Abstract
When responses are symmetrically distributed about the regression line, there are intuitively appealing reasons for expecting reduction in the variability of the responses from better experimental control to yield responses which cluster more closely about the regression line. This expectation is not always justified but may be appropriate for special cases.