In the Atlanta example, we could reject the null hypothesis of “no ^eating” because the pattern of test results was so wildly improbable in absence of foul play. But how implausible does the null hypothesis ~are to be before we can reject it and invite some alternative explanation?