Abstract
We use a set of 161 images, corresponding to a total of 23 individuals, each displaying one of six emotions (anger, sadness, fear, surprise, happiness and disgust) in addition to neutral. All images were of 80x120 pixels. We extended this set by reducing all the images to 40x50 pixels, yielding a total of 322 images. We then designed a staircase procedure as follows. A fixation cross appeared for 500 ms, followed by a randomly selected image from our set. The image was first displayed for a total of t=50ms. After a 500 ms mask, subjects were instructed to respond to the perceived emotion. If the subject response is correct, the exposure time t for that particular emotion is decreased. Otherwise, it is increased. To determine these increments/decrements, we assume that the value of t will converge to its right value (i.e., after several trials, the value of t will oscillate about the time threshold required to achieve recognition). The results show that happiness is the fastest to be recognized (23-28ms) and that this value does not change as the image size decreases. Neutral, disgust and surprise form a second group requiring additional time (3 to 4 times longer than happiness) and with a minimal increase of processing time as the size of the percept is reduced. Fear requires the same time as this second group, but its processing time increases dramatically as the percept decreases in size. Finally, sadness and anger constitute the group requiring the longest for recognition – about 10 times slower than happiness. These results show that the recognition of emotions has evolved differently for distinct emotions, suggesting an adaptation to some evolutionary needs.
National Science Foundation.