In a basic sense, reliability refers to the reproducibility of test scores. Imagine repeating a test over and over again and the only thing that changes with each administration is the selection of items on the test. Each set of items (i.e. each test form) represents a different way of measuring the same construct. They all conform to the same test specifications and measure the same content, but the specific items are different. We expect scores to change a little with each test form because of the different items on each one. However, we hope they do not change very much because that would adversely affect measurement of the construct and greatly increase the chance of making incorrect decisions about examinee performance. The extent to which scores from each form are similar is reflected in the reliability coefficient. If reliability is high, we can be assured that test scores from one test form will be similar to score from another. Reliable test scores are reproducible test scores.