Department of Psychology, Massey University, Palmerston North, New Zealand
Commonly, when we wish to carry out an experiment, we decide on the number of subjects to use based on (a) how many subjects there are available, (b) the cost to us or to them of the test procedure, and (c) some gut feeling about how many we need in order to make the sort of decision that we hope to make. In many situations it is desirable to keep the number of subjects used as low as possible, especially when the test procedures are stressful or painful. The "gut-feeling" estimation method usually greatly overestimates the number of subjects needed.
If you both (1) know the mean and variability of your population, and magnitude of effect of interest and (2) are able to test your subjects one at a time, then you can more economically carry out experiments using many fewer subjects. From prior research or from control group data, we often have an estimate of mean and variance, and can estimate an important magnitude of effect.
Sequential sampling techniques were developed for quality control engineers and are rarely used by behavioural scientists. Sequential sampling is preferable to commonly-used procedures as it is more powerful in that fewer subjects (around 50%) are required in order to arrive at a decision with the same degree of certainty.
Sequential sampling involves collecting data from single subjects in succession, plotting the subjects' scores on a predetermined graph, and then stopping the study when that plot exceeds a set level. The technique is simple to use, readily available in computer statistics packages (e.g. Statistica by StatSoft), makes statistical analysis simpler, is more powerful, and uses fewer subjects. It is a more ethical alternative in contentious experiments, and it should be more widely known and used.