This paper compares design and analysis techniques for mixtures. The main f
ocus is on comparing a statistical treatment taking the mixture aspect full
y into account and another where only a subset of the mixture variables is
considered both in the design phase and during analysis. Mixture techniques
are of special importance for product development and optimization of food
products and processes. The methods are tested on data from a baking exper
iment where three flours were mixed in different proportions and the differ
ent flours were baked in three different process conditions. The example is
of importance for understanding the relationship between flour quality, ba
king process and bread quality. The main conclusion is that treating the de
sign as a true mixture design and analyzing the data in terms of all variab
les involved is advantageous compared to a treatment where only a subset of
the mixture variables is considered. (C) 1999 Elsevier Science Ltd. All ri
ghts reserved.