''Suppose we were inspecting some newly discovered star, one that we could only see dimly because of some kind of hazy interference in outer space. Suppose further that through this haze we thought we could detect a system of planets orbiting the star. Then suppose that by some lucky circumstance the haze disappeared and our view of the star improved and that what we had previously thought was a planetary system could no longer be seen. Would we not feel justified in concluding that what we had thought was a planetary system was really only an artifact of the interference? We would feel this because we are suspicious of effects that depend on error [or stochasticity, risk or uncertainty]. If things that we think are there cannot be detected when the precision of our procedures improves [or if the imaginary richness assumptions of our theories fail], then we need additional evidence of their existence. Likewise, because the quantitative relationships that we think we can detect ... would disappear were error were eliminated, we require additional evidence of their existence. '' [Michell]