Abstract: Over the past decade, neuroimaging studies have provided detailed picture of the functional organization of the ventral visual pathway on object representation in the human brain. This study focused on the fundamental issue that lay in the neural representation in the cortex of the ventral visual pathway. The models in this field were firstly introduced in this study including the core idea of each model and the limitations of the models based on the comparison with each other, which took important roles in understanding the essence of the neural event in mapping the sensory information to the cerebral cortex in the ventral object-pathway. In detail, the modular representation opinion was firstly developed in this filed, many experimental evidences have been found to support the hypothesis of this model. For example, many studies have found the existence of the category specific brain areas in human brain including the fusiform face area (FFA), and the parahippocampal place area (PPA). However, other studies delivered some experiment findings which were not well interpreted on the aspect of the module representation model. Consequently, another model - distributed and overlapping representation model -was developed, which proposed that the voxel information content should be shared or distributed as a function of the shared attributes of objects, so similar object categories would share more voxels than dissimilar categories on the fact that the objects in these categories share more attributes. Although, many study findings supported the hypothesis of the distributed and overlapping model, moreover, the decoding method was raised based on the distributed and overlapping method in term of the machine learning method, which significantly influence the studies of the functional brain in neuroimaging field, it is still an important question how to reconcile the opinions of the module representation and the distributed and overlapping representation. In fact, many researchers have considered this question and verified it in the researches, in which the partial distributed representation model is one of them which tried to compensate the limitation of the distributed and overlapping representation. The partial distributed and overlapping model proposed that understanding partially distributed code is the critical to linking neural responses with the physical world they represent. Of course, it should be further studied that how the basic properties of the object was represented in the brain cortex, and how the neural network of the basic properties to combine the neural network of the whole object. The exploration of this issue may finally provide some solutions behind the category specific in object representation observed in cerebral cortex. On the other aspect, there exist a hierarchical structure in ventral visual pathway, and that the high level areas will modulate the neural activity underlying the object representation of the low level cortex. We suggested that the research of the object representation in ventral visual pathway cortex should consider the top-down modulation mechanism of the human brain. Especially, the lateral occipital complex cortex may take an important role in the top-down modulation process. Based on the consideration above, this study indicated that further origin of the neural representation in ventral visual pathway.

Key words: perception representation, perception modulation, fMRI, modular representation, distributed and overlapping representation

摘要： 大脑腹侧视觉通路知觉表征的神经机制是认知神经科学研究面临的基本问题。本文系统介绍了该问题研究中比较有影响力的理论模型，归纳分析了模型之间的分歧与各自的局限。文章分析指出大脑自上而下的调控机制是腹侧视觉通路神经表征机制问题研究的另一重要维度，如何有效整合知觉表征模型与大脑调控机制的相关研究是进一步深化知觉表征神经机制问题研究的关键。

关键词: 知觉表征, 知觉调控, 磁共振成像, 模块化表征, 分布式重叠表征