ABSTRACT
An experiment should involve the comparison of at least two sets of scores obtained under different experimental conditions. The idea is to vary the conditions (independent variable) and to examine the effects on the subjects’ behaviour; so we may need to compare two, three or even more samples in the course of an experiment; but what, then, would be the purpose of testing just one sample of scores? The answer is a simple one. Suppose we already know how subjects perform in a given condition and we wish to examine the effects of changing that condition. It would be tedious-and sometimes impracticable-to reassess performance in the basic condition when an established norm already exists. So we would simply test a random sample of subjects in the modified condition and compare their performance with the previously established norm. We know, for example, that children have a mean score of 100 on certain standardized tests of reading ability. If a new method of teaching is introduced, we can assess its effectiveness by comparing the reading scores of a random sample of children taught by the new method with the previous norm of 100. Thus a one-sample test could be used to determine whether the performance of the sample was significantly better than the well-established norm of 100. A significant difference would imply an effect of teaching method even though only one condition has been directly investigated.
