This paper is intended for designers of computer-assisted language learning (CALL) programs. It discusses the problems engendered by CALL materials for learners from a pedagogical viewpoint and suggests a rather idealistic solution.

The first part of the paper criticizes the CALL programs available, commercially or otherwise, then puts forth five basic principles for construction of worthwhile instructional CALL programs.

In the second part, the author relates how he applied the five principles in designing a CALL exercise called GENDERS, which is intended to teach the genders of French nouns. The program produced a disconcerting effect: working through the exercise made many learners feel they were less sure of the gender of the nouns that the system proposed than they had been before the exercise. As an explanation for this negative effect, the author suggests that formal learning actually harms the acquisition process, at least in this case. As a French speaker’s skill with grammatical gender seems to be an aspect of rule-governed behavior, one possible corrective strategy would be to approximate to a model of the competence of a 7-year-old native speaker.

In conclusion, this paper points out a relatively important pedagogical problem. This part of the discussion is pertinent. However, it seems rather naive to suggest that a model of a 7-year-old native speaker could be easily approximated and that this model would be easier to approximate than that of a 10-year-old or even an adult.