High-Level Influences on Low-Level Learning

描述（由申请人提供）：人类的能力对实践的效果有显著的反应。在实验室和自然环境中对练习效果的研究已经证明，人类在从运动到认知的任务中可以获得异常高水平的表现。在过去的二十年里，关于感知学习的快速增长的文献已经记录了一系列的任务，在这些任务中，对基本感知判断的练习可以产生显著的性能改进。这些研究表明，即使是人类信息处理的基本方面也可以大大改变，由此延伸，早期的感觉水平可能是高度可塑的。这一建议的具体重点是视觉任务，涉及刺激特征操纵在水平或接近观察者的检测阈值。虽然特定的任务可能需要分类、识别或辨别判断，但准确区分特征存在与否的需要是感兴趣的感知能力的一个基本方面；一个例子是熟练的放射科医生能够在x射线图像中检测和识别肿瘤的程度。关于这类技能的现有证据提出了三个假设：(a)知觉专门知识的获得和表达依赖于多层编码信息；(b)所有这些编码同时可用；(c)学习的时间过程涉及在越来越低的视觉系统层次上逐步加强编码。这些假设的检验将包括收集行为和神经生理学（EEG）数据，以及开发一种能够将这两种数据联系起来的计算建模方法。相关性：对需要检测和识别的感知能力的明确关注，在观察者必须获得可靠区分某些视觉属性存在和不存在的能力的领域中具有翻译应用的潜力，并且在观察者可获得的信息可能非常有限的情况下，根据该属性可靠地对物体进行分类或命名。例子包括针对视力障碍（如弱视）的行为干预，训练警察神枪手，或训练安全检查人员。

DESCRIPTION (provided by applicant): Human abilities are remarkably responsive to the effects of practice. Studies of the effects of practice, in both laboratory and natural settings, have documented the extent to which humans can obtain exceptionally high levels of performance in tasks ranging from motor to cognitive. Within the past two decades, a rapidly-expanding literature on perceptual learning has documented the range of tasks in which practice on basic perceptual judgments can produce dramatic improvements in performance. These studies suggest that even elementary aspects of human information processing can be substantially modified and, by extension, that early sensory levels may be highly plastic. The specific focus of this proposal is on visual tasks that involve stimulus characteristics manipulated at levels at or near observers' detection thresholds. While specific tasks may require categorization, identification, or discrimination judgments, the need to accurately distinguish between the presence and absence of a characteristic is a fundamental aspect of the perceptual abilities that are of interest; an example would be the extent to which a skilled radiologist is able to both detect and identify a tumor in an X-ray image. The available evidence on these types of skills suggests three hypotheses: (a) The acquisition and expression of perceptual expertise relies on multiple levels of encoded information, (b) All of these codes are available simultaneously, (c) The time-course of learning involves an incremental strengthening of codes at increasingly lower levels of the visual system. Tests of these hypotheses will involve the collection of both behavioral and neurophysiological (EEG) data, and the development of a computational modeling approach capable of linking these two types of data. Relevance: The explicit concern with perceptual abilities that require both detection and identification holds the potential for translational application in domains in which observers must acquire the ability to reliably distinguish between the presence and absence of some visual attribute, and to reliably categorize or name an object based on that attribute when the information available to observers may be extremely limited. Examples include behavioral interventions for visual impairments (e.g., amblyopia), the training of police sharp-shooters, or the training of security screeners.