這篇論文主要探討限定和非限定關係子句的結構，並以雙向優選理論作為理論架構，區別兩種關係子句的語意以及句型結構。本篇論文以傳統單向優選理論分析解釋客語名詞片語的基本結構，並進而主張傳統單向優選理論無法成功解釋限定和非限定關係子句的結構區別，而必須使用雙向優選理論方能將兩種關係子句與其限定和非限定語意做適當的連結。此外，本篇論文亦顯示由於雙向優選理論允許循環式的表格分析，兩種關係子句得以在同樣的限制條件下，將標記的句型連結標記的意義，未標記的句型連結未標記的意義。

This paper discusses the structure of two types of relative clauses, restrictive and non-restrictive, and provides an OT approach grounded in Weak Bidirectional Optimization to account for the syntactic construction of both types of clauses. The paper will show that while the Hakka basic NP structure favors the head to be positioned at the right edge, it is through the OT model of constraint interaction that different ordering patterns can be generated as grammatical outputs for a given input meaning. The first part of the analysis will explain the data by proposing constraints from the production-oriented perspective, as their ranking successfully yields the correct results in a unidirectional OT model. However, in the second part of the analysis, when the data grows more complex and requires syntactic and semantic distinctions between restrictive and non-restrictive relative clauses, the explanatory power of a unidirectional approach turns out to be limited. We find the necessity to adopt a bidirectional model, in which a form-meaning combination can be evaluated as a pair. Moreover, the fact that this innovative OT allows a recursive version of evaluation provides the possibility to generate optimal outputs successively for the two distinctive types of relative clauses under a single constraint ranking.