Abstract
Minimization of the variance of the estimated slope of a response surface maximized over all points in the factor space is taken as the design criterion. Optimal designs under the criterion are derived for second-order polynomial regression over hypercubic regions.