Symposium: Perceptual Learning

Organizer: Takeo Watanabe

List of confirmed speakers in alphabetical order:

Josh Gold, "Reinforcement, associative, and perceptual learning in a visual decision task"
Many perceptual tasks require the brain to weigh noisy evidence from sensory neurons to form categorical judgments that guide behavior. In my talk I will discuss recent experiments combining neurophysiology and behavior in monkeys that have begun to identify how experience can shape these decision mechanisms. For a visual motion discrimination task, improvements in perceptual performance, a phenomenon known as perceptual learning, does not appear to involve changes in the response properties of neurons that represent the sensory evidence. Instead, perceptual learning results from changes in how the sensory evidence is selected and weighed to form the decision. These changes are consistent with a reinforcement-driven process that first establishes the association between the neurons that represent the sensory evidence and those that prepare the behavioral response, and then further refines that association so that only the most informative sensory neurons drive the decision. I will discuss predictions, implications, and the generality of this scheme.
Zhonglin Lu, "Functions and Mechanisms of Perceptual Learning"
Perceptual learning – the improvement of performance through practice or training -- has been observed over a wide range of perceptual tasks in adult humans. The high degree of plasticity of the adult perceptual systems suggests that perception and perceptual learning cannot be studied separately. In this talk, I will review some major functions and mechanisms of perceptual learning, including specificity of perceptual leaning, the law of practice in perceptual learning, mechanisms of perceptual learning, the level and mode of perceptual learning, optimal training procedures, and computational models of perceptual learning. Studies of these various aspects of perceptual learning have greatly enhanced our understanding the information processing limitations of the human observer, and how the state of the observer changes with training, with strong implications for the development of potential noninvasive training methods for perceptual expertise in normal populations and for the amelioration of deficits in challenged populations.
Dov Sagi and Nitzan Censor, "Explaining training induced performance increments and decrements within a unified framework of perceptual learning"

Practicing sensory tasks could result in two main perceptual outcomes. The first, and more widely documented, is perceptual learning referring to long-lasting improvement of perceptual thresholds. The second is perceptual deterioration, which is observed when the number of trials is increased within a training session or between closely spaced sessions. Recent results with visual texture discrimination show that these two processes inversely affect each other: decremental effects interfere with further learning, while efficient short practice results in a long-term learning effect in which performance decrements are practically eliminated. Further results show that sleep is necessary to preserve learning effects following short training and facilitates the decay of deterioration that normally results from extensive training. We suggest a theoretical link between perceptual deterioration and learning, assuming a system with saturating connectivity, in which continuous learning leads to saturation unless connectivity is efficiently consolidated. Thus, best learning is achieved with short training sessions. Resistance to saturation is achieved by sleep-dependent consolidation of unsaturated connectivity. The different transfer properties of the performance decrements and increments allow us to identify local and global components of perceptual learning and their interactions. This suggests sleep-dependent consolidation mechanisms that induce modifications in higher brain areas that interact with local early visual networks to enable improvement of perceptual abilities.
Yuka Sasaki, "Perceptual learning and brain activation"

Practice and sleep improve perceptual learning (PL). However, the brain activation associated with practice and sleep has yet to be entirely clarified. Here, I will discuss our recent findings regarding brain activation related to PL of a texture discrimination task (TDT, Karni & Sagi, 1991) during practice and sleep, using fMRI.
First, we investigated whether activation of the primary visual cortex (V1) correlates with performance improvement of TDT. For the first week, we found that the activation increased as the performance improved. However, later, while the performance improvement retained, the increased activation disappeared.
Second, we investigated whether V1 activation is modulated during sleep. We found that the trained-region of V1 was enhanced compared to the untrained-region during sleep after TDT-practice. However, such regional difference was not found during sleep when TDT-practice did not precede sleep.
These results indicate two distinctive states of brain activation in different phases of the time course of PL. The first phase is characterized as gradual enhancement in both V1 activation and TDT performance. This occurs during both practice and the subsequent sleep. The second phase is characterized as retention of the enhanced performance without accompanying V1 activation.
Takeo Watanabe, "Perceptual learning as a result of an exposed feature being internally enhanced"

While it was once believed that no perceptual learning (PL) occurs without focused attention to a visual feature, we found that PL occurs on a feature that is task-irrelevant and subthreshold. (Watanabe, Nanez and Sasaki, 2001, Nature). We call this type of PL task-irrelevant PL. Later we have found that task-irrelevant PL occurs particularly when subjects are engaged in a task (Seitz and Watanabe, 2003, Nature). Here we present recent lines of research that has indicated that if a task-irrelevant feature is paired with reward, PL on the feature occurs without any task engagement (Seitz, Kim and Watanabe, 2009, Neuron) and when task-irrelevant PL occurs only on a weak feature (Tsushima, Seitz and Watanabe, 2008, Current Biology). Thus, there are at least three conditions in which PL on a feature occurs; (1) when the feature is task-relevant and subject to focused attention (Ahissar & Hochstein, 1992, PNAS), (2) when the feature is task-irrelevant and weak and, therefore, is not subject to attentional suppression (Tsushima, Sasaki and Watanabe, 2006, Science) and (3) when the feature is reinforced by reward. We conclude that irrespective of whether a feature is task-relevant or task–irrelevant, PL occurs when the feature is internally enhanced/reinforced.