EXPERIMENTAL DESIGN 213 



Laboratory biologists on the whole have been slow to adopt formal 

 experimental design. One reason is that laboratory biologists customarily 

 "design" their experiments without realizing it. It happens that experi- 

 ments with microorganisms, small animals, and single plants can easily 

 provide much information, so that very simple experimental designs are 

 adequate. One is not likely to despair over the failure of an experiment 

 to provide significant results if the experiment only cost an hour, one 

 cake of yeast, and a dime's worth of chemicals. In most experimental 

 laboratories, the experiments produce results faster than the experiment- 

 ers can calculate what the data mean. The experimental biologist, how- 

 ever, is rapidly reaching the point where the remaining problems are the 

 very difficult ones, and it may be necessary to use more complex ex- 

 perimental designs. 



The following sections briefly describe some of the more commonly 

 used experimental designs. The biologist is wise to seek the counsel of 

 a statistician instead of attempting to design his own experiments. The 

 statistician is likely to be able to foresee the failure of an experiment if 

 inadequate replications or repetitions are performed, but he might also 

 estimate that a proposed number of replications is more than enough and 

 therefore wasteful. 



Sampling and randomization 



It should be obvious by now that the organisms used and the measure- 

 ments made in an experiment represent samples of much larger popula- 

 tions. The selection of individual animals or plants to be used in an 

 experiment is thus an exercise in sampling. Regardless of the experi- 

 mental design, very careful consideration must be given to the choice of 

 samples because they must be random selections from the population. 



It might seem a simple matter to pick a few rats from a cage full of 

 rats and then to assign half of these to serve as controls and half to re- 

 ceive the experimental treatment. Unfortunately, it is virtually im- 

 possible to avoid personal bias. Maybe the rats selected are the ones that 

 were easiest to catch. If half the rats are to be controls and half are to 

 receive the experimental treatment, how does one decide which rats 

 receive which treatment? In which group do you place the one rat that 

 managed to bite your finger? 



The purpose of randomization is to minimize the natural and unavoid- 

 able bias. In the selection of the rats it would be safest to let the flip of 



