本文以Chomsky (2013) 的詞組標籤運算 (labeling algorithm) 為基準，探討論旨準則 (θ-criterion) 在當代句法理論中的必要性。本文從英語出發，指出句法理論中詞組標記系 統與論旨準則的功能有所重疊：經由討論日語論元重複 (argument doubling) 結構，我們 發現論旨準則在語料實證上確有不足。這顯示論旨準則應從當代句法理論中剔除，由標 籤運算取代其功能。

Chomsky’s (2013) labeling algorithm plays a fundamental role in the pursuit of minimalist syntax, restricting the structures that free Merge generates. I suggest that it also explains the effects of the θ-criterion to the extent that they are correct. This leads to the conclusion that the θ-criterion, as it is currently formulated, should be dispensed with. I first note that the θ-criterion and the labeling algorithm overlap in their effects in English. Then, I introduce the labeling mechanism for Japanese that I argued for in Saito (2014). It assumes that the language lacks φ-feature agreement altogether and claims that suffixal Case serves to aid labeling. With this background, I discuss examples of argument doubling from Kuroda (1988), where a verb assigns a single θ- role to two distinct arguments. I show that these and other similar examples are successful with labeling but violate the θ-criterion. Their grammaticality demonstrates then that the θ-criterion should be eliminated in favor of labeling. At the end, I examine the conditions on argument doubling and suggest that one of the doubled arguments must be construed as a focus and the other must serve to specify the set of alternatives for the focus in the sense of Rooth (1992). Based on this, I speculate that though the θ-criterion is untenable, Full Interpretation requires that each constituent play a unique role in the semantics.