ABSTRACT
For a test-item pool to support a variety of applications, a necessary requirement is that the items cover its subject area for a broad range of difficulty. Given the heavy burden placed by the requirement on the resources typically available for item-pool development, the use of item cloning techniques has become attractive. For textual items, such as word problems and cloze items, item cloning may involve random selection from replacement sets for some insignificant elements of an item template (Millman and Westman, 1989) or the use of more advanced linguistic rules (Bormuth, 1970). The result of item cloning are sets of items differing in surface features only, which can be assumed to have similar psychometric properties. Such sets are referred to as item families.
