A half-test reliability measurement can be performed by splitting the test questions into two groups, such as odd-numbered questions and even-numbered questions, and then comparing the results. The scores obtained from the two groups are correlated, and if the scores obtained from both halves demonstrate a high positive reliability coefficient, or correlation, it can generally be assumed that the test is reliable.
Reliability in a testing procedure refers to the degree of consistency that can be expected when the test is administered under varying circumstances. In addition to the split-half reliability method of measuring test consistency, other measurement procedures include the inter-observer, test-retest and parallel-forms methods.
Before relying upon a test to asses important research hypotheses and questions, researchers will subject their tests to one or more reliability measurement methods to help ensure the accuracy and consistency of the results.