Sei sulla pagina 1di 1

The Emotion Wheel A tool for the verbal report of emotional reactions

Tanja Bnziger, Vronique Tran, & Klaus R. Scherer


Introduction
The Emotion Wheel (see Fig. 1) was devised as a tool for the verbal report of emotions. It includes 16 emotion categories positioned in a circle. The 16 categories are ordered according to their postulated position in a 2 dimensional space. The two underlying dimensions are the level of perceived control in the situation that generates the emotion (vertical dimension) and the positive/negative (pleasant/unpleasant) quality of the situation and of the resulting feeling (horizontal dimension). Figure 1: Emotion Wheel
For each of the 16 categories, 4 adjectives with increasing emotional intensity are proposed. As shown in Fig. 1, the adjectives are not displayed but are represented by colored circles of different sizes. In the electronic version of the Emotion Wheel, the adjectives are displayed when the user is moving the cursor on top of the colored circles. The 64 adjectives proposed for each category in the original English version of the wheel as well as the 64 adjectives that are proposed for each category in a modified French version of the wheel can be obtained upon written request to the authors. (email: Tanja.Banziger@pse.unige.ch)

Results
Two Multidimensional scaling analysis were run on the similarity ratings (Task 1). Fig. 6 plots the emotional labels on the 2 dimensional solution obtained for the English labels (ordinal model, Euclidian distances, stress=.30, RSQ=.52). The results for the French labels (stress=.34, RSQ=.39) are displayed in Fig. 8. Figure 6: MDS analysis, Task 1 English version Figure 7: Average ratings, Task 3

Objectives
1. Verify the relevance of placing the emotion categories in a 2 dimensional space with underlying dimensions of 'control' and 'valence' 2. Verify that the adjectives fit in the postulated categories 3. Verify postulate of increasing intensity for the four adjectives in each category

Method
Three ratings tasks have been developed. A group of 28 native English speakers and a group of 31 native French speakers were recruited to complete the three tasks. Figure 2: Task 1 display Task 1 was to rate the similarity of the 16 emotion categories. The categories were presented in pairs on the computer screen (16 categories - 120 pairs). Participants reported a degree of similaritydissimilarity for each pair by moving the words apart (to indicate greater dissimilarity) or together (to indicate greater similarity). Fig. 2 shows the display that was presented to the participants. Figure 3: Task 2 display

Average ratings of valence and control obtained in Task 3 (grouped modalities) are represented on Fig. 7 (for English labels) and Fig. 9 (for French labels). Inter-rater reliabilities for all continuous ratings are displayed under those Figures. The variance for those ratings is substantial, especially for control ratings, and partly accounts for the low differentiation of negative labels on Fig. 7. The relative position of some labels in the space is shifted as compared with their predicted position on the Emotion Wheel (e.g. interest for the English version, soulagement and amour in the French version). The French category envie (theoretically equivalent to envy in the English version) was rated as positive by some participants (probably emphasizing its connotation with desire). The match between the dimensional solution based on the similarity ratings and the average ratings of valence and control is slightly better for English than for French labels. Figure 8: MDS analysis, Task 1 French version Figure 9: Average ratings, Task 3

Task 2 was to classify each of the 64 adjectives in the 16 provided categories. The categories were displayed along the sides of the computer screen in light yellow boxes (Fig. 3). The adjectives were sequentially displayed in the center of the screen. The participants used the computer mouse to drag and drop the adjectives into the yellow category boxes.
In Task 3, the participants of both linguistic groups were randomly split into two subgroups.

In Modality 1, 14 English speaking and 16 French speaking participants rated the 80 emotion categories/adjectives on 3 continuous scales: intensity, valence (positive vs negative quality), and control. The participants gave their answers for each emotion label on three sequentially presented visual analogue scales (see Fig. 4). In Modality 2, 14 English speaking and 15 French speaking participants rated the 3 continuous dimensions intensity, valence, and control in a two-step procedure. Figure 5: Task 3, Modality 2

Figure 4: Task 3, Modality 1

A large majority of adjectives were classified in the predicted categories by most participants (Task 2). Only 9 English labels and 8 French labels were not classified as predicted by a majority of participants. On the level of the 16 categories, there was substantial variation with only 55% of the adjectives (feeling distrust-4%, grudging-21%, envious-100%, jealous-96%) classified as expected into the category envy, and 94% of the adjectives (nervous-89%, apprehensive-93%, fearful-100%, panicked-93%) correctly classified into the category fear. For French labels, the category which contained the most ill-classified labels was contentement (with 52% correctly classified labels), the categories with the most well-classified labels were colre and peur (90%). The results obtained for the intensity ratings (Task 3) showed a less satisfactory pattern with respect to the predicted progressive increase of intensity for the 4 adjectives selected in each category. The average ratings only seldom presented the predicted trend.

Conclusions
To a large extent, the structure of the 'Emotion Wheel' (with underlying dimensions of control and valence) seems to match the spatial organisation of the categories in terms of similarity ratings and in terms of direct ratings of the underlying dimensions. The use of four adjectives with progressively increasing emotional intensity within each category is more problematic. It appears difficult to find such adjectives for all the categories. The theoretical advantage of including adjectives on the wheel is to provide a semantic field for the corresponding categories. But if users disagree on the belonging of some adjectives to postulated categories, this could in fact result in unexpected "blends" of the signification of the categories for different users.

(a) For each emotion category or adjective, participants had to increasedecrease the size of a red circle (Fig. 5) to report intensity. (b) They had to click on a colored circle defined by valence (horizontal dimension) and control (vertical dimension) to report their ratings of valence and of control for any given emotion label (Fig. 5).

Potrebbero piacerti anche