Abstract
The human visual system implicitly learns statistical regularities from the environment. Our previous study demonstrated that central tendency (prototype) and at least the first two principal components (eigenfaces) are more efficiently learned from a group of newly encountered faces than the actually studied faces (Gao & Wilson, 2013), which provides an efficient mechanism for encoding new facial identities at an individual level. However, it is not clear whether the prototype and eigenfaces also play an essential role in encoding familiar faces. In the current study, we investigate the effect of visual familiarity on the learning of the prototype and eigenfaces. Adult participants (N = 31, mean age = 20 ± 2.6 years, 13 males) studied 16 synthetic faces in four successive learning sessions. In each session, each face was studied for 20 seconds over four presentations. We measured participants memory performance after each learning session using an old/new recognition paradigm with the new faces sampled from an orthogonal volume of the face space relative to the studied faces. We also measured participants false memory for the unseen prototype face and eigenfaces of the first principal component of the studied faces after the first and the fourth learning sessions. Participants memory for studied faces improved from a moderate level (Hit = 0.60, FA = 0.22) after the first learning session to ceiling (Hit = 0.91, FA = 0.06) after the third learning session. However, the false recognition rates for the unseen prototype face and eigenfaces did not change between the first and the fourth learning sessions, and in both cases were higher than the recognition rates of the studied faces (ps<0.05). The results suggest that even with increased familiarity of the studied faces, prototype and principal components still play a crucial role in encoding individual facial identities.
Meeting abstract presented at VSS 2014