Some believe that longer multiple choice tests tend to be more reliable because more items automatically reduce the error of measurement.  Indeed, a sufficient number of items must be included to cover the content areas tested; however, there are other factors that contribute to how efficiently a test measures and separates candidate ability. 

The indication of reliability is candidate separation reliability, calculated as [SD² - SE²/ SD²] using Rasch logit measures.  This reliability index is appropriate because the goal of any certification examination is to distinguish between those candidates who are worthy of passing and those who are not.  The better the test items distinguish among candidate abilities, the less measurement error there is in the examination and the higher the candidate separation reliability, regardless of the number of items.  For example, deleting poorly performing items makes the test shorter, but also increases the accuracy of measurement because the items producing the most error are eliminated from the examination. 

In this simple study, we looked at test length compared to test reliability. Six different exams were included. The table shows the tests in reliability order along with the number of items in each test.  The number of items and reliability do not appear to be related. A test of 211 items had a reliability of .93 while another test of 233 items had a reliability of .78.  It appears that the quality of the items is as important, or even more important, than the absolute number of items on the test. Of course, a reasonable number of items must be included to insure accurate measurement and content coverage.

   Table of Test Length and Reliability