客体表征和客体关系表征都是知识表征中重要的部分，客体关系无法脱离客体单独存在，而客体关系对于客体的表征起到了重要的作用。目前对于客体表征的研究是基于客体的范畴 (category, Mahon & Caramazza, 2009)，范畴是一种描述客体分组的方法，分为分类范畴 (taxonomic category) 和主题范畴 (thematic category) ，两种范畴都是客体表征的重要依据。针对客体关系表征的研究大多针对范畴关系 (categorical relation) 和主题关系 (thematic relation) 展开，研究中对于范畴关系的定义是依据其分类关系 (taxonomic relation)。分类关系将具有相似的组成部分、相似感知觉特征、相似的功能特征的客体归为同类客体；主题关系将具有外部关联（但非同类）的客体定义为主题关系相关的客体，且主题关系包括了诸多子关系，例如时间关系（例如，春天和燕子），空间关系（例如，天空和飞机），和逻辑关系（例如，熊猫和竹子）等等。而目前对于客体关系的研究主要集中于分类关系与主题关系的对比。

本论文围绕客体关系表征展开，主要关注了两方面的内容： (1) 各类客体关系表征的规律和神经机制， (2) 客体类别对于分类关系和逻辑关系表征的影响和神经机制。针对这两个研究内容，共进行六个研究。

研究一和研究二针对分类关系表征和主题关系表征展开。首先，使用“目标客体驱动”的实验范式，对两种关系表征的粗加工任务和精加工任务分别进行研究，再通过个体在粗加工任务和精加工任务中的表现进行对比得到分类关系表征和主题关系表征的规律。随后，使用“目标关系驱动”的实验范式，对该规律进行验证。研究二在研究一的基础上，运用功能磁共振技术，考察分类关系和主题关系的精加工过程中，个体对分类关系与主题关系表征的神经机制。研究发现颞中回对于主题关系表征有特异性激活，推测主题表征可能需要更多的认知资源。

研究三和研究四，对主题关系子关系表征规律和神经机制分别进行研究。研究三共有四个实验，实验3-1和实验3-2分别考察了主题关系子关系的精加工任务和粗加工任务，通过对两个任务的对比，得出主题关系子关系的表征规律；实验3-3和实验3-4分别考察了分类关系子关系的精加工任务和粗加工任务，通过对两个任务的对比，得出分类关系子关系的表征规律。研究发现： (a) 主题关系子关系表征与分类关系子关系表征规律的差异在于：主题关系子关系的表征可能不存在粗加工的过程（即判断客体之间是否为主题关系的过程）。 (b) 主题关系子关系表征之间也存在差异，具体表现为主题关系子关系表征需要的反应时存在差异（反应时：时间关系 < 空间关系 < 逻辑关系），且个体对主题关系子关系进行选择性加工时，存在选择加工的优先级（优先级：时间关系 > 空间关系 ≥ 逻辑关系）。研究四在研究三的基础上，运用功能磁共振技术，考察主题关系子关系表征的神经机制，研究发现了时间关系和逻辑关系表征的特异性脑区。

研究五和研究六关注客体类别对于分类关系和逻辑关系表征的影响以及神经机制。实验5通过行为实验通过考察客体类型（自然物和人造物）对于分类关系和逻辑关系表征的影响。研究六在研究五的基础上，运用功能磁共振技术，主要关注自然物逻辑关系表征和人造物分类关系表征的神经机制。研究发现，自然物分类关系表征和人造物逻辑关系表征需要更少的反应时，而且自然物逻辑关系表征和人造物分类关系表征会激活更多的脑区，从而推测可能需要更多的认知资源。

本研究具有以下创新和贡献：首先，本研究通过定义关系加工的粗加工任务和精加工任务，一方面根据粗加工任务和精加工任务的对比，发现了主题关系子关系与分类关系子关系表征规律的差异；另外一方面根据精加工任务的结果，发现了主题关系和分类关系表征规律的差异，和主题关系子关系表征规律的差异；其次，本研究首次将主题关系细化分为时间关系、空间关系和逻辑关系，并首次从行为学和影像学两方面系统地考察个体对三种关系表征规律和神经机制的异同，为今后客体表征和客体关系表征的研究提供了新的研究方向、角度和思路；第三，本研究深入了客体类型对分类关系和逻辑关系表征的影响和神经机制。首先通过行为实验验证了客体类型对分类关系表征和逻辑关系表征的影响，从而发现自然物分类关系表征和人造物逻辑关系表征需要更短的反应时；然后，通过使用神经影像学技术，揭示了自然物逻辑关系表征和人造物分类关系表征的神经机制，并对已有客体表征理论进行补充。

Concept representation and conceptual relation representation are essential parts of knowledge representation. Studies on concept representation mostly built on the domain-specific hypothesis (Mahon & Caramazza, 2009) that the organization of conceptual objects in the brain relates to its semantic categories, mainly taxonomic categories (e.g., cats and dogs are both animals). Thematic relation, characterized by externality, in that the two concepts arise in the same scenario or event, is also a meaningful semantic relation. There are a few thematic subtypes, such as temporal relation (e.g., morning and coffee), spatial relation (e.g., sky and plane), and argumentative relation (e.g., panda and bamboo), to name a few. However, most recent studies focus on the dissociation of thematic relation and taxonomic relation only.

This thesis focuses on two parts: (a) the conceptual relation representation and its neural mechanism and (b) the influence of object types on conceptual relation representation. The thesis consists of six studies.

Study 1 and Study 3 investigated the potential processing differences between thematic and taxonomic relations and thematic and taxonomic subtypes. We defined a rough process and a precise process for each conceptual relation. By contrasting the rough and the precise processing tasks performance, we had three main findings. (a) A rough thematic relation identification takes longer than a rough taxonomic relation. (b) The rough thematic relation is represented in the forms of thematic subtypes instead. (c) Among thematic subtypes, argumentative relations take longer to process than spatial relations, and temporal relations take the shortest.

In Study 2 and Study 4, we investigated the neural mechanism of conceptual relation representation. We found that the thematic relation identification task required the additional recruitment of the left middle temporal gyrus. Our findings also show different neural networks representing temporal relation and argumentative relation.

Study 5 and study 6 focus on how object types affect taxonomic and argumentative relation representation. To start with, we used a knowledge-generation task. We investigated the ratio of three subtypes (thematic, taxonomic, and attributive) of knowledge representation within living and non-living objects. We found conceptual features and functional features were dominant features for living and non-living entities, respectively. Furthermore, we investigated how object types interact with taxonomic and argumentative representation. Our results showed that taxonomic relations were faster identified for living objects than non-living objects, whereas argumentative relations were faster identified for non-living objects than living objects. Further, with an fMRI experiment, we found that the activation of the left inferior frontal gyrus (IFG), the left middle frontal gyrus (MFG) and the left inferior parietal lobule (IPL) mediated living objects’ argumentative relation representation and non-living objects’ taxonomic relation representation, further suggesting that the unfamiliar relation representation requires more cognitive control than that of familiar relations.

Overall, the contribution and novelty of the thesis are as follows. Firstly, with a series of behavioural experiments, we investigated various conceptual relations, including the thematic relation (and its subtypes) and the taxonomic relation (and its subtypes). Secondly, thematic subtypes were found to require different cognitive control and have distinct neural bases. It suggested that future studies should pay more attention to thematic subtypes rather than the rough thematic relation. Finally, we investigated how object types interacted with taxonomic and argumentative relation representation and revealed the related neural basis. The approaches and results presented in this thesis provide new insights for future studies in conceptual relation representation.