Abstract
We perceive objects in the world as stable despite their constantly changing image properties due to factors like occlusion, visual noise, and eye movements. Recent results demonstrate that perception of low-level stimulus features and even more complex perception of human identity are biased towards recent percepts (Fisher & Whitney, 2014; Liberman, Fisher & Whitney, 2014). This effect is often called serial dependence. However, the existence of serial dependence in high-level evaluations such as attractiveness remains unclear. Here, we tested the existence of serial dependence in perception of attractiveness of objects and humans. In experiment one, we presented a random series of images of oranges drawn from a linear morph varying from ripe (attractive) to moldy (unattractive). Participants were asked to report the attractiveness of the oranges in each trial using a Likert scale. Results showed that attractiveness evaluations were systematically biased toward the attractiveness of the oranges seen up to several seconds prior, which suggests the existence of serial dependence in aesthetic judgments. In a second experiment, we extended the experiment to facial attractiveness, a task humans perform on a daily basis. In particular, we also wanted to test whether sequential dependence of attractiveness only holds for images of varying attractiveness of the same face or also for images of different faces. In an online survey, we presented forty-six subjects with a random series of television screenshots featuring female actresses. Each subject reported the facial attractiveness of the actress for each screenshot using a Likert scale. Results show serial dependence predominantly for two sequential images of the same television anchor. This result suggests that serial dependence of attractiveness may be modulated by identity perception. The perceived attractiveness for one person is sequentially dependent--stable over time, whereas for different people we are sensitive to differing levels of attractiveness.
Meeting abstract presented at VSS 2015