Short forms:
1. ROI (R): Occipito-temporal Left (Lh), Occipito-temporal Central (C), Occipito-temporal Right (Rh)
2 Emotion (E): Angry (A), Happy (H), Fearful (F), Neutral (N) 3. Emotional Valence (EV): Positive (Pos), Negative (Neg) 4. Presentation (P): Subliminal (Sub), Supraliminal (Sup) 5. Selective Attention (A): Target (T), Non target (NT) 6. Gender face (GF) : Female (F), Male (M)
Behavioral results Accuracy
The accuracy of facial expression discrimination in the sub trials was 45.17 % (-0.19 in z-score) and in the sup trials 87.27 % (1.32 in z-score). The binomial distribution shows that the probability that the sub trials are not at chance level is less than of 0.05; [p. < 0.05].
A repeated measure analysis of variance (ANOVA 2x2: P x EV) was applied. The analysis showed significant differences for the P. [F (1, 18) = 77.71, p. < 0.001, mean value: sub= 46.46, S.D. = 4.47; sup= 87.08, S.D. = 2.42].
The z-score normalization shows also a significant P effect. [F (1, 18) = 82.562, p.
< 0.001, mean value: sub= -0.15, S.D= 0.15; sup= 1.23, S.D. = 0.12].
To analyze the E effect an ANOVA 4x2x2: E x P x GF was applied.
We noticed a significant E effect. [F (1, 18) = 8.32, p. < 0.001. Mean values: H=
70.13, S.D. = 3.13, N= 53.15, S. D. = 4.22, F= 71.53, S.D. = 3.65, A= 65.56, S.D.
= 3.45]. More in details, the post-hoc analysis in the Sup condition (LSD test) shows a significant difference between H and F (p. < 0.05) and a tendency between H and A (p. < 0.05).
An interaction effect between E and P was also significant [F (3, 54) = 6.12, p. < y-coordinate represent the percentage of correct responses for each emotion presented in the experiment.
Reaction time (RT)
To analyze the emotional expression effect an ANOVA 4x2x2: E x P x GF was applied. The analysis showed significant differences for the P [F (1, 16) = 17.97, p. < 0.001. Mean value: Sub= 660 msec. S.D. = 34.41, Sup= 590 msec. S.D. = 22.17].
An E effect was also significant. [F (1, 18) = 19.96, p. < 0.001. Mean values: H=
600 msec., S.D. = 23.44, N= 700 msec., S. D. = 35.89, F =600 msec., S.D. =
29.78, A= 600 msec., S.D. = 26.87]. More in details, the post-hoc analysis (LSD test) shows a significant difference between N-H (p. < 0.001), N-F (p. <0.001) and N-A (p. <0.001).
Fig. 10. Interaction effect of emotion (E = emotion; H = happy, N = neutral, F = fearful, A = angry) and presentation factors (Sub = subliminal, Sup = supraliminal). The y-coordinate the average time (in msec.) per each emotion presented in the experiment.
Sub: H= 651 msec., S.D. = 28.80; N= 731 msec., S.D. = 44.76; F= 627 msec., S.D. = 39.5; A= 632 msec., S.D. = 34.09. More in details, the post-hoc analysis (LSD test) shows a significant difference between N-H (p. < 0.001), N-F (p.
<0.001), N-A (p. <0.001).
Sup: H= 549 msec., S.D. = 21.02; N= 669 msec., S.D. = 28.58; F= 574 msec., S.D. = 22.21; A= 567 msec., S.D. = 21.02. More in details, the post-hoc analysis
(LSD test) shows a significant difference between N-H (p. < 0.001), N-F (p.
<0.001), N-A (p. <0.001) and H-F (0.05).
EEG results
N170 component (135-190 msec.)
At the N170 latency, an ANOVA 3x4x2x2: R x E X A X P was applied.
An R effect resulted significant; [F (2, 19) = 12.59, p. < 0.001. Mean values: Lh=
-3.64 μV, S.D. = 0.41; C= -2.78 μV, S.D. = 0.31; Rh= -3.92 μV, S.D. = 0.4], (see fig.11).
The analysis showed significant differences for the E [F (3, 18) = 7.66, p. < 0.001.
Mean values: H= -3.37 μV, S.D. = 0.39; N= -2.90 μV, S.D. = 0.32; F= -3.8 μV, S.D. = 0.31; A= -3.79 μV; S.D. = 0.36], (see fig. 11). Furthermore, the post-hoc (LSD test) analysis shows a significant difference between H and F (p. < 0.05), H and A (p. < 0.001), N and F (p. < 0.05) and N and A (p. < 0.001).
An A effect was also observed [F (1, 18) = 26.14, p. < 0.05. Mean values: NT=
-3.11 μV, S.D. = 0.36; T= -3.44 μV, S.D. = 0.28], (see fig.14, below).
A P effect was observed [F (1, 18) = 46.37, p. < 0.001. Mean values: Sub= -2.79 μV, S.D. = 0.34; Sup= -3.77 μV, S.D. = 0.31], (see fig.14, above).
An interaction effect between E and P was significant [F (3, 18) = 3.22, p. < 0.05]
(see fig. 15).
More in details, a post-hoc comparison analysis (LSD test) show difference in sub condition between N-A (p. < 0.001), N-F (p. < 0.001) and N-H (p. < 0.05), (see fig. 12).
Fig. 11. ERPs of N1 and P2 components recorded on the right, left and central electrode sites. The x-coordinate represent the temporal evolution from the stimulus onset (origin, o msec.) to 350 msec. The y-coordinate represent the μV recorded respectively the three occipito-temporal ROIs together (average).
Fig. 12. ERPs of N1 and P2 components recorded on the right, left and central central occipito-temporal electrode sites in the subliminal condition. The x-coordinate represent the temporal evolution from the stimulus onset (origin, o msec.) to 350 msec. The y-coordinate represent the μV recorded respectively the three occipito-temporal ROIs together (average).
Fig. 13. ERPs of N1 and P2 components recorded on the right, left and central occipito-temporal electrode sites in the supraliminal condition. The y-coordinate represent the μV recorded respectively the three occipito-temporal ROIs together (average). The x- coordinate represent the temporal evolution from the stimulus onset (origin, o msec.) to 350 msec.
Fig. 14. Above: ERPs (N1 and P2) recorded for the occipito-temporal lobe (three ROIs together) for the presentation effect (sub = subliminal, sup = supraliminal). The y-coordinate represent the μV recorded on the three occipito-temporal ROIs together (average). The x- coordinate represent the temporal evolution from the stimulus onset (0 msec.) to 350 msec.
Below: ERPs (N1 and P2) recorded for the occipito-temporal lobe (three ROIs together) for the selective attention effect (T = target, NT = non target). The y-coordinate represent the μV recorded on the three occipito-temporal ROIs together (average). The x- coordinate represent the temporal evolution from the stimulus onset (0 msec.) to 350 msec.
Sub Sup
A F H N
E -4.5
-4.0 -3.5 -3.0 -2.5 -2.0 -1.5
μV
Fig. 15. Interaction effect of emotion (E = emotion; H = happy, N = neutral, F = fearful, A = angry) and presentation factors (Sub = subliminal, Sup = supraliminal). The y-coordinate represent the μV recorded on the three occipito-temporal ROIs together (average).
P2 component (200-270 msec.)
An ANOVA 3x4x2x2: R x E x A x P was applied.
An R effect resulted significant. [F (2, 18) = 13.15, p. < 0.001. Mean values: Lh
=2.31 μV, S.D. = 0.51; C= 3.93 μV, S.D. = 0.55; Rh= 1.96 μV, S.D. = 0.45], (see fig. 11)
A P effect was also observed [F (1, 18) = 118.86, p. < 0.001. Mean values: Sub=
4.32 μV, S.D. = 0.5; Sup= 1.15 μV, S.D. = 0.44], (see fig. 14, above).
An A effect was observed. [F (1, 18) = 5.99, p. < 0.05. Mean values: NT= 3.02 μV, S.D. = 0.45; T= 2.45 μV, S.D. = 0.47], (see fig. 14, below).
An interaction effect between R and P was significant [F (2, 36) = 4.13, p. < 0.05]
(see fig. 16).
Sub Sup
Lh C Rh
R -4
-2 0 2 4 6 8 10 12
μV
Fig. 16. Interaction effect between ROIs (R = ROI, Lh = left, C = central, Rh = right) and presentation factors (Sub = subliminal, Sup = supraliminal). The y-coordinate represent the μV recorded on the three occipito-temporal ROIs together (average).
Topographic maps over N 170 and P2 components Topographic maps displaying the N170 and the P2 components in the subliminal and subliminal conditions were observed (see fig. 17).
For the N170 component we observe the same activation mode for the two maps but a less intense activation for the subliminal condition (in line with N170 effect on ERPs).
For the P2 component, the subliminal condition presents more intense and more posterior positivity. The supraliminal condition presents a more centrally placed posterior positivity and also a frontal positivity not present in the subliminal condition.
Fig.17. Topographic maps which represent the distribution on the scalp of ERP recorded on electrodes.
Above: topographic maps representing on the left the subliminal and on the right the supraliminal conditions on the N170, peaking at 166 msec.
Below: topographic maps representing on the left the subliminal and on the right the supraliminal conditions on the P2, peaking at 227msec.
P300 component (440-540 msec.)
An ANOVA 4x2x2: E x A x P was applied. The R effect was not considered because we used a just one parietal ROI.
An A effect was observed. [F (1, 18) = 23.69, p. < 0.001, NT= 1.81 μV, S.D. = 0.30; T= 2.60 μV, S.D. = 0.32]
A P effect was observed. [F (1, 18) = 61.84, p. < 0.001. Mean values: Sub= 1.62 μV, S.D. = 0.27; Sup= 2.80 μV, S.D. = 0.34].
Source localization
We made a source localization to observe the brain activation (with a particular interest for the amygdala) for the four types of emotions (having collapsed all the other stimulus categories together).
Source analysis were performed on the grand-average on individual ERP data at the latency of the maximum N170 (135-190 msec. from the stimulus onset) for each subject and for each category (4 E) using a distributed and linear inverse solution LAURA.
Fig. 18. Represents the time frame of 135 to 190 msec. (red selection) which was considered as the one corresponding to the N170 component. The origin
represents the stimulus onset.
Fig. 19. Topographic maps evolution and source localization on the time frame from 135 to 190 msec. for the four emotions studied (Neutral, Happy, Angry and Fearful).
Although no statically analysis was made, we made some interesting observations.
The topographic maps show an earlier N170 for negative (A and F) emotions.
The source localization analysis presents an activation of temporal and limbic lobes for all the four emotions, with a stronger activation of the right part even if they are activated more strongly by negative emotions (F more than A) compared to neutral and happy ones.
An interesting amygdala activation is presented for H, A, F but not for N. The activation is stronger F than for A and H. Moreover, the amygdala activation is bilateral only for F.