对客体进行范畴化（categorization）有助于个体用有限的认知资源处理来自世界的复杂多变的信息，其在认知系统中处于核心地位。范畴化能力在个体成长过程中的发展规律一直是研究者们关心的问题。直接考察范畴化知识的研究通常聚焦于3岁及以上的儿童，研究者们采用匹配样例任务和自由联想等任务来考察儿童语义概念的组织规律，发现儿童依据客体共现频率（即同时空出现）进行分类的现象，早于他们依据类别关系来组织客体概念。基于这些发现，研究者们提出了同一范畴客体在视觉输入或者语言输入中类似的出现/使用方式（两客体和同一客体分别都有直接的共现规律）是人们形成客体范畴概念的主要原因之一。但是，另一方面，在婴儿研究中，研究者们通过测量婴儿在习惯化范式中的注视时长、连续触摸任务中的触摸顺序等范式发现婴儿对物体的范畴/类别信息敏感。同时，近年来采用核磁共振技术也发现了婴儿大脑中高级视觉加工皮层上存在与成人相似的范畴效应，提示知觉系统中范畴加工机制的早期发育。这些结果表明，不同于行为范式中的范畴效应的滞后性，儿童的范畴化概念可能基于知觉系统的加工特性在个体发育早期就已经开始发展。因此，为了进一步考察儿童的范畴化概念系统的早期形成机制，我们从自然学习情景中词汇习得的角度出发，通过幼儿（16-36月龄）词汇学习发展规律来考察概念范畴化现象的发生机制。而且，基于Wordbank，我们同时考察了21种语言（5种语系）中儿童的词汇发展数据，以探究儿童词汇习得过程中范畴化效应的跨语言规律。

研究一，本研究在21种语言中分析了成人语义空间稳定表征的动物、身体部位、场景和小型可操作性人造物4个范畴是否在儿童学习词汇的过程中表现出范畴现象。我们通过计算每一种语言内每个词汇在不同月龄儿童中的习得人数百分比来刻画每个词汇的习得发展曲线，然后比较同一范畴内词汇间的习得发展曲线的相似度与不同范畴间词汇的习得发展曲线的相似度的差异。结果发现，本研究考察的4个语义范畴在21种语言中普遍存在范畴内词汇习得发展曲线的相似性显著大于范畴间词汇习得发展曲线的相似性的现象，即都存在范畴效应。进一步验证性分析排除词汇间语音相似性这一混淆变量，得到了一致的范畴效应。这说明，幼儿在词汇学习的早期阶段（16-36月龄）就已经反映出了语义概念的范畴效应。此外，我们还用logistic曲线拟合了不同范畴词汇的学习轨迹，获得了不同语言中4个范畴各自发展过程的特征，相比于其他范畴，身体部位范畴的习得年龄更小，而场景范畴的习得年龄更大，动物范畴的习得速率更慢。                                                                                                                                                                                                                                                                                                                                       

研究二，在观察到幼儿词汇学习过程中的范畴效应后，我们进一步考察词汇间共现导致范畴效应产生的假说。我们基于儿童在自然情景下的语言产出文本（CHILDES），计算得到了7种语言中词汇间共享共现矩阵。然后，我们考察研究一所发现的范畴效应在（这7种语言中）排除了词汇共享共现频率后，范畴效应是否依旧存在。结果表明，在所考察的7种语言中，4个类别的范畴现象在排除了词汇共现频率后都稳定存在。该结果说明，在幼儿词汇学习过程中表现出的范畴化现象，并不（完全）是由于儿童自我产出文本中同一范畴词汇相似的使用方式所导致的。

综上，本研究从儿童实际学习词汇的过程出发，相比于前人研究，每个范畴不仅涉及了大量词汇，而且在多个语言中（甚至包括一些少数语种）发现了幼儿在词汇学习早期就表现出概念语义上的范畴效应现象，即范畴内词汇对之间的学习发展过程显著比范畴间词汇对之间的发展过程更相似，并且该范畴效应在排除词汇对间的语音相似这一混淆变量后，在考察的动物、身体部位、场景和小型可操作性人造物4个范畴上依旧存在效应。进一步探究促进范畴现象发生的机制，发现排除了儿童自我产出文本中的词汇间共享共现频率后，词语学习发展曲线上的范畴化效应仍显著存在。结合前人研究成果中婴儿对客体范畴信息敏感的行为和神经证据，本研究中所发现的幼儿在语言学习过程中的范畴化效应，揭示了人类的范畴化概念体系的早期发展机制。本研究为今后研究儿童更接近真实环境的概念发展的组织方式提供了新的研究角度。

Categorization helps individuals process world complex information using limited cognitive resources, and it plays a central role in human cognitive system. The categorization development in individual growth has long been a focal point for researchers. Studies on categorical knowledge usually target children aged 3 and above. Researchers often employ match-to-sample task and free-association task to study how children organize their semantic concepts. They find that children classify objects based on their co-occurrence (i.e., simultaneous appearance) earlier than organizing object concepts based on their taxonomic relation. Based on these findings, researchers have proposed theories suggesting that the formation of different domains in human semantic concepts originates from their similar usage patterns in visual or linguistic inputs (having direct co-occurrence with the same objects). However, in infant research, researchers have discovered infants' sensitivity to object category/domain information by measuring their gaze duration in habituation paradigms and touch order in continuous touch tasks. Moreover, recent studies utilizing functional magnetic resonance imaging (fMRI) have revealed category effects in infants' higher visual processing cortex, similar to adults, indicating the early development of categorical processing mechanisms in the perceptual system. These results suggest that, unlike the delayed category effects observed in behavioral paradigms, children's categorization in conceptual system may gradually form during early development based on the processing characteristics of the perceptual system. Therefore, to further investigate the early formation mechanisms of categorization in conceptual system, we approach from the perspective of vocabulary acquisition in natural learning scenarios and examine the mechanisms of conceptual categorization phenomena through the word learning trajectory in young children (16-36 months). Additionally, based on Wordbank, we investigate the word growth data in 21 languages (5 language families) to explore the cross-linguistic regularities of categorical effects during children's vocabulary acquisition.

Study 1: We examined whether children vocabulary learning trajectory shows categorical effect in 21 languages for four domains: animals, body parts, scenes, and small manipulable artifacts with stable representations in adult semantic space. We characterized the acquisition developmental curve of each word by calculating the percentage of learners at different monthly age groups within a language and then compared the similarities between the words acquisition developmental curves within the same domain versus those between different domains. We found that the four domains investigated in this study generally exhibited significantly greater similarities in word growth curves within the same domain than between domains in 21 languages, indicating the presence of semantic categorical effect. Further validation analyses, which ruled out the confounding variable of phonetic similarity between different words, yielded consistent categorical effect. This suggests that young children (16-36 months) already show categorical effects in semantic concepts during the early stages of word learning. Furthermore, we fitted logistic curves to the learning trajectories of different category words and obtained characteristics of the developmental process of the four categories in different languages. Compared to other categories, the age of acquisition for body parts words was younger, while the age of acquisition for scenes words was older, and the acquisition rate of animal words was slower.

Study 2: Observing the categorical effects in young children word learning, we further investigated the hypothesis that lexical co-occurrence leads to the emergence of categorical effects. Based on children's language product corpus in natural situations (CHILDES), we calculated the shared co-occurrence matrix of different words in seven languages. Then, we examined whether the category effects found in Study 1 still exist after excluding the vocabulary co-occurrence frequency in these seven languages. The results indicate that the categorical phenomena of the four domains remain stable in the seven languages after excluding the vocabulary co-occurrence frequency. This suggests that the categorization phenomena manifested during young children's vocabulary learning are not caused by the similar usage patterns of vocabulary items within the same category.

In summary, based on the process of actual children word learning, this study investigated the semantic categorical effects on words acquisition. And compared to previous research, it not only involves a large number of words but also discovers the phenomenon of conceptual semantic category effects in the early stages of word learning in young children across multiple languages (including some minority languages). That is, the words growth curve within a category is significantly more similar than that between categories. Moreover, this categorical effect persists after ruling out the confounding variable of phonetic similarity and is still present in animals, body parts, scenes, and small manipulable artifacts four domains. Further exploration into the mechanisms promoting the emergence of category phenomena found that the categorical effects on word growth curves remain significant even after excluding the shared co-occurrence between words. Combining the behavioral and neural evidence from previous studies on infants' sensitivity to object category information, the categorization effects discovered in young children words learning process in this study reveal the early development mechanisms of the human categorization concept system. This research provides a new perspective for future studies on the organization of children's concept development in more naturalistic environments.