Shared and specific mechanisms of auditory and visual category learning

项目总结/摘要 学习新的感知类别的能力使一些最复杂的人类行为成为可能， 感知到视觉物体识别。知觉范畴机制的研究进展 学习依赖于这样一个基本假设，即这种学习的基础过程是共享的。 感官然而，这项工作的绝大多数都集中在视觉模态上。因此， 关于人类如何学会将复杂的听觉信息分类的研究在很大程度上依赖于 在视觉领域的研究结论，而不测试这一关键假设。但最近的 来自注意力文献的证据表明，即使是看似领域一般的认知过程，如 工作记忆是通过额叶皮层的感觉偏向区域完成的。目前的调查将 直接比较支持听觉和视觉类别学习的计算和神经机制， 在功能磁共振成像扫描仪中，对同一个体进行两种模式的分类训练。目标#1 调查将确定共享和感觉偏置电路支持反馈处理听觉期间 和视觉类别学习。如果支持感知类别学习的神经回路在整个大脑中共享， 在反馈处理过程中，预计将在类似的区域以类似的程度征聘。如果 相反，神经回路对于特定的模态是不同的，预期感觉偏向区域将 出现支持类别学习的听觉和视觉方式。目标#2将使用先进的机器 学习技术（多变量模式分类和代表性相似性分析）来表征 在学习过程中出现类别级神经表征。目标#3将确定 电路的功能和结构连接，因为它们有助于感知类别学习。的 拟议中的研究将直接检验关于这一复杂问题性质的基本假设， 影响日常行为。这项研究有可能影响对模态- 特定的学习能力可能受损，如语音学习和语言相关的障碍， 阅读障碍自闭症和特殊语言障碍拟议的研究将提供培训基础 支持PI的长期目标，即发展受限制的感知类别学习理论， 通过神经生物学和行为，并将指定行为，计算，和神经机制，这种 学习该项目提供了直接测试理解的关键假设的机会， 知觉类别学习拟议的研究将在特殊的培训环境中进行， PI将由一组知识渊博和有成就的科学家指导。该研究将提供 PI在功能性磁共振实验设计和分析的培训，这将为她做好准备， 在计算认知神经科学领域成为一名独立的科学家。

PROJECT SUMMARY/ABSTRACT The ability to learn new perceptual categories enables some of the most complex human behaviors, from speech perception to visual object recognition. Current understanding of the mechanisms involved in perceptual category learning relies on the fundamental assumption that the processes underlying such learning are shared across the senses. However, the vast majority of this work has focused on the visual modality. As a consequence, the research regarding how humans learn to group complex auditory information into categories has relied greatly on conclusions from the research in the visual domain without testing this critical assumption. However, recent evidence from the attention literature suggests that even seemingly domain-general cognitive processes, such as working memory, are accomplished via sensory-biased regions in frontal cortex. The current investigation will directly compare the computational and neural mechanisms supporting auditory and visual category learning by training the same individuals on categories in both modalities while in an fMRI scanner. Aim #1 of this investigation will identify the shared and sensory-biased circuits supporting feedback processing during auditory and visual category learning. If the neural circuits supporting perceptual category learning are shared across the modalities, it is expected that similar regions will be recruited to a similar extent during feedback processing. If instead, the neural circuits are distinct for particular modalities, it is expected that sensory-biased regions will emerge as supporting category learning for auditory and visual modalities. Aim #2 will utilize advanced machine learning techniques (multivariate pattern classification and representational similarity analyses) to characterize the emergence of category-level neural representations over the course of learning. Aim #3 will identify the functional and structural connectivity of the circuits as they contribute to perceptual category learning. The proposed research will directly test the fundamental assumption about the nature of this complex problem that affects everyday behaviors. This research has the potential to impact understanding of cases where modality- specific learning abilities might be impaired, such as phonetic learning and language-related impairments in dyslexia, autism, and specific language impairment. The proposed research will provide the training foundation to support the PI’s long-term objective of developing theories of perceptual category learning that are constrained by neurobiology and behavior and will specify the behavioral, computational, and neural mechanisms of such learning. This project presents the opportunity to directly test a critical assumption underlying understanding of perceptual category learning. The proposed research will take place in an exceptional training environment and the PI will be mentored by a team of knowledgeable and accomplished scientists. The research will provide the PI with training in functional magnetic resonance experiment design and analysis which will prepare her well for a career as an independent scientist in computational cognitive neuroscience.