Emotion Recognition Test

1. Wait for the images to load, then start when you are ready.
2. Look at the portrait. Choose the closest of the seven labels. Tap a button or use its numbered key.
3. Continue through the set. The labels stay in the same positions. There is no speed bonus, and answers are not revealed during play.
4. Read the breakdown. The server compares your choices with the intended portrayals. Leaving the tab during play ends the attempt so you can restart.

Use a clear screen, comfortable lighting and your usual glasses if needed. Phone and desktop have the same portraits and scoring, but image size can affect how easily you see a subtle expression. This visual task cannot be reproduced through image descriptions for screen readers.

Accuracy is the share of portraits for which you selected the intended label. Intended matters: these are constructed portrayals, and an expression can look ambiguous even when its creator had a label in mind. The categories are choices in this game, not a claim that all human emotions fit seven boxes.

Each category and intensity breakdown contains only a few portraits. Differences between them may reflect the particular pictures and identities, not a lasting personal strength or weakness. Neutral is a separate category, not a low-intensity version of every emotion. Repeating the small set adds familiarity and practice.

When enough eligible players have completed this version, we show a below, typical or above comparison band using other players' first runs of the same version, excluding flagged runs. Before that, the result shows how many eligible runs have been collected. These visitors are not a representative population sample. The band is not an emotional-intelligence or empathy score.

Schlegel and Scherer's short Geneva Emotion Recognition Test uses audiovisual portrayals with facial, bodily and vocal cues, covering a wider range of emotions [1]. It is a useful example of why emotion recognition is broader than labeling still faces. Our game uses only still portraits and is not the GERT-S.

Barrett and colleagues review the limits of inferring emotion from facial movement: expressions vary across people, situations and cultures, and the same movement can serve different purposes [2]. In everyday life, context and what someone tells you matter.

Our synthetic faces are not the validated research stimuli. We use no research face database or real photographs. We visually reviewed the portraits for artifacts and intended expressions, but the bank and its intensity labels have not been independently validated. Published research norms do not apply here.