The chief tool of test development is the test specification. Davidson and Lynch (2002) define test specifications as generative blueprints for test design. Test specifications for classroom use can be a simple and practical outline of your tests. However, for tests that are intended for large-scale use and distribution, the test specification is much more formal and detailed. This article will provide a rationale for the use of test specifications, describe several well-known models and discuss issues related to the development of test specifications.
An Historical Perspective
Test specifications, alternatively termed plan, guideline, form or rubric, are not a new concept. In fact the earliest mention of test specification in the educational assessment literature was by Ruch in 1929 (as cited in Gopalan & Davidson, 2000). It is widely believed that the term ‘specification’ was probably derived from the industrial concept of a ‘specification.’
Rationale for the Use of Test Specifications
Regardless what the test score is used for, good tests involve clear thinking and should be iterative, consensus-based, and specification-driven (Davidson & Lynch, 2002). Tests should be iterative in a sense where there are cycles of feedback over time as the test grows and evolves. Tests should also be consensus-based in that they should result from dialog and debate, not from a top-down dictate. It is also recommended that tests be specification driven whereby the specs are recipes for tests that foster dialog and discovery at a higher level.
Test Specification Models
Davidson and Lynch Model (2002)
According to Davidson and Lynch (2002) there is no single best format or magic formula for test specifications. In fact, there are innumerable ways to design one. Davidson and Lynch base their model on an earlier model developed by Popham (1978).
Their model has five components: