Fig 1.
(A) The character on the right is performing the DA ‘to threaten’. This DA seems to be successful because the other person in the image shows fear. (B) The DA performed by the person on the right is ‘to comfort’. Here the person receiving the action still seems sad, meaning that the action has not yet taken effect. This DA may or may not work in the future. (C) Schematic of the basic unit of the survey in experiment 1. Online participants used a mouse to set a value on each of the continuous slider-scales (thus there is no default agreement value). The DA words were taken from List C (Table 1) using a pseudo-random order. The definitions of the words were taken from WordNet. See Figure C in S1 File for the full screenshot and more details. Cartoons reprinted from Shutterstock.com under a CC BY license, with permission from Shutterstock.
Table 1.
Categorized list of DA verbs used for experiment 1 (List C).
Fig 2.
Images and verbs clustered into groups according to the raters’ agreement.
Shown is the median score from 60 replies for each pair of images and DA verbs that exceed a statistical threshold (blue marks pairs below threshold). Images and verbs were ordered according to clustering, such that images that are close to each other have similar DA verbs, and DA verbs that are close to each other have similar images. The lower left block describes negative valence DAs, and the top right block represents positive valence DAs. Cartoons reprinted from Shutterstock.com under a CC BY license, with permission from Shutterstock.
Fig 3.
The DA stimuli set used in the experimental analysis.
Reprinted from Shutterstock.com under a CC BY license, with permission from Shutterstock.
Fig 4.
(A) Distribution of all answers to high-score-agreement questions of survey 1. (B) The distribution of high-score-agreement DA words per image. For example, two images had 7 high-score-agreement DA words.
Fig 5.
(A) Cartoon stimuli in the space of the first two principal components (PCs) show a V-shape. Each cartoon was described as a vector of responses in a 22-dimensional space of the DA words and projected on the PC1-PC2 plane. (B) Valence of the DA and the valence of the emotion of the actor are not identical. Cartoon stimuli organized by PC1 of DA and PC1 of emotions. Cartoons that didn't receive significant agreement in either DA labels or emotion labels were not included. There is a moderate correlation between the PCs (r = 0.49, p = 0.01). However, the cases where valence of the two PCs is opposite are not outliers, but instead are valid sub groups of the stimuli set. Reprinted from Shutterstock.com under a CC BY license, with permission from Shutterstock.
Fig 6.
Examples of stimuli where the DA valence is not correlated to emotion valence.
The same DA, ‘to support’, can be performed while being either happy or sad (right side of the image). Additionally, one can perform a negative DA such as ‘to bully’ while being either happy or angry. Inter-rater agreement in all cases was very high (median>64, p<10–4). Cartoons reprinted from Shutterstock.com under a CC BY license, with permission from Shutterstock.