In recent years, EFL teachers and linguists have evaluated the results of test materials to assess learners’ proficiency in languages. However, these approaches have several limitations, such as inappropriate test locations, lack of cost effectiveness, and insufficient test time. To address this issue, in the present study, the author proposes a novel and alternative method of determining L2 proficiency. Specifically, among 6 types of consonant clusters produced by Korean adult L2 learners, specific consonant clusters (e.g., voiceless stop+liquid) related to L2 proficiency were found, through which a series of equations were derived using discriminant analysis; furthermore, the participants’ scores in the onset clusters were applied to these discriminant equations to demonstrate the effectiveness of the ethod; the hit ratio of these equations for categorising L2 level was also examined. Depending on these induced equations, EFL teachers or researchers can assign a learner to an appropriate proficiency group. In future research, based on different data of EFL learners, the proposed approach can be meaningfully used to derive other useful discriminant equations.