In this paper, we introduce elements from the research we have done in connection with the design and experimentation of many hours of “classical” computer-assisted language learning (CALL) tutorials. We try to rely on learners’ spontaneous analyses and intuitions about language. We will show examples of learners’ messages, in particular in answer to “open” questions and of the way they can, or cannot, be interpreted or technically dealt with easily. Some messages show inadequate personal interpretations, misconceptions, circular explanations, or inconsistencies (sometimes apparent). Some difficulties seem to be induced by previous teaching, usual terminology, or inadequate phrasing of the instructions or questions. Various aspects of possible reactions of the CALL materials to errors will also be presented. We think that learners often have very adequate intuitions about language but they approach it with their words, their logic, their lack of scientific experience. Their capacities are often underestimated, but one should never overestimate their ability to behave like trained scientists.