Effects of Rewards on Perceptual Learning

描述（由申请人提供）：本提案的目标是了解奖励信号对低级视觉区域可塑性的影响。最近，已经发现，灵敏度增强不仅发生在主体执行任务的基础上的特征（任务相关学习），而且发生在与任务无关的呈现特征（任务无关学习）。一系列的心理物理学工作（Watanabe等人，2001，2002; Seitz & Watanabe，2003; Seitz等人，2004，2005 abc; Seitz和Watanabe，2005）指出，视觉学习是由内部因素（包括奖励）触发的扩散强化信号与来自呈现的视觉特征的自下而上的刺激信号之间相互作用的结果，而不管这些特征是任务相关的还是不相关的。在目前的建议，我们将测试的假设，任务无关的学习介导的奖励驱动的强化信号。首先，为了检验模型的有效性，我们将解决以下问题。(1)任务的处理是否是一种内部奖励？(2)这种强化信号是否调节低水平视觉区域的活动？(3)任务无关学习是否发生在任何预测试验结果的特征上？(4)在学习形式之后，是否以与强化学习相同的方式观察到具有预测错误的活动？其次，我们将使用外部液体奖励来直接测试奖励是否会导致感知学习。如果是这样的话，我们将改变奖励传递的时间和概率，以测试感知学习是否遵循强化学习中的规则。为了解决这些问题，我们将通过心理物理学测量行为表现，并通过功能性磁共振成像测量人类大脑的活动。据我们所知，没有研究直接研究奖励对低水平视觉加工和可塑性的影响，或者低水平视觉学习和强化学习之间的关系。因此，目前的研究是非常新颖的。同时，对结果的预测是可靠的，因为它们是在一系列心理物理研究的一致结果的基础上做出的。我们认为，R21是一个适当的轨道。拟议的研究具有临床应用的潜力，有助于科学知识，导致医疗工具的发展，以改善诊断和康复治疗，大脑疾病和病变。此外，研究结果可能有助于揭示学习缺陷的来源和治疗方法。

DESCRIPTION (provided by applicant): The goal of the present proposal is to understand effects of rewards signals on plasticity of low-level visual areas. Recently, it has been found that sensitivity enhancement occurs not only to features on the basis of which subjects perform tasks (task-relevant learning), but also to presented features that are irrelevant to the task (task-irrelevant learning). A on a series of psychophysical work (Watanabe et al, 2001, 2002; Seitz & Watanabe, 2003; Seitz et al, 2004, 2005abc; Seitz and Watanabe, 2005) indicates that visual learning occurs as a result of interactions between diffusive reinforcement signals triggered by internal factors including reward and bottom-up stimulus signals from presented visual features, irrespective of whether these features are task-relevant or irrelevant. In the present proposal, we will test the hypothesis that task-irrelevant learning is mediated by such reward-driven reinforcement signals. First, to test the validity of the model, we will address the following questions. (1), Does processing of a task work as an internal reward? (2), Does this reinforcement signal modulate activity in low-level visual areas? (3), Does task-irrelevant learning occur for any feature that is predictive of trial-outcome? (4), After learning forms, is activity with prediction errors observed in the same way as reinforcement learning? Second, we will use external liquid rewards to directly test whether reward leads to perceptual learning. If that is the case, we will vary the timing and probability of reward-delivery to test whether perceptual learning follows rules found in reinforcement learning. To address these questions, we will measure behavioral performance by psychophysics and activity in the human brain by means of fMRI. To our knowledge, no research has directly examined effects of reward on low-level visual processing and plasticity or relations between low-level visual learning and reinforcement learning. Thus, the present research is highly novel. At the same time, the predictions of the results are solid since they are made on the basis on consistent results of a series of psychophysical studies. We believe that R21 is an appropriate track. The proposed research has the potential for clinical applications by contributing to scientific knowledge leading to development of medical tools for improved diagnosis of, and rehabilitative therapies for, brain disorders and lesions. Moreover, the research results may help reveal sources of, and treatments for, learning deficits.