ABSTRACT
When trying to give an intuitive explanation of correlation, one frequently resorts to formulations such as, "lt is a measure of the extent to which two sets of scores are in the same order." Of course, in the case of Pearson correlation, that is not exactIy what a correlation represents. A more accurate verbal description would be that it is a measure of the simiIarity of standard scores on the two variables, simiIarity being measured as the averege product of standard scores. However, even that description suffers from the usual discrepancy between a verbal description and the numerical operations that go into a formula. One of the advantages of ordinal approaches to correlation is that they provide a doser match between the intuitive verbal description of correlation as simiIarity of order and the numerical operations that go into the quantification of degree of relation, which is what we are trying to do with a correlation coefficient.
