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?