This study illustrates how a synergy of two technologies--Intelligent Computer-Assisted Language Learning (ICALL) and corpus linguistic analysis--can produce a lasting improvement in L2 learners' ability to edit persistent grammatical errors from their writing. A large written English corpus produced by Korean undergraduate and graduate students enrolled at an American university was analyzed to determine four persistent grammatical errors in their writing assignments. An ICALL program was then designed to improve these L2 learners' awareness of these errors in texts and provide practice in correcting them. A pre-test/post-test experiment revealed a significant improvement in recognizing and correcting the four errors by Korean L2 learners who had taken the ICALL program over a matched control group that had received standard L2 writing instruction. This improvement held up on a second post-test administered five months later. The implications of the results for L2 instruction and the design of iterative ICALL tutors are discussed.