This paper investigates optimal designs for mixture experiments when there is uncertainty as to whether a polynomial regression model of degree one or two is appropriate. Three groups of novel results ...