Abstract: PURPOSE: Perceptual learning (PL), known as a basic and implicit learning form, has been demonstrated in almost all perceptual tasks, with pattern differed across tasks and learners. However, what determined learning ability and whether a general learning factor underlies this widespread plasticity remain largely unknown.
METHODS: We trained forty-nine subjects on seven different visual or audiovisual tasks in a continual learning paradigm. Training lasts for 35 sessions, with 5 sessions/task, allocated in separated days. Each session consisted of 7 blocks and each block has ~100 trials. In total, each subject completed ~24,500 trials. The training sequence was counterbalanced among subjects. Achievement of motivation, big-five personality, and non-verbal IQ were also assessed.
RESULTS: Learning varied with tasks, subjects and initial performance. Only extraversion was found to be correlated with average learning rate across tasks. We then developed a statistical model to dissect tasks-, subjects-, and initial performance- related components in learning ability, which explained 59.7% variance of the 343 drastically different learning curves. More importantly, the isolated estimators of subjects were now positively correlated with learners' score in non-verbal IQ test.
CONCLUSIONS: We successfully dissected learning ability and unveiled a general learning component that related to learner's inherent intellectual component in perceptual learning.

Key words: multi-task continual learning, general learning ability, IQ, initial performance