Learning beyond the prototype: Implicit learning of principal components in dot patterns
43.32, Monday, May 13, 8:30 am - 12:30 pm, Royal Ballroom 6-8
Xiaoqing Gao1, Hugh Wilson1; 1Centre for Vision Research, York University
Humans have the ability to implicitly learn the central tendency of a group of visual objects (the prototype effect, Posner & Keele, 1968). A recent study (Gao & Wilson, 2012) demonstrated that in addition to the prototype, adults also implicitly learn the feature correlations that capture the most significant variations among faces as defined by principal components (PC). However, it is unclear if the implicit learning of PC is specific to faces. In the current experiment, adults (n=13) studied 16 patterns each consisting of 9 dots. The 16 patterns deviated from a prototype in a systematic way so that the first PC explained 50% of the total variance. After participants studied the 16 patterns for 40 seconds each, their memories were tested in a studied/novel recognition task with the 16 studied patterns plus 16 new patterns that deviated from the prototype in orthogonal directions to the studied patterns. Participants also gave studied/novel judgments to the prototype pattern, and two patterns that deviated from the prototype on the positive and negative directions of the first PC of the 16 studied patterns. All the patterns have the same distance from the prototype. Participants recognized 59% (above chance, p <0.01) of the studied patterns and misidentified 43% (below chance, p <0.05) of the new patterns. As would be explained by the prototype effect, they recognized the unseen prototype 80% of the time. Interestingly, they also recognized the two unseen patterns representing the changes on the first PC 71% of the time. The recognition rates for the prototype and the two PC patterns were all higher than for the studied patterns (ps <0.05). The results suggest the implicit learning of the prototype and the most significant feature correlations as defined by PC is a general mechanism in visual object recognition.