A conservation problem of numerical equivalence which 80% of adults reliably fail and 40% of third graders pass was developed, and responses of 188 subjects (Grades 2, 3, 5, 7, 9, 11 and college) to it and related number conservation and probability problems indicated that the differences in nonconservation were rooted in subjects' different evaluations of the relevancy of the conservation question and not as much in qualitative differences in inferential competence. Still, there was evidence that adults may make exactly the same kind of nonconservation errors as have been attributed to preoperational and concrete operational children. (Author)