Unifying Perceptual Learning

What is the relation between perceptual learning (PL) in basic sensory discriminations and in more complex tasks, including real-world learning tasks? Most recent PL work focuses on the former, using simple sensory dimensions and a few specific stimulus values. In contrast, other PL research and virtually all real-world tasks involve discovery of invariance amidst variation, and may also involve PL working synergistically with other cognitive abilities. In this talk I will suggest that, despite superficial differences, low- and high-level PL tasks draw upon—and reveal—a unified type of learning. I will consider several arguments that have been advanced in favor of confining perceptual learning to plasticity at the earliest cortical levels along with models of PL based on receptive field change vs. selection. These analyses do not support the idea of a separate low-level process but do support both the abstract character of PL and its dependence on unifying notions of discovery and selection. In the final part of the talk, I will relate this unified view of PL to direct practical applications. Learning technology based on PL modules (PLMs) can address elusive aspects of learning, including pattern recognition, transfer, and fluency, even in high-level, symbolic domains, such as mathematics learning.