This article investigates an argument-adjunct asymmetry in English as a second language (ESL) learners' acquisition of inversion in "wh"-questions. A generative approach (DeVilliers, 1991; Stromswold, 1990) claims that inversion is acquired earlier in argument "wh"-questions than in adjunct "wh"-questions, the asymmetry resulting from their different structural properties. However, an input frequency approach (Rowland & Pine, 2000, 2003; Rowland, Pine, Lieven, & Theakston, 2005) proposes that individual "wh"-word + auxiliary combinations are acquired based on their input frequency, with no argument-adjunct asymmetry being predicted other than as an indirect consequence of a frequency difference. Two studies were conducted: one acquisition study with a grammaticality judgment task and one input study with a corpus analysis of the learners' input sources. The results show that learners are more successful at subject-auxiliary inversion in argument "wh"-questions than in adjunct "wh"-questions and that the argument-adjunct asymmetry cannot be traced to input frequency. The results of the studies here run counter to the input frequency account but conform to the prediction of a structure-based generative approach: There exists an argument-adjunct asymmetry in the acquisition of inversion in "wh"-questions because of the different syntactic properties of two types of "wh"-phrases.