After Darwin, a large number of studies on recognition of facial expressions of emotions have been carried out and some important models have been proposed. In such early studies by Feleky, Langfeld and others, the experimental paradigm itself had a serious problem. Pointing out the problem of those studies, Woodworth proposed his own model, and then, Schlosberg expanded Woodworth's model into a dimensional model for recognition of facial expressions of emotions. Since then, many studies confirmed the idea of the dimensional model, and recently, Russell proposed a circumplex model of affect. In contrast, Ekman and his colleagues developed a categorical model of facial expression based on their cross-cultural studies. This categorical model and the dimensional model offer different views of emotion and have been heated debates between them. Recently, Takehara and his colleagues found out a fractal structure in the recognition of facial expressions of emotions. The concept of fractal is novel in psychology and seems to offer an important addition to the research on recognition of facial expressions of emotions. In this article, I survey models for facial expressions of emotions and discuss in particular the importance of the fractal model.