A line of reasoning used in recent research on language acquisition assumes that a child acquiring the language has only two reliable sources of information available about the target grammar: a set of grammatical principles and the primary data of the language spoken around him. A third kind of evidence, negative evidence, would be helpful but is unavailable. However, with its current set of assumptions, this logic of language acquisition is not easily applicable to any problems involving overgeneralization in child language. Another kind of evidence, indirect positive evidence, is the combination of a principle stated in the form of a negative conditional or its equivalent, a disjunction, and data that allow such a principle to operate. This kind of evidence allows learners a way out of overgeneralizations, and is an indication of how positive evidence and principles of grammar can be turned around to substitute for the negative evidence that seemed to be the only answer to the problem of overgeneralization. (Author/MSE)