This study investigates the effectiveness of a computer-assisted pronunciation training (CAPT) system on second language learners' acquisition of three grammatical features. It presents a CAPT system on top of a phoneme-based, fine-tuned speech recognition model, and is intended to deliver explicit, corrective feedback on the pronunciation of the Arabic feminine grammatical morpheme tied /taa/ in three Arabic phrases: the construct state, the definite noun-adjective, and the indefinite noun-adjective. A total of 26 American college students studying Arabic were assigned to two groups. The control group received the traditional in-class instruction while the experimental group used the CAPT system. Prior to the experiment, a pretest was administered to examine learners' pronunciation performance, and a posttest was carried out after the experiment. The results showed that the experimental group performed significantly better in the posttest than the control group, suggesting that the explicit, corrective feedback helped learners improve their pronunciation of the three phrases. It also showed that learners were more knowledgeable about the three phrases based on grammatical cues, suggesting that CAPT systems are a promising method for supplementing pronunciation-based instruction and improving awareness of grammatical features reflected in the phonology of a language.