• Aucun résultat trouvé

3. EXPERIMENTAL PART

3.4. Experiments 3-4 - Choice both affects and reflects preferences

3.4.2. Experiment 3

3.4.2.1. Method Participants

One hundred female students (mean age = 24.03 ± 5.66 years) from the University of Geneva took part in this experiment. Before starting the experiment, each participant completed a consent form.

Pictures

In order to ensure interest in the task from our participants, we used infant pictures, which have been shown to be of particular relevance for adult humans (see, e.g., Brosch, Sander, Pourtois, & Scherer, 2008). Forty full-faced color pictures of infant faces were used that were selected from a previous study that examined attractiveness of infant faces (see Van Duuren, Kendel-Scott, & Stark, 2003, for details). The pictures were framed so that only the face of the baby could be seen. The presentation order of the faces was randomized for each participant.

Procedure

The experiment was divided into four parts: a first rating session, a first choice session, a second rating session, and a second choice session. For half of the stimuli, participants had to make choices between the two rating sessions and for the other half, participants had to make choices after the two rating sessions. As participants were asked to rate and choose between infant faces, we asked them to rate the “cuteness” rather than the “pleasantness” of the

pictures6. Participants were then first asked to rate the cuteness of 40 infants presented on a screen on a scale from 1 (not cute at all; left on the scale) to 10 (very cute; right on the scale = 10). Participants had to move a vertical marker with the mouse across a horizontal line and to click to indicate their rating. On the basis of these first ratings, a computer program created 20 pairs of pictures such that the difference of cuteness within each pair was minimized. Half of these pairs (10 pairs) were then presented for choice only in the first choice session, right after the first rating. In the choice session, participants were required, for each pair, to indicate the baby they considered cuter by clicking on the baby’s picture. Note that there was no consequence of the choice for the participants. Such a procedure has been shown to be efficient at producing postchoice rating modulation (e.g., Coppin et al., 2010; Sharot et al., 2009; Sharot, Shiner, & Dolan, 2010; Shultz, Léveillé, & Lepper, 1999). Despite the absence of real-life consequences from the choice, it still involves a reevaluation of the options at stake. The spreading of alternatives observed in such a nonconsequential choice condition can consequently be an even stronger argument for the importance of choice in modulating ratings in the case of a consequential choice. Next, participants were asked to rate the cuteness of all 40 babies again (second rating). Finally, participants were presented with the 10 remaining pairs in the second choice session (same task as in the first choice session). The order in which the different pairs were presented during the experiment for a given participant, both inside one choice session and between the two choice sessions, was random.

Data analyses

In the free-choice paradigm, choice-induced changes are typically reported when the choice is difficult (Brehm, 1956), that is, when the difference between the pleasantness ratings obtained for the two paired stimuli before the choice is small. In contrast, no modulation is assumed to occur in easy pairs (Festinger, 1957). To test for this, we split the 20 pairs of comparisons available in this experiment into two groups—the 10 easiest and the 10 most difficult pairs—on an individual participant basis. The analyses were conducted on each type separately.

Moreover, in a nonstandard version of the free-choice paradigm, recent results have raised the possibility that differential psychological processes underlie postchoice hedonic

6 Such a procedure allows the extension from traditional results found while measuring “pleasantness”

or “desirability” to the dimension of “cuteness,” which is equivalent to “attractiveness” in the context of infants’ evaluation (e.g., Hildebrandt, 1983).

modulation for chosen and rejected stimuli. Sharot, Velasquez, et al.’s (2010) blind-choice study showed divergent results according to choice: Although a significant modulation was observed for chosen stimuli, none occurred for rejected stimuli. In contrast, Izuma et al.

(2010) showed a true preference modulation induced by choice (in comparing the RCR and the RRC conditions) for rejected stimuli but not for chosen stimuli. To control for this, we performed repeated measures analyses of variance (ANOVAs) for chosen and rejected pictures separately.

Lastly, we did not want to include those participants whose ratings were inconsistent and could have biased the results. Thus, we excluded 12 participants (12%) from the analyses because the signed difference in their cuteness scores between Rating 2 and Rating 1 exceeded three standard deviations above or below the mean calculated across participants.

3.4.2.2. Results Chosen stimuli

First, we checked whether an increased preference was observed between the two rating sessions in the classical RCR sequence of measurement (Brehm, 1956), as well as in the control RRC sequence. To do so, we performed repeated measures ANOVAs on the cuteness pleasantness scores with the Rating (Rating 1, Rating 2) and Sequence of Measurement (RCR, RRC) as within-subject factors. For difficult pairs, cuteness ratings were significantly increased in the second rating session in comparison to the first one, both in the RCR and RRC conditions, Fs(1,87) = 110.05, 46.37, ps < .001, η2s= .56, .35 (see Figure 5).

For easy pairs, cuteness ratings for chosen stimuli were significantly increased in the second rating session for the RCR condition, F(1,87) = 3.99, p= .049, η2 = .04, but not in the RRC condition, F(1,87) = 2.67, p= .106.

Second, we checked whether the preference modulations observed between the two rating sessions varied as a function of the sequence of measurement. To do so, we conducted a repeated measures ANOVA with the within-subject factor Sequence of Measurement (RCR, RRC) on the signed difference between the cuteness score during Rating 2 minus the cuteness score during Rating 1. This signed difference was significantly higher in the RCR condition than in RRC condition, F(1,87) = 8.26, p = .005, η2 = .09 (see Figure 5). In easy choices, there was not such a difference, F(1,87) = 0.03, p = .874.

Figure 5. Cuteness signed difference scores between Rating 1 and Rating 2 from Experiment 3 according to the sequence of measurements, Choice-Rating2 (RCR) and Rating1-Rating2-Choice (RRC), and the difficulty (difficult, easy) of the choice. Error bars represent the standard error of the mean.

Rejected stimuli

For difficult choices, cuteness ratings were not significantly different between the two ratings in both RCR and RRC conditions, Fs(1,87) = 0.33, 0.87, ps = .567, .873. In contrast, for easy choices in both RCR and RRC conditions, cuteness ratings were increased in the second rating session, Fs(1,87) = 15.38, 7.74, ps < .001, < .007, η2s = .15, .08.

Was this modulation dependent on the sequence of measurement? For rejected pictures in difficult choices, the signed difference between the two cuteness scores was not significantly higher in the RCR condition in comparison to the RRC condition, F(1,87) = 0.03, p = .857. The same was true for rejected pictures in easy choices, F(1,87) = 1.15, p = .287.

3.4.2.3. Discussion

Our results showed that there is a significant rating modulation in the RRC condition for chosen items in difficult pairs. In this condition, choice cannot drive preference

modulation, because it occurs after the two rating sessions. Such results confirm the validity of Chen and Risen’s (2010) point that a rating modulation can be observed in the free-choice paradigm without it being driven by choice.

If there is no true preference modulation by choice, then one would not expect to observe any significant difference in the magnitude of rating modulation in the RCR and the RRC sequences. However, in difficult choices, the rating modulation was significantly more pronounced for chosen pictures in the RCR than in the RRC condition. The observation that, for chosen pictures in difficult choices, rating modulation was more pronounced in the RCR than in the RRC sequence suggests that there was truly a choice-induced preference modulation, rather than the rating modulation being a mere artifact of the free-choice paradigm. These results extend those found by Sharot, Velasquez, et al. (2010) of an increased preference for “chosen” stimuli in a blind-choice paradigm (stimuli that participants had the impression of choosing but actually did not).

However, this interpretation is limited by two observations. First, our results do not enable us to rule out that a mere exposure effect (described in detail below) may be at work.

This result deserves attention because preference modulation, according to cognitive dissonance theory, is thought to occur only for difficult choices, whereas no modulation should occur when choice is easy (see Festinger, 1957). However, in our RCR condition, preference modulation for chosen stimuli was observed not only for difficult choices, but also for easy choices. A mere exposure effect, constituted by a preference increase for a given stimulus following repeated exposure (Zajonc, 1968), could account for this. Note that this mechanism is not restricted to either difficult or easy pairs, but may apply to both alike.

During the second rating in RCR, participants had seen the stimuli one time more than they had in the second rating of RRC. Hence, in summary, the mere exposure hypothesis would predict increased ratings in RCR relative to RRC in both easy and difficult choices, which is what we observed. Therefore our data are consistent with both mere exposure and preference modulation induced by choice effects operating simultaneously for chosen stimuli in difficult pairs. Only if modulation was observed for difficult choices alone could we have clearly identified that preference modulation induced by choice was driving the effect.

Second, the results for rejected stimuli give further reason to consider the impact of the mere exposure effect. As in Sharot, Velasquez, et al.’s (2010) study, we did not find a

preference decrease between ratings 1 and 2 for rejected stimuli in difficult choices, either in the RCR or in the RRC condition. Moreover, in both RCR and RRC conditions, there was an increase in cuteness ratings for rejected stimuli in easy choices. This may be because the stimuli in this experiment were baby pictures. On a speculatory note, it might be evolutionarily counterproductive to devaluate baby cuteness, as cuteness elicits caretaking in adults (Glocker et al., 2009; Parsons et al., 2011) and could consequently be an important factor in babies’ survival. Conceivably, a mere exposure effect may explain the observation that, for rejected pictures, there was no significant decrease in difficult choices but there was a preference increase in easy choices.

To rule out this alternative and to counteract the putative specific rating bias when babies’ cuteness is evaluated, we conducted a second experiment using stimuli for which previous research has demonstrated devaluations after rejections. Following Coppin et al.

(2010), who obtained this result in a classical free-choice paradigm, we used olfactory stimuli.

If there is a true preference modulation by choice, the experimental procedure used in Experiment 1 with baby pictures, when applied to Experiment 2 with olfactory stimuli, should lead to a rating modulation for rejected smells in difficult choices, that is, a decreased pleasantness from the first to the second rating session. This decrease should be more pronounced in the RCR than in the RRC condition. Alternatively, if the difference between the RCR and RRC conditions found in Experiment 1 for chosen stimuli in difficult choices is due to a mere exposure effect only, an increase in rating modulation should be observed for chosen odors in difficult choices in Experiment 2, but no decrease in ratings for rejected odors in difficult choices should occur.