• Aucun résultat trouvé

3. E MPIRICAL PART

3.2. Temporal dynamics of neural evaluative processes: Event-related potentials

3.2.3. Materials and Methods

3.2.3.1. Participants

Twenty-four healthy and right-handed (mean of Edinburgh Handedness Inventory = 88.93, SD = 12.27) female undergraduate students of the University of Geneva participated for payment. 25 CHF (Swiss francs) was the minimal amount of guaranteed money for their participation. Depending on their performance on the task, participants could additionally obtain up to 16 CHF. Participants ranged in age from 18 to 30 years (M = 21.38, SD = 0.66).

All of them had normal or corrected-to-normal vision.

3.2.3.2. Task and Procedure

After participants arrived at the laboratory, they read and signed an informed consent form and filled out questionnaires about their current health and demographic characteristics.

Participants sat in a semi-dark and sound-attenuated room in front of a computer screen to perform the computerized gambling task (17, resolution 1280  1024). The distance between participants’ eyes and the computer screen was 60 cm. Participants gave their choices for each trial of the task via the numeric keyboard of a standard PC keyboard. Both the practice session and the gambling task contained only gray characters against a black background.

CHAPTER 3.2:CENTRAL PROCESSING OF APPRAISAL CHECKS II 67

Figure 2. Trial structure of the gambling task. Presentation times of each event in a trial are indicated below the corresponding screen. At feedback onset, the goal conduciveness and the power check information was simultaneously presented via filled or non-filled geometric shapes. RT = reaction time. See text for details.

The gambling task consisted of the following trial structure of subsequent events, presented in Figure 2. Each trial started with a fixation cross (randomized duration between 300 and 700 ms; 1° high, 1° wide) in the center of the screen, followed by two circles (Figure 2, screen “selection of circle”; 3.8° high, 4.6° wide). Participants were told that the possible outcomes of a trial (a win, +0.05 CHF or a loss, −0.05 CHF)2 were concealed under these circles. No cues were provided that allowed the participants to tell under which circle the win was hidden. The chosen circle was highlighted (300 ms) before the feedback stimulus

appeared at its center (Figure 2, screen “feedback”; 500 ms).

Feedback stimuli conveyed the information of both the goal conduciveness check (win vs. loss of money) and the power check (high [two choice options] vs. low [one choice option to decide about the gambling outcome]). For decoding the goal conduciveness check,

2 The break-even condition was not operationalized in Experiment 2 for two reasons: (1) Break-even outcomes are not necessarily evaluated as neutral, which is a problem since this effect cannot be explained conclusively. In order to explain the effects observed in response to break-even outcomes further experimental testing is necessary. However, (2) the scope of the thesis research was testing the sequence hypothesis for the goal conduciveness and the coping potential checks with a focus on the appraisal checks of coping potential.

selection of circle

CHAPTER 3.2:CENTRAL PROCESSING OF APPRAISAL CHECKS II 68 we used three different geometric shapes (hexagon, square, and diamond); for decoding the power check, these shapes had a solid fill or no fill (see Table 1 as an example). These associations were counterbalanced across participants. Moreover, in order to manipulate the control check, the amount of presented high and low power feedback was modified by creating two blocks of gambling sessions. In high control blocks, 75% of the gambling feedback consisted high power feedback. In the low control blocks, 75% of the feedback indicated low power. In these gambling blocks wins and losses occurred equally often (50%

of wins and 50% of losses). Overall gambling blocks, low and high power feedback was presented equally often. In the beginning of each block, participants were informed about their degree of control for the upcoming trials (high control block: “In the following sessions it will be you who decides most of the time about the gambling outcome.”; low control block:

“In the following sessions it will be the computer who decides most of the time about the gambling outcome.”).

Table 1

Example of Feedback Stimuli for the Operationalization of the Goal Conduciveness and the Power Checks

Power

Feedback conditions High (two choices) Low (one choice)

Goal conduciveness Win (+0.05 CHF)

Loss (−0.05 CHF)

Note. Goal conduciveness = goal conduciveness appraisal check; Power = power appraisal check.

The association of the geometric shapes with the levels of goal conduciveness as well as the meaning of filled and nonfilled shapes was counterbalanced across participants.

In total there were eight different feedback stimuli conditions: (1) “high control, high power loss”, (2) “high control, high power win”, (3) “high control, low power loss”, (4) “high control, low power win”, (5) “low control, high power loss”, (6) “low control, high power

CHAPTER 3.2:CENTRAL PROCESSING OF APPRAISAL CHECKS II 69 win”, (7) “low control, low power loss”, and (8) “low control, low power win”. The

minimum of repeated measures of one feedback stimulus condition was 57 times.

After feedback presentation, the screen went black (1 s), followed by a screen having one letter to the left and one to the right side (Figure 2, screen “choice about outcome”; A = accept, R = reject; 0.8° high, 6.6° wide; Arial font, size 28). Here, participants made their choice about the outcome of that trial: In high power trials, they could freely choose from two choice options of accepting or rejecting a win, or a loss (presentation of “A R” or “R A”:

randomized order with the same number of presentations); in low power trials, they had to accept the chosen choice option of either rejecting (presentation of “R R”) or accepting (“A A”) the outcome (randomized selection with the same number of presentations). Next, the letter of the participant’s decision was highlighted (Figure 2, screen “monetary and choice feedback”; 300 ms; Arial font, size 52 bold) and the total monetary outcome was

simultaneously presented. The total monetary outcome was updated at the end of each trial to the amount of money won (+0.05 CHF) or lost (−0.05 CHF). The next trial started

immediately after. The total amount of experimental trials was 864 trials (two blocks of each 144 trials, each repeated three times). The duration of the gambling task was about 50

minutes.

In order to familiarize participants with the gambling task, participants completed a practice session (48 trials of gambling, 5–7 min) to learn the gambling rules and the meaning of the filled and non-filled shapes (Table 1). Trials of the practice session were identical to experimental trials except for a longer duration of the feedback stimulus (900 ms), an additional presentation of explanatory labels (presented with each feedback stimulus: “win, choice”; “win, no choice”; “loss, choice”; “loss, no choice”). Participants were instruction that in the gambling task no explanatory labels will be shown. Participants responses were evaluated online by an implemented performance cutoff criterion (>80% of correct responses,

CHAPTER 3.2:CENTRAL PROCESSING OF APPRAISAL CHECKS II 70 i.e., accepting wins and rejecting losses). Only if the criterion was reached at the end of the practice session, the gambling task started immediately after a short break; otherwise, the practice session was repeated.

In the gambling task, at the end of each block participants were asked to answer four questions on a continuous scale. They rated how much they felt (a) positive or negative (valence), (b) aroused, (c) in control over the gambling task, and (d) how predictable they considered the occurrence of the feedback stimuli. Participants answered these questions by placing a mark at the appropriate position on a continuous line, anchored from -100 to 100 (1° high, 18° wide). In total participants answered these rating scales six times.

The amount of bonus money won during the gambling task depended on the

participant’s performance. Participants were told to maximize the amount of bonus money, without mentioning the maximum bonus amount possible (16 CHF). They were assured that they would not end up losing money (getting less than 25 CHF) or owing money to the experimenter. Participants were not informed that the type of feedback was selected at

random on each trial (feedback stimulus probability was balanced for all feedback conditions, with equal probability and without replacement across all trials); they were told only that they would play a gambling task. At the end of the experiment, participants were paid their

participation fee (25 CHF) plus the bonus money, and they were informed about the experimental manipulations.

3.2.3.3. Data Acquisition

Practice session, gambling task, and acquisition of behavioral data were administered using E-Prime 2.0 (Psychology Software Tools, Inc., Pittsburgh, PA).

Electroencephalography (EEG 64-channel electrode cap) and facial electromyography

CHAPTER 3.2:CENTRAL PROCESSING OF APPRAISAL CHECKS II 71 (EMG)3 data were recorded (bandwidth 0.1–417 Hz, sampling rate: 2048 Hz) with a

BIOSEMI Active-Two amplifier system (BioSemi Biomedical Instrumentation, Amsterdam, the Netherlands). The EEG data was preprocessed offline with Brain Vision Analyzer software (Brain Products, Gilching, Germany) and EEGLAB 11.0.4.3b (Delorme & Makeig, 2004), implemented in a commercial software package (MATLAB, 2012).

Prior to the horizontal and vertical eye movement correction based on individual component maps, extracted by ICA implemented in EEGLAB, the EEG raw data was down sampled to 256 Hz, high pass filtered (0.1 Hz), and noisy channels were removed. Next, the data were exported to Brain vision analyzer. Here, the removed channel were interpolated, the data was average re-referenced, filtered (30 Hz, roll-off 24db/octave), and segmented (200 ms of baseline period, and 1.5 s of after-stimulus-onset period). Next, the segmented data was cleaned from trials in which artifacts exceeded ±110 μV in all channels (2.62% total amount of excluded trials across all participants), baseline corrected, and averaged for each feedback stimulus condition.

3.2.3.4. Data Analyses

The ERP analyses were performed on mean microvoltage amplitudes (µV) of the averaged EEG data, time-locked to the onset of feedback stimulus presentation in the gambling task. A temporal principal component analysis was computed (PCA, for

methodological considerations see Picton et al., 2000; van Boxtel, 1998) to explore whether the extracted time intervals match with the predicted time intervals of the FRN and the P300.

Temporal PCA is a data-driven exploratory procedure that identifies systematic changes of variance over time. It extracts temporal factors (TFs) in a multichannel ERP data set by detecting systematic changes of variance, notably induced by the experimental

manipulation. The temporal dimensionality of the data is reduced by treating the time points

3 Facial EMG was concurrently recorded over frontalis, corrugator, and cheek regions. Results are reported in chapter 3.3.

CHAPTER 3.2:CENTRAL PROCESSING OF APPRAISAL CHECKS II 72 of the ERP segments as variables. The total number of cases (observations) depends on

sampling rate, epoch length, number of electrode sites, number of experimental conditions, and number of participants. Recent studies have shown that a temporal PCA is a sufficient linear reduction method for ERP analyses (Picton et al., 2000; van Boxtel, 1998), and that it is a data-driven alternative to arbitrarily setting time intervals to investigate ERP components (Dien, 2010; Dien, Beal, & Berg, 2005; Kayser & Tenke, 2003; Pourtois, Delplanque,

Michel, & Vuilleumier, 2008).

3.2.3.4.1. Temporal PCA

Data for the temporal PCA consisted of the feedback-locked ERP averages of 12,288 cases (64 electrode sites × 8 feedback stimulus conditions × 24 participants) within a time interval of 0 to 1.5 s following feedback stimulus presentation (equals 384 successive time points treated as variables). The baseline period (-200 ms) was removed prior to the

computation of the temporal PCA. Following recommendations from ERP simulation studies (Dien et al., 2005; Kayser & Tenke, 2003), the temporal PCA was computed based on the covariance matrix of the data and Kaiser normalization was applied to the factor loadings prior to their rotation. The TFs were rotated using an oblique Promax rotation (cf. Dien, 2010) and the number of retained factors followed the Scree plot criterion (Cattell, 1966).

The time intervals were identified using the loadings of TFs of the pattern matrix and the 0.707 criterion4 (Delplanque, Silvert, Hot, & Sequeira, 2005): A time interval started at the point where the rising factor loading curve intersected the criterion line for the first time and ended at the point where the descending curve crossed it the second time.

4 Factor loadings greater than or equal to 0.707 (equals square root of 0.5) indicate that a minimum of 50% of the total variance is explained by that factor at that time point. Within a time interval (located between loading values of greater than or equal to 0.707 and smaller than or equal to 0.707), this factor explains at least 50% of the variance. All the other factors explain the remaining variance for that time interval.

CHAPTER 3.2:CENTRAL PROCESSING OF APPRAISAL CHECKS II 73 3.2.3.4.2. ERPs

The grand averaged data were grouped into the eight feedback stimulus conditions with the factors control, goal conduciveness, and power: (1) “high control, high power loss”, (2) “high control, high power win”, (3) “high control, low power loss”, (4) “high control, low power win”, (5) “low control, high power loss”, (6) “low control, high power win”, (7) “low control, low power loss”, and (8) “low control, low power win”. The FRN was evaluated at the channels Fz and FCz, and the P300 at the midline channels Fz, FCz, Pz and POz. ERP analyses were based on the mean microvoltage amplitudes within the time range of the corresponding TF. For the FRN, the deflections of the mean amplitudes were investigated in a 2 (Control) × 2 (Goal Conduciveness) × 2 (Power) × 2 (Channel Location) repeated

measures analysis of variance (ANOVA). For the P300, they were examined in a 2 (Control)

× 2 (Goal Conduciveness) × 2 (Power) × 4 (Channel Location) repeated measures ANOVA.

Throughout all analyses Greenhouse-Geisser correction was applied (cf. Vasey & Thayer, 1987), the correction factor epsilon is provided (simply noted as ε). All statistical tests were performed at an alpha level of 5%. Multiple testing during post hoc analyses was considered by applying Holm’s stepwise correction procedure (Holm, 1979), the corrected p-values are labeled as padj. All reported effect sizes are partial 2 (in the results simply noted as 2). All statistical analyses were carried out with IBM SPSS Statistics 19.