Korean is classified into classifier languages, which require a classifier regardless of whether nouns are count or mass, while English is a typical non-classifier language, where only mass nouns require classifiers when counting. While most of the previous studies have focused on the acquisition order of classifiers based on their semantic features (e,g, human, shape, function, etc.), it should be noted that the correct use of classifiers requires syntactic knowledge as well as semantic knowledge (K. Lee and S. Lee, 2009). However, little is known about the L2 acquisition of syntactic knowledge related to classifier construction such as EPP feature (Byun and Ha, 2015). Against this background, this study aims to articulate what must be acquired by English-speaking learners for the correct use of Korean numeral-classifier constructions, and to investigate whether the syntactic properties are acquirable for them. 85 adult English-speaking learners at four proficiency levels (low intermediate; high intermediate; low advanced; advanced) and 31 native Korean speakers performed an elicitation task. The findings show that the advanced group produced the correct classifier constructions in the majority of cases in a similar way to the native speaker (NS) control group, while the lower proficiency groups supplied classifiers significantly less frequently than the NSs. The overall group results suggest that highly proficient learners have successfully acquired the required syntactic knowledge, even though its complete acquisition is delayed until advanced stages of L2 development.