Amundson, Ronald It is commonly held that scientists do not generate hypotheses randomly, but rather with the goal in mind of solving some scientiﬁc problem. Now, if a purposively generated hypothesis has a greater chance of scientiﬁc success than a randomly generated one (a supposition we must fervently hope is true) then Condition 2 fails fully to be met. There are degrees here, of course. Perhaps “insightful” hypotheses are only slightly more likely than random ones to be successful, and only a tiny bit of the success of science is to be explained by the insights of scientists. Selection would in this case retain much of its force. But if “insightful” hypotheses are much more likely to be successful, selection is much eroded.