardiogram (ECG) and thoracic impedance (impedance cardiogram, ICG) signals.

Four pairs of disposable spot electrodes (2 x 16-mm Ag/AgCl, Red Dot, 3M) were placed on the right and left side of participants’ neck and on the right and left side of the axillary line at the height of the xiphoid. The signals were assessed with a sampling rate of 1000 Hz and analyzed with BlueBox 2.V1.21 software (Richter, 2010a). The first derivative of the change in thoracic impedance was calculated and the resulting dZ/dt-signal was ensemble averaged over 1 min periods on the basis of the detected R-peaks (Kelsey &

Guethlein, 1990). To obtain cardiac PEP values—the time interval between R-onset and

99 B-point (Berntson et al., 2004)—B-point location was estimated based on the RZ interval as proposed by Lozano et al. (2007) and manually corrected as recommended by Sherwood et al. (1990). Additionally, we assessed systolic (SBP) and diastolic (DBP) blood pressure (in millimeters of mercury [mm Hg]) with a Vasotrac^® APM205A monitor (MEDWAVE^®, St. Paul, MN). This system uses applanation tonometry by a pressure sensor placed on the top of the radial artery on the wrist. Systolic and diastolic pressure values were recorded each 10-15 heart beats, i.e., 4-5 measures per minute (see Belani et al., 1999 for a validation study). Values were offline averaged over 1 min periods.

II.4.3.d. Procedure

The study was conducted in accordance with the ethical guidelines of the University of Geneva; the procedure was approved by the local ethical committee. Data were collected in individual sessions. After having obtained signed consent, participants were seated in a comfortable chair in front of a personal computer. The experimenter attached the electrodes and the blood pressure cuff and went to an adjacent control room.

The whole procedure was computerized (E-Prime, Psychology Software Tools, Pittsburgh, PA). After some biographical questions (age, gender, study major etc.) participants indicated their momentary affective state with 2 items of the positive (happy, joyful) and 2 items of the negative (sad, depressed) affect scales of the UWIST mood checklist (Matthews, Jones, & Chamberlain, 1990). Two adjectives assessing anger (angry, irritated), also used in previous research (Freydefont et al., 2012), were added.

Participants responded to the 6 items (“Right now, I’m feeling…) on 7-point scales (1 - not at all, 7 - very much) to assess their affective state before exposure to the affect primes.

Next, cardiovascular baseline values were recorded during a habituation period (8 min) in which participants watched a hedonically neutral documentary film showing landscape. Then participants received instructions (“Please respond as quickly and accurately as possible”) for a short-term memory task adapted from Sternberg (1966). In study by Freydefont et al. (2012), the here-administered version of the task had been evaluated as being high in objective task difficulty. Task trials started with a fixation cross (1000 ms), followed by a picture of a facial expression of the AKDEF database (26 ms; i.e., 2 frames on a 75 Hz monitor). In each condition 1/3 of the faces showed an

100 emotional expression (anger vs. sadness) while the other 2/3 showed emotionally neutral faces.² This procedure has been found to be the most effective in the present priming paradigm (Silvestrini & Gendolla, 2011a). The facial expression pictures were immediately backward masked with a random dot pattern (130 ms), which was followed by a string of 7 letters, presented for 750 ms and backward masked by a row of the letter

"X" (2 sec.). A target letter appeared in the center of the screen and participants indicated by pressing a "yes" or a "no" key if that letter had been included in the previously presented string. The letter remained on the screen until participants gave a response (maximal response time: 2 sec.). After participants had responded, the message “response entered” appeared. In case of no response within 3 sec., the message "please answer more quickly" was presented. To keep performance time and number of presented primes constant for all participants, the respective message appeared for 4 sec. minus participants’ reaction time. The inter-trial interval randomly varied between 2 and 5 sec.

In order to assure that participants were presented with emotional expression from task onset on, randomization was effectuated in portions of 6 trials. The task comprised 32 trials and started after a training session of 10 trials with correctness feedback and only neutral facial expressions as primes.

Following the training session, participants were informed that they could win additionally 1 Swiss Franc (i.e., USD 1.10) in the low-incentive condition vs. 15 Swiss Francs (i.e., USD 17) in the high-incentive condition if they met the success criterion of at least 90 % correct responses (“If you succeed in at least 90 %, you will win….”). That way low vs. high incentive was contingent upon success. The program gave no correctness feedback during the task to prevent disengagement after some wrong responses and to avoid affective reactions (Kreibig, Gendolla, & Scherer, 2010) that could interfere with the affective primes’ impact. No participant attained the success criterion, which is not surprising, because we had administered a difficult version of the task. In the Freydefont et al. (2012) study we had administered an easy version with strings of 4 letters and the present difficult version with stings of 7 letters for 750 ms. The difficulty manipulation was based on the idea that processing a string of 7 letters needs more working memory capacity than processing a 4-letter string. In the literature, the capacity in short term memory is to 4/5 elements (Cowan, 2000; Miller 1950). Thus, processing more than 5 elements is considered as difficult.

101 After the task, participants retrospectively rated subjective demand (“Was it difficult for you to succeed on the task?”), and subjective incentive value (“How valuable was the reward for you?”) on 7-point scales (1 – very low, 7 – very high). Then, participants rated the same affect items that had been assessed at the beginning of the experiment to assess possible affect prime effects on conscious feelings. Subsequently, the experimenter interviewed the participants in a standardized funnel debriefing procedure about the study’s purpose and what they had seen during the trials. Participants who mentioned “flashes” were asked about their content. Finally, participants were debriefed, thanked, and received their payment.

II.4.3.e. Data Analyses

We tested our theory-based predictions about effort-related cardiac response with contrast analysis, which is the most powerful and thus most appropriate statistical tool to test complex interactions (Rosenthal & Rosnow, 1985; Wilkinson and the Task Force for Statistical Inference, 1999). Our hypothesis that high performance-contingent incentive can compensate the effort-mobilization deficit of participants exposed to sadness primes during a difficult task predicts the following cardiovascular response pattern: Low response in both anger-prime conditions, because subjective demand during performance should be low and high reward should thus not increase effort by justifying high engagement (contrast weights = -1). Even lower response in the sadness-prime/low-incentive condition, because here subjective demand during performance was anticipated to be very high, but not justified due to the low incentive, leading to disengagement (contrast weight = -2). Finally, and most important, very high effort in the sadness-prime/high-incentive condition, because high incentive justifies the very high effort that is necessary due to the very high subjective demand during performance (contrast weight = +4). Other variables, for which we had no theory-based a priori predictions, were analyzed with conventional exploratory ANOVAs. Unless indicated otherwise, the alpha-error level for all tests was 5%. For reasons of consistency and easier comparability of effect sizes, we transformed effect size coefficients r of cell comparisons to coefficients η².

102 II.4.4. Results

II.4.4.a. Cardiovascular Baselines

Baseline values of PEP, SBP, DBP, and HR, which appear in Table 7, were calculated by averaging values assessed during the last 3 min of the habituation period (Cronbach’s αs ≥ .98).¹¹. According to 2 (Prime) x 2 (Incentive) between-persons ANOVAs, there were no significant a priori differences between the experimental conditions in the baseline values of PEP (ps > .776) and HR (ps > .059), but significant differences between the low and high incentive conditions in SBP and DBP baselines.

Both were higher in the low-incentive than in the high-incentive condition (SBP: M = 129.98, SE = 3.06 vs. M = 118.54, SE = 2.95; DBP: M = 72.92, SE = 2.18 vs. M = 65.67, SE = 2.15), Fs(1, 58) > 5.61, ps < .021, η² > .09, (other ps > .19). Below we will deal with these findings in ANCOVAs. Preliminary 2 (Prime) x 2 (Incentive) x 2 (Gender) ANOVAs found no gender effects on baseline values of PEP (p > .089) and HR (p >

.107). But men had higher SBP (M = 131.47, SE = 2.23 vs. M = 119.99, SE = 2.67) and DBP (M = 74.29, SE = 2.40 vs. M = 66.35, SE = 1.96) baselines than women, Fs(1, 54) >

7.83, ps < .007, η² > .12, which is a common finding (Wolf et al., 1997).

11 We calculated the cardiovascular baseline values from the three last minutes of the habituation period, because values were stable and did not differ significantly from one another (ps > .38).

103 Table 7. Means and Standard Errors (in Parentheses) of the Cardiovascular Baseline Values.

Sadness Primes Anger Primes

Low incentive High incentive Low incentive High incentive

PEP 105.85

Note: PEP = pre-ejection period (in ms), SBP = systolic blood pressure (in mmHg), DBP

= diastolic blood pressure (in mmHg), HR= heart rate (in beats/min).

II.4.4.b. Cardiovascular Reactivity

The 1-min scores of cardiovascular activity assessed during task performance showed high internal consistency (αs ≥ .92) and were averaged to create task scores.

Reactivity scores were then determined by subtracting baseline scores from those task scores. Preliminary contrast analyses of the reactivity scores with their respective baseline scores as covariate found no significant associations between baseline values and reactivity scores for any cardiovascular index (ps > .126). Thus, all reactivity scores were analyzed without baseline correction. Furthermore, there were no significant Gender effects on reactivity scores in preliminary 2 (Prime) x 2 (Incentive) x 2 (Gender)

104 ANOVAs (ps > .058) and, most relevant, considering Gender as covariate in the contrast analyses of cardiovascular reactivity did not moderate any of the effects reported below.

PEP Reactivity

Our theory-based a priori contrast was significant, F(1, 58) = 6.19, p = .016, η² = 10, while the test of the residual variance was not (F < 1), reflecting that the a priori contrast explained all significant variance. As depicted in Figure 10, the PEP responses showed the anticipated pattern.

Figure 10: Cell means and standard errors of cardiac pre-ejection period reactivity (in ms) during task performance.

The weakest reactivity occurred in the sadnessprime/lowincentive cell (M = -2.68, SE = 0.84), where effort intensity was anticipated to be low because the subjectively high necessary effort was not justified by the low incentive. Most relevant, a focused cell contrast found that PEP response in the sadnessprime/highincentive condition (M = -7.02, SE = 1.34) was, as expected, significantly stronger than the modest response in the

105 sadness-prime/low-incentive condition, t(58) = 2.80, p = .007, η² = .12.¹² This was expected because the high incentive justified the subjectively necessary high effort in this cell, resulting in the expected compensation of the effort mobilization deficit of participants in the sadness-prime/low-incentive cell. Moreover, PEP responses in the anger-prime/low-incentive (M = -4.42, SE = 1.14) and anger-prime/high incentive cells (M = -5.51, SE = 1.01) did not differ from one another, t(58) = 0.66, p = .515, η² < .01, and fell in between the above-discussed sadness-prime cells. This pattern was expected because the anger primes should lead to subjectively low necessary effort even if higher effort was justified in the high-incentive condition. Comparing the low reactivity in the sadness-prime/low-incentive cell against the higher reactivity in the two anger-prime cells revealed the expected difference, t(58) = 1.65, p = .051, η² = .04. Finally, PEP reactivity in the sadness-prime/high-incentive condition was also stronger than in the anger-prime/low-incentive condition, t(58) = 1.65, p = .052, η² = .04, while the difference to the anger-prime/high-incentive condition fell short of significance (p = .176). In summary, the PEP response pattern supports our hypothesis about a joint effect of affect primes and incentive on effort-related cardiac response during the performance of an objectively difficult task and shows that high incentive can compensate the effort-mobilization deficit of individuals exposed to sadness primes during a difficult task.

HR Reactivity

The a priori contrast was significant, F(1, 58) = 11.25, p = .001, η² = .16, and explained all significant variance (residual F < 1). As visible in Figure 11, the pattern of cell means largely corresponds to that of PEP reactivity and our effort-related predictions.

A focused cell contrast found that the high HR response in the sadness-prime/high-incentive condition (M = 9.15, SE = 1.27) was significantly stronger than the modest response in the sadness-prime/low-incentive condition (M = 4.48, SE = 1.42), t(58) = 2.64, p = .011, η² = .11. This reflects again the expected compensation of the effort mobilization deficit of participants in the sadness-prime condition by high incentive.

Moreover, HR responses in the anger-prime/low-incentive (M = 3.99, SE = 1.07) and

12 Given our directed a priori predictions about PEP reactivity, we applied one-tailed cell contrast tests.

106 anger-prime/high-incentive cells (M = 3.97, SE = 1.37) were also modest and did not differ from one another (p = .988), and the high HR reactivity in the sadness-prime/high-incentive condition was also significantly stronger than in both anger-prime conditions, ts(58) > 2.82, ps < .007, η² = .12.

Figure 11. Cell means and standard errors of heart rate reactivity (in beats/min) during task performance.

SBP and DBP Reactivity

Cell means and standard errors appear in Table 8. Neither the a priori contrasts for SBP and DBP reactivity, Fs(1, 58) < 0.89, ps > .35, nor tests of the residuals (Fs < 1.34), revealed any significant effects.¹³

13 Silvestrini and Gendolla (2011a) had found significant effects of affect primes and task difficulty on SBP that were partly determined by effects on total peripheral resistance (TPR) in the vasculature. Given that the present study did not reveal any SBP and DBP effects, TPR effects were most unlikely. Indeed, calculation of TPR (in dynes second per centimeter to the 5^th power) based on measures of cardiac output (Cardioscreen system) and mean arterial pressure (Vasotrac monitor) revealed neither TPR baseline differences (ps > .16; average M = 1232.70, SE = 42.03), nor significant contrast or residual tests of TPR reactivity (ps > .29; average M = 36.17, SE = 34.73).

107 Table 8. Means and Standard Errors (in Parentheses) of Blood Pressure Reactivity During Task Performance.

Note: SBP = systolic blood pressure (in mmHg); DBP = diastolic blood pressure (in mmHg).

II.4.4.c. Task Performance

Neither a 2 (Prime) x 2 (Incentive) ANOVA of the percentage of correct responses in the Sternberg task (all ps > .14; average M = 52.23%, SE = 1.17), nor an ANOVA of the reaction times for correct responses (ps > .26; average M = 1103 ms, SE = 34.98) found any significant effects. However, the high error rates indicate a successful manipulation of an objectively difficult version of the short-term memory task.

Sadness Primes Anger Primes

Easy Difficult Easy Difficult

SBP 2.00

(3.01)

7.68 (1.40)

9.70 (2.66)

7.13 (1.83)

DBP 3.13

(1.89)

6.01 (1.08)

6.86 (1.99)

6.44 (1.33)

108 II.4.4.d. Affect Ratings

We calculated average mood scores for the positive and negative affect subscales of the UWIST scale assessed at the beginning of the session and after the task (α range .66 to .87). Given that the correlation between the ratings of the two anger items was modest (α = .44), we analyzed them separately¹⁴. A 2 (Prime) x 2 (Incentive) x 2 (Time) mixed model ANOVA of the negative affect scores revealed no significant effects (ps >

.08). The ANOVA of the positive affect scores found a time main effect, F(1, 58) = 4.04, p = .049, η² = .07, due to higher scores before the task (M = 2.30, SE = 0.07 vs. M = 2.17, SE = 0.08), which was qualified by a time x incentive interaction, F(1, 58) = 5.87, p = .019, η² = .09. In the low-incentive condition, positive affect decreased from before (M = 2.40, SE = 0.08) to after the task (M = 2.12, SE = 0.12), t(30) = 3.63, p < .001, η² = .30, while it did not in the high-incentive condition (before task: M = 2.19, SE = 0.11; after task: M = 2.22, SE = 0.10; p = .802). Moreover, the prime main effect was nearly significant, F(1, 58) = 3.82, p = .054, η² = .06, reflecting slightly higher positive mood scores in the anger-prime condition (M = 2.37, SE = 0.08) than in the sadness-prime condition (M = 2.11, SE = 0.10). However, neither the prime x time interaction nor any other effect approached significance (ps > .23). A 2 x 2 x 2 ANOVA of the item “angry”

revealed only a significant time main effect, F(1, 58) = 8.29, p = .006, η² = .13. Ratings were slightly lower before (M = 1.42, SE = 0.11) than after the task (M = 1.73, SE = 0.15). The only other effect that approached significance (p = .069) was a time x incentive interaction (other ps > .17). No significant effects emerged in the analysis of the item

“irritated” (ps > .09, average M = 1.57, SE = 0.09). Thus, in summary, there was no evidence for a moderation of conscious affect from before to after the task due to the affect primes.

14 For the sake of comparability with our previous studies, we also calculated global mood scores by summing the ratings of the positive and reverse-coded negative affect items assessed at the beginning of the experimental session and after the task (αs > .69). Correspondingly, we summed the scores of the two anger items for both times of measure (αs > .44). To obtain comparable scores, we divided the sum scores by the number of items of the general mood (4 items) and anger (2 items) scales. A 2 (Prime) x 2 (Incentive) x 2 (Time) mixed model ANOVA of the global mood scores revealed only a significant Time x Incentive interaction, F(1, 58) = 4.71, p = .034, η² = .08, (other ps ≥ .097). In the low-incentive condition, mood decreased from before the task (M = 5.40, SE = 0.14) to after the task (M = 5.16, SE = 0.17), t(30) = 2.27, p = .031, η² = 0.14. The mood scores did not change in the high-incentive condition (before task: M = 5.13, SE = 0.22; after task: M = 5.30, SE = 0.17; p = .270). A 2 x 2 x 2 ANOVA of the anger scores revealed no significant effects (ps > .073; average M = 1.54, SE = 0.08).

109 II.4.4.e. Task Ratings

The subjective difficulty ratings in the sadness-prime condition (M = 4.48, SE = 0.28) tended to be higher than in the anger-prime condition (M = 4.14, SE = 0.25). But in a 2 (Prime) x 2 (Incentive) ANOVA neither the Prime main effect (p = .373), nor the other effects (ps > .439) approached significance. A 2 x 2 ANOVA of participants’

subjective incentive value ratings found a significant Incentive main effect, F(1, 58) = 5.76, p = .020, η² = .09, reflecting higher subjective incentive value in the high (M = 3.84, SE = 0.30) than in the low-incentive condition (M = 2.81, SE = 0.30). Interestingly, also the Prime main effect was significant, F(1, 58) = 6.65, p = .012, η² = .10, due to higher

subjective incentive value ratings found a significant Incentive main effect, F(1, 58) = 5.76, p = .020, η² = .09, reflecting higher subjective incentive value in the high (M = 3.84, SE = 0.30) than in the low-incentive condition (M = 2.81, SE = 0.30). Interestingly, also the Prime main effect was significant, F(1, 58) = 6.65, p = .012, η² = .10, due to higher