• Aucun résultat trouvé

1 Chapter

1.6 Supplementary Information

3.4.1 Behavioral Responses

As first step we analyzed the appropriateness ratings from both experiments (Figure 25). For each experiment, for each subject, and for each dilemma, appropriateness ratings were fitted with a linear mixed model with Unpleasantness (high vs. low), Modality (pain vs. disgust) and subjects’ Gender (male vs. females) as fixed factors, and subjects’ identity as random factor. Furthermore, the experiment was designed to include also the Dilemma Type (moral vs. non-moral) as additional fixed factor; however, data from our pilot study suggest a considerable variability within our dilemma database which goes beyond a dichotomic categorization (see SI Figure 32). Therefore, to fully account for the idiosyncratic properties of each stimulus, we replaced the categorical factor Dilemma Type, with the median appropriateness values from the pilot study (covariate “appropriateness” – See SI Figure 32), which should account for subtle inter-dilemma differences in moral content. The p-value of the parameter estimated in the linear mixed model was calculated with the Satterthwaite approximation, as implemented in the lmerTest package555 of R 3.3.1 open-source software (https://cran.r-project.org/).

We found that the appropriateness ratings from both our experiments matched those found in the pilot study (Experiment 3: t(668) = 13.04, p < 0.001; Experiment 4: t(498) = 10.302, p < 0.001), thus providing a reliable sanity check for our stimuli database. Most importantly, participants’

appropriateness ratings were modulated by the expectancy condition (see Figure 25). Indeed, for Experiment 3, we found a significant main effect of Unpleasantness (Experiment 3: t(668) = -2.09, p = 0.04; Experiment 4: t(497) = 1.22, not significant [n.s.]) and a significant effect of Modality (Experiment 3: t(668) = -2.56 , p = 0.01; Experiment 4: t(497) = 0.34, [n.s.]), whereas for Experiment 4 we found a significant Unpleasantness*Modality interaction (Experiment 3: t(668) = 0.21, [n.s.]; Experiment 4: t(497)

= -2.07, p = 0.032). No other main effect or interaction was found to be significant (│t│s ≤ 1.52, [n.s.]).

To further explore the role of each expectancy condition on appropriateness evaluations, we run three targeted planned contrast (see Figure 25). First, in line with previous studies linking physical disgust to moral judgment479,478,556, we tested whether expectancy of high disgust lead to harsher judgments

129

than its tailored control (HD vs. LD); this was found to be the case for both experiments (Experiment 3: t(668) = -1.79, p (1-tailed) = 0.04; Experiment 4: t(497) = -1.70, p (1-tailed) = 0.04). Second, we tested whether expectancy of high disgust let to harsher judgments than the expectancy of equally-unpleasant painful stimulations (HD vs. HP); this was the case for both experiments (Experiment 3:

t(668) = -2.25, p = 0.02; Experiment 4: t(497) = -2.58, p = 0.01), thus highlighting the prominent role of disgust in moral cognition. Finally, we tested whether expectancy of high pain, let to harsher judgments than its tailored control (HP vs. LP); this was the case for Experiment 3 (t(668) = -2.03, p = 0.04) but not for Experiment 4 (t(497) = 1.22, [n.s.]). Overall, although the two experiments do not provide identical results (especially for the LP condition), they converge in arguing that expectancy of high disgust leads to harsher judgments than the expectancy of both low disgust and high pain, thus suggesting that link between physical disgust and moral cognition found in previous studies cannot be explained exclusively in terms of unpleasantness.

Figure 25 - Mean appropriateness ratings for dilemmas as a function of the predictive cues in Experiment 3 (left side) and in Experiment 4 (right side).

Data averaged across 22 subjects for Experiment 3, and across 25 subjects for Experiment 4. The more values are negative, the more inappropriate the dilemma judged. The different expectancy conditions preceding the dilemmas are labelled as follows: LP – expectancy for low pain, HP – expectancy for high pain, LD – expectancy for low disgust, and HD – expectancy for high disgust. Error bars refer to standard errors of the mean. * refer to conditions eliciting differential appropriateness at p < 0.05 and ^ at p (1-tailed) < 0.05 (see results).

Finally, similar to Experiment 1, we also tested for any influence of individual scores, using measures from questionnaires on personality traits or other ad hoc items in the post-experiment debrief questions. For both Experiments 3 and 4, these data were submitted to exploratory ANCOVAs. In none of these analyses, the individual covariates were associated with significant effects, neither under a rigorous Bonferroni-correction for the number of ANCOVAs employed, nor when adopting a more liberal uncorrected approach.

130 3.4.1.2 Pain & Disgust evaluations

We then focused on those trials in which participants rated the unpleasantness of a stimulus event following a moral judgment (Figure 26). As for the case of moral evaluations, unpleasantness ratings were fitted with a linear mixed model with Unpleasantness (high vs. low), Modality (pain vs. disgust) and subjects’ Gender (male vs. females) as fixed factors, the appropriateness of the previous dilemma as a covariate, and subjects’ identity as random factor.

The analysis revealed in both experiments a main effect of Unpleasantness (Experiment 3: t(668) = -12.63, p < 0.001; Experiment 4: t(495) = -12.24, p < 0.001). Furthermore, in neither experiment the effect of Unpleasantness changed significantly across Modality (Unpleasantness*Modality interaction – Experiment 3: t(668) = -0.68, [n.s].; Experiment 4: t(497) = 0.85, [n.s.]), thus suggesting that, also in the trials following dilemmas, pain and disgust were well matched, at least when each was contrasted with its tailored control (Figure 26). However, it should be stressed that in Experiment 3 a Modality main effect was found (Experiment 3: t(668) = 3.45, p < 0.001 ; Experiment 4: t(495) = 0.24, [n.s.]), suggesting that thermal stimuli (both HP and LP) were rated as more unpleasant than olfactory events (Figure 26).

Furthermore, following previous studies who found that exposure to immoral behavior led to harsher evaluation of disgusting stimuli480 and raised the need to physical cleansing540, we then tested if unpleasantness ratings were influenced by the previous’ dilemma appropriateness. In neither experiment, we found a significant effect (or interaction) associated with this factor (│t│s≤ 1.44, [n.s.]).

Finally, we found that participants’ gender affected unpleasantness ratings (especially in Experiment 3) in the shape of a Unpleasantness*Gender (Experiment 3: t(668) = -3.20, p = 0.001 ; Experiment 4:

t(495) = -1.65, p (1-tailed) = 0.05) and in a Unpleasantness*Modality*Gender interactions (Experiment 3: t(668) = 2.36, p = 0.019 ; Experiment 4: t(495) = -0.23, [n.s.]), reflecting female participants associated with harsher ratings for high unpleasant events, especially if painful. No other main effect or interaction was found to be significant (│t│s ≤ 1.48, [n.s.]).

131

Figure 26 - Mean unpleasantness ratings for thermal pain and olfactory disgust stimuli following dilemmas, as a function of the predictive cues in Experiment 3 (left side) and in Experiment 4 (right side).

Data averaged across 22 subjects for Experiment 3, and across 25 subjects for Experiment 4. The more values are negative, the more unpleasant the experience. The different expectancy conditions preceding the different stimuli are labelled as follows: LP – expectancy for low pain, HP – expectancy for high pain, LD – expectancy for low disgust, and HD – expectancy for high disgust. Error bars refer to standard errors of the mean. *** refer to conditions eliciting differential unpleasantness at p < 0.001.