Experimental psychologists love a complete factorial design. Generating data in all possible treatment combinations provides a wealth of information. Complete factorials also offer practical advantages. One is ease of statistical analysis. With off-the-rack statistical programs, results are a few mouse clicks away. Graphs and tables almost make themselves. In a graph, every filled symbol has an open-symbol counterpart, every cell in a table is full. Aesthetic virtues are elegance and symmetry. For these reasons I love factorials as much as the next experimental psychologist – but only as a starting place. My ethics tale has to do with responsible experimental design and its real-world consequences.

Design Matters

Most work in our lab over the last 20 years has concerned individual differences in taste and their correlates. A key tool is selective breeding. Selectively breeding rats for high- and low-saccharin solution intake (respectively, HiS and LoS rats) ensures robust taste differences and thus is a powerful means of examining taste correlates in, say, emotionality and social status. Years back, we wanted to know whether stress would differentially affect HiS and LoS rats’ “jumpiness” (startle amplitude) and, if so, what neurochemical processes might be involved. We set about designing an experiment to find out, bearing in mind scientific objectives, practicality, and ethics. We began – of course – with a 2 × 2 × 2 complete factorial design. Filling out the design would mean collecting data in eight conditions. Detecting effects with rats generally requires 8–12 observations per condition, and none of the manipulations could be repeated on the same rats. Therefore, a complete factorial design would involve roughly 80 rats.