Fig 1.
Maps of the difference between a happy response and the other emotional responses.
The colors represent t-values. The different scalp maps show the contrast (expressed in t-values) between activity representing a happy response, and activity representing the other emotional responses for the specific frequencies that are indicated below the maps (the delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–128 Hz) frequency range respectively), and across the head for the 64 electrodes.
Fig 2.
Maps of the difference between a sad response and the other emotional responses.
The colors represent t-values. The different scalp maps show the contrast (expressed in t-values) between activity representing a sad response, and activity representing the other emotional responses for the specific frequencies that are indicated below the maps (the delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–128 Hz) frequency range respectively), and across the head for the 64 electrodes.
Fig 3.
Maps of the difference between a fear response and the other emotional responses.
The colors represent t-values. The different scalp maps show the contrast (expressed in t-values) between activity representing a fear response, and activity representing the other emotional responses for the specific frequencies that are indicated below the maps (the delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–128 Hz) frequency range respectively), and across the head for the 64 electrodes.
Fig 4.
Maps of the difference between a disgust response and the other emotional responses.
The colors represent t-values. The different scalp maps show the contrast (expressed in t-values) between activity representing a disgust response, and activity representing the other emotional responses for the specific frequencies that are indicated below the maps (the delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–128 Hz) frequency range respectively), and across the head for the 64 electrodes.
Table 1.
Mean and percentiles for the distributions of out of sample generalization accuracies across 500 repetitions.
Fig 5.
Dynamics of posterior probabilities for Up.
Averaged across participants and 500 repetitions, with different observations used in the feature selection and classifier training stages across repetitions. The shaded areas indicate the standard deviation across repetitions. The vertical lines illustrate six examples of scenes at different moments in time, with moments (3), (5) and (6) indicating parts of the video that contain relatively more sad content. (1) Carl and Ellie just got married and are renovating the house: the posterior probabilities for a happy response are high. Once in a while they go on a picnic and look at the sky full of clouds. First, they see a cloud turn into an animal, then they see a cloud turn into a baby, and eventually at (2) all the clouds start to look like babies. (3) The sad part in de middle is elicited by the moment Ellie “gets told” in the hospital that she cannot have a baby: the posterior probabilities for a sad response rise briefly above chance level. After a short while, the couple picks up where they left off, and the distinction between the probabilities of a happy and sad response increase again. Over time, Carl and Ellie grow old and although they are still very happy with their lives together (4; Carl and Ellie dance together-scene), Carl realizes after having looked at an old photo (5) that much time has passed and their lives have not turned out the way they had hoped for. Eventually, we see Ellie in a hospital bed and at (6) she has just given back to Carl the book in which they had saved all their planned adventures. The posterior probabilities of a sad response rise above chance level from time to time and the happy response does not clearly dominate anymore.