Long-term benefits of seeding the knowledge base.

Exposure to numerical examples (seed facts) produced a substantial long-term reduction in domain-specific innumeracy. In particular, learning the populations of 24 seed countries improved accuracy of estimates of the populations of 75 untrained countries, both at the time of learning and 4 months later. Consistent with abstraction-based theories of learning and memory, the benefits of having been exposed to the seed facts were as large 4 months after the exposure as immediately after it, despite the specific populations of the seed countries having been forgotten during the interval.