We present a new feedback strategy implemented in a natural language generation-based e-learning system for German as a second language (L2). Although the system recognizes a large proportion of the grammar errors in learner-produced written sentences, its automatically generated feedback only addresses errors against rules that are relevant at the learner's current L2 acquisition stage. This approach is motivated by the results of two recent studies into German as L2 teaching in classroom situations. Both studies observed that the acquisition stages that L2 learners go through during L2 acquisition are similar to developmental stages for German as first language (L1), with only slight differences depending on the learner's native language (French or Italian). The individual acquisition stages often deviate from the organization of the L2 instruction lessons used in the classroom. They also found that attempts to teach L2 rules which surpass the learner's current acquisition stage are futile. Our system emulates the observed acquisition stages, and the feedback it provides only addresses errors that are 'teachable' at the learner's acquisition stage; the learner is not notified of errors that are beyond this stage. The computational approach to obtain the desired error-diagnostic behavior is based on underspecifications of the system's grammar rules. [For full proceedings, see ED565087.]