Considerations Regarding an Experimental Study of Peaked and Unpeaked Tests
- Author(s):
- Lord, Frederic M.
- Publication Year:
- 1951
- Report Number:
- RM-51-15
- Source:
- ETS Research Memorandum
- Document Type:
- Report
- Page Count:
- 5
- Subject/Key Words:
- Difficulty Level, Test Forms, Test Items, Test Reliability
Abstract
A "peaked" test is a test composed entirely of items at (or near) a single difficulty level. It may be considered as a recognized fact that such a test will have optimal properties whenever the test is to be used solely to divide examinees into "passing" and "failing" or "accepted" and "rejected" groups. Brogden, Cronbach and Warrington, Gulliksen, and others have produced theoretical considerations leading to the conclusion that in many other typical test situations the peaked test will have more over-all reliability or validity than will the usual unpeaked test. RB-51-13 and RB-51-24 deal at some length with certain aspects of this problem. During the past several months considerable thought has been given to an attempt to plan an empirical study that would lead to fairly conclusive results regarding the relative over-all reliability of peaked and unpeaked tests in certain situations. It was finally decided that even a carefully designed study would probably lead to somewhat inconclusive results. Consequently, plans for conducting an empirical study have been dropped for the present. Some of the more important problems considered in this connection are recorded here for reference in case the desirability of such a study is reconsidered at some future time.
Read More
- Request Copy (specify title and report number, if any)