III.4 Résultats des expériences

Expérience 5 : Effet des différences interindividuelles sur le traitement des arguments MP et AC (Conditionnels basiques) : une étude combinant la mesure du temps de lecture et l’EEG.

Methods

A. Behavioral experiment

Participants

Forty-one, healthy, native French-speaking volunteers with no history of neurological or psychiatric disorders participated in the study. Participants were separated into two groups based on the way they responded to the Affirmation of the Consequent (AC): participants who endorsed the AC argument were considered “pragmatic” (15 participants) and those who generally rejected the argument were considered “normative” (13 participants before further data analysis). Six participants were excluded from the analyses due to inconsistent responses to the AC argument (whom we refer to as “mixed” participants; 13 participants). All subjects were right-handed as measured by the Edinburg Handedness Inventory

Participants were presented with 48 conditional arguments involving letters of the alphabet. As an example, consider If A then C, where A refers to the antecedent and C the consequent. We purposely excluded the letters Y and W, which require more than two syllables when pronounced in French and, as in the example, conditional statements never included two letters that appear consecutively in the alphabet.

The experiment was designed so that any given conditional could provide a Modus Ponens (MP)argument that was valid (If A then C; A//C) or not (If A then C;A;D). Likewise, conditionals allowed for the validation of Affirmation of the Consequent arguments (If A then C;C//A) or not (If A then C;C//F). Thus, the experiment can be viewed as a 2 (minor premise: conditional’s antecedent vs. consequent) by 2 (conclusion: matching other mentioned item vs. non matching) design. Altogether, we were concerned with six experimental conditions. Two were at the premise level (If A then C; A versus If A then C;C) and correspond to the premises for MP and AC (respectively, 24 trials each). Four further conditions were nested within these two: the Matching MP condition (If A then C; A; C ); the Mismatching MP condition (If A then C; A; D ); the Matching AC condition (If A then C; C; A ) and the Mismatching AC condition (If A then C; C; X ). There were 12 trials in each of these four conditions.

There were two response keys corresponding to two evaluations of the conclusion: “CORRECT” and “INCORRECT”. For the “normative” group, 75% of the trials ought to be considered “INCORRECT” (for the Mismatching MP, and the two AC conditions) while for the “pragmatic” participants, 50% of responses are expected to be “INCORRECT” (for the Mismatching MP and the Mismatching AC conditions). Figure 1 summarizes the design of this experiment.

---------------------------------------

Insert Figure 1 about here

---------------------------------------

Procedure

Stimuli were presented with Presentation 10.2 software (Neurobehavioral Systems, http://www.neurobs.com/ ) on a computer screen. In the instruction set, participants were informed that they would have to make two sorts of conclusion evaluation and were given an example of each argument they would have to process during the experiment. Each trial started with the presentation of a visual central mark (a cross) in the center of the screen for 800 ms. The major premise was then presented. The participant had to press the “space” button to continue the task. The minor premise was then presented. Again, the participant had to press the “space” button to continue the task. Eventually, the conclusion appeared and the participants had to press the response key. Participants were instructed to respond as quickly and as accurately as possible.

Data analysis

We applied a log transformation to the reaction time data to improve the conformity of the data to the standard assumptions of ANOVA (e.g., Howell, 1997).

We ran an independent t - test to compare the mean reading time of the major premise of the two groups and an ANOVA using the two argument conditions (Modus Ponens and Affirmation of the Consequent) as a within-subject factor and the reading time of the minor premise as a between-subject factor.

B. EEG experiment

Participants

Forty-six, healthy, native French-speaking volunteers with no history of neurological or psychiatric disorders participated in the study. Participants were separated into two groups based on the way they responded to the AC argument: participants who endorsed the AC argument were considered “pragmatic” (17 participants before further data analysis) and those who generally rejected the argument were considered “normative” (23 participants before further data analysis). Six participants were excluded from the analyses due to inconsistent responses to the AC argument (whom we refer to as ‘mixed’ participants) and five due to excessive eye movement artifacts in the EEG signal. Additionally, four participants were randomly removed from the “normative” group in order to obtain a comparable number of participants in both groups (15 “normative” and 15 “pragmatic”). These 30 participants (24 females) were aged between 20 and 30 (mean age: 24 years). All subjects were right-handed as measured by the Edinburg Handedness Inventory.

Participants were presented with 144 conditional arguments involving letters of the alphabet. The same constraints in the choice of stimuli as in the behavioral experiment were applied to this experiment. As in the behavioral experiment, we were concerned with six experimental conditions in this experiment. Two were at the premise level (If A then C; A versus If A then C;C) and correspond to the premises for MP and AC (respectively; 72 trials each). Four further conditions were nested within these two: the Matching MP condition (If A then C; A; C ); the Mismatching MP condition (If A then C; A; D ); the Matching AC condition (If A then C; C; A ) and the Mismatching AC condition (If A then C; C; X ). There were 36 trials in each of this condition. We also added 30 filler trials that were designed to block expectations about the two sorts of minor premise presented above. These used letters in the minor premise that were unmentioned in the conditional, as in (If A then C; R ; S).

There were two response keys corresponding to two evaluations of the conclusion: “CORRECT” and “INCORRECT”. For the ‘normative’ group, 75% of the trials ought to be considered “INCORRECT” (for the Mismatching MP, and the two AC conditions) while for the “pragmatic” participants, 50% of responses are expected to be “INCORRECT” (for the Mismatching MP and the Mismatching AC conditions). Figure 2 summarizes the design of this experiment.

---------------------------------------

Insert Figure 2 about here

---------------------------------------

Procedure

Stimuli were presented with Presentation 10.2 software (Neurobehavioral Systems, http://www.neurobs.com/ ) on a computer screen. In the instruction set, participants were informed that they would have to make two sorts of conclusion evaluation and were given an example of each argument they would have to process during the experiment. Each trial started with the presentation of a visual central mark (a cross) in the center of the screen for 800 ms. The conditional premise was presented for 1300 ms whereas the minor premise and the conclusion were presented for 900 ms. The two premises and the conclusion appeared successively and were separated by a cross displayed in the center of the screen for 800 ms. Subjects had to wait for the display of the phrase “your response” before pressing the response key. Participants performed the experiment in 3 blocks. Trial order within each block was randomized and the block order within each task was counterbalanced across participants. Participants were asked to avoid making eye movements or vocal articulations (audible or inaudible) during the trial. Participants were instructed to respond as quickly and as accurately as possible. The task began with 10 training trials.

Electroencephalogram (EEG) recording

Subjects were seated in a dimly lit, electrically shielded, sound-attenuating chamber. EEG was recorded with a 64 channel NetAmps.200 system (Electrical Geodesics Inc, see figure 3). Amplified analogue voltages (0.1-200 Hz bandpass) were sampled at 500 Hz. Electrode impedance was kept below 40 KΩ. All channels were referenced to Cz during recording, and off-line re-referenced to the average mastoids. ERP analyses were conducted using ELAN-Pack software developed at INSERM U821 (Lyon, France). They consisted in averaging the EEG segments in synchronization with the onset of the minor premise and the conclusion in each condition over a 800 ms period including a 100 ms pre-stimulus interval. The signals were low-pass filtered (20 Hz) and a baseline correction was calculated from the 100 ms preceding the presentation of the letter in each condition. Trials contaminated by eye blinks or eye movements (threshold: ±100 µV) were not included in the analyses. Also, analyses were restricted to trials on which subjects made correct responses (according to their general profile as “normative” or “pragmatic”).

Data analysis

We distinguished two types of components: transient components and sustained components, the latter of which corresponds to slow waves, lasting more than 100 ms. For transient components, we computed the mean value of the signal in a 60 ms time-window centred around the peak of each analysed component. To overcome effects that could result from the difference in amplitude of preceding components, we corrected the signal so that the first value of the window was equal to zero. For sustained components, we extracted the mean value of the amplitude during a longer time window.

All analyses were run with the same electrode positions. Twelve representative electrodes of the 10-20 system were chosen to define different scalp regions (frontal: F3, Fz and F4; central: C3, Cz and C4; centro-parietal: CP3, CPz and CP4 and parietal: P3, Pz and P4). We ran multiple ANOVAs using repeated measures including Conditions (defined for each analysis) and two levels of Electrode Site: laterality (Left, Midline and Right) and anterior-posterior location (Frontal, Central, Centro-parietal, Parietal) as within-subject factors. The group was used as between-subject factors. Relevant post-hoc comparisons were computed with Tukey HSD tests.

---------------------------------------

Insert Figure 3 about here

---------------------------------------

Results

A. Behavioral experiment

The independent t-test run to compare the mean reading time of the major premise of the two groups revealed that the ‘normative’ participants read this premise slower than the ‘pragmatic’ participants (mean = 2791 ms vs 1849 ms; t(26) = 2.03 p < .05).

The reading time for the minor premise was slower for AC condition in both groups, and generally ‘normative’ participants (MP mean = 1021 ms ; AC mean = 1207 ms) were slower than ‘pragmatic’ (MP mean = 797 ms ms ; AC mean = 918 ms) participants

Our ANOVA using MP and AC as a within-subject factor and the reading time of the minor premise as a between-subject factor revealed a main effect of the condition (F(1,25) = 15.38, p < .001) and a main effect of the group (F(1,25)= 4.42, p < .05).

Finally, we also run a paired-t test to compare the reaction time of ‘pragmatic’ participants to the conclusion of both arguments. This paired t-test revealed no differences between the conditions (t (15) = 0.32, p = 0.7).

B. EEG experiment

In what follows, we present the results in the order of stimuli presentation. The components investigated are the N2, the P3b and the PSW.

Introduction of Minor Premise

N2 component

As shown in Figures 3 and 4, the frontocentral N2 was larger in the AC condition than in the MP condition for both the ‘normative’ and ‘pragmatic’ groups. Moreover, this difference was larger among ‘normative’ participants compared to ‘pragmatic’ participants.

We analysed the effect of the argument by comparing MP to AC. We entered the mean value obtained in the N2 window into a four-way ANOVA with argument type (MP vs AC), laterality and anterior-posterior location as within-subject factors and the group (‘normative’ vs ‘pragmatic’ participants) as a between-subject factor. This analysis showed main effects of anterior-posterior location (F(3,84) = 17.97, p < .001), laterality (F(2,56) = 19.06, p < .001) and a significant interaction between these two factors (F(6,168) = 10.8, p < .001). It also revealed a significant interaction between these factors and the group (F(6,168) = 2.37, p < .05), between anterior-posterior location and the argument type (F(3,84) = 4.53, p < .01) and between the four factors (F(6,168) = 2.21, p < .05).

Post-hoc tests revealed that the N2 was larger in the AC condition than in the MP condition in midline and left frontal electrodes (p < .01 and p < .05 respectively) for the ‘pragmatic’ participants, and in the frontal area (p < .001 on the three electrodes) as well as midline, right central and centroparietal electrodes (p < .001 for each) for the ‘normative’ participants. The N2 component was also larger in the MP condition for the ‘pragmatic’ participants than it was for the ‘normative’ participants.

P3b component

Visual inspection of figure 3 and 4 reveals a larger centroparietal P3b component for the MP condition than for the AC condition in both groups. The difference was larger in the ‘normative’ group.

We carried out a four-way ANOVA on the amplitude of the P3b component, using argument type, anterior-posterior location and laterality as within-subject factors and group as a between-subject factor. We found a main effect of anterior-posterior location (F(3,84) = 20.06, p < .001) and argument type (F(1,28) = 10.7, p < .01). We also found a significant interaction between these two factors (F(3,84) = 14.67, p < .001) as well as between anterior-posterior location and laterality (F(6,168) = 3.01, p < .01). The ANOVA also revealed a significant interaction between anterior-posterior location, argument type and group (F(3,84) = 2.53, p < .05).

Post-hoc tests showed that the P3b was larger in the MP condition for the ‘normative’ participants compared to the ‘pragmatic’ participants (p < .05). The two groups also differed with respect to the P3b location. While the MP condition prompted a larger P3b (with respect to the AC condition) in the centroparietal and parietal sites (p < .001 for both sites) among the ‘normative’ participants, the P3b in the MP condition was larger than in the AC condition in the parietal site (p < .01) among the ‘pragmatic’ participants.

Presentation of the Conclusion

N2 component

Visual inspection of Figure 5 shows that, for the ‘pragmatic’ participants, the frontocentral N2 seem not to differ between the Matching MP conclusion condition and the Matching AC conclusion condition. The inspection of Figure 6 reveals that ‘normative’ participants produced a larger frontocentral N2 component for the Matching AC condition compared to the Matching MP condition.

We computed a four-way ANOVA argument type (MP vs AC), laterality and anterior-posterior location as within-subject factors and the amplitude of the N2 component as a between-subject factor. This analysis revealed a main effect of anterior-posterior location (F(3,84) = 37.26, p < .001) and laterality (F(2,56) = 16.68, p < .001). It also showed a significant interaction between anterior-posterior location and laterality (F6, 168) = 9.89, p < .001), between anterior-posterior location and the argument type (F(3,84) = 3.68, p < .01) and between anterior-posterior location, argument type and group (F(3,84) = 3.5, p < .01).

Post-hoc tests revealed that the N2 was larger in the AC conditions than in the MP conditions only for the ‘normative’ participants in frontal areas (p < .001).

P3b component

As shown on figures 5, the centroparietal P3b component did not appear different between the Matching MP and AC conditions for the ‘pragmatic’ participants. The inspection of Figure 6 reveals that ‘normative’ participants produced a larger frontocentral P3b component for the Matching MP condition compared to the Matching AC condition.

We entered the mean value obtained in the P3b window into a four-way ANOVA with argument type (MP vs AC), laterality and anterior-posterior location as within-subject factors and the group as a between-subject factor. This analysis revealed a main effect of anterior-posterior location (F(3,84) = 90.39, p < .001), laterality (F(2,56) = 15.79, p < .001), and argument type (F(1,28) = 4.73, p < .01). We also observed a significant interaction between anterior-posterior location and laterality (F(6,168) = 6.63, p < .001) and between anterior-posterior location, argument type and group (F(3,84) = 3.51, p < .01).

Post-hoc tests revealed that the latter interaction effect resulted from a larger P3b component in MP conditions compared to AC conditions in centroparietal and parietal electrodes (p < .001) for ‘normative’ participants only.

Complementary results at the premise level

PSW component

Visual inspection of Figure 3 reveals that the centroparietal PSW seems larger in the AC condition than in the MP condition for the ‘pragmatic’ participants. More precisely, two distinct components can be observed. A first component with its maximum value on the midline central, centroparietal and parietal electrodes, and a later component with its maximum value in the left central, centroparietal and parietal electrodes.

As shown in Figure 4, participants in the ‘normative’ group prompted a PSW in the MP condition and a long lasting positive wave in the AC condition, which do not have the same properties as the PSW observed in the MP condition for this group or in both conditions for the ‘pragmatic’ group. This long-lasting wave is also observed in Mismatching MP (and AC for the ‘pragmatic’ group) conditions (cf Figures 5 and 6).

Because of the overlap between the PSW in the MP condition and the positive wave in the AC condition, we analyzed the effect of the argument on the amplitude of the two PSW components only for the ‘pragmatic’ group. We computed two three-way ANOVA on the amplitude of the two components, using argument type (MP vs AC), anterior-posterior location and laterality as within-subject factors.

The first analysis revealed a main effect of anterior-posterior location (F(3,42) = 23.26), p < .001) and argument type (F(1,14) = 6.61, p < .05). It also showed a significant interaction between anterior-posterior location and laterality (F(6,84) = 3.42, p < .01), as well as between these two factors and argument type (F(6,84) = 4.16, p < .001).

Post-hoc tests showed that this last effect resulted from a larger PSW component in the AC condition in the right and midline centroparietal and parietal electrodes (p < .001 for each electrode except the right parietal electrode with p < .01).

The second analysis showed a main effect of laterality (F(2,28) = 6.67), p < .01) and a significant interaction between anterior-posterior location and laterality (F(6,84) = 2.63, p < .05). It also showed a marginally significant interaction between laterality and argument type (F(2,28) = 2.74, p = .07).

Figures

Figure 1 :Experimental design of the behavioural experiment. The numbers correspond to the number of trials in each condition.

Figure 2 :Experimental design of the EEG experiment. The numbers correspond to the number of trials in each condition.

Figure 3: Electrode layout of the 64 channel geodesic sensor net. Sites of interest are stained in blue. Frontal electrodes are shown at the top of the figure.

Figure 4 : ‘Pragmatic’ participants. Stimulus locked grand-average waveforms evoked by the appearance of the letter in the minor premise of the MP (black line) and AC (red line) conditions across the 12 sites of interest. Left electrodes are shown in the left column, midline electrodes in the middle column, and right electrodes in the right column. The components of interest are shown by an arrow; these components are indicated only on electrodes showing significant statistical effects.

Figure 5 : ‘Normative’ participants. Stimulus locked grand-average waveforms evoked by the appearance of the letter in the minor premise of the MP (blue line) and AC (green line) conditions across the 12 sites of interest. Left electrodes are shown in the left column, midline electrodes in the middle column, and right electrodes in the right column. The components of interest are shown by an arrow; these components are indicated only on electrodes showing statistical effects.

Figure 6 : ‘Pragmatic’ participants. Stimulus locked grand-average waveforms evoked by the appearance of the letter in the conclusion of the Matching MP (dashed dark line) and Matching AC (dashed red line) conditions across the 12 sites of interest. Left electrodes are shown in the left column, midline electrodes in the middle column, and right electrodes in the right column. The components of interest are shown by an arrow; these components are indicated only on electrodes showing significant statistical effects.

Figure 7 : ‘Normative’ participants. Stimulus locked grand-average waveforms evoked by the appearance of the letter in the conclusion of the Matching MP (dashed blue line) and Matching AC (dashed green line) conditions across the 12 sites of interest. Left electrodes are shown in the left column, midline electrodes in the middle column, and right electrodes in the right column. The components of interest are shown by an arrow; these components are indicated only on electrodes showing significant statistical effects.

Figure 1
Figure 1
Figure 2
Figure 2
Figure 3
Figure 3
Figure 4
Figure 4
Figure 5
Figure 5
Figure 6
Figure 6
Figure 7
Figure 7