For a long period, Computational Linguistics (CL) and Computer-Assisted Language Learning (CALL) have developed almost entirely independently of each other. A brief historical survey shows that the main reason for this state of affairs was the long preoccupation in CL with the general problem of Natural Language Understanding (NLU). As a consequence, much effort was directed to fields such as Machine Translation (MT), which were perceived as incorporating and testing NLU. CALL does not fit this model very well so that it was hardly considered worth pursuing in CL. In the 1990s the realization that products could not live up to expectations, even in the domain of MT, led to a crisis. After this crisis the dominant approach to CL has become much more problem-oriented. From this perspective, many of the earlier differences disadvantaging CALL with respect to MT have now disappeared. Therefore the revolution in CL offers promising perspectives for CALL.