语言迁移（Language Transfer）指目标语言和其他已习得语言之间的共性和差异所造成的影响，一直是二语习得及外语教学等领域的研究热点。相较二语学习，三语学习中的语言间互动更为复杂，于是很多学者提出三语习得现象应作为一个独立的习得现象进行研究，三语习得成为近年来二语习得研究领域中较受关注和成果较为集中的研究方向。

语言相近模型是近年研究者最为关注的多语迁移模型之一。该模型由Westergaard（2017）在累积强化模型和类型优先模型的基础上提出，主要观点认为：第一，三语学习中，L1和L2都是语言迁移的来源。第二，抽象语言属性的相似性是迁移发生的主要原因，而非单纯的语言类型和语言距离。当某种语言输入中的特定属性表现出与先前所学语言的属性具有抽象的结构相似性时，就会产生语言迁移。第三，迁移可能是促进性的或非促进性的，取决于学习者能否正确分析Ln输入。目前，有关该模型的论证多以印欧语系语言的三语学习为研究对象，缺乏来自其他语言的证据,此外该模型作者对语言熟练度影响的阐释似乎仍值得思考。因此，本研究拟以德、英母语者习得汉语三语为例，对该模型的有效性进行检验，并在此基础上对语言熟练度及语言迁移的关系问题进行探索。

本研究采用在线组间实验的研究形式，以E-prime2.0为实验工具，以汉语话题-述题结构（Topic-Comment）及从句中的VO/OV结构设计实验材料，采用语法判断及改正任务对中级汉语水平的德-英-汉和英-德-汉三语学习者就语言相近模型的基本假设进行检验，考察语言熟练度与语言迁移的关系。结果发现：（1）两被试组在实验中的整体表现不存在统计学差异，被试同时受到一语和二语迁移的影响，一语和二语都是迁移来源，抽象语言属性的相似性是语言迁移的原因；（2）两组被试都受到语言负迁移的影响，正向迁移和负向迁移同时存在，被试对三语输入的错误感知是导致负向迁移产生的原因；（3）两组被试都在熟练度更高的题类中表现更好，但即使某句法结构已经被习得，仍会受到来自背景语言的非促进性迁移影响；（4）两组被试分别在不同题类中具有比较优势，熟练度更高的背景语言知识更容易在三语习得过程中进行迁移。

综上，本研究结果基本支持语言相近模型的假设，否定了母语因素假说、二语地位因素假说、类型优先模型等其他多语迁移模型。此外本研究还发现，随着三语熟练度的提高，学习者受到来自背景语言的负迁移影响越小，但这种影响不会消失；背景语言中熟练度更高的语言更容易进行语言迁移。以上结果一定程度上丰富了三语习得语言迁移机制的理论成果，对汉语三语教学实践具有一定的借鉴意义。

Language Transfer refers to the influence caused by the similarities and differences between the target language and other acquired languages, which has always been a research focus in the fields of Second Language Acquisition (SLA) and Foreign Language Teaching. Due to the fact that the interaction between three or more languages is much more complex than it is in SLA, researchers have begun to treat the Third Language Acquisition (TLA) as an independent phenomenon. Therefore, TLA has naturally become one of the most concerned and concentrated research areas in TLA in recent years.

The Linguistic Proximity Model is one of the key concerns among multilingual language transfer models, proposed by Westergaard (2017) on the basis of the Cumulative Enhancement Model and the Typological Primacy Model, which has three main hypotheses. First of all, both L1 and L2 are the sources of transfer in TLA. Secondly, the similarity of abstract language attributes is the main reason for the transfer, instead of the language type or the language distance. Language transfers occur when the specific language attributes in input show abstract structural similarities with the attributes of previously learned languages. Thirdly, transfers may be facilitative or non-facilitative, depending on whether learners can correctly analyze Ln inputs. However, there is currently little empirical support and a lack of evidence from non-Indo-European languages. Besides, Westergaard seems to ignore the further interpretation of the influence of language proficiency on transfers which is still worth thinking. This study intends to put the effectiveness of the model to test by taking German and English Chinese learners as an example. Based on this, the relationship between language proficiency and transfers is also expected to be explored.

This article examines the hypothesis of the Linguistic Proximity Model as well as discusses the relationship between the language proficiency and language transfers through a horizontal inter-group experiment paradigm among two groups of intermediate learners of Chinese as a third language (Group 1: German- English –Chinese learners; Group 2: English- German– Chinese learners) with the Grammatical Judgment and Correction Task on the platform of E-prime 2.0. The experimental materials consist of Chinese sentences constructed by 2 main structures. One is called Topic-Comment, and the other is the VO/OV structure of the clause. The result finds: (1)there is no statistical difference in the overall performance of the two groups in the experiment. Both the first language and the second language are the sources of transfer. The similarity of abstract language attributes is the main reason for transfers. (2) Both groups are affected by non-facilitative language transfer. The wrong perception of inputs of the third language could have led to the negative transfer. (3) Both groups perform better in VO part which they have better command. However, even if the certain syntactic structure of the target language has been acquired, it is still subject to non-facilitative transfers from other languages, which doesn’t stand for the statement that Westergaard held in his experiment about language proficiency. (4) Each group has a comparative advantage in a certain structure in the test. The background language with higher proficiency is more likely to transfer in TLA.

To summarize, the study is in support of the basic hypotheses of the Linguistic Proximity Model and denies the others such as the L1 Factor Hypothesis, the L2 Status Factor Model and the Cumulative Enhancement Model. Furthermore, it also finds that the higher the proficiency level of the target language, the less affected by the negative transfer from the background language, but this influence always exists. Apart from that, the background language with higher proficiency is more likely to transfer. These research findings enrich the theoretical results of the language transfer mechanism in TLA to some extent and have a certain significance of reference for the practice of TLA of Chinese.