In the production of science, experimental design is one of the hardest steps in the chain. It’s really where a researcher sets himself up for success or failure. In “How To Avoid Seven Deadly Sins in the Study of Behavior” Manfred Milinski [Advances in the Study of Behavior, Vol. 26 (1997), pp. 159-180] offers his opinion on what he sees as the most common errors. He lists:
- Unjustified conclusions are made from observational (i.e., correlational) data
- Data are not independent (“pseudoreplication”)
- Treatments are confounded by time and sequence effects
- No effort is made to avoid observer bias.
- Potential artifacts arise when animals are not accustomed to experimental procedures
- Unsuitable controls are used
- An attempt is made to “prove” the null hypothesis with small samples.
A later article will explore why the layperson should also be aware of these sins before coming to conclusions about dog behavior.