It is now relatively straightforward to program software to detect words and phrases containing errors in the free-form writing of L2 learners of English. This article, however, reports on progress in the development of software which not only detects errors but also provides feedback explaining the nature of each error and how to correct it. Such feedback was prepared from a corpus of approximately 257,000 words and phrases containing errors either found in the writing of Spanish speaking A2 and B1 level learners, or errors, which, while not actually found, were typical of learners at this level of competence. For a sample of 68 student compositions comprising a total of 12,063 words, the software delivered error-specific feedback on approximately 46 per cent of the errors in the students' work. Adding feedback on a further 103,000 words and phrases to the diagnostic corpora subsequently increased the error-specific feedback rate to 59 per cent. The feasibility of further increasing this percentage is outlined, and the pedagogical significance of the findings discussed.